Luigia Binda-Learning From Failure - Long-Term Behaviour of Heavy Masonry Structures - WIT Press (2008) PDF

Learning from Failure
Long-term Behaviour of Heavy

Masonry Structures
WITPRESS
WIT Press publishes leading books in Science and Technology.
Visit our website for the current list of titles.
www.witpress.com
WITeLibrary
Home of the Transactions of the Wessex Institute, the WIT electronic-library provides the
international scientific community with immediate and permanent access to individual
papers presented at WIT conferences. Visit the WIT eLibrary at
http://library.witpress.com
International Series on Advances in Architecture
Objectives
The field of architecture has experienced considerable advances in the last few
years, many of them connected with new methods and processes, the development
of faster and better computer systems and a new interest in our architectural heritage.
It is to bring such advances to the attention of the international community that this
book series has been established. The object of the series is to publish state-of-the-
art information on architectural topics with particular reference to advances in new
fields, such as virtual architecture, intelligent systems, novel structural forms, material
technology and applications, restoration techniques, movable and lightweight
structures, high rise buildings, architectural acoustics, leisure structures, intelligent
buildings and other original developments. The Advances in Architecture series
consists of a few volumes per year, each under the editorship - by invitation only -
of an outstanding architect or researcher. This commitment is backed by an illustrious
Editorial Board. Volumes in the Series cover areas of current interest or active
research and include contributions by leaders in the field.
Managing Editor
F. Escrig
Escuela de Arquitectura
Universidad de Sevilla
Spain
Honorary Editors
C. A. Brebbia P. R. Vazquez
Wessex Institute of Technology Estudio de Arquitectura
UK Mexico
Associate Editors
C. Alessandri K. Ishii
University of Ferrara Yokohama National University
Italy Japan
F. Butera W. Jger
Politecnico di Milano Technical University of Dresden
Italy Germany
J. Chilton M. Majowiecki
University of Lincoln University of Bologna
UK Italy
G. Croci S. Snchez-Beitia
University of Rome, La Sapienza University of the Basque Country
Italy Spain
A. de Naeyer J. J. Sendra
University of Ghent Universidad de Sevilla
Belgium Spain
W. P. De Wilde M. Zador
Free University of Brussel Technical University of Budapest
Belgium Hungary
C. Gantes R. Zarnic
National Technical University of Athens University of Ljubljana
Greece Slovenia
K. Ghavami
Pontificia Univ. Catolica, Rio de Janeiro
Brazil
This page intentionally left blank
Long-term Behaviour of Heavy
Masonry Structures
Editor:
L. Binda
Politecnico di Milano, Italy
Editor:
L. Binda
Politecnico di Milano, Italy
Published by
WIT Press
Ashurst Lodge, Ashurst, Southampton, SO40 7AA, UK
Tel: 44 (0) 238 029 3223; Fax: 44 (0) 238 029 2853
E-Mail: witpress@witpress.com
http://www.witpress.com
For USA, Canada and Mexico
WIT Press
25 Bridge Street, Billerica, MA 01821, USA
Tel: 978 667 5841; Fax: 978 667 7582
E-Mail: infousa@witpress.com
http://www.witpress.com
British Library Cataloguing-in-Publication Data
A Catalogue record for this book is available

from the British Library
ISBN: 978-1-84564-057-6
ISSN: 1368-1435
Library of Congress Catalog Card Number: 2007922340
The texts of the papers in this volume were set

individually by the authors or under their supervision.
No responsibility is assumed by the Publisher, the Editors and Authors for any injury and/
or damage to persons or property as a matter of products liability, negligence or otherwise,
or from any use or operation of any methods, products, instructions or ideas contained in the
material herein. The Publisher does not necessarily endorse the ideas held, or views expressed
by the Editors or Authors of the material contained in its publications.
WIT Press 2008
Printed in Great Britain by Athenaeum Press Ltd.
All rights reserved. No part of this publication may be reproduced, stored in a retrieval
system, or transmitted in any form or by any means, electronic, mechanical, photocopying,
recording, or otherwise, without the prior written permission of the Publisher.
Contents
Preface ............................................................................................................ xiii
Chapter 1
Failures due to long-term behaviour of heavy structures........................... 1
L. Binda, A. Anzani & A. Saisi
1.1 Introduction ............................................................................................. 1
1.2 The collapse of the Civic Tower of Pavia: search for
the cause .................................................................................................. 2
1.2.1 Description and historic evolution of the tower ......................... 3
1.2.2 First experimental results and interpretation
of the failure causes .................................................................... 4
1.2.2.1 Structure and morphology of the walls....................... 4
1.2.2.2 Geotechnical investigation.......................................... 6
1.2.2.3 Physical, chemical and mechanical tests
on the components ...................................................... 7
1.2.2.4 Compression tests on masonry prisms........................ 8
1.2.3 Long-term tests ........................................................................... 9
1.2.3.1 Fatigue tests ................................................................ 9
1.2.3.2 Constant load tests ...................................................... 11
1.3 Long-term behaviour of masonry structures ........................................... 12
1.3.1 Deformation during mortar hardening........................................ 12
1.3.2 First, secondary and tertiary creep in rock and
hardened masonry....................................................................... 15
1.4 Collapse and damage of towers due to long-term heavy loads............... 16
1.4.1 St. Marco bell-tower and St. Maria Magdalena tower
in Goch ....................................................................................... 16
1.4.2 The bell-tower of Monza Cathedral and the Torrazzo
of Cremona ................................................................................. 16
1.5 The role of investigation on the interpretation of the
damage causes ......................................................................................... 17
1.5.1 The bell-tower of the Cathedral of Monza ................................. 18
1.5.2 The Torrazzo of Cremona........................................................ 21
1.6 Comparison between the two towers ...................................................... 25
1.7 Conclusions ............................................................................................. 26
Chapter 2
Experimental researches into long-term behaviour of
historical masonry........................................................................................... 29
A. Anzani, L. Binda & G. Mirabella Roberti
2.1 Introduction ............................................................................................. 29
2.2 Tests on the masonry of the Civic Tower of Pavia ................................. 31
2.2.1 Characterization by sonic tests ................................................... 33
2.2.2 Monotonic tests on prisms of different dimensions ................... 33
2.2.3 Fatigue tests ................................................................................ 35
2.2.4 Creep tests on prisms of 300 300 510 mm........................... 36
2.2.5 Pseudo-creep tests on prisms of 100 100 180 mm............... 39
2.2.6 Pseudo-creep tests on prisms of 200 200 350 mm............... 39
2.3 Tests on the masonry of the crypt of the Cathedral of Monza ................ 42
2.3.1 Preparation of prisms of 200 200 350 mm........................... 42
2.3.2 Characterization by sonic tests ................................................... 44
2.3.3 Monotonic tests........................................................................... 45
2.3.4 Fatigue tests ................................................................................ 45
2.3.5 Creep test on one prism of 300 300 510 mm ....................... 48
2.3.6 Pseudo-creep tests, first series .................................................... 48
2.3.7 Pseudo-creep tests, second series ............................................... 50
2.4 Comments................................................................................................ 52
Chapter 3
Collapse prediction and creep effects............................................................ 57
P.B. Loureno & J. Pina-Henriques
3.1 Introduction ............................................................................................. 57
3.2 Short-term compression: failure analysis and collapse
prediction using numerical simulations .................................................. 58
3.2.1 Experimental results ................................................................... 58
3.2.2 Continuum model ....................................................................... 59
3.2.3 Particle model ............................................................................. 62
3.2.4 Discussion of the results ............................................................. 63
3.3 Long-term compression: experimental assessment................................. 66
3.3.1 Tested specimens........................................................................ 66
3.3.2 Standard compression tests......................................................... 68
3.3.3 Short-term creep tests ................................................................. 69
3.3.4 Long-term creep tests ................................................................. 71
3.3.5 Discussion of the results ............................................................. 74
3.4 Conclusions and future work .................................................................. 78
Chapter 4
Effects of creep on new masonry structures................................................. 83
N.G. Shrive & M.M. Reda Taha
4.1 Introduction ............................................................................................. 83
4.2 The step-by-step in time approach to modeling
time-dependent effects ............................................................................ 84
4.3 Case 1: An axially loaded column .......................................................... 85
4.3.1 Creep model................................................................................ 85
4.3.2 Effect of coupling creep and damage in concentrically
loaded columns........................................................................... 89
4.3.3 Examining the effect of rehabilitation ........................................ 91
4.4 Case 2: Composite structural element subject to bending ...................... 92
4.4.1 Development of model ............................................................... 92
4.4.2 Application to a beam................................................................. 97
4.4.3 Masonry walls subject to axial load and bending....................... 103
4.5 New mathematical approaches to modeling creep.................................. 103
4.6 Discussion ............................................................................................... 104
4.7 Conclusions ............................................................................................. 105
Chapter 5
Experimental study on the damaged pillars
of the Noto Cathedral ..................................................................................... 109
A. Saisi, L. Binda, L. Cantini & C. Tedeschi
5.1 Introduction ............................................................................................. 109
5.2 The collapse and the decision for reconstruction.................................... 109
5.3 On-site investigation on the remaining parts of the collapsed
pillars....................................................................................................... 110
5.3.1 Layout of the section and of the masonry morphology.............. 111
5.3.2 General characterisation of the materials ................................... 111
5.3.3 Damage description .................................................................... 114
5.3.4 Laboratory testing....................................................................... 114
5.3.4.1 Mortars........................................................................ 115
5.3.4.2 Stones.......................................................................... 115
5.3.4.3 Injectability tests ......................................................... 117
5.3.5 On-site tests ................................................................................ 117
5.3.5.1 Flat-Jack tests.............................................................. 117
5.3.5.2 Application of sonic pulse velocity test to
pillars........................................................................... 118
5.3.6 Design decisions ......................................................................... 119
5.3.7 The dismantling of the remaining pillars.................................... 120
Chapter 6
Monitoring of long-term damage in long-span
masonry constructions.................................................................................... 125
P. Roca, G. Martnez, F. Casarin, C. Modena, P.P. Rossi,
I. Rodrguez & A. Garay
6.1 Introduction ............................................................................................. 125
6.2 Monitoring and long-term damage.......................................................... 125
6.3 Role of monitoring in the study of ancient constructions ....................... 127
6.4 Monitoring: methodology and requirements........................................... 128
6.4.1 Technology ................................................................................. 128
6.4.2 Distinction between dynamic and static monitoring .................. 129
6.4.3 Requirements .............................................................................. 131
6.5 Measuring damage and deformation related to historical
or long-term processes ............................................................................ 133
6.5.1 Monitoring and long-term damage ............................................. 133
6.5.2 Structural deformation................................................................ 133
6.5.3 Tensile damage in arches and vaults .......................................... 135
6.5.4 Damage of compressed members ............................................... 135
6.5.5 Fragmentation............................................................................. 139
6.6 Structural modelling and monitoring ...................................................... 140
6.7 Case studies ............................................................................................. 141
6.7.1 Dynamic monitoring of Mallorca Cathedral .............................. 141
6.7.2 S. Maria Assunta Cathedral, Reggio Emilia, Italy ..................... 145
6.7.3 Vitoria Cathedral ........................................................................ 148
6.8 Conclusions ............................................................................................. 151
Chapter 7
Modelling of the long-term behaviour of historical
masonry towers ............................................................................................... 153
A. Taliercio & E. Papa
7.1 Introduction ............................................................................................. 153
7.2 A continuum damage model for masonry creep ..................................... 154
7.2.1 Unidimensional viscoelastic model with damage ...................... 154
7.2.2 Three-dimensional viscoelastic model with damage.................. 157
7.2.3 Identification of the model parameters and comparisons
with experimental results............................................................ 160
7.3 Structural analyses of two masonry towers............................................. 166
7.3.1 The Civic Tower of Pavia........................................................... 166
7.3.2 The bell-tower of Monza Cathedral............................................ 167
7.4 Remarks and future perspectives ............................................................ 171
Chapter 8
Repair techniques and long-term damage of massive
structures ......................................................................................................... 175
C. Modena & M.R. Valluzzi
8.1 Introduction ............................................................................................. 175
8.2 The bed reinforcement technique............................................................ 176
8.3 The experimental campaigns................................................................... 178
8.3.1 Laboratory tests on the use of stainless steel bars ...................... 179
8.3.2 Laboratory tests on the use of CFRP bars and thin strips........... 183
8.4 Case studies ............................................................................................. 197
8.4.1 The bell-tower of the Basilica of S. Giustina in Padua .............. 197
8.4.2 The pillars of S. Sofia church in Padua ...................................... 199
8.4.3 The bell-tower of S. Giovanni Battista Cathedral
in Monza (Milan)........................................................................ 199
8.5 Final remarks........................................................................................... 201
Chapter 9
Simple checks to prevent the collapse of heavy historical structures
and residual life prevision through a probabilistic model .......................... 205
L. Binda, A. Anzani & E. Garavaglia
9.1 Introduction ............................................................................................. 205
9.2 The safety of ancient towers ................................................................... 205
9.2.1 A survey on Italian cases ............................................................ 206
9.2.2 Comments on the observed crack patterns ................................. 206
9.2.3 Elaboration of the collected data ................................................ 209
9.3 A probabilistic model for the assessment of historic buildings .............. 210
9.4 Fragility curves from the experimental data ........................................... 215
9.4.1 Fragility curve F versus applied to creep tests ...................... 215
9.4.2 Comparison between vertical and horizontal strain-rate ............ 215
9.4.3 Fragility curve F versus applied to pseudo-creep tests ......... 216
9.4.4 Comparison between vertical and horizontal strain-rate ............ 218
9.5 Application to the bell-tower of Monza .................................................. 219
9.6 Conclusions ............................................................................................. 221
Conclusions...................................................................................................... 225
Preface
On March 17 1989, the Civic Tower of Pavia collapsed without any apparent warning
signs killing four people. Subsequently, L. Binda, together with four colleagues
from DIS, Politecnico of Milan, was nominated a member of a Committee that had the
aim of helping the Prosecutor of the Procura della Repubblica in Milan find the
causes of the collapse. After an experimental and analytical investigation lasting
nine months, the collapse cause was found. Progressive damage dating back many
years, due mainly to the heavy dead load put on top of the existing medieval tower
with the addition of a massive bell-tower in granite, was to blame.
This type of long-term behaviour of masonry structures was not as well researched
as it was for concrete and steel structures and for rocks. Experimental research
aimed at showing the reliability of this interpretation was carried out, and is still
continuing, that is more than sixteen years of research since 1989. After careful
interpretation of the experimental results, also based on experiences from rock
mechanics and concrete, the modelling of the phenomenon for massive structures,
such as creep behaviour of masonry, was implemented by collaboration with E.
Papa and A. Taliercio from the same department.
Other case histories were collected such as the collapse of the Sancta Maria
Magdalena bell-tower in 1992 in Dusseldorf, the damage to the bell-tower of the
Monza Cathedral, Italy, and to the Torrazzo in Cremona, Italy. Later on, in 1996 the
collapse of the Noto Cathedral, Italy, showed that similar progressive damage can
take place in pillars of churches and cathedrals.
Collaborations on the topic first started with the University of Padua (C. Modena)
and later on with the University of Minho, Portugal (P. Lourenco). Then the University
of Calgary, Canada (N. Shrive) and the University of Barcelona (P. Roca) were
involved.
The editor would like to thank the technicians Mr Antico, Mr Cucchi and
Mr Iscandri for their collaboration in the experimental research and Mrs C. Arcadi
for her help in the editing of the chapters.
The Editor
2007
CHAPTER 1
Failures due to long-term behaviour of

heavy structures
L. Binda, A. Anzani & A. Saisi

Department of Structural Engineering, Politecnico di Milano,
Milan, Italy.
1.1 Introduction
The authors interest towards the long-term behaviour of heavy masonry struc-
tures started after the collapse of the Civic Tower of Pavia in 1989, when L. Binda
was involved in the Committee of experts supporting the Prosecutor in the trial,
which involved the Municipality and the Cultural Heritage Superintendent after
four people died under the debris of the tower.
The response required by the Committee concerned the cause of the failure;
therefore an extensive experimental investigation on site, in the laboratory and in
the archives was carried out and the answer was given within the time of nine
months. Several hypotheses were formulated and studied before finalizing the
most probable one, from the effect of a bomb to the settlement of the soil caused by
a sudden rise of the water-table, to the effect of air pollution, to the traffic vibration
and so on.
Several documents were collected concerning the sudden collapse of other tow-
ers even before the San Marco tower failure and the results of the investigation
were interesting. In fact, the failure of some towers apparently happened a few
years after a relatively low intensity shock took place. In other cases, the collapse
took place after the development of signs of damage, such as some crack patterns,
for a long time. This suggests that some phenomena developing over time had prob-
ably to be involved in the causes of the failure, combined in a complex synergetic
way with other factors.
As for the experimental investigation carried out on some prisms cut out from
the large blocks of the collapsed walls of the Pavia tower found on the site, the
2 Learning from Failure
attention was more and more concentrated on the dilatancy of the masonry under
compressive monotonic and creep tests and on the fatigue behaviour of masonry
under cycling loads.
This chapter discusses the investigation carried out on the materials of the Civic
Tower of Pavia and the conclusion reached by the previously mentioned Commit-
tee. Furthermore, the phenomena of early and retarded deformations of historic
masonry structures will be described together with the results of an investigation
carried out on other damaged structures.
Finally the research campaign carried out on site and in laboratory on the bell-
tower of the Cathedral of Monza and the bell-tower of the Cathedral of Cremona.
The investigation shows that the damaged state of the structures or of structural
elements can be precociously detected by the recognition of the typical crack pat-
terns, based on simple visual investigation.
Collapses may be prevented by detecting the symptoms of structural decay, par-
ticularly the crack patterns, through on-site survey, monitoring the structure move-
ments for long enough periods of time, choosing appropriate analytical models
and appropriate techniques for repair and strengthening the structures at recog-
nized risk of failure.
1.2 The collapse of the Civic Tower of Pavia:

search for the cause
The Civic Tower of Pavia, an eleventh-century tower apparently made of brickwork
masonry, suddenly collapsed on 17 March 1989 (Fig. 1.1). Several hypotheses were
Figure 1.1: The ruins after the collapse, seen from the arcade opposite to the
Cathedral.
Failures due to Long-Term Behaviour of Heavy Structures 3
made about the causes of that sudden failure, from soil settlements to the presence
of a bomb, from vibrations caused by traffic to the passage of super sonic jets.
For a thorough understanding of the real causes of the collapse, an experimental
investigation was carried out on site and in the laboratory, on the large amount of
material coming from the remains of the tower.
1.2.1 Description and historic evolution of the tower
The tower, about 60 m high with a square base measuring 12.3 12.3 m was
located close to the north-west corner of the Cathedral. Each of the four facades
was divided horizontally into six orders (Fig. 1.2a and b). The first four from the
bottom were divided into five parts by four pilaster strips topped by two small
arches. The third and fourth orders had no pilaster strips, but were topped by simi-
lar hanging arches. The fifth order terminated in a cornice. A large mullioned win-
dow with two apertures opened out on each side of the sixteenth-century belfry.
Inside the tower two timber floors were situated at a height of approximately 11
and 23 m.
(a) (b)
Figure 1.2: (a) The Civic Tower and Cathedral of Pavia, Italy. (b) Geometry of the
Civic Tower and Cathedral of Pavia, Italy.
According to the few historical documents found, the first order and half of the
second order can be dated between 1060 and 1100 AD [1, 2], the part from the
middle of the second order and the fifth perhaps were built between the twelfth and
thirteenth centuries; the tower was surmounted by a brick belfry and a timber roof.
Between 1583 and 1598 the granite belfry weighing 3,000 tons, designed by the
famous architect Pellegrino Tibaldi was set on top of the tower.
A staircase built into the wall ran along all four walls from the south-west corner
up to the belfry. The staircase was covered by a small barrel vault apparently made
of conglomerate.
1.2.2 First experimental results and interpretation of the failure causes
The few documents available at the time of the collapse [3] were insufficient to
give an accurate geometric configuration of the tower.
Consequently, in order to draw prospects and sections of the tower the following
operations, described in detail in [4, 5], were carried out:
topographic survey of the remains of the tower (Fig. 1.3), and partial rectifica-
tion of existing photographs to define the precise plan and the thickness and
morphological features of the cross-section of the masonry;
reconstruction of the geometry of the belfry from a survey of the granite parts,
practically all recovered from the internal portion of the remaining part of the
tower;
assessment of the overall height of the tower from an existing aerial photogram-
metric survey;
perspective plotting from existing photographs to reconstruct the geometry of
the staircase and the arrangement of the architectural elements.
1.2.2.1 Structure and morphology of the walls

The medieval walls, built according to the techniques normal at that time for tow-
ers, were characterized by two external brick cladding ranging from 120 to 400 mm
Figure 1.3: Photogrammetric survey of the remains of the tower.

with an average of 150 mm, with the intermediate portion of the walls consisting
of irregular courses of large pebbles of brick and stones alternated with mortar,
constituting a sort of conglomerate (Fig. 1.4). The walls of the second building
phase were characterized by a much more irregular filling and by thinner external
facings.
Figure 1.5 shows one of the large blocks among the remains revealing part of
the section of the wall with the external cladding. Figure 1.6 shows a complete
cross-section of the wall of the present remains of the tower (south side), the ratio
between the thickness of the external leaf of the wall and the internal one was
approximately 1:16.
The section of the wall near the staircase was composed by an external wall
similar to the one described above, but 1400 mm thick, a stairwell 800 mm wide,
and an internal wall 600 mm thick. The latter wall was of the rubble type and was
particularly heterogeneous.
Figure 1.4: Cross-section of the wall of the Civic Tower of Pavia.
Figure 1.6: View of the complete

section of the bearing wall. Note
Figure 1.5: Part of the section of the how thin the external facing is in
bearing wall (2.8 m thick), showing the comparison with the total thickness
external brick cladding. of the wall.
1.2.2.2 Geotechnical investigation

The remains of the lower part of the tower left standing reached a height from 0.1 to
5 m still visible at the moment, since it was decided to leave the remains as they were
without reconstructing the tower. The outside main wall is continuous and does not
show any appreciable signs of dislocation or displacement from its original position.
This, as well as the behaviour of the tower over time (there is no evidence of any
specific surveys, but no appreciable settlement appears ever to have been reported),
suggests that the collapse cannot be attributed to failure of the foundation soil. At
most, possible differential settlements, caused by abnormal variations in the ground-
water level in the previous years, could have worsened the stressstrain distribution
within the structure, leading to the collapse. The settlements were, however, very
limited and their effect is considered to be negligible. These qualitative consider-
ations have been confirmed by calculation [5].
The soil consists mainly of sandy deposits, sometimes silty or with lithoid ele-
ments, intercalated with highly impermeable strata of clayey, sandy silt. The most
important clayey, silty strata are found between 7.5 and 10 m, 13.5 and 15 m and
between 29.5 and 32 m below ground level.
These are over consolidated materials with a medium to low degree of plasticity.
On-site and laboratory geotechnical surveys were carried out to obtain the
mechanical parameters of the soil [5]. The on-site survey consisted of two geog-
nostic drillings in which undisturbed samples were taken and measurements made
with a standard penetrometer (SPT); two seismic cone penetration tests (SCPT)
and four cone penetration tests (CPT) were also carried out. The samples taken
during the drillings were subjected to identification, three-axial and oedometric
compressibility tests.
The penetrometric resistances give a similar picture of the pattern of the resis-
tance of the soil. The values measured are as follows: from the base of the tower to
a depth of approximately 14 m Nspt ranging from 8 to 30 blows/foot, Nscpt ranging
from 5 to 23 blows/foot, Qc ranging from 4.0 to 9.0 N/mm2; from 14 m down to
the maximum depth reached, Nspt ranging from 34 to 65; Nscpt ranging from 26 to
56; Qc ranging from 14.5 to 28.0 N/mm2.
The shearing strengths were determined: (i) for sandy soils on the basis of the
correlations presented in the literature between Nspt or Qc, effective vertical pres-
sure sv and friction angle j ( was subsequently suitably reduced to take into
account the presence of silt); (ii) for silty soils by means of laboratory tests (three-
axial tests and direct shear tests).
To calculate the bearing capacity, for safetys sake, the soil from a depth of 4 to
14 m was considered. This depth range showed mean j values of 34 and 33,
respectively, depending upon the correlations adopted.
To get at least an indicative value of the ultimate capacity, the foundation was
initially considered in the two extreme situations of a continuous beam with width
equal to 2.8 m (thickness of the foundation walls) and of a square foundation with
a side equal to 12.3 m (base of the tower).
By adopting the smallest shear resistance angle j = 33, and prudently assuming
the groundwater to be at the foundation level, an ultimate capacity of 2788 kN/m2
and the second case 4583 kN/m2 was calculated. The unit load on the soil was
1161 kN/m2 and the safety factor was therefore 2.4 and 3.95, respectively. The
effective safety factor will lie between these and even the lower value can be
considered sufficient to guarantee the stability of the foundation.
In the period from January 1987 to February 1989, the maximum measured
variation in level was 400 mm.
Even though there are no reasons to believe that the variations around the tower
were greater, the effect of an abnormal drop in level of 3 m was examined. The soil
was considered deformable down to the depth at which sv the variation in pore
pressure stale is about 0.2 of the sv geostatic pressure. The average settlement
calculated was 8 mm. Since the ground around the tower is relatively uniform, it
must be assumed that the differential settlements are negligible. In order to evalu-
ate the maximum theoretical distortion possible, penetrometric profiles were cal-
culated at opposite sides using all the minimum and maximum values recorded
during the various tests at various depths. Maximum settlements of 11 mm and a
minimum of 6 mm were obtained. The ultimate differential settlement would, there-
fore, be 5 mm and consequently of negligible effect on the stressstrain condition
within the structure.
1.2.2.3 Physical, chemical and mechanical tests on the components

To determine the effect of any possible chemical or physical degradation of
the masonry, numerous samples of mortar were taken from the large blocks of
masonry. The bricks and stones showed no signs of degradation except in the out-
ermost area; in fact, even in the most deteriorated areas of the examined blocks,
the degradation did not penetrate any deeper than 80100 mm.
Chemical and mineralogical/petrographic analyses were performed on 22 sam-
ples of mortars. The chemical analyses revealed that the binder used for the mor-
tars during the first building phase consisted chiefly of lime putty (soluble silica
0.280.40%) and that the aggregate was mainly siliceous (unsoluble residue
between 69.94 and 82.04%). The binder/aggregate ratio varied from 1:3 to 1:5.
Similar values were obtained for the mortars of the second and third building
phases. The porosity was around 1213% and the bulk density about 18.5 kN/m3.
In most cases, the sulphur trioxide content was negligible (around 0.06). Optical
inspection of thin sections of the mortar revealed numerous porous areas which
were sometimes covered by a layer of carbonates of relatively recent formation,
thus making the surface of the mortar far more resistant. This could be the result
of calcareous matter being deposited by flows of water. Similar deposits have been
found in different areas of the masonry [6] and in each case the covering layer
strengthened the surface of the mortar.
Thin section mineralogical/petrographic analysis also confirmed the total car-
bonation of the mortars and the siliceous nature of the aggregate and revealed
corrosion along the surface of contact between certain aggregates (pebbles of
stained quartz and flintstone, etc.) and the binder. As it is quite common [7], how-
ever, the reaction products cause no fissures inside the mortars. The adhesion
between mortar, bricks and stone was also fairly good (except in cases where the
building techniques had left large voids).
The mortars were consistent, as the mechanical tests confirmed, had a low con-
tent of sulphates and did not show heavy deterioration except for the outermost
ones.
The possibility of any significant reduction in structural strength of the masonry
due to the chemical or physical degradation of the mortars or other materials was,
therefore, excluded.
Since the collapse was not caused by the degradation of the building materials
or sudden or differential settlement, attention was turned to how dead and live
loads might have affected the mechanical behaviour of the materials over time.
Compression tests were performed on small cubes of mortar [5] (with sides
ranging from 2.7 to 3.5 mm) taken from the mortar joints of the inner conglomer-
ate. The strength was 2.9213.37 N/mm2, with a mean value of 6.45 N/mm2 and
SD 49% (n = 11). Since the specimens are very small these results are merely
indicative. Nevertheless it can be said that the results confirm the chemical and
physical analyses; in general, the mortar was consistent despite its heterogeneity
and very hard and strong when sampled.
The compressive strength of the bricks, on the other hand, as tested on cubes
with sides of 4050 mm was rather low: the mean value was 13.37 N/mm2, with
SD 26% (over 50 specimens). The elastic modulus between 20 and 60% of the
peak stress was 1973 N/mm2 for the bricks and 905 N/mm2 for the mortars [5].
Tests reported later show that the strength of the masonry was less than that of
the mortar, suggesting that the low carrying capacity of the masonry was mainly
due to the construction technique.
1.2.2.4 Compression tests on masonry prisms

Compression tests were performed on prisms of masonry, cut from large blocks
that had remained intact, in order to obtain the stressstrain curve up to and beyond
failure [5].
Fatigue tests were then performed using a load value reproducing the stress
induced by the dead load and applying a cyclic load, the amplitude of which repro-
duced the stress variations due to the effects of the wind.
Lastly, a survey of the effects of the dead load of the tower on the behaviour of
the materials over time was carried out by means of constant load tests.
Prisms measuring 4000 600 700 mm approximately were obtained from the
recovered blocks. These dimensions were chosen so as to simulate the behaviour
of the masonry, which was very thick (2.8 m) compared to height (60 m) and plan
form (12.3 12.3 m) of the tower. The load-control or displacement-control com-
pression tests were carried out with a 2250 kN, servo-controlled MTS hydraulic
press, with programmed cycles.
Monotonic compression tests to failure were conducted on seven masonry
prisms from the first two building phases. The tests were carried out under
displacement-control, at rates of 3.85 103 mm/s and 9.62 104 mm/s.
Figure 1.7: se curves of Figure 1.8: se curve obtained for a cyclic

prisms subjected to monotonic compression test.
compression tests.
Figure 1.7 shows the curves obtained for the seven prisms, two of which (102A
and 102B) were tested by applying the load in the direction of the horizontal joints.
The peak strengths and ultimate strain values vary quite considerably: low ulti-
mate strains appear to correspond to higher strengths. Strength varies from 2.0 to
4.1 N/mm2, ultimate strains from 3.0 to 5.5 103 and the modulus of elasticity,
defined between 20 and 40% of the peak stress, varies from 719 to 1802 N/mm2.
Five prisms were tested to failure by means of loading and unloading cycles
applied every 0.5 N/mm2 under displacement-control conditions up to and beyond
the peak stress. Strength varied from 1.8 to 3.3 N/mm2, the elastic modulus from
544 to 1455 N/mm2 and the ultimate strains from 3.6 to 8.5 103.
A typical curve is shown in Fig. 1.8. Although on average the strength is lower
than that of the seven prisms subjected to the monotonic tests, the cycles appear
not to have any great influence on the se curve, the peaks of which at each load-
ing approximate well to points of the monotonic curve.
1.2.3 Long-term tests
The behaviour detected from cycling tests and particularly the evident increase
in deformation while the stress was kept constant (Fig. 1.8) led to a study of the
effects of fatigue and long-term tests at constant load. The experimental research
is described in the following sections.
1.2.3.1 Fatigue tests

It is well known that repeated load cycles cause damage to the material. The dam-
age originates from imperfections in the material itself, such as small cracks which
get larger as the cycling load is applied. Generally, failure occurs at peak load
value lower than that measured when the load is statically applied.
In the case of a masonry structure, fatigue may be caused by the repeated action
of horizontal loads such as wind or seismic loads. In the particular case of the
tower, no appreciable seismic effects have been recorded, whereas the effects of
the wind must certainly have been felt over the centuries, causing significant vari-
ations in the stress due to the static load of the tower.
As mentioned above, the damage caused by repeated cycles of loading and
unloading is not very high when the average load applied is low and cycles are not
frequent.
However, significant damage may be caused if the cycles are repeated at an
average stress close to the ultimate capacity [8]. Three prisms were subjected to
cyclic loads corresponding to calculated stress variations of 0.2 N/mm2 starting
from very high compression values similar to those produced by the dead load as
calculated at the most loaded points of the structure. Cyclic loads corresponding to
repeated wind effects, simulated according to the Italian Code, produced no appre-
ciable damage except for higher strains (Fig. 1.9), when the loads starting from
stress values very close to failure were applied. The fact that the load history
included one or more cycling phases did not reduce the failure values (which
remained 1.7, 2.7 and 4.4 N/mm2) obtained during the monotonic tests.
Figure 1.9: se curve obtained during Figure 1.10: se and et curves

the fatigue test performed to simulate obtained during a step-by-step
wind effects. constant load test.
1.2.3.2 Constant load tests

Displacements measured on the prisms during loading showed a tendency to
increase at constant load suggesting that the behaviour of the material could be
time-dependent.
During the constant load tests, almost all the prisms were tested under load control
up to 1.01.5 N/mm2. The stress was then increased in steps of 0.14 N/mm2, at inter-
vals of at least 15 min. The increase in the effects of strain was on average 1.6
103 for each 15-min interval at constant load.
At higher stresses close to the ultimate strength of the material, the time-depen-
dent effects of the constant load evolved more rapidly. No further load increases
were made until the increase in strain stopped. At the last step the strain rate con-
tinued to increase rapidly until failure occurred suddenly after a period of time
varying between 10 min and 2 or 3 h. It seems reasonable to assume that the time
needed to reach collapse is a function of the ratio between the load applied and the
maximum load the specimen is able to withstand. The curves of strain as a function
of time, (Fig. 1.10) clearly show the type of behaviour described earlier.
At loads over approximately 70% of the ultimate compression strength, a small
number of cracks appeared on one of the sides of the tested specimens. The cracks
were found chiefly on the bricks and at the surfaces of contact between the mortar
and the stone or bricks of the inner face of the wall.
The cracks, which were always vertical, were hardly visible right up to the
moment of collapse (see Fig. 1.11a and b).
Although the tests were carried out under load control, it was almost always
possible to keep the collapse under check and thus prevent the sudden spalling,
which often characterizes the failure of prisms of solid bricks arranged in regular
courses, and/which is often an indication of a brittle failure of the bricks.
Structural analysis carried out by a FE (finite element) elastic model [5] revealed
that some parts of the tower were subjected to severe stress, very often close to the
failure limits found experimentally. This probably led to the gradual evolution of
micro-fissures which, over the centuries, may have contributed to the sudden col-
lapse of the material for no apparent immediate cause and without the appearance
(a) (b)
Figure 1.11: Width of cracks after (a) 22 min and (b) 60 min under a constant
stress of 2.0 N/mm2 (peak stress).
Figure 1.12: Vertical cracks on the external wall of the tower (1968).
of warning signs, such as large cracks or spalling, even during the days immedi-
ately preceding the collapse.
A careful examination of the photographs taken by archaeologists for the Civic
Museums in 1968 reveals that even at this time there were thin vertical cracks
largely diffused on the outer face of the wall. These cracks were extremely difficult
to see from the Piazza and were similar to those that appeared on the specimens
during the tests (see Fig. 1.12).
1.3 Long-term behaviour of masonry structures

Masonry is a composite material and its mechanical and physical behaviour strongly
depends on that of its components (mortar and brick/stone). Mortar influences
mainly the deformability and bricks or stones influence mainly the strength.
Historic buildings are very often characterized by high values of deformations,
which may have taken place in the past or may still be in progress and may lead
the building even to unexpected failure. Early deformation due to delayed harden-
ing of hydrated mortars based on carbonation is typical of ancient buildings pre-
senting thick mortar joints; multiple leaf masonry is usually characterized by
differential creep displacements induced by the different deformability of the
leaves; persistent and cyclic loads can give rise to a creepfatigue interaction and
to greatly retarded strain.
1.3.1 Deformation during mortar hardening
It has been shown that early creep behaviour, due to the carbonation process of
fresh mortar, can last for a long time particularly when mortar was made with
hydrated lime [9]. This can be the case, for instance, of very thick ancient walls
or of masonry characterized by very thick mortar joints like those of St. Vitale in
Ravenna shown in Fig. 1.13 [10].
Increasing deformation due to heavy dead or cyclic loads can vary the geometry
of masonry walls in a visible way already during the construction. These modifica-
tions can occur locally, or involve a whole structural element. Large displacements
and deformations frequently involve piers and columns like those of gothic cathe-
drals (Fig. 1.14) [11] due to the horizontal thrust exerted by vaults and arches or
due to soil and structure settlements.
Generally speaking, old or ancient structures are continuously subjected to
modifications concerning their geometry and their state of stress and strain. J.L.
Taupin [12] says that time moulds the structure of towers, cathedrals, bridges etc.
which we would like to consider immutable. Time plays a role both in the short
and in the long run dispersing and returning energy in three ways: through defor-
mations and settlements, through vibration, and through material modification or
deterioration.
Figure 1.15 shows a detail of the well-known Hagia Sophia at Istanbul, where
the rotation of a column and the deformation of an arch in the north gallery can be
clearly observed [13]. Figure 1.16 shows a much less famous small Romanesque
church, St. Maria la Rossa in Milan, dating from the tenth to thirteenth century.
This single aisled brickwork masonry building is covered by a timber gable roof,
with a chancel comprised by two small chapels besides the choir, terminating with
a semicircular apse. The church was subjected to different transformations during
centuries, and its present aspect is due to the restoration works done in the 1960s.
In the picture the tilt of the lateral walls and the deformation of the central arch can
be seen [14].
Figure 1.13: View of the Basilica of San Vitale in Ravenna (Sixth century AD).
Figure 1.14: Pillar of the Cathedral of Salisbury [11].
Figure 1.16: Church of St. Maria la

Figure 1.15: Hagia Sophia, north Rossa, view of the central nave
gallery looking west. looking east.
1.3.2 First, secondary and tertiary creep in rock and hardened masonry
The influence of time on the mechanical behaviour of stiff clays, soft rocks, fresh
cement mortar, concrete and hardened concrete becomes evident when both uni-
axialtriaxial compressive test at different rate of loading and compressive test at
vertical constant load are carried out. On the one side, when testing compression
of soft porous materials a decrease of the rate of loading produces a decrease of
the vertical peak stress and of the stiffness of the material (Fig. 1.17 v. Cook). On
the other side, if a constant load is applied an increase of deformation develops
which is commonly subdivided into three phases: the so-called primary, second-
ary and tertiary creep (Fig. 1.18 v. Cook) [15]. The appearance of one or more
of these phases and the strain rate of the secondary creep phase depend on the
stress level.
Very poor research was done before the collapse of the Civic Tower in Pavia, on
the creep behaviour of masonry structures, apart from the papers published by
Lenczner [16].
Figure 1.17: Dependency of strain on the loading rate.
Figure 1.18: First, secondary and tertiary creep.

The influence of time on the mechanical behaviour of masonry structures under

high states of stress became evident after the collapse of the Tower of Pavia, when the
identification of a time-dependent behaviour, probably coupled in a synergetic way to
cyclic loads [17], was identified as a possible explanation of the sudden collapse.
1.4 Collapse and damage of towers due to long-term

heavy loads
The failure of monumental buildings is fortunately an exceptional event; neverthe-
less, when their safety assessment is required, any risk factor that may affect the
integrity of the buildings has to be taken into account. Ancient buildings often show
diffused crack patterns, which may be due to different causes in relation to their
original function, to their construction technique and to their load history. In many
cases it is simply the dead load, usually very high in massive monumental buildings,
which plays a major role into the formation and propagation of the crack pattern.
The only way to prevent the occurrence of these failures is continuous observa-
tion and maintenance of these structures.
1.4.1 St. Marco bell-tower and St. Maria Magdalena tower in Goch
The first well-studied example of the collapse of a tower in Italy was certainly the
one of the bell-tower of St. Marco in Venice in 1902 (Fig. 1.19) [18]. The tower
collapsed suddenly with no previous evident signs of heavy damage. In the long
debate following the tower collapse and in the long report made by Luca Beltrami
in [18], the settlement of foundation was excluded from the causes and it was
clearly described that the damage was interesting considering the structure and the
repairs made 45 years earlier by confining the bearing corners of the tower with
steel reinforcements (Fig. 1.20).
A similar situation occurred in June 1902 with the collapse of a bell-tower in
Corbetta, near Milan, in the same year. The tower was under modification, being
elevated from the original 24 m to 41 m in 1860 and by adding a spire in 1900.
In 1993, the bell-tower of Sancta Magdalena church in Goch collapsed suddenly
during the night (Fig. 1.21); it had already been decided a few years earlier to start a
repair intervention due to the extent of the damage detected. Probably the waiting
time was too long, taking into account that the tower was badly cracked for a long
time and also damaged during the last world war.
1.4.2 The bell-tower of Monza Cathedral and the Torrazzo of Cremona
The bell-tower of the Cathedral of Monza is a sixteenth-century building made of

solid brick masonry, at present subjected to a repair intervention. Its walls were
showing large vertical cracks crossing the whole transversal section of the walls
on the west and east (Fig. 1.22a and b), which were continuously widening at a
constant rate [19, 20]. These cracks were certainly present before 1927, when they
were roughly monitored. Wide cracks were also present in the corners of the tower
Figure 1.19: The bell-tower Figure 1.20: Detail of the collapse with the
of the St. Marco Basilica in reinforced pillar [18].
Venice.
at a height of 30 m, together with a damaged zone at a height of 1125 m with a

multitude of very thin and diffused vertical cracks.
A similar crack pattern is visible on the Torrazzo, a medieval brickwork tower adja-
cent to the Cathedral of Cremona (Fig. 1.23a and b) [21]. The precise date of construc-
tion is not known but is assumed to be around the thirteenth century. It belongs to a
group of monuments, including the Cathedral, the Baptistery, the Town Hall Palace,
the Militia Loggia, which forms one of the most beautiful Italian squares.
The external load-bearing walls of the tower, which is about 112 m tall, have
been showing several cracks for many years [21]; since the crack pattern has expe-
rienced an evolution, a time-dependent behaviour of the material may possibly be
assumed to cause the phenomenon.
1.5 The role of investigation on the interpretation of the

damage causes
On the basis of the previous experience the authors have developed investigation
procedures for the safety of these structures; the idea came first when studying the
collapse of the Civic Tower in Pavia.
Figure 1.21: The church of St. Magdalena in Goch (Germany) after the collapse of
the bell-tower, 1993.
The procedure is based on the following steps: (i) historic research to know the
evolution of the structure over time, (ii) geometrical and crack pattern surveys,
which allow one to understand the evolution of the structure, to calculate weights
and give a first interpretation of the crack pattern, (iii) geognostic investigation and
monitoring, to understand the soilstructure interaction, (iv) on-site mechanical and
non-destructive testing (radar, sonic, etc.) to define local states of stress and stress
strain behaviour of the material, (v) chemical, physical and mechanical tests on
mortars, brick and stones to find their composition and their characteristics, (vi) if
necessary, passive and active dynamic tests on site to survey the overall structural
behaviour and (vii) monitoring system applied to the structure when necessary.
1.5.1 The bell-tower of the Cathedral of Monza
The bell-tower of the Cathedral of Monza, is a masonry structure 70 m high, with

a square plan (a side is 9.7 m long) with solid brick walls 140 cm thick. The tower
construction started in 1592, probably following the design of Pellegrino Tibaldi,
the architect of the Pavia tower belfry, and ended in 1605 [19, 22]. The only dam-
age to the tower reported by the documents occurred in 1740 and was due to a fire
which started in the bell-tower and caused the collapse of the belfry dome and
roof and the fall of the bells with their supporting frame down to the vault of the
Figure 1.22: Survey of the crack Figure 1.23: Survey of the crack pattern
pattern for the bell tower of the for the Torrazzo tower: (a) west and
Cathedral of Monza: (a) west and (b) east sides.
(b) east sides.
first floor at 11 m. No damages were reported in other known calamities, such as

lightning or thunderstorms through the centuries. Nevertheless cracks are present
since 1927 or even before, as mentioned above. From 1978 the cracks have been
surveyed with removable extensometers: they show a slow increase of their open-
ing through time. From 1988 the rate of opening seems to be increasing faster.
The trend of widening of the three main cracks was calculated as 30.6, 31.3 and
39.7 m/year from 1978 to 1995. Actually if this trend is considered from 1988 to
1997 the values change, respectively, into 41.2, 35.2 and 56.2.
The first step of the investigation procedure [19] was the geometrical survey
[20]. A geodetic network set up in the square of the Cathedral in 1993, was used
as support. No relevant leaning was measured due to the small subsidence which
is taking place in the square. Two distinct products were obtained: (i) a detailed
three-dimensional model from which the external and internal prospects and the
vertical sections were obtained and (ii) a simplified model for which only the
essential aspects of the geometry were preserved for the structural analysis.
The survey of the crack pattern showed that the tower walls have a dangerous distri-
bution of passing-through cracks on the western and eastern load-bearing walls for
more than 50 years, and of a net of thin vertical cracks from a level of 11 m up to 30 m
(Fig. 1.22). Other cracks can be seen on the internal walls of the tower; they are very
thin, vertical and diffused along the four sides of the tower and deeper at the sides
of the entrance where the stresses are more concentrated. The thin diffused cracks run
450 mm deep inside the section, reducing its total working thickness from 1400 mm
to no more than 900 mm. From laboratory tests it was found that the mortar is very
weak and made with putty lime and siliceous aggregates; also the bricks were of poor
strength (between 4 and 12 N/mm2 measured on 40 mm-side cubes).
On-site single flat-jack tests were carried out at different heights of the tower
(5.4, 5.6, 13.0, 14.0, 31.5 and 38.0 m) and the stress values against the height are
plotted in Fig. 1.24. The maximum compressive stress acting in the tower, mea-
sured on site by the flat-jack test, is about 2.2 N/mm2. The most interesting infor-
mation came from the double flat-jack test results, where it was possible to see the
real risky situation if compared with the local state of stress measured by the single
flat-jack (Fig. 1.25). Passive dynamic tests using the bell ringing were also carried
out monitoring the dynamic excitation of the extensometers applied across the
Figure 1.24: Single flat-jack tests of Monza.

(a) 3.00 (b) 3.00
2.50 2.50
Stress [N/mm2]
2.00 2.00
Local state
Stress [N/mm2]
Local state
of stress
of stress
1.50
1.50
1.00
1.00
0.50
0.50
h v
h v 0.00
0.00 -2.00 -1.00 0.00 1.00 2.00 3.00 4.00 5.00
-2.00 -1.00 0.00 1.00 2.00 3.00 4.00 5.00
Strain [m/mm]
Strain [m/mm]
Figure 1.25: Monza tower stressstrain plot at (a) 5 m and (b) 13 m height.
main cracks, giving under these cycling stresses a maximum peak to peak (open-
ing to closing) of 28 m that has to be compared with a daily variation of 100 m
due to the temperature effects. The diagnosis based on the experimental survey
and on the FE modelling lead to the conclusion that the bell-tower was a high risk
building and needed a quick intervention.
In Chapter 8 the preservation and repair intervention which is still being carried
out is illustrated.
1.5.2 The Torrazzo of Cremona
The bell-tower of the Cathedral of Cremona, an interesting historic town not far
from Milan (Italy), is known by the nickname il Torrazzo from long time ago.
The tower is situated at the northern side of the Cathedral and it is connected to
it by a Loggia called Bertazzola. The geometry of the tower is rather complex,
being composed of a lower part (Romanesque tower) with a square plan of 13 m
side and 70 m high, an upper part, the Ghirlandina, with an octagonal plan (2.5 m
side), more than 40 m high. The Torrazzo is known as the tallest medieval bell-
tower in Europe being 112 m high [23].
The lower part of the tower, with a square plan, is a massive construction
with few openings localized on the western and eastern sides. The upper Ghir-
landina appears as a light structure with arches and large openings on all the
four sides.
The staircase from the lowest level up to the Ghirlandina level was built within
the thickness of the walls (approximately 3.3 m thick). Along the staircase, cov-
ered with a barrel vault, the thickness of the external wall is approximately 1 m,
while the thickness of the internal wall is 0.71 m with the span of the staircase
measuring 1.31.6 m. The staircase allows one to reach some internal vaulted
rooms.
Archive research did not clarify the date of construction; nevertheless the high-
est number of reference data collected locates the date of construction between the
eighth and the thirteenth centuries. In 1491, the porch of the Bertazzola was added
connecting the Torrazzo with the Cathedral and in 1519 the Loggia was built
resting on the arches of the porch. Maintenance works were carried out starting
from the fifteenth century. These works mainly concerned the highest part of the
tower damaged by storms and lightening, especially the stone and brick columns
which were sometimes substituted. The last intervention at the Ghirlandina was
carried out in 1977. The works performed were the following: connection of struc-
tural and decorative elements, construction of a concrete frame sustaining the twin
columns of the Stanza delle Ore (at 85 m height) and surface treatments of stone
and brick elements with an epoxy resin.
The first step of the investigation carried out in 1998 was the geometrical sur-
vey. A principal network defining fixed points in the horizontal and vertical plan
was set up having 21 nodes inside and around the tower made with fixed nails. The
co-ordinates of the nodes were determined with a T2000 WILD equipment. The
vertical and horizontal profiles were determined by rays starting from the network
nodes, using a TC1600 DIOR system and an auto scanning Laser System MDL. A
photogrammetric survey of the external prospects was also carried out using
TC1600-DIOR and T460* DISTO equipment. The prospects were obtained by a
Rollei special software, MSR. The survey enabled the finding of some irregulari-
ties of the structure: (i) a 21 cm horizontal displacement of the centre of the tower
in direction north-east, calculated from the ground level to the top at 112 m, (ii) a
non-symmetrical reduction of the plan dimensions from the ground level to the top
at 31 cm for the north-east corner and 66 cm for the south-west corner, (iii) the
Ghirlandina not being perfectly centred on the square part of the tower, but with a
slight counter-clockwise rotation toward west.
The presence of a diffused crack pattern particularly on the western and the
eastern sides of the tower and on the Ghirlandina can indicate high states of stress
due to the dead loads, the temperature variations and/or to a slight leaning. The
survey was carried out on the outer surfaces by reaching the height of 60 m thanks
to a special crane. The crack pattern is certainly also influenced by differential
movements due to temperature variation between one side and the other of the
tower. The highest variations certainly occur between the north and the south side.
The west side has a diffused fissuration with passing-through cracks; the cracks
are mostly vertical and start from approximately 20 m. Important cracks appear
also between 48 and 60 m from the ground level (Fig. 1.23a). The north side is
cracked in the centre between 27 and 40 m and at the north-east corner. The east
side is cracked between 6 m and 20 m from the ground level and between 35 and
60 m (Fig. 1.23b). The south side has few cracks located between 14 m and 27 m.
The Ghirlandina shows the most important cracks, on the buttress and on the brick
columns particularly on the south-west corner. Also the internal part of the tower,
along the staircase and inside the rooms shows a diffused crack pattern with some
passing-through cracks. Three thresholds were established concerning the measure
of the opening of the cracks: <3 mm, between 3 and 10 mm, >10 mm. The crack
pattern survey helped one to understand and interpret roughly the mechanical
damage and to locate the position for the monitoring system.
The inspection of the masonry surface and the inside of the walls leads to the
following description: (i) the walls are made with solid bricks and no rubble was
used for the inner part of the section; (ii) the bricks are regular with varying dimen-
sions: 240280 1001200 5570 mm; (iii) the external walls of the Ghir-
landina are irregularly scaled and tooled; (iv) the colour of the bricks is variable
red, dark red, yellow, orange etc.; (iv) the mortar joints are regular with thickness
variable from 10 to 30 mm; (v) different techniques of jointing and pointing can be
found and often the vertical joints seem to be void or recessed; (vi) the masonry
texture is also regular with header and stretcher alternatively positioned; (vii) an
external leaf one brick thick with a weak collar joint is certainly present along the
staircases, in the internal rooms and at the level of the Bertazzola and more research
is needed to test the real extension of this leaf along the tower; (viii) in the stair-
case walls a row composed of 422 headers is repeated at rather regular intervals
as if it should represent a connection of the external leaf to the internal one; (ix)
several scaffolding holes externally closed can be seen along the masonry walls;
and (x) numerous restorations by brick substitution can also be seen externally and
internally.
Together with the geometric survey an accurate survey of the material decay
was carried out.
Concerning the reinforced concrete frame (85 m level) the columns between the
north and north-east and the east and south-east side show washout of the binder,
formation of carbonates near the stirrups with partial detachment of the reinforce-
ment cover (no more than 1 cm thick) and reinforcement corrosion.
Sixteen samples of bricks and mortars were collected from the masonry: five
from the facades, four inside from the walls of the internal rooms, four along the
staircase and three from the external and internal walls of the Ghirlandina. The
maximum depth of sampling was 300 mm. All sampling operations were docu-
mented graphically or photographically.
Laboratory tests were carried out on mortar and bricks. Chemical analyses
showed that the mortar binder was hydrated lime (probably lime putty) and the
aggregates were mainly siliceous. Two types of bricks were used, which differ in
colour (red and brown) and also in properties. The red bricks have high absorption
(2128.8%), low strength 812.4 N/mm2 in compression and in tension (0.11.6
N/mm2) and low modulus of elasticity (10002175 N/mm2); the brown bricks
have lower absorption (18.521.7%), higher compressive strength (9.425.43 N/
mm2) and tensile strength (2.22.6 N/mm2) and modulus of elasticity (17254417
N/mm2). The two types of bricks are present everywhere in the tower, so an aver-
age between the two bricks can be considered as reference.
In order to detect the suspected existence of an external cladding in use during
the Middle Age as a false curtain to hide the roughness of the real wall, bricks were
sampled from the external wall of the Bertazzola at 6 m level and from the walls
of the Stanza dei Contrappesi at 13.6 m level. This external leaf was confirmed
and its thickness is around 12 cm. The sampling allowed one to find large areas
where the leaf seems to be detached from the rest of the wall; following these
results the application of NDT technique was required in order to map the detached
areas which represent structurally a reduction of the wall section to be taken into
account when modelling. All the areas from where samples were taken were then
repaired with similar bricks and mortars.
Flat-jack tests: Single and double flat-jack tests were carried out on the Torrazzo.
The single flat-jack test was also used to study the behaviour of the external leaf
of the wall. Different types and dimensions of flat-jacks were used: (i) 240 mm
12 mm rectangular jacks where the detachment of the external leaf was suspected,
(ii) 400 mm 200 mm rectangular jacks and (iii) 350 mm 240 mm semicircular
jack where no detachment was suspected and for the double jack-test.
Twenty-one tests were carried out, of which 19 were with single flat-jack and 2 with
double flat-jack: 3 single flat-jack tests at between 1 and 5 m from the ground, 7 single
flat-jack tests at 7 m, 10 single flat-jack between 15 and 18 m and 1 single flat-jack at
22 m. The double flat-jack tests were carried out at 7.2 and 19 m from the ground.
The results of the single jack tests are reported in Fig. 1.26 and show clearly two
situations: a state of stress varying between 0.4 and 0.9 N/mm2 where the test
found a detached leaf and a state of stress varying from 1.01 and 1.81 N/mm2
where no detachment was found.
Also double flat-jack tests were performed and Fig. 1.27 shows the stressstrain
plots. It was impossible to carry out tests at higher levels due to the lack of scaf-
folding and of appropriate means for carrying the jack equipment. In future other
tests will be carried out.
3.50
masonry section 3.3 m
masonry section 1 m
3.00 15.2-16.5 m
presence of veneer
2.50 inner walls (rooms)
16-17.8 m
Stress [N/mm2]
16.6-17.8 m
2.00 7.2-7.7 m
7.2-7.4 m
15.4-19.1 m
7-7.7 m
1.50 5m
1.7 m
1.00
0.50
0.00
Figure 1.26: Single flat-jack tests of Cremona.

(a) 4.00 (b) 4.00
3.50 3.50
3.00 3.00
Stress [N/mm2]
Stress [N/mm2]
2.50 2.50
2.00 2.00
Local state
Local state
of stress
of stress
1.50 1.50
1.00 1.00
0.50 0.50
h v h v
0.00 0.00
-3.00 -2.00 -1.00 0.00 1.00 2.00 3.00 -3.00 -2.00 -1.00 0.00 1.00 2.00 3.00
Strain [m/mm] Strain [m/mm]
Figure 1.27: Stressstrain plot at (a) 7.2 m and (b) 19 m height.
1.6 Comparison between the two towers

Since the bell-tower of Monza is considered a building with high risks of collapse,
a comparison between the data collected on both towers seems to be useful to
understand better the real situation of the Torrazzo.
As mentioned above, the mortar composition of the two bell-towers does not dif-
fer much from one another, though the Torrazzo mortar seems to be more consistent.
The bricks of the Monza tower are generally weaker than those of the Torrazzo
(Fig. 1.28a and b) except for the brown type, which is mainly used on the outside
surface of the bearing walls and very seldom used in the interior. On the contrary
the brown and the red bricks are evenly distributed in the Torrazzo walls.
It is also interesting to compare the results of single and double flat-jack tests
carried out on the two towers.
The results of four tests, two for each tower are discussed. In Fig. 1.27a and b
maximum stress reached with the double flat-jack test on the Torrazzo together
with the values obtained with the single one, respectively, at 7 and 19 m height are
considered, showing an elastic linear behaviour up to, respectively, 2.45 and 2.7 N/
mm2. The maximum stress level when cracks clearly appear is, respectively, 3.77
and 3.77 N/mm2 and the state of stress measured is 1.5 and 1.5 N/mm2. So in these
two cases the safety coefficient at collapse is certainly more than 3.
In Fig. 1.25a and b the results of two tests at the height of 5 and 13 m, out of the
four carried out on the walls of the Monza tower, are considered. Here the linear
elastic behaviour stops at, respectively, 1.65 and 1.1 N/mm2 and the maximum stress
reached before cracks propagated was 2.62 and 1.87 N/mm2. The measured local
state of stress was, respectively, 1.67 and 0.98 N/mm2. In these two cases the safety
(a) 34 (b) 26
32 red brick
24 brown brick
30 22
28 20
26 18
24
Stress [N/mm2]
16
22
Stress [N/mm2]
14
20
12
18
10
16
14 8
12 6
10 4
8 red brick 2 h v
brown brick
6 0
4 -20 -16 -12 -8 -4 0 4 8 12 16 20
Strain [m/mm]
2 h v
0
-20 -16 -12 -8 -4 0 4 8 12 16 20
Strain [m/mm]
Figure 1.28: Stressstrain plot for (a) Monza tower bricks and (b) Torrazzo
bricks.
coefficient at failure is much lower than in the first one and certainly less than 2.
Furthermore in the case of the Torrazzo the modulus of elasticity is much higher and
the Poisson ratio much lower than in the case of the Monza tower.
1.7 Conclusions
The investigation carried out on the specimens cut from the walls of the Pavia tower
after its collapse allowed formulating for the first time on an ancient masonry the
hypothesis of a collapse due to the long-term behaviour of the material. Probably
since the construction of the bell-tower in the sixteenth century the structure was
under a high state of stress and the damage very slowly but continuously increasing
until the collapse. The creep behaviour of the material was shown clearly during
the experimental research which started in 1989 and is still developing, as will be
shown in Chapter 2. Examples of other similar situations were found in the history
of collapses of towers and damages or collapses of churches (Noto cathedral).
The two experiences of investigation on tall towers allow some concluding
remarks:
the on-site and laboratory tests carried out following the methodology described
in the first section allowed one to detect situations of danger and to characterize
the materials and calculate input parameters for the structural analysis;
the laboratory tests were able to show the difference of properties of the ma-
sonry in the two buildings and that where the materials used are weaker, the
damage is more;
the flat-jack test is a powerful tool to calculate the actual state of stress in com-
pression and to detect the mechanical behaviour of the masonry, so that two
different situations (Torrazzo and Monza towers) can be compared;
the investigation allowed the authors to state that the situation of the Monza
tower is very difficult and that a quick intervention has to be started;
for the Torrazzo, even if the state of damage is not considered dangerous, a
monitoring system has been set up, and the tower will be under control for 45
years at least in order to study its further evolution; in the meantime some
repairs are being done for the external surfaces.
FE numerical models were used for the static and dynamic analysis of the two
towers [19, 23]. The results of the experimental research were used to calibrate the
FE models.
References
[1] Panazza, G., Campanili Romanici a Pavia, Arte Lombardia, pp. 1827, 1956.
[2] Ward-Perkins, B., Scavi nella Torre Civica di Pavia, Sibrium, 12, pp. 177
185, 197375.
[3] Milano, F. & Toscani, X., Il fond di documenti relativi alla Torre Civica esist-
ence nellArchivio Comunale di Pavia, Sibrium, 12, pp. 467493, 197375.
[4] Anti, L. & Valsasnini, L., Indagini preliminari allanalisi strutturale ed
alle prove sui materiali della Torre Civica di Pavia, TEMA J., LArsenale,
Venezia, 1991.
[5] Binda, L., Gatti, G., Mangano, G., Poggi, C. & Sacchi Landriani, G., The
collapse of the Civic Tower of Pavia: a survey of the materials and structure.
Masonry International, 6(1), pp. 1120, 1992.
[6] Knoffel, D.F.E. & Wisser, S.G., Microscopic investigation of some historic
mortars, Proc. 10th Int. Conf. Cement Microscopic, S. Antonio, Texas, 1988.
[7] Baronio, G. & Binda, L., Reazioni di aggregati in intonaci antichi, Conv.
Intonaco: Storia, Cultura e Tecnologia, Bressanone, pp. 269276, 1985.
[8] Binda, L., Anzani, A. & Mirabella Roberti, G., The failure of ancient
towers: problems for their safety assessment, Int. Conf. on Composite
Construction - Conventional and Innovative, IABSE, Insbruck, pp. 699
704, 1997.
[9] Ferretti, D. & Bazant, Z.P., Stability of ancient masonry towers: moisture
diffusion, carbonation and size effect, Cement and Concrete Research, 36,
pp. 13791388, 2006.
[10] Binda, L., Lombardini, N. & Guzzetti, F., St. Vitale in Ravenna: a survey
on materials and structures, Int. Conf. Historical Buildings and Ensembles,
invited lecture, Karlsruhe, pp. 113124, 1996.
[11] Gordon, J.E., Strutture, ovvero Perche le cose stanno in piedi, Edizioni
Scientifiche e Tecniche, Mondadori, 1979.
[12] Taupin, J.L., Rflexions sur la cathedrale Saint-Pierre de Beauvais.

ANAGKH, 12, pp. 86100, 1995.
[13] Mainstone, R.J. Haghia Sophia: Architecture, Structure and Liturgy of the
Justinians Great Church, Thames and Hudson, 1985.
[14] Binda, L., Mirabella Roberti, G. & Guzzetti, F., St. Vitale in Ravenna: a
Survey on materials and structures, International Symp. on Bridging Large
Spans (BLS) from Antiquity to the Present, Istanbul, Turkey, pp. 8999,
2000, ISBN 975-93903-02.
[15] Jaeger, J.C. & Cook N.G., Fundamentals of Rock Mechanics, 2nd edn,
Chapman & Hall: London, 1976.
[16] Lenczner, D. & Warren, D.J.N., In situ measurement of long-term move-
ments in a brick masonry tower block. Proceedings of the 6th IBMaC,
Rome, pp. 14671477, 1982.
[17] Anzani, A., Binda, L. & Mirabella Roberti, G., The behaviour of ancient
masonry towers under long-term and cyclic actions, in Proc. Computer
Methods in Structural Masonry 4, Computer & Geotechnics: Swansea,
pp. 236243, 1998.
[18] Fradeletto, A., et al., Il campanile di S. Marco riedificato. Studi, ricerche,
relazioni, ed. Comune di Venezia, Carlo Ferrari: Venezia, 1912.
[19] Binda, L., Tiraboschi, C. & Tongini Folli, R., On site and laboratory inves-
tigation on materials and structure of a bell-tower in Monza. Int. Zeitschrift
fr Bauinstandsetzen und baudenkmalpflege, 6, Jahrgang, Aedification
Publishers, Heft 1, pp. 4162, 2000.
[20] Binda, L., Tongini Folli, R. & Mirabella Roberti, G., Survey and investigation
for the diagnosis of damaged masonry structures: the Torrazzo of Cremona.
12th Int. Brick/Block Masonry Conf., Madrid, Spain, pp. 237257, 2000.
[21] Binda, L. & Poggi, C., Ricerca volta a stabilire le condizioni statiche ed
il comportamento meccanico della muratura del campanile del Duomo di
Cremona. Relazione Finale, Contratto Consiglio della Chiesa Cattedrale di
Cremona, 1999.
[22] Scotti, A., Let dei Borromei in Monza. Il Duomo nella storia e nellarte,
Electa: Milano, 1989.
[23] Binda, L., Falco, M., Poggi, C., Zasso, A., Mirabella Roberti, G.,
Corradi, R. & Tongini Folli, R., Static and dynamic studies on the Torrazzo
in Cremona (Italy): the highest masonry bell tower in Europe. Int. Symp.
on Bridging Large Spans from Antiquity to the Present, Istanbul, Turkey,
pp. 100110, 2000.
CHAPTER 2
Experimental researches into long-term

behaviour of historical masonry
A. Anzani1, L. Binda1 & G. Mirabella Roberti2

1Department of Structural Engineering, Politecnico di Milano,
Milan, Italy.
2Department of History of Architecture, University Iuav of Venice,
Venice, Italy.
2.1 Introduction
The time-dependent behaviour of ancient masonry structures, often characterized
by non-homogeneous load-bearing sections, is considered among the factors
affecting the structural safety of monumental buildings. Together with other
synergetic aspects, this has proved to be involved in collapses, which occurred
during the last thirty years.
Exploiting the ancient (from the Middle Ages to the sixteenth century) masonry
coming from the ruins of the collapsed tower of Pavia, several experimental
procedures have been adopted to understand the phenomenon, from creep to
pseudo-creep tests at different time intervals, and various rheological models have
been applied to describe the creep evolution and creep-induced damage, as
explained later in this book in Chapter 7.
Purpose of the testing activity has been initially the identification of the creep
behaviour as a possible cause of the collapse of buildings, then the study of factors
affecting creep (rate of loading, stress level, etc.) and the set-up of the most suit-
able testing procedures to understand the phenomenon, and finally the individua-
tion of significant parameters (e.g. strain rate of secondary creep phase) that may
be referred to as risk indicators in real structures.
After the first tests carried out on prisms of dimension 400 600 700 mm
described in Chapter 1, long-term tests on six prisms coming from the ruins of
the tower of Pavia and one from the crypt of Monza were performed, some
of which lasted 1000 days. Considering that long-term tests require constant
thermo-hygrometric conditions and especially designed testing apparatus, a more

rapid and therefore more convenient testing procedure was subsequently preferred.
The so-called pseudo-creep tests were carried out applying the load by subsequent
steps corresponding to a constant value (generally 0.25 or 0.3 MPa) kept constant
for a specific time interval. Different durations of the time interval have been
experimented (from 300 to about 30,000 seconds) that allowed one to indirectly
observe the influence of the rate of loading. In fact, these tests characterized by a
regular load history tend to simulate, by discrete load steps, monotonic tests where
the load increases continuously at an equivalent rate that can be calculated. They
give the opportunity to satisfactorily catch the limit between primary and second-
ary creep phase.
Considering tertiary creep, pseudo-creep tests imply a disadvantage. In fact, the
load value that in a monotonic test at the equivalent rate of loading would cause
failure may not correspond exactly to one of the applied load steps, but to an inter-
mediate value (Fig. 2.1). Therefore, if the applied load is higher than that which
would have caused tertiary creep and failure, the specimen collapses instanta-
neously, without showing the tertiary creep phase. The latter is not particularly
important in itself; what is interesting is the secondary creep strain rate just before
failure. The problem could be solved by simply prolonging the time interval of the
last load step so as to reach the failure limit curve (see Fig. 2.1) in constant load
conditions instead of at increasing load. Actually, before reaching it, the stress at
failure is not known; however, the failure conditions may be roughly previewed by
estimating the ultimate stress through sonic tests and by controlling accurately the
strain rate in real time during the pseudo-creep test. In some of the cases described
below tertiary creep was therefore recorded.
failure limit curve

v
viscosity limit curve
Figure 2.1: Pseudo-creep testing: simulation of monotonic test at an equivalent

rate of loading.
LONG-TERM BEHAVIOUR OF HISTORICAL MASONRY 31
In the case of the masonry from the Pavia tower, different prism dimensions were
adopted in order to use as much as possible the material coming from the irregular
blocks taken from the ruins, compatibly with the capacity of the testing machine. On
the contrary, from the crypt of Monza two big blocks were purposely extracted from
which prisms of the same dimensions were obtained. When relevant, comments on
the influence of the specimen dimensions on the test results will be given.
The mechanical test series were carried out on the prisms previously capped
with 1:3 cement mortar; PTFE sheets were interposed between the sample bases
and the machine platens (Fig. 2.2a); a hydraulic compressive machine MTS (2500
KN) was used, connected with a control unit for data-acquisition, a plotter produc-
ing load-displacements diagrams, a PC for storing data. If not differently specified,
vertical and horizontal displacements were measured directly on the prisms using
four LVDT with a base of 150 mm and four LVDT with a base of 180 mm, respec-
tively; two overall vertical readings were also taken from plate to plate of the
machine in case the other LVDTs had fallen during the tests (Fig. 2.2b and c).
2.2 Tests on the masonry of the Civic Tower of Pavia

After the sudden collapse of the Civic Tower of Pavia (built from eleventh to
sixteenth century), during the investigation into the causes, many prisms of differ-
ent dimensions were obtained out of the large blocks coming from the ruins of the
tower and constituting the medieval trunk of the structure (Fig. 2.3). The prisms,
subjected to mechanical tests, had mainly been obtained from the conglomerate
forming the very thick inner core of the 2800 mm three-leaf walls (Fig. 2.4); a
few of them were coming from the fairly regular external layers made of roman
brick masonry of thickness varying between 150 and 490 mm; no specimens were
initially sampled from the plain masonry belonging to the sixteenth-century belfry,
as it was not involved in the initiation of the collapse [1] although this addition
may be suspected as a remote trigger responsible for the collapse [2].
(a) (b) (c)
Figure 2.2: Preparation and instrumentation of the prisms for mechanical testing.
Sixteenth
century
Eleventh
twelfth
centuries
Figure 2.3: Civic Tower of Pavia before failure.
Inner leaf
Outer leaf
Figure 2.4: Detail of the cross-section Figure 2.5: Cutting the sixteenth-
of the 2800 mm thick medieval century plain masonry.
masonry.
Only recently, it was decided to study also the sixteenth-century plain masonry
belonging to the upper part (Figs 2.5, 2.6). In fact, its behaviour may provide a
useful comparison for other structures of the same age and constructive technique,
being a historical masonry not usually available for mechanical testing.
Prisms of larger dimensions were cut first and progressively smaller specimens
were also obtained subsequently in order to exploit the historical material as much
as possible. The 400 600 700 mm masonry prisms were identified by a number
(a) (b) (c)
Figure 2.6: Civic tower of Pavia: (a) medieval outer leaf, (b) medieval inner leaf,
(c) sixteenth-century plain masonry.
(indicating the block which they came from) and a letter, (since more than one
specimen had been cut from the same block). Smaller specimens were named in a
similar way, including a second number after the first figure, which indicates the
dimension of the base of the prism [3].
2.2.1 Characterization by sonic tests
Before designing the load history to which the prisms were to be submitted, sonic
tests were carried out for non-destructive strength estimation. In the case of prisms
of 400 600 700 mm and in the case of prisms of 300 300 510 mm, eight
trajectories in each of the two horizontal directions were adopted to test the prisms
by transparency; in the case of prisms of 200 200 350 mm four trajectories in
each of the two horizontal directions were adopted (Fig. 2.7).
Figure 2.8 shows a direct relationship between the mean sonic velocity and the
compressive strength of the medieval inner leaf, tested monotonically as described
below.
2.2.2 Monotonic tests on prisms of different dimensions
Monotonic uniaxial compression tests on the masonry of the inner leaf have been
carried out, respectively, on seven prisms of dimensions 400 600 700 mm [1],
on one prism of dimensions 200 200 350 mm and on five prisms of dimen-
sions 100 100 180 mm [4]. The test results are summarized in Table 2.1. As
expected, the smaller prisms exhibited higher values of the peak vertical stress and
lower values of the ultimate deformation.
4.0
3.5
3.0
m [N/mm2]
2.5
2.0
1.5
1.0
1000 1200 1400 1600 1800 2000
sonic velocity [m/s]
Figure 2.7: Apparatus for Figure 2.8: Compressive strength vs. sonic
sonic tests. velocity: tests on 400 600 700 mm masonry
prisms.
Table 2.1: Results of monotonic tests on the masonry of the inner leaf.
Dimensions s fv e fv s fv, ave e fv, ave
Sample (mm) (MPa) (103) (MPa) (103)
67C 400 600 700 2.0 4.0
94B 3.1 3.0
94D 2.5 3.0
100C 3.0 3.9 2.8 4.14
100D 2.5 5.5
102A 2.4 5.2
102B 4.1 4.4
Y 200 200 350 2.26 2.5
18-10M 100 100 180 3.86 2.20 3.83 2.63
19-10A 3.60 3.90
19-10H 3.00 2.90
102-10B 3.90 2.14
102-10A 4.80 2.04
Two prisms of the sixteenth-century plain masonry of dimension 200 200

350 mm were tested, the results of which are reported in Table 2.2. A higher peak
stress and a lower value of maximum vertical strain than in the case of the inner
leaf have been registered, indicating an important influence of the construction
technique on the masonry mechanical behaviour.
The stressstrain diagrams of the larger size prisms are shown in Figs 2.9 and
2.10. Considering the vertical strains, the marked initial locking branch appearing
in Fig. 2.9 is due to the fact that, in this case, the plotted strain has not been read
directly on the specimens, but calculated from the displacement between the
machine platens. In all cases, a linear part can be seen during which no visible
Table 2.2: Results of monotonic tests on the sixteenth-century plain masonry.

Dimensions s fv e fv s fv, ave e fv, ave
Sample (mm) (MPa) (103) (MPa) (103)
R 200 200 350 5.80 1.79 5.9 1.945
X 6.00 2.1
7 7
X
6 6
5 X
5
v [N/mm2]
f [N/mm2]
4 4 R R
3 3
100C
2 2 Y Y
67C
1 1
h v
0 0
0 2 4 6 8 10 -20 -15 -10 -5 0 5 10
v [m/mm] [m/mm]
Figure 2.9: Monotonic tests on Figure 2.10: Monotonic tests on prisms

prisms of 400 600 700 mm of 200 200 350 mm: (---) inner leaf,
from the inner leaf. () plain masonry.
cracks appear, but probably steady diffusion of micro-cracks takes place; in the case
of specimens R and X belonging to the plain masonry, this basically continues up
to the peak stress, showing a brittle behaviour. Vice-versa, in the case of the masonry
of the inner leaf a non-linear part up to the peak stress value shows, corresponding
to first visible crack appearance. Finally a softening branch, characterized by a
decrease of the vertical stress with increasing strain, can be observed in all cases.
2.2.3 Fatigue tests
Following the monotonic tests, cyclic tests were performed. Two cycles of loading-
unloading were applied every 0.5 MPa under displacement control and diagrams
similar to those presented in Fig. 2.11 were obtained.
A behaviour similar to that shown by the prisms tested monotonically can be
observed. An interesting tendency of the samples to deform under constant load
can also be seen. This is indicated by an almost horizontal line in the stressstrain
diagram just before the unloading phase initiates, corresponding to the time elapsed
while measuring the horizontal displacements [5].
Values of secant modulus have been calculated on the linear part of the unloading
diagrams: this is particularly interesting because it describes the elastic stiffness
referred to recoverable strain only, whereas the secant modulus based on the loading
branch of the diagram is influenced by elastic and permanent strain. The obtained
5 5
4 4
v [N/mm2]
v [N/mm2]
3 3
2 2
1 1
0 0
0 2 4 6 8 10 0.0 1.0 2.0 3.0
v (103) v (103)
Figure 2.11: Results of a cyclic test Figure 2.12: Results of a cyclic test
on a 400 600 700 mm masonry on a 200 200 400 mm masonry
prism. prism.
values followed an increasing trend up to the peak stress, and in general were higher
than the values of the tangent modulus calculated for monotonic tests [6].
More specific tests were then carried out on 100 100 200 mm prisms to
understand the effect of cyclic actions, e.g. that of wind or that of thermal cycles.
Figure 2.12 shows the results obtained on a sample which was loaded monotoni-
cally and submitted to cycles of 0.5 Hz frequency and 0.05 N/mm2 amplitude, at
different stress levels. Cycles acting within a relatively small load range proved
capable of provoking noticeable material strain. When applied at a high stress level,
relative to the material strength, load cycles appeared particularly effective, having
induced failure at a lower stress level than the estimated peak stress [3].
2.2.4 Creep tests on prisms of 300 300 510 mm
Six prisms of dimensions 300 300 510 mm were tested in compression in con-
trolled conditions of 20C and 50% RH at ENEL-CRIS Laboratory (Milan), using
hydraulic machines capable of keeping constant a maximum load of 1000 KN. The
dimensions adopted for the prisms were the maximum compatible with the testing
machine. The load was applied in subsequent steps, kept constant until either the creep
strain reached a constant value or a steady state was attained. The first stress level was
chosen between 40 and 50% of the static peak stress of the prisms, estimated by sonic
tests. The test results are reported in Table 2.3 and in Fig. 2.13a and b.
From the experimental data, the development of all the creep phases was evi-
dent, with secondary creep showing even at 41% of the estimated material peak
stress and tertiary creep showing at about 70%; material dilation took place under
severe compressive stress corresponding to high values of the horizontal strain due
to slow crack propagation until failure [3].
Table 2.3: Results of long-term tests on inner leaf prisms of 300 300 510 mm.
Sample Test duration (days) s fv (MPa) e fv (103)
19-30A 1170 2.0 4.75
19-30B 1082 2.0 2.92
67-30B* 163* 1.3* 0.70*
47-30A 524 2.0 3.50
41-30B 894 2.3 2.70
102-30A 894 2.9 2.90
Average 2.24 3.35
*Collapsed by premature failure; not included in average calculation.
(a) 4 Secondary creep
3 Tertiary creep
Primary creep
v (103)
-1
-2
h (103)
-3 (b) 3.0
dilation
-4 2.5
v [N/mm2]
2.0
-5
1.5
-6
1.0
-7
0.5
-8 0.0
0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200
time [days] time [days]
Figure 2.13: Results of creep tests on prisms of 300 300 510 mm from the
inner leaf.
The strain vs. time values of one of the prisms tested are reported in Fig. 2.14a
and b. Due to technical problems, after 630 days from the beginning of the test the
load was unintentionally lowered to zero for 90 days. Undesired unloading caused
only partial strain recovery without affecting very much the test results.
Despite the apparent scatter due to the sensitivity of the calculated value to ran-
dom reading errors, the volumetric strain seems nearly constant during the first load
steps of the test. Subsequently it starts to decrease markedly: the slope of the plot
(a) 6 (b) 6
4 4
v (103)
2 2 'v
evol
(103)
0
0
'h
-2
-2
h (103)
-4
-4
2.4 MPa
2.8 MPa
2.2 MPa
-6
2.4 MPa
2.6 MPa
2 MPa
0 MPa
2.2 MPa
2.8 MPa
2.4 MPa
-6
2.6 MPa
2.4 MPa
0 MaP
2 MPa
-8
-8 0 100 200 300 400 500 600 700 800 900 1000
0 100 200 300 400 500 600 700 800 900 1000
time [days]
time [days]
Figure 2.14: (a) Vertical and horizontal strain vs. time on prism 102-30A. (b) Deviatoric
and volumetric strain vs. time on prism 102-30A.
(a) (b)
Figure 2.15: Prism of the inner leaf: (a) face A at the beginning of the test; (b) crack
pattern of face B at the end of the test.
(i.e. the strain rate) increases and the curve becomes negative until collapse is reached;
negative deformation corresponds to dilation due to fracturing and cracks opening. The
creep behaviour is evident since the beginning from the deviatoric strain plots [7].
Two faces of the prism at the beginning and at the end of the test are shown in
Fig. 2.15a and b. The highly irregular texture of the wall is evident, with a great
part of the masonry being occupied by mortar. The crack pattern is characterized
by the presence of vertical and sub-vertical cracks.
2.2.5 Pseudo-creep tests on prisms of 100 100 180 mm
Twelve prisms of dimensions 100 100 180 mm were tested in compression by

subsequent load steps corresponding to 0.3 MPa kept constant for different time
intervals, respectively, of 300, 900, 3600 and 10800 s, as indicated in Table 2.4. An
initial step of 0.6 MPa was first applied. Before the application of any new load step,
the sample was completely unloaded so the unloading Young modulus could be eval-
uated [4]. Looking at Table 2.4, it is interesting to notice that the average peak stress
tends to decrease and the corresponding strain tends to increase at extending the time
interval, indicating that the stressstrain behaviour is strongly time-dependent.
The results of a test carried out at 10800 s are reported, as an example, in
Fig. 2.16: at each load step primary creep occurred and, during the last load step,
secondary and tertiary creep.
2.2.6 Pseudo-creep tests on prisms of 200 200 350 mm
A first series of four prisms, three coming from the external layers of the masonry
and one coming from the inner part constituting the trunk of the tower of Pavia,
was tested applying constant load steps of 0.25 MPa and keeping them constant for
10800 s. The test results are reported in Table 2.5 and in Fig. 2.17.
Strictly speaking, a direct comparison between these results and those previously
presented could not be done, due to various factors: different testing procedures,
different dimensions of the specimens, non-homogeneity of the material, different
texture of these prisms with respect to the previous ones. Anyway, this last aspect
seems to have played a major role, since higher values of peak stress and lower
values of the corresponding strain have on average been obtained in this case.
Table 2.4: Results of pseudo-creep tests on inner leaf prisms of 100 100 180 mm
at different time intervals.
Time s fv e fv s fv, ave e fv, ave
Sample interval (s) (MPa) (103) (MPa) (103)
18-10A 300 3.24 3.19 3.68 3.17
18-10B 4.18 3.60
18-10C 3.63 2.73
18-10D 900 2.69 4.71 2.73 4.20
18-10E 2.99 4.15
18-10F 2.52 3.75
18-10G 3600 2.89 5.48 2.47 4.12
18-10H 2.39 3.94
18-10I 2.15 2.96
18-10J 10800 2.59 5.78 2.68 5.87
18-10K 2.88 10.21
18-10L 2.58 1.65
3.0
2.5
2.0
v [N/mm2]
1.5
1.0
0.5 v (103)
0 3 6 9 12
0.0
60
t [min]
120
180
Figure 2.16: Pseudo-creep tests on a 100 100 180 mm prism of the inner leaf.
Table 2.5: Results of compression tests on medieval outer leaf prisms of 200 200
350 mm, at constant load step.
Sample Test interval (s) s fv (MPa) e fv (103)
40-20B 10800 5.25 1.17
40-20C 7.00 1.27
57-20A 4.50 1.80
Average 5.583 1.413
102-20A* 3.50 0.90
*Inner leaf, not considered for average calculation.
The horizontal strain takes higher absolute values than the vertical strain, indi-
cating that at failure dilation takes place. The results of a test on a single specimen
of the first series are shown in Fig. 2.18a and b. Considering the trend of the stress
strain plot (Fig. 2.18a), a linearly elastic behaviour can be observed below a stress
value of 3.25 MPa. Correspondingly, the straintime plot shows that within this
interval only primary creep develops. After that level, the stressstrain diagram
indicates non-linear behaviour; and the straintime plots exhibit the steady-state
(or secondary creep) and, eventually, the tertiary creep phases. The volumetric
strain (Fig. 2.18b) keeps almost naught values approximately during the first
twelve load steps and subsequently gets decreasing negative values until collapse.
As appears also from the data previously presented, decreasing volumetric strain
can be certainly interpreted as a sign of increasing material damage.
8
40-20b
v (103)
57-20a 40-20c
4 102-20a
0
-4
57-20a
-8
vol (103)
-12
102-20a 40-20c
-16
40-20b
-20
0 100000 200000 300000 400000
time [sec]
Figure 2.17: Pseudo-creep tests of the first series.
(a) 6 (b) 20
'v
v (Mpa)
4
10
2
v (103)
(103)
0
0 1 2 3 4 5 6 7 8
'h
-10
12000 8000 4000
time [sec]
-20
evol
-30
0 40000 80000 120000 160000 200000 240000
time[sec]
Figure 2.18: (a) Vertical stress vs. vertical strain and vertical strain vs. time on prism
40-20B. (b) Deviatoric and volumetric strains vs. time on prism 40-20B.
Figure 2.19 shows the crack pattern of a prism after the test. A more regular
texture than that appearing in Fig. 2.15 characterizes this sample, with the pres-
ence of whole bricks lying horizontally; nevertheless a great difference between
the four faces of the same prism has to be pointed out. The vertical cracks tend to
follow the directions corresponding to the interfaces between mortar and bricks
and split the prism faces from bottom to top.
A second series of pseudo-creep test was then carried out on additional prisms
recently obtained from the ruins of the tower. A total of four prisms coming from
the inner medieval masonry (labelled Q, B, M and S) and four coming from the
sixteenth-century solid masonry (labelled Ec, W, K and In) were tested applying
subsequent load steps of 0.3 MPa kept constant for intervals of 28800 s (Table 2.6).
On the average, higher peak stresses (sf) and lower strains at failure (evmax, ehmax)
were registered on the plain masonry.
Face A Face B Face C Face D
Figure 2.19: Prism of the outer leaf: crack pattern at the end of the test.
Table 2.6: Results of pseudo-creep tests of the second series.

sf evmax ehmax
Specimen Masonry (MPa) (m/mm) (m/mm)
Prism Q Sixteenth-century 6.07 9.52 20.39
Prism B solid masonry 4.36 5.98 26.46
Prism M 5.27 4.67 15.56
Prism S 4.13 7.19 17.23
Average 4.96 6.84 19.91
Prism Ec Medieval 3.59 8.43 11.84
Prism W inner leaf 2.75 4.00 7.89
Prism K 1.56 3.99 45.35
Prism In 2.47 4.78 18.30
Average 2.59 5.30 20.85
In Fig. 2.20 the results of the pseudo-creep tests of the second series on the inner
leaf and on the plain masonry are compared. As expected, the prisms of the inner
leaf reached failure well before the prisms of the plain masonry.
In Figs 2.21 and 2.22 the results of a test carried out on the masonry of the inner
leaf and those of a test carried out on the plain masonry are, respectively, shown.
In both cases, all the creep phases are visible.
2.3 Tests on the masonry of the crypt of the Cathedral

of Monza
2.3.1 Preparation of prisms of 200 200 350 mm
After the plane rearrangement of the Museum of the Cathedral of Monza in 1994,
the removed material, resulting from a door opening on the northern wall of the
10
Ec S B Q
v (103)
5 K In W M
0
-5
-10 W
Ec
-15
M
-20 In S
Q
-25
B
h (103)
-30
-35
K
-40
0 200000 400000 600000 800000
time [sec]
Figure 2.20: Pseudo-creep curves: (---) inner leaf, () solid masonry.
5 5
4 4
v [N/mm2]
v [N/mm2]
3 3
2 2
1 1
0 0
10000 10000
20000 20000
t [sec]
30000 30000
t [sec]
h v h v
40000 40000
-45 -40 -35 -30 -25 -20 -15 -10 -5 0 5 10 -45 -40 -35 -30 -25 -20 -15 -10 -5 0 5 10
[m/mm] [m/mm]
Figure 2.21: Results obtained on prism Figure 2.22: Results obtained on

K of the inner leaf. prism B of the solid masonry.
crypt, was collected for experimental testing. According to historical informa-

tion, the construction of the crypt (concluded in 1577) is nearly contemporary
to the construction of the bell tower (15921605), a building made of solid brick
masonry that was badly damaged by compression and is now undergoing a repair
intervention [8]. As it has become clear after the collapses of the last fifteen years,
towers as well as pillars of the cathedrals turn out to be particularly vulnerable to
the effects of persistent loading; therefore, achieving a better experimental knowl-
edge on their creep behaviour became crucial. Since considerable amounts of his-
torical masonry are not normally available, it was a good opportunity of gaining
original sixteenth-century masonry (Fig. 2.23).
Two large blocks were extracted by coring their perimeter in order to obtain as
much as possible undisturbed material. Subsequently, they were cut by a diamond
Figure 2.23: Sampling of the masonry from the crypt of the Cathedral of Monza.
(a) (b) (c)
Figure 2.24: Cutting scheme of the masonry sampled from the crypt of the
Cathedral of Monza.
saw into prisms of dimensions 200 200 350 mm to be subjected to different

mechanical testing (Fig. 2.24).
As shown in Fig. 2.24c, the crypt is mainly made of solid brick masonry appar-
ently regularly laid which nevertheless includes stones, voids and cracks, and has
therefore to be considered as a non-homogeneous material. In fact, the different
prisms appeared damaged to rather different extents, those subjected to monotonic
tests showing the most evident signs of damage.
2.3.2 Characterization by sonic tests
Before mechanical tests, the prisms were characterized by sonic tests as described
in Section 2.2.1. The results are reported in Fig. 2.25.
m/s
2500
2000
1500
1000
500
0
I5 II 8 II 2 I8 II 5 II 16 II 4 II 12 I4 II11
Figure 2.25: Results of sonic tests.
Table 2.7: Results of monotonic tests.

Sample sm (MPa) ev at sm Ei (MPa)
IIp6 3.70 4.41 2810
IIp9 3.75 5.21 2110
IIp13 3.50 5.38 1775
2.3.3 Monotonic tests
Monotonic tests on three prisms were carried out initially, to have a first indication
on the static compressive strength of the masonry; the tests were performed
in displacement control with a velocity of 1 m/sec and data acquisition every
1020 sec. The results obtained are reported in Table 2.7, where the peak stress
(sm), the vertical strain at peak stress, the initial elastic modulus and the test
duration to reach sm are indicated. In Fig. 2.26 stress vs. vertical and volumetric
strain diagrams are plotted for all the tested samples.
2.3.4 Fatigue tests
Three fatigue tests were carried out in load control. The samples were loaded
monotonically up to a stress value of 2.25 MPa, equal to 65% of the average static
compressive strength previously obtained with monotonic tests; cyclic actions
of 0.15 MPa at 1 Hz were then applied for a period of 5400 s. After this, the
vertical stress was increased of 0.25 MPa, a new cyclic phase was applied and the
sequence repeated until failure. When a single test lasted more than one day, the
sample was unloaded at night for safety reasons and reloaded the day after.
The results obtained are reported in Table 2.8 where the maximum vertical stress,
the vertical strain at failure, the initial elastic modulus calculated during the mono-
tonic phase and the test duration are shown. The ratio between the vertical stress at
4
IIp9
IIp6 IIp9
IIp6
3
IIp13
IIp13
v [MPa]
0
-15 -10 -5 0 5 10 15
vol (103) v (103)
Figure 2.26: Monotonic tests.
Table 2.8: Results of fatigue tests.

sm ev (103) Test duration
Sample (MPa) at failure Ei (MPa) sm/sm* (no. of cycles)
Ip1 5.00 9.21 3118 0.97 110286
Ip6 5.00 6.13 3774 0.81 51507
Ip7 4.25 4.14 2805 0.92 24243
failure and the estimated static maximum stress (sm*) is also indicated the latter hav-
ing been calculated on the basis of the initial elastic modulus. In fact, the average
ratio between the initial elastic modulus and the maximum vertical stress obtained
with monotonic tests on samples IIp6 and IIp9 was used for the calculation; the
results obtained on prism IIp13 were not included in calculating the average, as the
stressvolumetric strain diagram of this sample showed a very dilatant behaviour,
probably due to some local effects of the material. Of course, the estimated values of
sm* cannot be considered statistically relevant, but still they can give an indication on
the severity of the testing procedure with respect to the strength of the material [9].
In Fig. 2.27 vertical stress vs. vertical and volumetric strain diagrams obtained
on prism Ip1 are shown as an example. It can be observed that during the applica-
tion of the cyclic load a deformation takes place. Moreover, considering the volu-
metric strain, it appears that dilation occurs to the material after very low stress
values are reached.
The strain rate per cycle was calculated for the prisms tested cyclically, as pre-
sented by Taliercio and Gobbi [10] relatively to cyclic tests on concrete specimens.
4
v [MPa]
0
-24 -20 -16 -12 -8 -4 0 4 8 12 16
vol (103) v (103)
Figure 2.27: Results of a fatigue test carried out on prism Ip1.
2.5
Ip1
2.0
Ip6
v/ n (106)
1.5 Ip7
1.0
0.5
0.0
0 2000 4000 6000 8000 10000
cycles
Figure 2.28: Strain rate vs. number of cycles in the last series of cycles.
It was interesting to notice that plotting the strain rate vs. the number of cycles allows
the primary, secondary and tertiary creep phases to be distinguished quite clearly.
Figure 2.28 shows the results relative to the last series of cycles obtained on each
prism: though the number of test results is not particularly significant, a relationship
between the strain rate of the secondary creep phase, corresponding to the portion of
the diagram with horizontal tangent, and the fatigue life of the material, correspond-
ing to the total number of cycles at failure, can be found. In particular, the higher the
strain rate of the secondary creep phase, the shorter the fatigue life.
2.3.5 Creep test on one prism of 300 300 510 mm
The creep test was carried out at the ENEL-CRIS laboratory (Milan) in an espe-
cially designed apparatus, in controlled conditions of 20C temperature and 50%
RH. During the 630-day test, three load increments of 1.4, 2 and 2.25 MPa and an
unloading phase were applied. In Fig. 2.29 the vertical and volumetric strain are
plotted vs. time.
It appears that after load removal, it took more than 100 days for the material to
completely recover the accumulated creep strain.
2.3.6 Pseudo-creep tests, first series
Compression tests in displacement control were carried out loading the prisms
monotonically until a stress value of 2.25 MPa equal to 65% of the average short
term strength obtained by monotonic tests and then applying the load in subse-
quent steps kept constant for periods of about 5400 s, during which creep strain
took place. In this case also unloading reloading cycles had been necessary over-
night. Table 2.9 shows the results obtained with this test series: the maximum
vertical stress, the vertical strain at failure, the initial elastic modulus calculated
during the monotonic phase, the test duration and the ratio between the maximum
vertical stress and the static maximum stress are reported.
Figure 2.30 shows the diagrams obtained from all the tested prisms and
Fig. 2.31 shows the results obtained on a single prism. Results similar to those
obtained by the pseudo-creep tests previously presented can be seen. A clear dila-
tancy phenomenon is evident, with considerable volumetric strain developing
when approaching failure, as a typical feature of brittle materials.
4
v (103)
-1
vol (103)
-2
2.25 MPa
2 MPa
1.4 MPa
0 MPa
-3
-4
0 90 180 270 360 450 540 630
time[days]
Figure 2.29: Creep test.

Table 2.9: Results of pseudo-creep tests, first series.

sm ev (103) eh (103) Ei
Sample (MPa) at failure at failure (MPa) sm/sm*
IIp4 2.75 7.57 7.57 2122 0.79
Ip5 2.80 5.07 5.07 2159 0.79
Ip4 4.25 8.87 8.87 3024 0.85
IIp11 5.30 3.98 3.98 4298 0.75
10
I 4p
II 4p
I 5p
5 II 11p
v (103)
-5
vol (103)
-10
-15
I 5p
I 4p
II 4p II 11p
-20
0 20000 40000 60000 80000 100000 120000 140000 160000 180000 200000
tempo [sec]
Figure 2.30: Pseudo-creep tests: first series.
Comparing the strength values obtained on the masonry of Monza with the dif-
ferent test series, it appears unexpectedly that the prisms tested monotonically
showed the lowest strength values, which apparently is not coherent with viscous
behaviour. Actually, some aspects which may have influenced the results have
to be considered: first of all, masonry does not fulfil the conditions of continu-
ity, homogeneity and isotropy which have normally to be assumed in continuous
mechanics. In addition, the masonry studied here is an ancient one; at present it is
damaged. Moreover, the amount of cracks already present before testing was not
the same for all the samples, those tested monotonically showing more signs of
damage. However, if the ratio between the actual maximum vertical stress and the
static strength (sm/sm*) is evaluated through the initial elastic modulus as shown
v [MPa] 4
2
v(103)
1 2 3 4 5 6 7 8 9
0
2000
time [sec]
4000
6000
8000
Figure 2.31: Results of a compression test with constant load steps on prism Ip4.
Table 2.10: Results of pseudo-creep tests, second series.

Sample sm (MPa) ev (103) at failure eh (103) at failure
II8pN 2.00 13 16
II2pN 3.25 4 9
I8pN 4.00 6 9
II16pS 4.00 4 12
II12pS 4.25 8 9
II5pS 4.75 6 9
in Tables 2.8 and 2.9, it appears that cyclic and constant load step tests damaged
the material lowering its strength on average by 85%.
2.3.7 Pseudo-creep tests, second series
The second series followed a more regular procedure: no monotonic phase was
carried out; the load steps were applied since the start of the test and kept constant
for a time interval of 10800 s (Table 2.10). As shown in Figs 2.32 and 2.33 more
regular data were obtained.
Considering the trend of the stressstrain plot, it can be noted that in this case
also the behaviour can be considered linearly elastic below a stress value of 2.5
MPa. Correspondingly, the straintime plot shows that within this interval only
primary creep develops. After that level, the stressstrain diagram indicates
10
II 12
v (103)
5
II 8 II 2 II 16 I 8 II 5
-5
II 16 II12
-10 II 5
vol (103)
I8
II 2
-15
II 8
-20
0 40000 80000 120000 160000 200000
Time [sec]
Figure 2.32: Pseudo-creep tests: second series.
4
v[MPa]
0
3600
7200
Time [sec]
10800
14400
18000
21600
-10 -8 -6 -4 -2 0 2 4 6 8 10
vol (103) v(103)
Figure 2.33: Results obtained on prism II 12.
non-linear behaviour, and correspondingly the straintime plots exhibit the steady-
state (or secondary) and, eventually, the tertiary creep phases. In Fig. 2.34 the
crack pattern across the specimen at the end of the test is represented.
It is apparent from the drawings that the prism was characterized by the pres-
ence of a large portion of stone, occupying most of face D and large parts of faces
A and C. The crack pattern has basically developed in sub-vertical direction, with
II 8 II 2 II 16 I 8 II 8 II16 I 8 II 8 II2 II 16 II 8 II 2 II 16 I 8
II2 I8
II 16 II 16 I II II 16
II 16
I8 I8 I8 I8
II 2 II 2 II 2 II 2
II8 II 8 II 8 II8
Figure 2.34: Crack pattern of prism II 12 at the and of the test, faces A, B, C and D.
fissures opening preferably along discontinuities already present at start of the test,
whereas bricks were not cracked.
2.4 Comments
The effect of persistent loads on the damage of ancient masonry has been experi-
mentally studied and their effects on the mechanical properties of the material
have been shown. Constant load step tests turned out to be a suitable procedure for
analysing creep behaviour, having the advantage of being carried out more easily
than long-term tests.
Primary, secondary and tertiary creep phases have been clearly observed,
together with their relationship with the stress level, a damage development being
associated to an increase of the stress level.
The action of cyclic and persistent loads has proved to cause a severe damage
on the mechanical properties of ancient masonry. The fatigue life of masonry
under uniaxial cyclic compression is related to the secondary creep strain rate,
which is the strain rate during the phase of stable cyclic damage growth.
Having tested the masonry coming from two historical buildings and built by
different constructive techniques, a comparison can be made by using different
mechanical parameters. In Fig. 2.35 the compressive peak stress obtained through
pseudo-creep tests have been plotted vs. the sonic velocity.
A strong linear correlation can be clearly observed; the prisms obtained from
the outer leaf of the medieval part of the tower of Pavia achieved the highest
strength values, whereas those from the inner leaf present the lowest ones, being
the sixteenth-century plain masonry in between. In fact, this is coherent with the
texture characteristics of the three materials (Figs 2.19 and 2.24). The outer leaf is
the highest quality one, built to be the visible part of the structure; the inner leaf is
constituted by rubble masonry. The values of the crypt of Monza, the texture of
which shows an intermediate pattern between the previous two, with the presence
of bricks and stones, are overlapped to those of the inner leaf and those of the plain
masonry.
10
Monza
Pavia
8
peak stress [MPa]

6
0
0 500 1000 1500 2000 2500 3000
sonic velocity[m/sec]
Figure 2.35: Peak stress of pseudo-creep tests vs. sonic velocity.
1E+001
Monza
Pavia
v /t [x106/sec]
1E+000
1E-001
1E-002
1E-003
100 1000 10000 100000
Time [sec]
Figure 2.36: Secondary creep strain rate before failure vs. duration of the last load
step.
Considering the last load step for each specimen tested with pseudo-creep tests,
the secondary creep rate, which is the strain rate during the phase of stable damage
growth, has been calculated before collapse and then related to the duration of the
last load step, which can be regarded as the residual life of the material. In Fig.
2.36 these values for the masonry of Monza and the ruins of the tower of Pavia
have been plotted.
Though the number of test results is not particularly large, an interesting inverse
relationship can be found, which seems to apply to both the materials considered,
as well as to other brittle materials subjected to creep and fatigue tests [10, 11]. In
this respect a useful comparison with concrete behaviour could be made since the
analysed relationship is well known in the case of concrete [12, 13]. A strong cor-
relation exists between creep time to failure and secondary creep rate which,
accordingly, can be used as a reliable parameter to predict the residual life of a

material subjected to a given sustained stress. In view of preserving the historical
heritage, it would be useful to define similar relationships to evaluate, for instance,
the results of a monitoring campaign on a massive historic building subjected to
persistent load, to judge whether the creep rate indicates a critical condition in
terms of safety assessment. Of course, the precocious recognition of a critical state
will allow one to design a strengthening intervention to prevent total or partial
failure of the construction.
Acknowledgements
Architects C. Curallo, M. Garau, P. Garau, L. Giannecchini, M.G. Paccapelo,
R. Tassi and S. Sironi are gratefully acknowledged for their assistance in the
experimental work and data elaboration; Mr Marco Antico for his technical
support in the laboratory. The research has been partially supported by COFIN
2002 MIUR funds.
References
[2] Anzani, A., Binda, L. & Taliercio, A., Application of a damage model to
the study of the long term behavior of ancient towers. Proc. 1st Canadian
Conference on Effectiveness Design of Structures, Hamilton, Ontario
1013/07/2005, CD ROM.
[3] Anzani, A., Binda, L. & Melchiorri, G., Time dependent damage of rubble ma-
sonry walls. Proceedings of the British Masonry Society, 2(7), pp. 341351,
1995.
[4] Anzani. A., Mirabella Roberti. G. & Binda, L., Time dependent behav-
iour of masonry: experimental results and numerical analysis. Structural
Repair and Maintenance of Historical Buildings, Vol. III, Bath: STREMA,
pp. 415422, 1993.
[5] Anzani, A. & Mirabella Roberti, G., Experimental research on the creep
behaviour of historic masonry. Struct. Studies Repairs and Maintenance of
Heritage Architecture VIII, ed. C.A. Brebbia, WIT Press: Southampton and
Boston, pp. 121130, 2003.
[6] Binda, L. & Anzani, A., The time-dependent behaviour of masonry prisms:
an interpretation. The Masonry Society Journal, 11(2), pp. 1734, 1993.
[7] Anzani, A., Binda, L. & Mirabella Roberti, G., The behaviour of ancient
masonry towers under long-term and cyclic actions. Proc. Computer Meth-
ods in Structural Masonry, 4, pp. 236243, 1998.
[8] Modena, C., Valluzzi, M.R., Tongini Folli, R. & Binda, L., Design choices
and intervention techniques for repairing and strengthening of the Monza
Cathedral bell-tower. Proc. 9th Int. Conf. and Exhibition, Structural Faults +
Repair, CD-ROM, 2001.
[9] Mirabella Roberti, G., Binda, L. & Anzani, A., Experimental investigation
into the effects of persistent and cyclic loads on the masonry of ancient
towers. Proc.7th Int. Conf. and Exhibition, Structural Faults + Repair 97,
Edinburgh, Vol. 3, pp. 339347, 1997.
[10] Lenczner, D., The effect of strength and geometry on the elastic and creep
properties of masonry members. Conf. of North American Masonry, Boulder,
CO, 1978, 23-1/23-15.
[11] Xiao, X. & Shrive, N.G., Compressive fracture of masonry. Experimental
study. 10th Canadian Masonry Symposium, Banff, Alberta, 812 June 2005.
[12] Taliercio, A. & Gobbi, E., Effect of elevated triaxial cyclic and constant
loads on the mechanical properties of plain concrete. Magazine of Concrete
Research, 48(176), pp. 157172, 1996.
[13] Taliercio, A. & Gobbi, E., Fatigue life and change in mechanical proper-
ties of plain concrete under triaxial deviatoric cyclic stresses, Magazine of
Concrete Research, 50(3), pp. 247256, 1998.
CHAPTER 3
Collapse prediction and creep effects
P.B. Loureno & J. Pina-Henriques

Department of Civil Engineering, University of Minho, Portugal.
3.1 Introduction
The collapse of the Pavia Civic Tower in 1989 was an event of great concern for the
public authorities and for the technical community. The collapse rapidly became
a focus of interest among masonry researchers and several masonry blocks were
recovered from the ruins for mechanical and physical/chemical laboratory testing
[1]. Such tests permitted to identify the time-dependent mechanical damage of the
tower walls due to high sustained loading as the possible main cause of collapse.
The tower of Pavia is not an isolated case and several other famous examples
can be mentioned, such as the collapse of the St. Magdalena bell-tower in Goch,
Germany, in 1993, the partial collapse of the Noto Cathedral, Italy, in 1996, and
the severe damage exhibited by the bell-tower of the Monza Cathedral, Italy.
Masonry creep depends mainly on factors such as the stress level and the
temperature/humidity conditions but cyclic actions, such as wind, temperature
variations or vibrations induced by traffic or ringing bells, in the case of bell towers,
have a synergetic effect, increasing material damage. For these reasons, high towers
and heavily stressed columns are the structural elements where severe time-dependent
damage can occur [2].
Traditionally, three creep stages can be recognized: a primary stage where the
creep rate decreases gradually, a secondary stage where the creep rate remains
approximately constant and a tertiary stage where the creep rate increases rapidly
towards failure. A sufficiently high stress level must be applied so that the last two
stages are initiated. In the secondary stage, diffuse and thin vertical cracking prop-
agates and coalesces into macro-cracks that may lead, possibly, to creep failure of
the material. Creep of cementitious materials is generally attributed to crack
growth and interparticle bond breakage due to moisture seepage. In fact, under
sustained loading, forced moisture redistribution can occur in the pore structure of
the material causing debonding and rebonding of the micro-structure particles [3].
In the case of concrete or new masonry, if drying shrinkage is occurring simultane-
ously to creep, the time-dependent deformation is increased due to a coupled effect
known as the Pickett effect [4]. Time-dependent deformation in a constant hygral
and thermal environment, and in the absence of cracking, is denominated by basic
creep (see, e.g., Neville [5], and the reader is referred to Van Zijl [6] for a compre-
hensive discussion on the viscous behaviour of masonry).
For low stress levels, below 40 to 50% of the compressive strength, only pri-
mary creep is present and creep deformation can be assumed proportional to the
stress level. References on masonry creep within the elastic range are rather abun-
dant in literature, [79]. On the contrary, creep under high stresses, even in the
case of concrete, is not a sufficiently debated issue [1012]. The fact that standard
design methods for new structures are based on linear elastic material hypothesis
has contributed to the diminished interest of researchers on this topic. However,
ancient masonry structures are often functioning under low safety margins accord-
ing to modern safety regulations. This can be due to inadequate knowledge of
mechanics or due to the structural modifications that occurred along centuries,
resulting in overweighting of the structure and rendering importance to non-linear
creep.
Detailed modelling approaches in which units and mortar are individually rep-
resented are of great interest in understanding the phenomena occurring at the
constituent level. In particular, insight into stress redistribution and damage growth
occurring under sustained loading can be provided. However, before introducing
long-term effects, knowledge of the short-term behaviour and governing failure
mechanisms is of fundamental importance. Nevertheless, numerical prediction of
the short-term response of masonry based on the properties of the constituents
remains unresolved [13]. In Section 3.2, an attempt to provide reliable predictions
of the compressive strength of masonry and a discussion of the observed failure
mechanisms are discussed. A standard continuum model, based on plasticity and
cracking, and a particle model developed in discrete settings have been considered
to represent units and mortar. In Section 3.3, an experimental investigation into the
behaviour of ancient masonry under high compressive stresses is described and its
results are carefully analysed. Standard uniaxial compression tests, short-term
creep tests and long-term creep tests were considered with the aim of presenting a
comparative discussion.
3.2 Short-term compression: failure analysis and

collapse prediction using numerical simulations
3.2.1 Experimental results
Binda et al. [14] carried out deformation controlled tests on masonry prisms
with dimensions of 600 500 250 mm3, built up with nine courses of
Collapse Prediction and Creep Effects 59
Table 3.1: Mechanical properties of the masonry components [14].

Component E (N/mm2) v fc (N/mm2) ff (N/mm2)
Unit 4865 0.09 26.9 4.9
Mortar M1 1180 0.06 3.2 0.9
Mortar M2 5650 0.09 12.7 3.9
Mortar M3 17760 0.12 95.0 15.7
Table 3.2: Mechanical properties of the masonry prisms [14].

Prism type Mortar type E (N/mm2) fc (N/mm2)
P1 M1 1110 11.0
P2 M2 2210 14.5
P3 M3 2920 17.8
250 120 55 mm3 solid soft mud bricks and 10 mm thick mortar joints. Three
different types of mortar, denoted as M1, M2 and M3, have been considered and
testing aimed at the evaluation of the compressive properties of the prisms. For
each type of mortar, a total of three prisms were tested.
The tests were carried out in a uniaxial testing machine MTS 311.01.00, with
non-rotating steel plates and a maximum capacity of 2500 kN. The applied load
was measured by a load cell located between the upper plate and the testing
machine, while the average vertical displacement was recorded with the machines
in-built displacement transducer, permitting one to capture the complete stress
strain diagram, including the softening regime.
The characteristics of the masonry components in terms of compressive strength
fc, flexural tensile strength ff, elastic modulus E and coefficient of Poisson v are
given in Table 3.1. The results obtained for the prisms are given in Table 3.2.
Prisms P1, P2 and P3 were built with mortars M1, M2 and M3 of increasing
strength, respectively. The experimental failure patterns found were rather similar
despite the type of mortar used [15]. Figure 3.1 depicts the typical failure pattern.
3.2.2 Continuum model
The simulations were carried out resorting to a basic cell, i.e. a periodic pattern
associated to a frame of reference (see Fig. 3.2), in which units and mortar were
represented by a structured continuum finite element mesh. However, in order to
reduce computational effort only a quarter of the basic cell was modelled assuming
symmetry conditions for the in-plane boundaries (see Fig. 3.3). The dimensions
of the components are equal to the ones used in the experiments. It is empha-
sized that the adopted approach is only an approximation of the real geometry
Figure 3.1: Typical experimental failure patterns [15]. The shaded areas indicate
spalling of material.
130
260
(a) (b)
Figure 3.2: Definition of basic cell: (a) running bond masonry and (b) geometry.
Unit
Mortar
Figure 3.3: Continuum model used in the simulations (only the indicated quarter
was simulated, assuming symmetry conditions).
and that the obtained numerical response is phenomenological, which means that a
comparison in terms of experimental and numerical failure patterns is not possible.
In particular, splitting cracks usually observed in prisms tested under compression
[16], boundary effects of the specimen and non-symmetric failure modes are not cap-
tured by the numerical analysis. Nevertheless, most of these effects control mainly
the post-peak response, which is not the key issue in the present contribution.
Three different plane approaches have been considered taking into account the
out-of-plane boundaries, namely: (a) plane-stress (PS), (b) plane-strain (PE) and (c)
an intermediate state, here named enhanced-plane-strain (EPE). This last approach
consists in modelling a thin out-of-plane masonry layer with 3D elements, imposing
equal displacements in the two faces of the layer. Full 3D analyses with refined
meshes and softening behaviour are unwieldy, and were not considered. Moreover,
recent research indicated that enhanced-plane-stress analysis provides very similar
results [17]. EPE response is always between the extreme responses obtained with
PS and PE. For this reason, EPE is accepted as the reference solution for the con-
tinuum simulations and only its results are considered in this paper. A complete
description of the continuum simulations can be found in Pina-Henriques and
Loureno [18].
Modelling of the cell in EPE was carried out using approximately 900 20-noded
brick elements with 6650 nodes, totalling 13,300 degrees of freedom (note that the
tying adopted for the out-of-plane degrees of freedom mean that, basically, a 2D
model is used). The integration used was 3 3 3 Gauss. The material behaviour
was described using a composite model including a traditional smeared crack
model in tension, specified as a combination of tension cut-off (two orthogonal
cracks), tension softening and shear retention [19], and a Drucker-Prager plasticity
model in compression [20]. The inelastic behaviour exhibits a parabolic harden-
ing/softening diagram in compression and an exponential-type softening diagram
in tension. The material behaves elastically up to one-third of the compressive
strength and up to the tensile strength.
In order to reproduce correctly the elastic stiffness of the masonry prisms, the
experimental elastic modulus of the mortar E must be adjusted by inverse fitting.
In fact, the mortar experimental stiffness leads to a clear overstiff response of the
numerical specimens. This can be explained by the fact that the mechanical prop-
erties of mortar inside the composite are different from mortar specimens cast
separately. This is due to mortar laying and curing and represents a severe draw-
back of detailed micro-models. The material properties adopted, including the
adjusted mortar stiffness values E* are fully detailed in Table 3.3. Here, c is the
cohesion, ft is the tensile strength, f is the friction angle, y is the dilatancy angle,
Gft is the tensile fracture energy and Gfc is the compressive fracture energy. The
value adopted for the friction angle was 10 (a larger value in plane-stress would
implicate an overestimation of the biaxial strength) and, for the dilatancy angle, a
value of 5 was assumed [21]. The values assumed for the fracture energy have
been based on recommendations supported by experimental evidence [22, 23], and
practical requirements to ensure numerical convergence.
Table 3.3: Inelastic properties given to masonry components.

E* c ft Gft Gfc
Component (N/mm2) (N/mm2) (N/mm2) sin f sin y (N/mm) (N/mm)
Unit 4865 11.3 3.7 0.17 0.09 0.190 12.5
Mortar M1 355 1.3 0.7 0.17 0.09 0.350 2.7
Mortar M2 735 5.3 3.0 0.17 0.09 0.150 10.0
Mortar M3 1065 39.9 12.0 0.17 0.09 0.600 23.0
*In the case of mortars, the values refer to the adjusted stiffness values.
3.2.3 Particle model
The 2D particle model supposed to represent the micro-structure of units and

mortar consists of a phenomenological discontinuum approach based on the
finite element method including interface elements. The discontinuous nature of
the masonry components is considered by attributing a fictitious micro-structure
to units and mortar, which is composed by linear elastic continuum elements of
polygonal shape (hereafter named particles) separated by non-linear interface ele-
ments. All inelastic phenomena occur in the interfaces and the process of fractur-
ing consists of progressive bond-breakage. This is, of course, a phenomenological
approach, able, nevertheless, to capture the typical failure mechanisms and global
behaviour of quasi-brittle materials. For a detailed discussion of the model, includ-
ing proposals for selection of numerical data, sensitivity studies, fracture pro-
cesses and failure mechanisms, and size effect studies, the reader is referred to
Pina-Henriques and Loureno [13]. There, it is also shown that the compressive
and tensile strength values yielded by the model can be considered as independent
from particle size and particle distortion for practical purposes.
The constitutive model used for the interface elements was formulated by Lou-
reno and Rots [24] and is implemented in the finite element code adopted for the
analyses [20]. The model includes a tension cut-off for tensile failure (mode I), a
Coulomb friction envelope for shear failure (mode II) and a cap mode for com-
pressive failure. Exponential softening is present in all three modes and is pre-
ceded by hardening in the case of the cap mode. Micro-structural disorder is
considered in the model by the irregular geometry of the particles and by attribut-
ing to particles and interfaces randomly generated material properties, according
to a Gaussian distribution, for given values of the average and coefficient of varia-
tion of the material parameters.
The particle model simulations were carried out employing the same basic cell
used for the continuum model (see Fig. 3.2). The particle model is composed by
approximately 13,000 linear triangular continuum elements, 6000 linear line inter-
face elements and 15,000 nodes (see Fig. 3.4). Macro homogeneous symmetry
conditions have been assumed.
The material parameters were defined by comparing the experimental and numeri-
cal responses of units and mortar considered separately. Each material was modelled
Mortar
Unit
Figure 3.4: Particle model of the masonry cell (only the quarter indicated was
simulated, assuming symmetry conditions).
resorting to specimens with the same average particle size, mesh distortion and
dimensions of the masonry components used in the composite model (basic cell).
Given the stochastic nature of the model, five simulations were performed for
each masonry component assuming equal average values for the model material
parameters. The parameters were obtained, whenever possible, from the experimen-
tal tests described in Section 3.2.1, but most of the inelastic parameters were unknown
and had to be estimated. For the particles, average elastic modulus E values larger
than the experimental ones had to be adopted due to the contribution of the interfaces
deformability, characterized by kn and ks, to the overall deformability of the speci-
men. This correction is necessary despite the high dummy stiffnesses assumed.
On the contrary, the values adopted for the interfaces tensile strength ft are slightly
lower than the experimental tensile strength of the specimens, given the contribution
of the interfaces shear strength due to the irregular fracture plane. The cohesion c
was taken, in general, equal to 1.5 ft [25]. However, quite low experimental ratios
between the compressive and tensile strengths were reported for the units and mor-
tars considered here, with values ranging between four and eight. Due to this reason,
cohesion values lower than 1.5 ft had to be adopted for mortars M1 and M2.
The values for the friction coefficient tanf were adopted so that the numerical
compressive strength showed a good agreement with the experimental strength.
The values assumed for mode I fracture energy GfI have been based on recommen-
dations supported in experimental evidence [23, 26]. For mode II fracture energy
GfII, a value equal to 0.5c was assumed, with the exception of the very high strength
mortar M3, for which a lower value equal to 0.3c was adopted. The complete
material parameters adopted are given in Table 3.4 and, for such input, the response
obtained is given in Table 3.5.
3.2.4 Discussion of the results
The numerical results obtained for the masonry prisms considering the mortar
experimental Num_E and adjusted Num_E* stiffnesses are given in Table 3.6,
Table 3.4: Values assumed for the material parameters (in brackets, the coefficient
of variation is given in %).
Unit M1 M2 M3
Particles E* (N/mm2) 6000 (30) 355 (30) 750 (30) 1200 (30)
v 0.09 (0) 0.06 (0) 0.09 (0) 0.12 (0)
Interfaces kn (N/mm3) 1 104 (0) 1 104 (0) 1 104 (0) 1 104 (0)
ks (N/mm3) 1 104 (0) 1 104 (0) 1 104 (0) 1 104 (0)
ft (N/mm2) 3.40 (45) 0.75 (45) 3.50 (45) 10.50 (45)
GfI (N/mm] 0.170 (45) 0.038 (45) 0.175 (45) 0.525 (45)
c (N/mm2) 5.10 (45) 0.30 (45) 0.70 (45) 15.75 (45)
GfII (N/mm) 2.55 (45) 0.15 (45) 0.35 (45) 3.15 (45)
tan f 0.10 (45) 0.00 (0) 0.00 (0) 0.10 (45)
*In the case of mortars, the values refer to the adjusted stiffness values.
Table 3.5: Numerical response obtained for the masonry components (in brackets,
the coefficient of variation is given in %).
Unit M1 M2 M3
fc (N/mm2) 27.2 (2.7) 3.2 (5.0) 12.7 (5.4) 95.8 (4.4)
ft (N/mm2) 3.61 (1.4) 0.64 (4.7) 2.70 (4.2) 11.62 (6.6)
E (N/mm2) 4786 (1.9) 1309 (1.4) 5632 (3.0) 17176 (3.1)
Table 3.6: Experimental results (Exp) and numerical results using experimental
Num_E and adjusted Num_E* mortar stiffness values.
Continuum model Particle model
Prism type P1 P2 P3 P1 P2 P3
fc (N/mm2) Exp 11.0 14.5 17.8 11.0 14.5 17.8
Num_E 19.8 24.2 31.0 15.5 19.3 30.8
Num_E* 18.2 24.1 30.0 15.4 17.3 24.6
ep (103) Exp 10.5 7.9 6.6 10.5 7.9 6.6
Num_E 10.6 9.7 8.4 5.4 4.6 6.2
Num_E* 19.9 16.0 33.5 11.8 8.1 8.9
where fc is the compressive strength and ep is the peak strain. In addition, the prisms
experimental results are shown for a better comparison. It is noted, however, that the
reference solution for the numerical simulations is the solution provided by Num_E*.
Figure 3.5 depicts the experimental and numerical stressstrain diagrams.
From the given results, it is clear that the experimental collapse load is overesti-
mated by the particle and continuum models, and that the predicted strength is
affected by the mortar stiffness, especially in the case of the particle model. How-
ever, a much better agreement with the experimental strength and peak strain has
been achieved with the particle model, when compared to the continuum model. In
fact, the numerical over experimental strength ratios ranged between 165 and 170%
in the case of the continuum model while in the case of the particle model, strength
ratios ranging between 120 and 140% were found. The results obtained also show that
the peak strain values are well reproduced by the particle model but large overesti-
mations are obtained with the continuum model. For this last model, experimental
over numerical peak strain ratios ranging between 190 and 510% were found.
Failure patterns are an important aspect when assessing numerical models. The
(incremental) deformed meshes near failure using the continuum and particle
models are depicted in Fig. 3.63.8 for prisms P1 to P3, respectively. In case of the
continuum model, the contour of the minimum principal plastic strains is also
given for a better interpretation of the mechanisms governing failure. It is noted
that despite the fact that only a quarter of the basic cell has been modelled the
results are shown in the entire basic cell to obtain more legible figures.
(a) 20.0
CM
16.0 PM
Stress [N/mm2]
12.0
8.0
Exp
4.0
0.0
0.0 5.0 10.0 15.0 20.0 25.0
Strain [10-3]
(b) 30.0
CM
24.0
Stress [N/mm2]
PM
18.0
12.0
6.0 Exp
0.0
0.0 4.0 8.0 12.0 16.0 20.0
Strain [10-3]
(c) 35.0
28.0 PM CM
Stress [N/mm2]
21.0
14.0
Exp
7.0
0.0
0.0 4.0 8.0 12.0 32.0 36.0
Strain [10-3]
Figure 3.5: Numerical and experimental stressstrain diagrams, using adjusted

mortar stiffness values, for prisms: (a) P1, (b) P2 and (c) P3. In the
diagrams CM stands for continuum model, PM for particle model and
Exp for experimental data.
The numerical failure patterns obtained are similar for both continuum and par-
ticle models. Even if the proposed particle model approach is phenomenological,
the failure patterns resemble well typical compression experimental patterns
observed in the face of masonry specimens. In the case of prism P1, failure occurs
mainly due to the development of vertical cracks in the centre of the units and along
the head-joints, being the mortar in the bed-joints severely damaged. Prism P2
fails due to diffuse damage developing in units and mortar in a rather uniform
manner. In the case of prism P3, diffuse damage is also present but localized crush-
ing of the units can be clearly observed at one-half and one-sixth of the length of
the masonry units.
(a) (b)
(c)
Figure 3.6: Results at failure for prism P1 using the continuum model: (a) deformed
(incremental) mesh and (b) minimum principal plastic strains; and
using the particle model: (c) deformed incremental mesh.
(a) (b)
(c)
3.3 Long-term compression: experimental assessment

3.3.1 Tested specimens
The experimental investigation was carried out on ancient masonry prisms due
to the difficulty of producing laboratory specimens that correctly represent the
(a) (b)
(c)
Table 3.7: Number of specimens n

for each type of test.
n
Compression 4
Short-term creep 4
Long-term creep 6
material typically found in historical masonry structures. Major obstacles to fab-

ricate specimens are mortar carbonation, hardening or setting, which have a sig-
nificant influence on the viscous behaviour of masonry and cannot be adequately
reproduced in new specimens. On the other hand, the high cost and very limited
number of ancient masonry specimens available for destructive testing are obvi-
ous. As previous experience with similar materials in the scientific community is
not frequent [2], the current testing program was much relevant as it represents a
learning process. In particular, recommendations for testing such specimens could
only be given at the end of the testing program.
The experimental investigation carried out focuses on regular coursed brick
masonry specimens PRe recovered from the ruins of the belfry of the Pavia Civic
Tower. The dimensions of the specimens were (200 5) (200 5) (330
20) mm3. Before subsequent testing under compression, the loaded faces of the
prisms were regularized with a cement based mortar layer approximately 10 mm
thick. In all tests, Teflon sheets were introduced between the prisms and the loading
plates to minimize restraining frictional stresses. A summary of the tests performed
is given in Table 3.7.
3.3.2 Standard compression tests
Standard compression tests were conducted on four specimens. The tests were
partly carried out in University of Minho (specimens PRe_1 and PRe_2) and in
Politecnico di Milano (specimens PRe_3 and PRe_4). The specimens had to be
tested with different test setups according to the conditions locally available at
each laboratory. In this way, the tests performed in University of Minho were car-
ried out in a uniaxial hydraulic testing machine with non-rotating steel plates and a
maximum capacity of 2000 kN. The load was monotonically increased under dis-
placement control at the rate of 4 m/s. The applied load was measured by a load
cell located between the upper plate and the testing machine, and displacements in
the specimens were recorded by two vertical inductive displacement transducers
HBM (10 mm range), positioned at two different faces of the prisms and by two
horizontal transducers positioned at the other two faces.
The tests performed in Politecnico di Milano were carried out using a uniaxial
servo-controlled MTS 311.01.00 testing machine, with non-rotating steel plates
and a maximum capacity of 2500 kN. Loading was applied under displacement
control at a rate of 1 m/s. The applied load was recorded by a load cell and dis-
placements were measured with one vertical and one horizontal displacement
transducers GEFRAN PY2-10 (10 mm range) positioned at each face of the prisms.
For all tested specimens, longitudinal displacements were measured over approxi-
mately 200 mm span and transversal displacements over about 150 mm span.
The results obtained are illustrated in Fig. 3.9. Here, the negative sign is adopted
for contraction (longitudinal or vertical strains ev) and the positive sign is adopted
for elongation (transversal or horizontal strains eh). It is noted that the null hori-
zontal deformations exhibited up to the peak load in the case of specimen PRe_1
can be explained by the fact that only two horizontal transducers per specimen
were used. Table 3.8 gives a summary of the test results in terms of the elastic
modulus E, compressive strength fc and peak strain ep. The elastic modulus was
9.0
7.5
Stress [N/mm2]
6.0
4.5
3.0 PRe_1
PRe_2
PRe_3
1.5 PRe_4
h v
0.0
12.0 8.0 4.0 0.0 -2.0 -4.0 -6.0 -8.0
Strain [10-3]
Figure 3.9: Stressstrain diagrams obtained from standard compression tests.

Table 3.8: Results obtained from standard compression tests (in

brackets, the coefficient of variation is given).
Specimen E (N/mm2) fc (N/mm2) ep (103)
PRe_1 4980 8.0 2.7
PRe_2 4515 6.3 2.9
PRe_3 2510 5.7 2.2
PRe_4 2720 6.2 3.0
Average 3680 (34%) 6.6 (15%) 2.7 (13%)
calculated as the average slope of the stressstrain diagram between 30 and 50%
of fc. It is noted that the elastic modulus is the parameter showing the largest coef-
ficient of variation, approximately the double of the values found for the strength
and peak strain.
3.3.3 Short-term creep tests
Experimental set-up
Short-term creep tests were carried out at Politecnico di Milano using, again, the
uniaxial servo-controlled MTS 311.01.00 testing machine. The experimental tests
considered in this section are part of an extensive testing program under develop-
ment at the Politecnico di Milano, which is thoroughly described in Chapter 2.
The displacements in the specimens were recorded by a vertical and a horizontal
displacement transducer GEFRAN PY2-10 (10 mm range) positioned in each face
of the prisms, in a total of eight transducers per specimen. Vertical transducers
measured the average longitudinal deformation over approximately 200 mm span
and horizontal transducers measured the average transversal deformation over
approximately 150 mm span.
Testing program
A total of four specimens were tested for short-term creep. In standard creep tests,
a specimen is subjected to a constant load and strain is recorded at subsequent
times. Reproduction of the test with a series of different loads gives a family of
creep curves, which characterize the creep behaviour of the material. However, in
the case of ancient masonry, this procedure has severe drawbacks due to the high
scatter of the material strength and the limited number of specimens available. To
overcome these problems and to obtain as much information as possible from each
specimen, a stepped load-time diagram has been applied to the specimens.
The specimens were tested by applying successive load steps of 0.30 N/mm2 at
intervals of eight hours. In this way, failure could occur either during the loading
phase (short-term failure) or during sustained loading (tertiary creep). An attempt
to obtain creep failure of the specimens was pursued by increasing the duration of
the last steps whenever the strain rate was similar to the values observed in previ-
ously tested specimens. This issue will be further addressed in Section 3.3.5.
Test results
Figure 3.10a depicts the average vertical (longitudinal) and horizontal (transver-
sal) strains obtained, respectively ev and eh. Figure 3.10b illustrates, as an example,
the timestressstrain diagram for specimen PRe_5, which provides a detailed
description of the results. In addition, Table 3.9 gives a summary of the experi-
mental results in terms of the elastic modulus E, peak stress fc and time to failure
T, which corresponds to the duration of the creep test. The values for the elastic
modulus E were calculated as an average from the second to fourth load steps
(0.301.2 N/mm2).
The sample is too small to extract any conclusion. Nevertheless, the comparison
between the average standard compression strength ( fc = 6.6 N/mm2) and the
(a) -6.0 (b)

PRe_5
PRe_8 PRe_7
-6.0
-4.0
v [10-3]
-4.5
v [10-3]
PRe_6
-2.0 -3.0
-1.5
0.0
0.0
PRe_6
4.0
5.0
h [10-3]
h [10-3]
8.0 10.0
12.0 PRe_5 15.0

PRe_8
PRe_7
16.0 20.0
0.0 2.0 4.0 6.0 8.0 6.0 4.0 2.0 0.0 3.0 6.0 9.0
Time [days] [N/mm2] Time [h]
Figure 3.10: Results obtained from short-term creep tests: (a) straintime diagrams
for all tested specimens and (b) timestressstrain diagram for specimen
PRe_5.
Table 3.9: Results obtained from short-term creep tests (in brackets,
the coefficient of variation is given).
Specimen E (N/mm2) fc (N/mm2) T (days)
PRe_5 2700 4.50 4.7
PRe_6 3185 5.70 7.0
PRe_7 4075 5.40 6.1
PRe_8 3815 3.90 4.1
Average 3445 (18%) 4.9 (17%) 5.4
(a) (b) (c) (d)
Figure 3.11: Failure pattern for specimen PRe_7. Shaded areas indicate spalling/
loss of material.
average short-term creep strength ( fc = 4.9 N/mm2) seems to indicate that damage
growth due to sustained loading influenced the results. In terms of average elastic
modulus, the difference is rather small.
With respect to crack patterns, thin and diffuse vertical cracks developed in the
specimens during testing but large cracks and spalling were observed only at failure.
This failure mode is particularly dangerous as it can lead to erroneous conclusions
about the safety level of existing structures. Figure 3.11 illustrates, as an example,
the failure pattern for specimen PRe_7.
3.3.4 Long-term creep tests
Experimental set-up
Long-term creep tests require specific testing equipment able to keep the load
constant for long periods. In this study, three steel frames were specially designed
and built to perform the tests conducted at University of Minho, see Fig. 3.12a
and b. Each frame includes two loading steel plates, a hydraulic jack, a pressure
gauge and a gas reservoir to stabilize the applied load. The lower steel plate was
fixed, while the upper plate was hinged. The equipment was designed to test two
prisms simultaneously, separated by a steel plate. Upon failure of one of the speci-
mens, the equipment is unloaded to remove the failed specimen and re-loaded
with the remaining specimen. Further references on long-term creep tests can be
found in [27].
Longitudinal and transversal deformations were measured on each face of the
prisms with a removable strain-gauge Laser electronic TP (see Fig. 3.12c).
Longitudinal deformations were measured over three mortar bed-joints with an
approximate span of 250 mm, while transversal deformations were measured over
one head-joint with an approximate span of 145 mm. In addition, one inductive
transducer HBM (10 mm range) per specimen was employed in the longitudinal
direction to act as control of the strain-gauge measurements. It is noted that in the
face of the specimen where the transducer was placed, the transversal displacement
was not measured. In this way, the average longitudinal displacement of each
specimen results from four strain-gauge measurements, while the transversal
(a) (b) (c)
Figure 3.12: Testing apparatus: (a) hydraulic frame, (b) specimens under testing
and (c) removable strain gauge and contact seats glued to the specimen.
displacement results from three strain-gauge measurements. The tests were carried
out under controlled conditions of temperature (22 2C) and humidity (55
10%), which were recorded by a data logger Testostor 175-2.
Testing program
The tests were conducted on six specimens. As in short-term creep tests, the load
was applied by successive steps and kept constant for a given period. Two dif-
ferent load histories have been considered in order to better define future testing
programs in similar specimens. A total of two prisms were tested by applying an
initial stress of 1.50 N/mm2 and successive steps of 0.65 N/mm2. The initial load
step corresponds, approximately, to 25% of the compressive strength fc obtained
from the standard compression tests described in Section 3.3.2, while further load
steps correspond, approximately, to 10% of fc. The duration of each period under
constant load was of three months.
The other four specimens were initially loaded at 4.10 N/mm2 (approximately
60% of fc) with subsequent load increases of 0.65 N/mm2 (about 10% of fc), applied
at intervals of six months. Both load histories adopted have been defined in order
that the estimated duration of the tests would be of about two years.
Test results
Figure 3.13a illustrates the average vertical (longitudinal) and horizontal (transver-
sal) strains obtained for prisms tested with constant load periods of three months.
Table 3.10 gives a summary of the experimental results obtained.
For specimens tested with constant load periods of six months, Fig. 3.13b shows
the average straintime diagrams obtained for all tested prisms. In addition, Table
3.11 gives a summary of the results. Figure 3.14 illustrates, as an example, the
strain evolution at each face of specimen PRe_12 and, also, the timestressstrain
diagram for the same specimen. From Fig. 3.14a it is possible to observe that the
strain evolution is different at each face of the prism. This behaviour is typical of
compression tests in quasi-brittle materials but, in the present experiments, such
(a) (b)
-3.0 -4.5
?= 1.5 ?= 2.2 ?= 2.8 ?= 3.5 ?= 4.1 ?= 4.1 ?= 4.8 ?= 5.4 ?= 6.1
?= 4.8 ?= 6.7
-2.0 -3.0
v [10-3]
v [10-3]
-1.0 -1.5
89 days 95 days 91 days 91 days 97 days 184 days 187 days 188 days 180 days
0.0 0.0
1.0 2.0
h [10-3]
h [10-3]
2.0 4.0 PRe_11
PRe_9 PRe_12
PRe_10 PRe_13
PRe_14
3.0 6.0
0 100 200 300 400 500 0 200 400 600 800
Time [days] Time [days]
Figure 3.13: Straintime diagrams obtained from long-term creep tests: (a) constant
load periods of three months and (b) constant load periods of six
months. s stands for applied stress in N/mm2.
Table 3.10: Results obtained from long-term creep tests with constant
load periods of three months.
PRe_9 5055 4.75 465
PRe_10 4380 4.75 464
Average 4718 4.8 464
feature is more salient due to the hinged upper loading plate. Another important
aspect is that in some specimens cracks suddenly arise during constant load
steps, resulting in a strain jump in the straintime diagram, see, for example, the
diagrams of specimen PRe_10 at 325 days or PRe_11 at 450 days shown in
Fig. 3.13a and b, respectively.
The values obtained for the compressive strength are within the range obtained
for the standard compressive strength and short-term creep tests. Displacements
recorded with the transducers employed (one per specimen) were found to be in
agreement with the strain-gauge measurements.
Figure 3.15 depicts the crack pattern evolution for specimen PRe_13 as an
example. Again, diffuse vertical cracks developing during testing have been
Table 3.11: Results obtained from long-term creep tests with constant
load periods of six months (in brackets, the coefficient
of variation is given).
PRe_11 3720 6.05 559
PRe_12 5055 6.70 742
PRe_13 4345 6.70 749
PRe_14 3270 4.75 184
Average 4100 (19%) 6.0 (15%) 558
(a) (b)
-8.0
= 4.1 = 4.8 = 5.4 = 6.1
= 6.7 -4.0
-6.0
v [10-3]
-3.0
-4.0
v [10-3]
-2.0
-2.0
-1.0
0.0
0.0
2.0
1.0
h [10-3]
h [10-3]
4.0 2.0
A
6.0 B 3.0
C
D
8.0 4.0
0 200 400 600 800 8.0 6.0 4.0 2.0 0.0 40 80 120 160 200
Time [days] [N/mm2] Time [days]
Figure 3.14: Results from long-term creep tests on specimen PRe_12 (constant
load periods of six months): (a) strain evolution for each face and (b)
timestressstrain diagram. s stands for applied stress in N/mm2.
observed, with large cracks and spalling occurring near failure. It is noted that the
specimens with lower values of fc presented the most diffused crack patterns.
Severe non-uniform distribution of damage can be observed along the four faces
of the specimens, confirming the results shown in Fig. 3.14a.
3.3.5 Discussion of the results
The short-term compressive strength fc of each prism tested in creep is, of course,
unknown and can only be estimated. In this section, the peak stress values fc
(a) (b) (c) (d)
Figure 3.15: Failure pattern for specimen PRe_13. Shaded areas indicate spalling/
loss of material.
(a) (b)
0.20 0.20
0.16 0.16
Creep coefficient [-]
Creep coefficient [-]
0.12 0.12
0.08 0.08
0.04 0.04
0.00 0.00
0.0 2.0 4.0 6.0 8.0 0 20 40 60 80 100
Time [h] Time [days]
Figure 3.16: Variation of the creep coefficient with time obtained from: (a) short-
term creep results and (b) long-term creep results (constant load
periods of three months).
obtained from the creep tests are considered as a close estimate of the compressive
strength fc. Even if, in reality, the compressive strength fc does not correspond to
fc, such values remain the closest estimate in a material as heterogeneous as the
one addressed in this study.
Figure 3.16 illustrates the evolution of the creep coefficient, defined as the ratio
between the creep strain and the elastic strain, calculated from the short-term and
long-term creep tests results. For each specimen, the creep coefficient was calcu-
lated considering all creep diagrams at low stress levels (below 45% of fc). It is
further noted that the creep coefficient obtained from short-term creep tests was
calculated from the average of the four tested specimens while the values obtained
from long-term creep tests result from the average of the two specimens tested
with constant load periods of three months. In the remaining four specimens tested
in long-term creep, a first load step of approximately 60% of fc was applied and,
thus, such tests cannot be used to calculate creep coefficients.
Creep coefficients of approximately 0.10 and 0.15 were found at the end of
8 hours and 90 days of sustained loading, respectively, confirming that most creep
strain occurs at an early stage. Another important aspect is that the creep coeffi-
cient found at the end of 90 days is significantly lower than the values recom-
mended by EC6 [28] for masonry made with clay units, which range from 0.5 to
1.0. This can be explained by the fact that EC6 values refer to new masonry, where
maturation of mortar is in an initial stage and, also, because the specimens tested had
already been under service loads for approximately five centuries prior to testing.
. .
Figure 3.17 shows the strain rate evolution, vertical e v and horizontal e h, versus
the applied stress over strength ratio s/fc for specimens tested in short-term creep.
Strain-rate values were calculated between the sixth and eighth hour of each con-
stant load step. It is expected that vertical strain rate values would be negative and
horizontal strain rate values positive but some exceptions were found. This can be
explained by minor variations in the applied load or changes in the environmental
conditions. Such values have been considered equal to zero in the strain-rate
diagrams shown in the rest of this section.
In Fig. 3.17a three phases can be distinguished: for low stress levels (up to 50%
of fc), the vertical strain rate is approximately constant and rather low; for medium
stress levels (between 50 and 80% of fc), the vertical strain rate increases at a
moderate pace; and, for high stress levels (over 80% of fc), a remarkable growth
of the strain rate stress can be observed. The existence of three distinct phases had
also been reported by Mazzotti and Savoia [29] on short-term creep tests performed
on concrete specimens. Figure 3.17b shows that beyond 50% of fc, crack growth
initiates, influencing the creep behaviour of the material.
Figures 3.18 and 3.19 illustrate the strain rate evolution versus stress over
strength ratio for long-term creep tests with constant load periods of three and six
months, respectively. Strain rates were calculated from the average results over the
last 30 days in the case of the tests with constant load periods of three months and
over the last 90 days in the case of tests with constant load periods of six months.
It is noted that the number of results is scarce and further testing is needed to
better fundament the observations made. Nevertheless, the difference between the
strain rate values obtained from short-term creep tests and long-term creep tests
(a) (b)
-5.0E-001 2.5E+000
-4.0E-001 2.0E+000
v [year-1]
h [year-1]
-3.0E-001 1.5E+000
-2.0E-001 1.0E+000
-1.0E-001 0.5E-001
0.0E+000 0.0E+000
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
/fc' [-] /fc' [-]
Figure 3.17: Strain rate evolution versus applied stress over strength ratio for short-
term creep tests: (a) vertical strain rate and (b) horizontal strain rate.
(a) (b)
-1.5E-003 1.5E-003
-1.2E-003 1.2E-003
h [year-1]
v [year-1]
-9.0E-004 9.0E-004
-6.0E-004 6.0E-004
-3.0E-004 3.0E-004
0.0E+000 0.0E+000
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
/fc' [-] /fc' [-]
Figure 3.18: Average strain rate over the last 30 days versus applied stress over
strength ratio for long-term creep tests (three months steps): (a) verti-
cal strain rate and (b) horizontal strain rate.
(a) (b)
-4.0E-004 8.0E-004
-3.0E-004 6.0E-004
4.76 . 10-5
h = + 1.19 . 10-4
h [year-1]
v [year-1]
1 /fc
-2.0E-004 4.0E-004
2.70 . 10-5
v = 6.76 . 10-5
/fc 1
-1.0E-004 2.0E-004
0.0E+000 0.0E+000
0.4 0.5 0.6 0.7 0.8 0.9 1.0 0.4 0.5 0.6 0.7 0.8 0.9 1.0
/fc' [-] /fc' [-]
Figure 3.19: Average strain rate over the last 90 days versus applied stress over
strength ratio for long-term creep tests (six months steps): (a) vertical
strain rate and (b) horizontal strain rate.
is striking. In fact, strain rates ranging from zero to 5.0 101 year1 were
observed in short-term creep tests while in long-term creep tests, values ranging
from zero to 1.0 103 year1 were found. The results obtained from the two
types of test seem therefore not comparable. Furthermore, primary creep seems
not extinguished at the end of 8 h under sustained loading and, thus, secondary
creep rates measured from short-term creep tests must be interpreted carefully.
Another important aspect is that secondary creep was observed to initiate
between 60 and 70% of fc. It is further noted that larger strain rate values were
obtained for the prisms tested with constant load periods of three months, stressing
the scattered nature of the masonry tested.
A hyperbolic least squares fit of the experimental data obtained from the long-
term creep tests with constant load periods of six months was computed, which
(a) (b)
-8.0E-003 2.5E-002
Average values Average values
2.0E-002
-6.0E-003
h [year-1]
v [year-1]
1.5E-002
-4.0E-003
1.0E-002
-2.0E-003
0.5E-003
0.0E+000 0.0E+000
0 40 80 120 160 200 0 40 80 120 160 200
Time [days] Time [days]
Figure 3.20: Strain-rate evolution in time for applied stresses larger than 60% fc:
(a) vertical strain rate and (b) horizontal strain rate.
can be quite useful in calibrating non-linear creep models. The hyperbolic curve
adopted is in the following form
. 0.4 a
e = +a
1 s fc (1)
which yields zero for s/fc = 0.6 and has a vertical asymptote for s/fc = 1.0.
From the least squares method, a = 6.76 105 for the vertical strain rate and
a = 1.19 104 for the horizontal strain rate.
The striking difference between strain rate values in short-term and long-term
creep tests draws attention to what should be the minimum duration of constant
load periods when conducting creep tests at high stress levels. A reasonable crite-
rion is believed to be keeping the load constant until only secondary creep is pres-
ent, i.e. until a fairly constant strain rate is attained. For this purpose, the vertical
and horizontal strain rates were calculated for each 15-day period of the total 180
days constant load steps, as illustrated in Fig. 3.20. It is noted that only results cor-
responding to load levels larger than 60% of fc were considered. The results
obtained indicate that the strain rate gets approximately constant after 7080 days
in the case of longitudinal strains and after 3040 days in the case of transversal
strains.
3.4 Conclusions and future work

The ability of standard continuum models, based on plasticity and cracking, and
of a particle model, consisting in a phenomenological discontinuum approach, to
reproduce the experimental compressive behaviour of masonry has been addressed.
The comparison between the obtained numerical results and experimental results
available in literature allows one to conclude that discontinuum models show
clear advantages when compared to standard continuum models in predicting the
compressive strength and peak strain of masonry prisms from the properties of
the constituents. Further investigation on models able to provide reliable predic-

tions of masonry compressive strength, accounting for the discrete nature of the
masonry components, is therefore suggested.
The creep behaviour under high stresses of ancient regular masonry specimens
recovered from the collapsed Civic Tower of Pavia has also been analysed. Stan-
dard compression tests, short-term creep tests and long-term creep tests have been
conducted. From experimental practice, it is possible to conclude that creep tests
on ancient masonry prisms should be carried out by applying the load in succes-
sive steps, at a given time interval, starting from a low stress level. In this way, a
thorough description of the viscous behaviour of the material can be obtained.
Creep tests in which the load is applied in a single step are inadequate in the case
of ancient masonry due to the high scatter in the mechanical properties and to the
small number of specimens usually available.
The time period between successive load steps should be sufficiently long to
extinguish primary creep. In fact, the evolution for different stress levels of the
strain rate associated with secondary creep can only be evaluated in such a way.
From the results obtained on the regular masonry prisms tested, a minimum time
period under sustained loading of 7080 days should be adopted. For this reason,
remarkable differences were observed between secondary creep rates calculated
from short-term or long-term creep tests. Short-term creep results should, there-
fore, be interpreted carefully. Finally, it should be stressed that secondary creep
was found to initiate at 6070% of the compressive strength. A hyperbolic fit to
describe the evolution of secondary creep rate with the applied stress-level has
been suggested in the present study.
Suggestions for future work include further creep tests so that an adequate char-
acterization of the material can be obtained, given the wide scatter associated with
ancient masonry. With respect to the masonry constituents, experimental results on
the non-linear creep behaviour of units and mortar are nearly absent in literature
[30], meaning that further investigation is required. In terms of numerical model-
ling, the development of suitable 3D models for viscous inelastic behaviour is
needed in order to include time-dependent effects in the numerical simulations of
failure.
References
[1] Binda, L., Gatti, G., Mangano, G., Poggi, C. & Landriani, G.S., The collapse
of the Civic Tower of Pavia: a survey of the materials and structure.
Masonry Int., 6(1), pp. 1120, 1992.
[2] Anzani, A., Binda, L. & Mirabella Roberti, G., The effect of heavy persis-
tent actions into the behaviour of ancient masonry. Materials and Struc-
tures, 33(228), pp. 251261, 2000.
[3] Bazant, Z.P., Material models for structural creep analysis. Mathemati-
cal Modelling of Creep and Shrinkage of Concrete, ed. Z.P. Bazant, John
Wiley & Sons: New York, pp. 99215, 1988.
[4] Pickett, G., The effect of change in moisture content on the creep of
concrete under a sustained load. ACI J., 38, pp. 333355, 1942.
[5] Neville, A.M., Properties of Concrete, John Wiley & Sons: New York, 1997.
[6] Van Zijl, G.P., Computational Modelling of Masonry Creep and Shrinkage.
Dissertation, Technical University of Delft, Delft, The Netherlands, 2000.
[7] Ameny, P., Loov, R.E. & Shrive, N.G., Models for long-term deformation
of brick work. Masonry Int., 1, pp. 2736, 1984.
[8] Lenczner, D., Creep and prestress losses in brick masonry. The Structural
Engineer, 64B(3), pp. 5762, 1986.
[9] Brooks, J.J., Composite modelling of masonry deformation. Materials and
Structures, 23, pp. 241251, 1990.
[10] Bazant, Z.P., Current status and advances in the theory of creep and interac-
tion with fracture. Proc. 5th Int. RILEM Symp. on Creep and Shrinkage of
Concrete, Barcelona, Spain, pp. 291307, 1993.
[11] Papa, E., Taliercio, A. & Gobbi, E., Triaxial creep behaviour of plain
concrete at high stresses: a survey of theoretical models. Materials and
Structures, 31, pp. 487493, 1998.
[12] Mazzotti, C. & Savoia, M., Nonlinear creep damage model for concrete
under uniaxial compression. J. Engineering Mechanics, 129(9), pp. 1065
1075, 2003.
[13] Pina-Henriques, J. & Loureno, P.B., Masonry compression: a numerical
investigation at the meso-level. Engineering Computations (accepted for
publication).
[14] Binda, L., Fontana, A. & Frigerio, G., Mechanical behaviour of brick
masonries derived from unit and mortar characteristics. Proc. 8th Int. Brick
and Block Masonry Conf., Dublin, Ireland, Vol. 1, pp. 205216, 1988.
[15] Frigerio, G. & Frigerio, P., Influence of the Components and Surrounding
Environment in the Mechanical Behaviour of Brick Masonry (in Italian).
Graduation thesis, Politecnico di Milano: Milan, Italy, 1985.
[16] Mann, W. & Betzler, M., Investigations on the effect of different forms of
test samples to test the compressive strength of masonry. Proc. 10th Int.
Brick and Block Masonry Conf., Calgary, Canada, pp. 13051313, 1994.
[17] Berto, L., Saetta, A., Scotta, R. & Vitaliani, R., Failure mechanism of ma-
sonry prism loaded in axial compression: computational aspects. Materials
and Structures, 38(276), pp. 249256, 2005.
[18] Pina-Henriques, J. & Loureno, P.B., Testing and modelling of masonry
creep and damage in uniaxial compression. Proc. 8th Int. Conf. Structural
Studies, Repairs and Maintenance of Heritage Architecture, Halkidiki,
Greece, pp. 151160, 2003.
[19] Rots, J.G., Computational Modelling of Concrete Fracture. PhD Dissertation,
Delft University of Technology, Delft, The Netherlands, 1988.
[20] DIANA Finite Element Code, version 8.1. TNO Building and Construction
Research, Delft, The Netherlands.
[21] Vermeer, P.A. & de Borst, R., Non-associated plasticity for soils, concrete
and rock. Heron, 29(3), pp. 164, 1984.
[22] Loureno, P.B., A User/Programmer Guide for the Micro-Modelling of

Masonry Structures. Report 03.21.1.31.35, Delft University of Technology,
Delft, The Netherlands, 1996. Available from http://www.civil.uminho.pt/
masonry.
[23] CEB-FIP Model Code 1990; Bulletin DInformation No. 213/214, Comite
Euro-International du Beton, T Telford: London, 1993.
[24] Loureno, P.B. & Rots, J.G., A multi-surface interface model for the analysis
of masonry structures. J. Engineering Mechanics, ASCE, 123(7), pp. 660
668, 1997.
[25] Loureno, P.B., Computational Strategies for Masonry Structures. PhD
Dissertation, Delft University of Technology, Delft, The Netherlands, 1996.
Available from www.civil.uminho.pt/masonry.
[26] van der Pluijm, R., Out-of-plane Bending of Masonry: Behaviour and Strength.
PhD Dissertation, Eindhoven University of Technology, Eindhoven, The
Netherlands, 1999.
[27] Anzani, A., Binda, L. & Melchiorri, G., Time dependent damage of rubble
masonry walls. Proc. British Masonry Society, 2(7), pp. 341351, 1995.
[28] Eurocode 6: Design of masonry structures; prEN 1996-1-1:2002, CEN:
Brussels, Belgium, 2003.
[29] Mazzotti, C. & Savoia, M., Nonlinear creep, Poissons ratio, and creep-
damage interaction of concrete in compression. ACI Materials J., 99(5),
pp. 450457, 2002.
[30] Papa, E., Binda, L. & Nappi, A., Effect of persistent loads in masonry struc-
tures. Proc. 3rd Int. Masonry Conf., London, pp. 290294, 1992.
CHAPTER 4
Effects of creep on new masonry structures
N.G. Shrive1 & M.M. Reda Taha2

1Department of Civil Engineering, University of Calgary, Calgary,
Canada.
2Department of Civil Engineering, University of New Mexico,
Albuquerque, USA.
4.1 Introduction
Creep can affect structures in two ways: deformations typically increase and loads
(stresses) can be redistributed among structural components and, within a mem-
ber, the constituent materials [1]. The effects of creep can be beneficial, neutral, or
detrimental for a structure: beneficial, for example through the relief of stress con-
centrations, detrimental through increasing deformations. The latter can lead to a
structure no longer meeting serviceability criteria. Stress redistribution can cause
cracking, especially in cases where there is deterioration in strength over time due
to environmental factors in that element of the structure which carries increasing
load due to creep effects.
Sometimes the two effects work together. Creep buckling is one example. An
initial lateral imperfection in a column subject to axial load, or an initial eccentric
load, causes an initial lateral displacement of the column. Consequently, there are
higher compressive stresses on the inner curvature than on the outer curvature of
the column. The side of the column under the higher stress creeps more than the
other side, under the lower stress. The creep strains result in increasing lateral
displacement. In turn, the secondary moment (the axial load times the lateral dis-
placement) increases, increasing the stress on the inner curvature. Creep increases
with the higher stress, so the lateral displacement increases more and more rapidly
with the end result being a buckling failure of the column. An initial applied load
less than the Euler buckling load for the end constraints of the column can thus
cause a buckling failure sometime after load application. For some materials, the
stress to initiate such failure can be as low as 60% of the failure stress in a monotonic
compression test. Binda et al. [2] indicate that for some types of masonry this
number may be in the order of 4550%.
Crack growth caused by creep in tension tests has been recognized as damage
in the same context as that caused by fatigue [3, 4]. Cracks also grow under com-
pression-induced creep, parallel to the direction of compression. The cause is
similar to that in creep buckling in that the presence of the crack disturbs the local
stress field, with a higher than average compressive stress parallel to the edge of
the crack. Increased creep there accentuates the bending component of the stress
and deformation beside the crack, increasing the crack width, and thus the tensile
stresses at the crack tip. The crack extends when the stress and energy conditions
favor such growth [5]. Creep crack propagation and its cracking rate dependence
on bond breakage at the fracture process zone were discussed and modeled [68].
Binda et al. [9] remark on the existence of such cracks in the civic tower of Pavia
at least 20 years prior to its collapse, and the appearance of such cracks in many
historic structures at various ages after construction [2].
It is now well established that masonry creeps. The pioneering work of Lenc-
zner [10, 11] has led to the realization that creep can be expected with any masonry
units [2, 1215] perhaps with the exception of some dry stacked stonework. The
potential effects of creep therefore need to be considered in new construction and
in rehabilitation. In rehabilitation interventions in historic masonry structures,
essentially a new structure has been created, in which one component has already
undergone some creep and the new component has yet to creep. Some masonry
codes of practice now recognize the effect of creep in terms of increased deforma-
tion. Such codes advise designers to use the effective modulus technique to esti-
mate long-term displacements [16, 17]. This technique, however, ignores what
may happen in the structure between the initial and long-term states.
4.2 The step-by-step in time approach to modeling

time-dependent effects
We demonstrate here, using a simple step-by-step in time technique, that elements
in a structure may see increasing, then decreasing, proportions of load over time:
that by calculating only the initial and long-term states, the designer may miss a
peak stress occurring at an intermediate time. We recognize that masonry is com-
plex, multi-component material: an outer skin of brickwork or blockwork may be
filled with grout or rubble masonry, and may contain reinforcing bars. Modern
techniques of rehabilitation may involve use of fiber reinforced polymers (FRPs),
some of which are known to creep [18]. Alternatively, the epoxy binding the FRP
to the underlying masonry may relax over time under the shear stresses transfer-
ring load between the FRP and the masonry: such behavior would cause a redistri-
bution of load between these two components.
The step-by-step in time method of creep analysis relies on the applicability of
the principle of superposition. In relation to creep, the principle requires that for a
material subjected to stresses at different times, the creep strain produced at any
Effects of Creep on New Masonry Structures 85
time due to a previously applied stress is independent of the effects of any other
stresses applied before or after that particular stress. The creep response to a set of
stresses applied at different times is thus the summation of the creep effects of
each stress. We demonstrate the technique and the consequences for load redistri-
bution and increasing deformations in structures through the examples of an axi-
ally loaded masonry column and a beam subjected to pure bending. We also
demonstrate the effect of damage of masonry on stress redistribution. A wall sub-
jected to both axial and bending loads could be analyzed similarly. The complexity
of the analysis can be increased by including plasticity constitutive equations as in
[15] or by adopting variable adaptive time-stepping when damage is considered,
as recommended by [19]. At the end of the chapter we introduce the concept of
using Artificial Neural Networks (ANNs) to predict creep effects, as they can be
substantially more accurate than explicit equations that best fit with regression to
experimental data.
4.3 Case 1: An axially loaded column

4.3.1 Creep model
Consider a concentric axial load P applied to a symmetric column made of two

materials A and B each symmetrically distributed about the column axis. Effects
due to eccentricity can therefore be neglected. The two materials have different
time-dependent properties. The two materials have cross-sectional areas: AA being
that of material A and AB that of material B. Equilibrium requires:
AAsA + ABsB = P (4.1)
where P is the applied concentric axial load and sA and sB are the stresses in mate-
rials A and B, respectively.
Compatibility requires the axial strain in the column and each respective mate-
rial is the same
e = eA = eB (4.2)
Next we assume that each material creeps such that the creep can be expressed as
a compliance function in time, where DA and DB are the compliances of materials
A and B, respectively. This creep function (eqn (4.3)) is shown in Fig. 4.1.
t
tA

eA (t ) = DA (t ) sA = DA fA sA 1 e (4.3)

The creep strain eA(t) is obtained by simply multiplying the time-dependent

creep coefficient, (fA(t)), by the initial strain as in eqn (4.3). fA is the creep coef-
ficient for infinite time. The time constant tA denotes the time when 63% of the
creep has occurred. Material B is assumed to creep with a mathematical form
Asymptotic for t =
fA sA DA
A
sA D A
tA
Time (t)
Figure 4.1: The simple creep function.
similar to material A, but with DB, fB(t) and tB representing its compliance, creep
coefficient, and the 63% creep time, respectively. Other formulations of the creep
function as suggested by other researchers [e.g. 11, 12, 20, 21] can be employed in
lieu of eqn (4.3) and will have a slight effect in the overall conclusion. Using the
equilibrium and compatibility considerations above, the initial stresses in the two
materials are:
DB P
sA = (4.4)
AB DA + AA DB
DA P
sB = (4.5)
AB DA + AA DB
Since both materials are stressed, both will want to creep. We therefore permit a
small increment in time to occur from t = 0 to t = t1. In this first time increment,
material A will want to creep an amount of strain ecrA as presented in eqn (4.6).
t1

ecrA (1) = sA fA DA 1 e tA (4.6)

Material B will also have a creep increment of similar form. However, the incre-
ments in creep strain will be different. Hence, compatibility will be violated. In
order to restore compatibility, the material that wants to creep more will have its
stress reduced by the one that wants to creep less, while the latter will have its
strain increased by the former. Compatibility therefore requires:
ecrA (1) + sA (1)DA = ecrB (1) + sB (1)DB (4.7)
where the ss are the incremental changes in stress. Since there is no increase in
the overall axial force, equilibrium requires:
sA (1)AA + sB (1)AB = 0 (4.8)
The incremental stress changes are therefore

AB ( ecrA (1) ecrB (1))
sA (1) = (4.9)
AB DA + AA DB
AA ( ecrA (1) ecrB (1))
sB (1) = (4.10)
AB DA + AA DB
The stress in A (and similarly in B) at the end of the first time step is therefore
sA (1) = sA + sA (1) (4.11)
The materials, however, want to continue to creep. We therefore invoke superpo-
sition since we have made the materials linear viscoelastic. The amount of creep
strain that material A would like to creep in the second time step can therefore be
expressed as:
t t1 t2

ecrA (2) = sA fA DA e e tA
A

( t2 t1 )

+ sA (1)fA DA 1 e tA (4.12)

The specific creep (creep strain per unit stress) curves for the different stress
increments are the same; they just start at each time step, as shown in Fig. 4.2.
Assume c1 is the specific creep that material A would like to creep in time step 1
due to the initial stress. c1 is also the specific creep for the stress increment sA
(4.1) between time t1 and t2, whereas the influence of the initial stress in that time
interval will be c2. When multiplied by their respective stresses, the influences are
simply added.
Equilibrium and compatibility can be enforced again and the third step consid-
ered. This leads to the formula for the (nth) time step where (n 2).
tn1 tn

ecrA (n) = sA fA DA e tA e tA

n ( tnt1 ti1 )
( t n ti 1 )

+ sA (i 1)fA DA e A
e tA (4.13)
i=2
with the stress in material A at the end of the nth time step being
n
sA (n) = sA + sA (i ) (4.14)
i =1
With a similar equation for material B, the total strain is

n n
eA (n) = eA + ecrAn (i ) + DA sn (i ) (4.15)
i =1 i =1
Figure 4.2: The specific creep for the different stress increments.
Example
Consider, for example, a blockwork column (A) filled with grout (B). We consider
the following values for the following parameters: AA = 0.6Atotal, AB = 0.4Atotal,
DA = 1/15 GPa1, DB = 1/22 GPa1, fA = 5.0, fB = 2.5, tA = 500 days and tB = 1000
days. The blockwork is modeled to creep more and in a relatively shorter time than
the grout. The results in Fig. 4.3 show the stress changes in both the blockwork
and the grout over time.
Based on their relative stiffnesses, the initial stress in the blockwork is 12.6 MPa
while that in the grout is 18.5 MPa. As the blockwork wants to creep faster than
the grout, the blockwork initially offloads to the grout. However, as the grout creeps
more than the blockwork at later ages, the blockwork will be re-stressed and the
grout stress will be reduced during this phase. The effective modulus method
[1, 22] is applied to estimate the final stress in the blockwork using eqn (4.16):
DB (1 + fB )
s A ( ) = (4.16)
AB DA (1 + fA ) + AA DB (1 + fB )
The final stress in the grout can be evaluated similarly. Final stresses of 9.3 and
23.5 N/mm2 are determined for the blockwork and grout, respectively. Similar
stresses are attained using the step-by-step analysis (Fig. 4.3). However, a designer
using the effective modulus method will only detect that creep of the two materi-
als will result in the grout stress rising by 25%, while in reality the grout will be
overstressed by 36% from the initial value during an intermediate time period. The
model presented here does not include the possibility that the material in which the
stress rises (here the grout) might be damaged (degraded) during the overloading
period. Typically, as a quasi-brittle material, the grout might be cracked during
the overloading, reducing its stiffness. The stress distribution would therefore be
Figure 4.3: Stress development in blockwork and grout due to creep (Case 1).
changed from what is shown here and the blockwork re-stressed to a level higher
than shown in Fig. 4.3. The example also demonstrates that the initial and final
stress states are not the extremes. Higher stresses occur at intermediate stages.
4.3.2 Effect of coupling creep and damage in concentrically

loaded columns
Another interesting effect that is difficult to correlate is the coupling of creep and
damage. As mentioned, models are being developed for tensile creep and fatigue
[3, 4], but there are other mechanisms which can induce damage. External dete-
rioration can begin on the surface of masonry from actions like freeze-thaw or
weathering. Mirza [23] for example, discussed how the resistance of a member
in the context of limit states design can decline with time after construction. Val-
luzzi et al. [24] and Baant [25] presented methods to account for this coupling
effect in finite element modeling of historical masonry. Various damage models
are described in the literature [2628]. We have chosen a simple model as our
objective is to show what can happen, rather than to develop a model for a par-
ticular case or material. We consider the analysis above but we also introduce a
damage model to account for the effect of damage in one material of the column
and couple it with creep. Damage is assumed to accumulate in the form
n
t
h t
DM(t ) =
i = tDstart tDM tDM
(4.17)
tD is the damage time constant which refers to the time where most damage would
occur. The coefficients are taken here as tD = 800, h = 0.3, and n = 10. DM(t)
represents the level of damage accumulated from the time at which damage starts,
tDstart, to the time of evaluation t. In the calculations here, damage is assumed to
begin at 400 days. The rate of damage accumulation with this model is slow ini-
tially, but accelerates over time, as shown in Fig. 4.4. For the sake of showing the
trends and effects, quite considerable damage is assumed to occur in a relatively
short time in this example. Following [26] and [29], the modulus of the material
changes over time with this model as
EA(t) = (1 DM(t))EA(tDstart) (4.18)
where EA(tDstart) is the material stiffness at the time when the damage begins. When
DM is zero, there is no damage and when DM equals 1, the material is unable to
bear any load.
Figure 4.4: The damage model showing the non-linear change of damage ratio
with time. Here the blockwork will have a damage factor of about 0.33
after 1000 days with the damage starting at 400 days. Damage initially
accumulates at a very low rate but then increases rapidly.
The effect of combined creep and damage on the Youngs moduli of both materials
is shown in Fig. 4.5, while the stresses variations with time are shown in Fig. 4.6.
These stresses are different to those in Fig. 4.3 in that they begin to change quite
rapidly at later ages, as the damage accumulates in the outer blockwork, causing
redistribution to the stiffer, undamaged grout. It thus becomes possible that the
grout could now begin to fail, leading to collapse of the whole structure in time.
This problem has been analyzed further elsewhere [30], with consideration of sev-
eral possible combinations of creep and damage.
4.3.3 Examining the effect of rehabilitation
Much effort has been spent on rehabilitating and strengthening structures, both of
historic and of simply practical value. Fiber reinforced polymers have been used
on various occasions as they offer distinct advantages over steel in terms of being
light weight and thus adding little mass to a structure, and highly durable if pro-
tected from sunlight. Some FRPs creep while others do not, unless the stress is a
Figure 4.5: The change in effective moduli of the two materials with time due to
combined creep and damage. Note the change of the slope of stiffness
reduction of the blockwork due to the combined effect of creep and
damage after 400 days of age while the slope of grout stiffness reduc-
tion due to creep only is almost constant after 400 days.
Figure 4.6: The change in stress in the two materials over time with creep and
damage.
very high proportion of ultimate [18, 31]. Thus when an FRP is used to strengthen
a structure, the longer term consequences should be evaluated. If the structure is
historic, creep may well have substantially run its course for the current loading.
However, if the FRP is prestressed or changes the stresses in the structure in some
other way then creep will start anew. Essentially a new structure has been created
and the consequences of time-dependent effects need to be evaluated. One poten-
tial effect that needs to be considered is the reduction in load carried by the FRP
from flow of the bonding epoxy from the shear that transfers load between the FRP
and the underlying substrate. For example, if an FRP strip is applied to a structure
to help carry a dead load, or is applied pretensioned to counteract a dead load, then
there is the potential for the bonding epoxy to flow [32]. The FRP strip offloads
over time. The effect is demonstrated with the next problem. Rigid mechanical
anchorage of the strip would be required to avoid the consequence shown.
4.4 Case 2: Composite structural element subject to bending

4.4.1 Development of model
Consider a reinforced concrete beam where a layer of external reinforcement is

bonded to the bottom of the beam as shown in Fig. 4.7. The external reinforcement
Figure 4.7: Schematic representation of a reinforced concrete beam cross-section

with one layer of steel reinforcement and externally applied FRP.
is added to the model such that the model is general and the effect of any externally
applied strengthening material can be considered. We consider the case of pure
bending for simplicity. Prestressing or axial dead load can be included. Again, for
simplicity, we have assumed there is no compression reinforcement.
Strain compatibility, with plane sections remaining plane, gives
d kd
es = ec (4.19)
kd
1 k
es = ec (4.20)
k
h kd
ef = e (4.21)
kd c
where kd is the depth of the neutral axis. Force equilibrium requires
Tf + Ts C = 0 (4.22)
1
Af sf + As ss bkd sc = 0 (4.23)
2
Moment equilibrium requires
kd kd
Ts d + Tf h = M (4.24)
3 3
kd kd
As ss d = M Af sf h (4.25)
3 3
Solving the equations provides

E 1 k 1
As sc s bkd sc = Af sf (4.26)
Ec k 2
Af sf
sc = (4.27)
1 Es 1 k
bkd As
2 Ec k
and from the moment equations
3 k Es 1 k kd
As d s = M Af sf h (4.28)
3 Ec k c 3
Thus at t = 0 (no creep)
E h kd
sf = sc f (4.29)
Ec kd
sf h kd Af
= = (4.30)
sc kd 1 E 1 kd
bkd As s
2 Ec k
leaving the following equation to be solved for k:
1 2 2
Ec bd k + d ( As Es + Af Ef )k ( Af Ef h + As Es d ) = 0 (4.31)
2
Thus the initial stresses in the concrete, FRP, and steel are
M
sci = (4.32)
v
3 k Es 1 k
v = As d
k Ec k
kd 1 E 1 k
+ h bkd As s (4.33)

3 2 Ec k
E h kd
sfi = sci f (4.34)
Ec kd
E 1 k
ssi = sci s (4.35)
Ec k
Now, for t > 0, the concrete creeps and
ei
E c (t ) = Eci (4.36)
e (t )
Further, we need to solve for k as a function of sf as sf will reduce with time

3 k Es 1 k Af sf
As d
3 E c (t ) k 1 Es 1 k
bkd As
2
E c (t ) k
kd
= M Af sf h
3 (4.37)
kd 1 E 1 k Es 1 k
Af sf h bkd As s As d
3 2 Ec (t ) k Ec (t ) k
1 E 1 k
= M bkd As s (4.38)
2 E c (t ) k
E
( Af sf d 2 b)k 3 + 3bd ( M Af sf h)k 2 + 6 As s ( M Af sf (h d )) k
Ec ( t )
E
6 As s ( M Af sf (h d )) = 0 (4.39)
Ec ( t )
We let the FRP stress decline with time as the epoxy creeps (flows) under the shear
it is transmitting. We consider a simple classical formula to represent FRP stress
relaxation due to binding matrix creep [33]. At the end of the first time step, the
stress in the FRP is
t1

sf (1) = sfi a + (1 a )e tr (4.40)

ei
Ec (1) = Eci (4.41)
e(t1 )
Equation (4.39) is solved for k(1) and this value is substituted into eqn (4.27) for
sc(1)
sc (1)
ec (1) = (4.42)
Ec (1)
h k (1)d
ef = ec (1) (4.43)
k (1)d
Ef h k (1)d
sf (1) = (sc (1) sci ) (4.44)
Ec (1) k (1)d
sf(1) = sf (1) + sf (1) (4.45)
With the increment in concrete stress, force and moment equilibrium are checked.
If equilibrium is not obtained, we reiterate the calculation of k. After k has been
computed, sc(1) is calculated from eqn (4.27)
sc1 = sc (1) sci (4.46)

sf1 = sfi sf (1) (4.47)
for the nth step

n
t n 1 n i
t t
sf (n) = sfi a + (1 a )e tn + sfi a + (1 a )e t (4.48)

i =1
n sci n i
t n 1 t t
ec (n) = e 1 + f 1 e tn + 1 + f 1 e t (4.49)
i =1 Ec (ti )
ei
E c (n ) = Eci (4.50)
e (n )
The cracked beam curvature at any section j along the beam at any time t denoted
2j(t) can be computed as
Mi
2j ( t ) = (4.51)
Ec (t )I tj (t )
where Itj(t) is the transformed cracked moment of inertia at the jth section at time
t, computed as
b [k (t )d ]
3
E
I tj (t ) = + As s [d k (t )d ]2
3 E c (t )
E (t )
+ Af f [h k (t )d ]2 (4.52)
E c (t )
The uncracked beam curvature at any section along the beam at time t denoted
1j(t) can be computed as
Mj
1j (t ) = (4.53)
Ec (t )I gj (t )
Given the effect of tension-stiffening on beam deflection, an effective curvature
can be determined by interpolating between the cracked and uncracked section
curvatures using the moment-curvature approach recommended by the CEB-FIP
[34] and Ghali and Favre [1] as
ej (t ) = x j 2 (t ) + (1 x ) j1 (t ) (4.54)
where x is the tension stiffening coefficient that can be determined using the CEB-
FIP [34] method. Here the tension stiffening coefficient x was determined such
that the error between the experimentally measured (Section 4.4.2) and model pre-
dicted mid-span deflections is a minimum. The mid-span deflection of the beam
mid(t) can be predicted by integrating the curvature along the span (S) or esti-
mated approximately by considering the geometrical relation using three sections
at the ends and at mid-span as
S2
mid (t ) = (left (t ) + 10emid (t ) + right (t )) (4.55)
96
where left and right can be assumed to be equal to zero as no end curvature due
to restrained shrinkage is expected to occur and emid (t) is the mid-span curvature
computed as in eqn (4.54).
4.4.2 Application to a beam
The model is now applied to reinforced concrete beams with the dimensions
shown in Fig. 4.8 and loaded as shown in Fig. 4.9. The material and load proper-
ties, including the concrete creep are taken as listed in Table 4.1. Two beams were
tested. Both beams have similar properties (cast from the same concrete batch, at
the same time and cured in a similar way) one without FRP strengthening strips
and one with FRP strengthening strips. The cross-section of the FRP strengthened
beam is shown in Fig. 4.8.
Figure 4.8: Cross-section dimensions of the concrete beam reinforced with one
layer of steel reinforcement and externally applied FRP.
Figure 4.9: Schematic representation of the beam loading set-up.

For the beam without FRP strengthening, the predicted increase in deflection at
the center of the beam is compared with experimental data (presented in [35, 36])
in Fig. 4.10.
It can be observed in Fig. 4.10 that the model was capable of predicting the
creep deflection of the beam fairly well at early and late times. It is worth mention-
ing this prediction is achieved using the creep properties listed in Table 4.1 and by
adjusting the tension stiffening coefficient to reduce the model error. A tension
stiffening coefficient x = 0.9 is needed. This high coefficient indicates that the sec-
tion will be very close to fully cracked in this case (x = 1.0 indicates fully cracked
section). In fact, the beam has numerous flexural cracks. Now, if we consider the
beam strengthened with the FRP strips to have similar material properties, and
assume that the only time-dependent effect is the concrete creep, we get the result
Table 4.1: Material properties of reinforced concrete beam.

Concrete compressive strength (MPa) 34.3
Concrete initial modulus of elasticity (GPa) 21.1
Reinforcing steel modulus of elasticity (GPa) 200
CFRP modulus of elasticity (GPa) 165
Concrete creep coefficient f (t, t0) 4.2
Concrete ultimate creep time tc (days) 2000
Figure 4.10: Predicted versus experimentally measured deflections of reinforced

concrete beam including creep effect (Beam 1: no FRP is used).

concrete beam (Beam 2: FRP is used) including creep effect only. Ten-
sion stiffening coefficient similar to that of Beam 1 is used (x = 0.9).
shown in Fig. 4.11. It is obvious that the predicted mid-span deflection does not
meet the measured mid-span deflection.
The significant difference between the measured and predicted deflections in
Fig. 4.11 can be attributed to two factors: the tension-stiffening effect and the
effect of the creep of the epoxy binding matrix. As installing the FRP strips
increased the cracking capacity of the beam, a lower tension stiffening coefficient
(representing a cracked section) is required compared to that used in computing
the mid-span deflection of the unstrengthened beam. Changing the tension-stiffen-
ing coefficient in the case with FRP we obtain the curve shown in Fig. 4.12. It can
be observed that predicted deflection in this curve is closely related to the experi-
mentally measured one. The deflection predicted in Fig. 4.12 assumes no FRP
binder creep has occurred (a = 1.0). If relaxation of stress in the FRP strips due to
the creep in the epoxy binding the FRP strip to the concrete (a = 0.9 and t = 600 days)
is included, we get Fig. 4.13. While higher creep coefficient of epoxy at the concrete
FRP interface might be assumed, recent experimental investigations showed that
creep of epoxy at the concreteFRP interface would result in little but fast stress
loss in the FRP sheets [37]. Therefore the reduction of FRP stress in the analysis
presented here due to creep of epoxy was limited to 90% of the original stress.
Figures 4.12 and 4.13 look very similar, but it is misleading to judge the effect
of FRP binder creep from these time-deflection diagrams. This is because the
section is close to uncracked and the influence of FRP stress-relaxation on the

concrete beam (Beam 2: FRP is used) including creep effect of
concrete and no FRP stress relaxation effect (a = 1.0) and tension
stiffening effect (x = 0.27).

concrete beam (Beam 2: FRP is used) including creep of concrete
effect, FRP stress relaxation effect (a = 0.9 and t = 600 days) and
tension stiffening effect (x = 0.27).
transformed section properties is not reflected in the change in beam deflection.

The influence of changing the FRP stress-relaxation coefficient from no relaxation
(Case 1: a = 1.0) to significant relaxation (Case 2: a = 0.9) on the transformed
inertia is shown in Fig. 4.14.
Figure 4.14: Change in transformed cracked moment of inertia of the FRP strength-
ened beam with two cases: one with no FRP stress relaxation (Case
1: a = 1.0) and the other with FRP stress relaxation (Case 2: a = 0.9).
Both cases include tension stiffening effect (x = 0.27).
The results shown in Fig. 4.14 demonstrate the fact that the transformed cracked
section inertia reduces due to FRP stress relaxation. This reduction would result in
a considerable change in the beam deflection if the beam was cracked, particularly
if the beam were cracked prior to the application of the FRP strengthening strips.
This change (although happening) did not affect the beam deflection in the case
study because the beam was uncracked and thus its deflection is dominated by the
uncracked section inertia. The step-by-step in time analysis also allows the change
in the stress in the concrete top fibers, the reinforcing steel bars, and the FRP strips
to be followed. Exemplar result for the change in the concrete stresses is shown
in Fig. 4.15.
The stresses shown in the Fig. 4.15 are based on the FRP stress decreasing due
to creep of the epoxy at the FRPconcrete interface. With the current model, we
force this reduction in stress through our imposition of relaxation in the FRP.
However, such a reduction is likely not to be correct over the full time domain if
the concrete creeps extensively, as would occur if the concrete were young. Such
creep will increase the curvature of the beam resulting in increasing stress in the
FRP after the initial decrease. This issue represents a modeling challenge because
the concrete and epoxy have different rates of creep. Creep in the epoxy occurs in
a considerably shorter time compared with that of creep in concrete. Therefore, the
analysis above can be used to model the beam deflection and to give a reasonable
estimate of the stresses in the concrete but not in the steel or the FRP. The small
Figure 4.15: Change of concrete top fiber stresses with time due to combined
concrete creep and FRP stress relaxation (a = 0.9 and t = 600 days).
Figure 4.16: Change of stresses in FRP strips with time due to creep of epoxy at
the concreteFRP interface (a = 0.9 and t = 600 days) while creep of
concrete is neglected.
effect on the accuracy of the concrete stresses in this example is related to the
relatively small change in the FRP stress over the entire analysis time due to stress
relaxation (only 3%) given the geometry of the beam analyzed and the relaxation
parameters considered (a = 0.9 and t = 600 days). If the creep of concrete is mini-
mal, as would be the case in strengthening an old concrete beam under its own
weight, then the FRP stress would be accurately estimated from the above analysis
as shown in Fig. 4.16.
4.4.3 Masonry walls subject to axial load and bending
Masonry walls are typically subject to both axial load (example 1) and bending
(example 2). Hence to solve for the effects of creep for such loading, it is neces-
sary to combine the two examples above, invoking the need for equilibrium and
compatibility in the process. The equations for this situation are being developed.
The consequences are well known, in that out-of-plane deformation of the wall
will increase, just as the central deflection of the concrete beam in example 2. The
stress on the inside curve of the wall will increase while that on the outside will
decrease. Two possible long-term consequences are that failure by cracking or
crushing may develop on the inner curve of the wall, or the wall may buckle. The
situation which should be carefully analyzed is one where a cracked wall, vault, or
beam is strengthened in situ, and the load increased. The FRP is likely to off-load
itself over time if not anchored mechanically (The FRP strips in the beam as above
were anchored with U-shaped FRP sheets).
4.5 New mathematical approaches to modeling creep

The above analysis demonstrated the importance of considering time-dependent
analysis in evaluating serviceability of masonry structures. It also showed the
importance of predicting creep with a good accuracy. Here we present findings of
recent research efforts to enhance the accuracy of predicting creep using means
of artificial intelligence. The inspiration for ANNs came from the desire to pro-
duce artificial systems capable of performing sophisticated (or perhaps intelligent)
computations that mimic the routine performance of the human brain. Artificial
neural networks are networks of many simple processors (neurons) operating in
parallel, each possibly having a small amount of local memory. Artificial neu-
ral networks resemble the brain in two respects: first, knowledge is acquired by
the network through a learning process and second, the interneuron connection
weights are used to store the knowledge [38, 39]. No closed-form solution for the
problem is provided by ANNs. However, ANNs offer a complex, accurate solution
based on a representative set of historical examples of the relationship [38]. The
units (neurons) are connected by weighted channels which are adjusted on the
basis of learning data. Artificial neural networks learn from examples (of known
input/output sequences) and exhibit some capability for generalization beyond the
training data. Artificial neural networks normally have great potential for paral-
lelism, since the computations of the components are largely independent of each
other [39, 40].
The creep deformations of masonry prisms subjected to different stress levels,
representing approximately (12, 24, 36, and 48%) of the prisms compressive
strength, respectively, and exposed to different environmental conditions [12]
were used to develop and assess the ANN model. A series of unloaded prisms
subjected to environmental conditions similar to their counterparts loaded
prisms was also measured at the same time intervals to account for shrinkage and
800 800
(1/MPa)
(1/MPa)
700 700
600 600
-6
-6
Predicted J(t,t 0) x 10
Predicted J(t,t 0) x 10
500 500
400 400
300 300
200 200
100 100
100 200 300 400 500 600 700 800 100 200 300 400 500 600 700 800
Measured J(t,t 0) x 10-6 (1/MPa) Measured J(t,t 0) x 10
-6
(1/MPa)
(a) (b)
15% RTN1 50% GERBER
Figure 4.17: Prediction of creep compliance using (a) an ANN model [41] and (b)
a regression analysis (modified Burger) model [12].
thermal changes. Test results from fourteen testing groups were included in train-
ing of the network. A series of multi-layer ANNs for predicting creep performance
of masonry structures was developed [41, 42].
The creep prediction neural network consists of an input layer, one hidden layer,
and an output layer. The network utilizes a log-sigmoid transfer function and a
linear output function. A backpropagation training algorithm was used as the learn-
ing rule for the network. A learning matrix including 47 training samples drawn
from the 14 testing groups was used in training the network. The Leveneberg-
Marquardt training criterion [42] was utilized during the learning process of the
network with a training goal of achieving a mean square error of 0.0002.
The network was then tested against groups of data that had not been used in
training the network. The creep compliance was computed by the network and the
output of the network was compared to the measured creep compliance. A com-
parison between the creep compliance prediction using ANN and a classical model
based on regression analysis [12] is shown in Fig. 4.17. While the ANN predic-
tions lie within 15% accuracy, regression analysis prediction lies within only 50%
accuracy. Research investigations showed that ANN accuracy can be further
enhanced by optimizing the network architecture [42] and by considering time-
delaying effects in the model [43].
4.6 Discussion
The step-by-step in time technique demonstrated here allows stresses at interme-
diate stages to be calculated. However, creep in masonry is a function of the age
at loading and the environment [4], so more complex analyses will be needed to
simulate reality. The step-by-step method can be used with the specific creep as the
input [44] and aging functions can be expressed in integral form [1, 21]. Shrinkage
and thermal effects can also be included. However, the number of factors that affect
masonry creep (age of loading, environment, unit type, mortar type, and stress
level) suggests that creep for a particular structure will be very difficult to predict.
Few data are available, particularly, long-term data. In these circumstances, meth-
ods using means of artificial intelligence (e.g. ANNs) which can deal with high
levels of stochastic variation may prove useful in predicting both the creep and the
range of possible outcomes from time-dependent effects. It becomes obvious that
using means of artificial intelligence in modeling creep have two advantages over
classical models, first: it allows incorporating a large number of interdependent
factors that affect creep without adding further complexity. Second, it provides a
systematic approach for model improvement as new data become available.
4.7 Conclusions
The step-by-step in time analysis is a powerful tool to investigate the change of the
stresses due to creep over a large time range. Stresses in a material can rise and fall
due to the effects of creep (or fall then rise).
Although the effective modulus method using the final creep coefficient can
accurately estimate the final stresses in the components of a composite material
(e.g. masonry) due to creep, the method may not predict intermediate peak stresses
accurately: one needs to know when they will occur.
Rehabilitation with FRPs may introduce additional creep mechanisms with
undesirable effects. Particularly, FRP strips may unload a part of any additional
dead load applied to a structure after strengthening. Adequate anchorage must be
designed.
The use of the step-by-step time-dependent analysis in creep stress redistribu-
tion can be made more accurate by incorporating ANNs where all factors affecting
creep deformations can be included in the modeling process.
References
[1] Ghali, A. & Favre, R., Concrete Structures: Stresses and Deformations, 2nd
edn, E&FN Spon: London, 1994.
collapse of the civic tower of Pavia: a survey of the materials and structure.
[3] Shenoi, R.A., Allen, H.G. & Clark, S.D., Cyclic creep and creep-fatigue in-
teraction in sandwich beams. Journal of Strain Analysis, 32(1), 118, 1997.
[4] Oh, Y.J., Nam, S.W. & Hong, J.H., A model for creep-fatigue interaction in
terms of crack-tip stress relaxation. Metallurgical and Materials Transac-
tions A, 31(7), pp. 17611775, 2000.
[5] Wang, E.Z. & Shrive, N.G., Brittle fracture in compression: mechanisms,
models and criteria. Engineering Fracture Mechanics, 52(6), pp. 1107
1126, 1995.
[6] De Borst, Feenstra, P.H., Pamin, J. & Sluys, L.J., Some current issues in com-
putational mechanics of concrete. Computational Modelling of Concrete Struc-
tures, eds H. Mang et al., Pineridge Press: Sansea, UK, pp. 283302, 1994.
[7] Wu, Z.S. & Baant, Z.P., Finite element modelling of rate effect in concrete
fracture with influence of creep. Creep and Shrinkage of Concrete, eds Z.P.
Baant et al., E. & FN Spon: London, pp. 427432, 1993.
[8] van Zijl, G.P.A.G., de Brost, R. & Rots, J.G., The role of crack rate depen-
dence in the long-term behaviour of cementitious materials. International
Journal of Solids and Structures. 38, pp. 50635079, 2001.
[9] Binda, L., Saisi, A., Messina, S. & Tringali, S., Mechanical damage due
to long term behaviour of multiple leaf pillars in Sicilian churches, Proc.
Historical Constructions, eds Lourenco & Roca, Guimaraes, Portugal,
pp. 707718, 2001.
[10] Lenczner, D., Creep in model brickwork. Proceedings of Designing Engi-
neering and Construction with Masonry Products, ed. F.B. Johnston, Hous-
ton, USA, pp. 19581969, 1969.
[11] Lenczner, D., Creep in brickwork piers. Structural Engineer, 52(3),
pp. 97101, 1974.
[12] Shrive, N.G., Sayed-Ahmed, E.Y. & Tilleman, D., Creep analysis of clay
masonry assemblages. Canadian Journal of Civil Engineering, 24(3),
pp. 367379, 1997.
[13] Brooks, J.J. & Abdullah, C.S., Composite modelling of the geometry influ-
ence on creep and shrinkage of calcium silicate brickwork, Proc. of the
British Masonry Society, No. 4, pp. 3638, 1990.
[14] Ameny, P., Jessop, E.L. & Loov, R.E., Strength, elastic and creep prop-
erties of concrete masonry. Int. Journal of Masonry Construction, 1(1),
pp. 3339, 1980.
[15] Van Zijl, G.P.A.G., A Numerical Formulation for Masonry Creep, Shrink-
age and Cracking, Series 11, Engineering Mechanisms 01, Delft Univ.
Press: The Netherlands, 1999.
[16] CSA, Masonry Code. Masonry Design for buildings (limit states design)
structures (design). CSA-S304.1-94, Canada, Ontario, 1994.
[17] SAA, Masonry in Buildings. Revisions of Australian Standard-SAA-AS
3700/1988. Standards Association of Australia, Sydney, Australia, 1995.
[18] Scott, D.W., Lai, J.S. & Zureick, A.H., Creep behaviour of fibre reinforced
polymer composites: a review of technical literature. Journal of Reinforced
Plastics Composites, 1(6), p. 588, 1995.
[19] Van den Boogaard, A.H., De Borst, R. & Van den Bogert, P.A.J., An adap-
tive time-stepping algorithm for quasistatic processes. Communication in
Numerical Methods in Engineering, 10, pp. 837844, 1994.
[20] Harvey, R.J. & Hughes, T.G., On the representation of masonry creep
by rheological analogy. Proceedings of the ASCE Structural Congress
Restructuring: America and Beyond. 1, pp. 385396, 1995.
[21] Neville, A.M., Dilger, W.H. & Brooks, J.J., Creep of Plain and Structural
Concrete, Construction Press: UK, 1983.
[22] Baant, Z.P., Prediction of concrete creep effects using age-adjusted effec-
tive modulus method. ACI Journal, 69(4), pp. 212217, 1972.
[23] Mirza, S., A framework for durability design infrastructure, Proc. First
Canadian Conf. on Effective Design of Structures, Hamilton, Canada,
pp. 67103, 2005.
[24] Valluzzi, M.R., Binda, L. & Modena, C., Mechanical behaviour of his-
toric masonry structures strengthened by bed joints structural repointing.
Journal of Construction and Building Materials, 19, pp. 6373, 2005.
[25] Baant, Z.P., Asymptotic temporal and spatial scaling of coupled creep,
aging diffusion and fracture process. Creep, Shrinkage and Durability
Mechanics of Concrete and Other Quasi-Brittle Materials, eds Ulm et al.,
pp. 121145, 2001.
[26] Lemaitre, J.A., Course on Damage Mechanics, 2nd edn, Springer Verlag:
Berlin, 1996.
[27] Sukontasukkul, P., Nimityongskul, P. & Mindess, S., Effect of loading rate on
damage of concrete. Cement and Concrete Research, 34, pp. 21272134, 2004.
[28] Garavaglia, E., Lubelli, B. & Binda, L., Two different stochastic approaches
modelling the deterioration process of masonry wall over time. Materials
and Structures, RILEM, 35, pp. 246256, 2002.
[29] Lland, K.E., Continuous damage model for load-response estimation of
concrete. Cement and Concrete Research, 10, pp. 395402, 1980.
[30] Reda Taha, M.M. & Shrive, N.G., A model of damage and creep interac-
tion in a quasi-brittle composite materials under axial loading. Journal of
Mechanics, 22(4), 2006.
[31] Saadatmanesh, H. & Tannous, F.E., Relaxation, creep and fatigue behaviour
of carbon fibre reinforced plastic tendons. ACI Materials Journal, 96(2),
pp. 143153, 1999.
[32] Gauthier, C., Bonnet, A., Gaertner, R. & Sautereau, H., Creep behaviour of
polymer blend based in epoxy matrix and intractable high Tg thermoplastic.
Polymer International, 53(5), pp. 541549, 2004.
[33] Shackelford, J.F., Introduction to Materials Science for Engineers, 6th edn,
Pearson, Prentice Hall: Upper Saddle River, NJ, 2005.
[34] CEB-FIP Model Code 90, Model Code for Concrete Structures, Comit
Euro-International du Bton (CEB)Fdration Internationale de la Prcon-
trainte (FIP), Thomas Telford Ltd.: London, UK, 1993.
[35] Masia, M.J., Shrive, N.G. & Shrive, P.L., Creep performance of reinforced
concrete beams strengthened with externally bonded FRP strips. Proceed-
ings of Int. Conference on Performance of Construction Materials, eds
El-Dieb et al., Cairo, Egypt, Vol. 2, pp. 12951304, 2003.
[36] Masia, M., Reda Taha, M.M. & Shrive, N.G., Investigations of serviceabil-
ity of reinforced concrete beams strengthened with FRP. Journal of Com-
posites in Construction, 2006 (in preparation).
[37] Meshgin, P., Creep of epoxy at the concretefiber reinforced polymer (FRP)
interface. MSc Thesis, Department of Civil Engineering, University of New
Mexico, 2006.
[38] Haykin, S., Neural Networks: A Comprehensive Foundation, Prentice Hall:

New York, 1999.
[39] Luger, G., Artificial Intelligence: Structures and Strategies for Complex
Problem Solving, 5th edn, USA: Addison Wesley, 2004.
[40] Tsoukalas, L.H. & Uhrig, R.E.. Fuzzy and Neural Approaches in Engineer-
ing, Wiley: New York, USA, 1997.
[41] Reda Taha, M.M., Noureldin, A., El-Sheimy, N. & Shrive, N.G., Artificial
neural networks to predict creep with an example application to structur-
al masonry. Canadian Journal of Civil Engineering, 30(3), pp. 523532,
2003.
[42] Reda Taha, M.M., Noureldin, A., El-Sheimy, N. & Shrive, N.G., Feedfor-
ward neural networks for modelling time-dependent creep deformations in
masonry structures. Proc. of the Institution of Civil Engineers, Structures in
Buildings, UK, 157(SB4), pp. 279292, 2004.
[43] El-Shafie, A., Noureldin, A. & Reda Taha, M.M., On investigating recur-
rent neural networks for predicting masonry creep, Proceedings of Third
International Conference on Construction Materials: (CONMAT 05), Van-
couver, Canada, eds Banthia et al., August 2005 [CD-ROM].
[44] Shrive, N.G. & England, G.L., Elastic, creep and shrinkage behaviour of
masonry. International Journal of Masonry Construction, 1(3), pp. 103109,
1981.
CHAPTER 5
Experimental study on the damaged pillars

of the Noto Cathedral
A. Saisi, L. Binda, L. Cantini & C. Tedeschi

Milan, Italy.
5.1 Introduction
The partial collapse of the Noto Cathedral required the choice to repair and reuse
it. This seems in contrast with the need for preserving the authenticity of the
monument.
Nevertheless a number of difficult problems, from social to safety concerns, sug-
gested the reconstruction of the lost parts. This chapter describes the on-site and
laboratory investigation carried out on the remaining parts of the Noto Cathedral, in
order to verify their state of conservation in view of reconstruction.
5.2 The collapse and the decision for reconstruction

On December 1990 an earthquake hit the eastern part of Sicily damaging old and
contemporary buildings in different towns. Noto, known as the Baroque city was
among them and several of its most beautiful buildings were seriously damaged.
Also the Church of St. Nicol, the Cathedral had damages to the vault, the lateral
domes and to the pillars, apparently no more than other buildings. Provisional
structures and scaffoldings were set up to sustain the damaged parts waiting for
the repair and strengthening intervention. The partial sudden collapse occurred on
March 13, 1996, fortunately without any casualty, and left the Noto community
astonished by the loss of one of its most famous buildings.
The church had been built in different phases suffering several casualties from
1764, over a previous smaller church opened in 1703 to the public and demolished
in 176970 as the new cathedral was growing. The Cathedral was opened in 1776.
In 1780 the dome collapsed and the church was reopened in 1818. In 1848 the
dome collapsed again under an earthquake and then it was rebuilt and the church
reopened again in 1862 but the dome was not completely finished until 1872. In
1950, the cathedral was restored with new renderings and paintings and the timber
roof substituted with a concrete structure; the work continued until 1959.
The losses caused by the collapse were the following: 4 pillars of the right part
of the central nave and one of the 4 pillars sustaining the main dome and the tran-
sept, the complete roof and vault of the central nave, three quarter of the drum and
dome with the lantern, the roof and vault of the right part of the transept and part
of the small domes of the right nave (Fig. 5.1) (see Chapter 1).
The extensive experimental and numerical investigation carried out after the
removal of the ruins by a team of experts together with the designers R. De
Benedictis and S. Tringali [1, 2] clearly showed that the collapse started from
one of the pillars, due to the damaged condition they were in before the earthquake
(see Chapter 1).
Taking into account the clear weakness of the collapsed pillars, the designers
asked for a further careful investigation on the remaining pillars of the central nave
which also were damaged by the collapse. The first idea was to repair and preserve
these pillars during the reconstruction of the cathedral. This chapter describes the
on-site and laboratory investigations carried out and how the designers had to take
the decision of demolishing them.
Other cases of similar damages that occurred in Sicily will also be described
showing the importance of investigation in order to prevent future failures.
5.3 On-site investigation on the remaining parts of the

collapsed pillars
After the study of the collapse mechanism, carried out on site during the removal
of the ruins [3], the attention of the consultants was focused on a careful study of
the peculiar features of the collapsed pillars.
M1B
P1A P1B P1C P1D P1E
PA PB PC PD PE
MB
Figure 5.1: Plan of the remains and of the tested elements.

STUDY ON THE DAMAGED PILLARS OF THE NOTO CATHEDRAL 111
5.3.1 Layout of the section and of the masonry morphology
The removal by layers of the components of the collapsed pillars allowed one to
understand the poor technique of construction used. Layers of large round river
stones with thick mortar joints, where the mortar appeared very weak and dusty,
were found in the core of the structure, surrounded by an external leaf made with
regular blocks of more compact limestone at the base of the pillars (Fig. 5.2).
Since only the base had remained after the collapse and the symmetric pillars were
still covered by plaster, the hypothesis was made at first that limestone had been
used for the external part of the whole pillars.
This material, compact but not very strong, came from sedimentary carbonatic
depositions, which can be found in the area and are still used as quarries for the
building industry [4]. Inside the rubble filling also pieces of a material full of voids
were found, which was called travertine; this material is of the same nature as the
limestone, but deposited in the presence of turbulent waters and it is rich in voids
of various shape and dimensions, which previously contained vegetarian and
organic parts that later on dissolved.
The height of the blocks varies from 25 to 30 cm and the thickness, small compared
to the pillar dimensions, is ranging from 25 (stretcher) to 40 cm (header). No really
effective connection was realised between the external leaf and the core. The stones
of the pillar strips supporting the arches, vault, and domes have no connection either
to the internal masonry or to the other parts of the external leaf (Figs 5.3 and 5.4).
The inner part of the pillars represents 55% of the entire section, while in the
pillars sustaining the dome it is 58%. This part is a rubble masonry made with
irregular stones. It could be seen from the ruins, that it was made with large round
river pebbles approximately up to half of the total height. The courses of these
stones are rather irregular without any transversal connection or small stones to fill
the voids and with thick mortar joints. Nevertheless every two courses of the exter-
nal leaf (about 50 cm) a course made with small stones and mortar was inserted in
order to obtain a certain horizontality (Figs 5.2 and 5.4). Scaffolding holes were
left everywhere, some crossing the whole section.
5.3.2 General characterisation of the materials
The mortar appeared to be very weak made with lime and a high fraction of very
small calcareous aggregates. Also the bond between the mortar and the stones was
very weak; in fact it was possible to remove stones and pebbles from the interior of
the pillars without any difficulty and with the stones being completely clean.
This poor technique of construction and the use of the weak limestone (actually
called Noto stone) typical in the Noto region, was probably the cause of the dam-
ages to the pillars of the cathedral, even if a clear crack pattern was reported to have
appeared only after the 1990 earthquake. The walls were built similarly; neverthe-
less, the internal part was made with smaller sharp stones alternated with a slightly
stronger mortar, in some ways a better masonry. Some stones were sampled from
pillars and walls and mortar samples were taken from horizontal, vertical joints and
Figure 5.2: A collapsed pillar.
Figure 5.3: Horizontal pillar section.
from the interior of the masonry (Fig. 5.2). The samples were sent to the DIS Labo-
ratory in Milan and tested in order to find the material characteristics [1, 5].
The investigation (Fig. 5.5) has shown that the foundations of pillars and walls
were sufficiently well constructed; rubble walls but with enough load carrying capac-
ity for the weight of the structures above. The soil was a sort of natural compact silt
and thick layer of clay from where also the aggregates of the mortars were taken.
Figure 5.4: Reconstruction of a pillar section.
Figure 5.5: Detail of the foundation.

Up to this point of the investigation, even if the weakness of the material used
seemed to be the cause of the high damage suffered from the earthquake it was not
still clear why the pillars reached the point of collapse.
5.3.3 Damage description
The pillars on the left, still covered with a thick plaster, seemed to have suffered
minor damages; only small and diffused vertical cracks were present on the plas-
ter. Nevertheless the doubt that the damage could be deeper inside and perhaps
even present before the 1990 earthquake, led one to carry out on these pillars a
more accurate survey.
As the plaster applied at the end of the repair works in the fifties was partially
removed, a series of vertical large cracks was found, some of which were filled
with the gypsum mortar used for the plaster (Fig. 5.6).
This finding gave the authors the first evidence that the damage was already
present in the fifties. The pre-existing crack pattern was clearly damage from com-
pressive stresses, a long-range damage dating probably from a long time before
the fifties. The lesson after the collapse of the Civic Tower in Pavia and the subse-
quent research taught the authors that the damage would probably have progressed
even without the earthquake, which only accelerated the collapse. After the recog-
nition of the damages, the removal of the plaster from all the pillars was planned
in order to survey the crack pattern.
Figure 5.7 shows a reconstruction of the crack pattern of the pillar before remov-
ing the plaster from all the pillars. As it is evident, the cracks are diffused and
interesting the whole prospect, with a concentration in the corners.
5.3.4 Laboratory testing
On the materials sampled on site physical, chemical, petrographic- mineralogical

and mechanical tests were carried out in Milan at the DIS Laboratory. The aim
was to characterise the materials of a typical (CC) transversal section (Fig. 5.8)
of the cathedral [5].
Figure 5.6: Large crack in a pillar and example of a crack filled with mortar.
Figure 5.7: Prospect of pillar P1B and Figure 5.8: Transversal Section CC.
survey of the crack pattern.
5.3.4.1 Mortars
The chemical and mineralogical analyses were carried out following a procedure
set up in [6] on the mortars sampled from all the pillars and walls at different
heights. The mortars contain a high percentage of CaCO3 showing that they are
based on hydrated lime with a slightly high content of soluble silica but with fine
aggregate size distribution. Table 5.1 gives the results on the mortar of pillar P1A
together with the crack repair composition. In the last column, the composition
and the high percentage of SO3 shows the presence of gypsum.
Figure 5.9 gives an example of grain size distribution of aggregates. The soil
was also examined and it appears of being composed by more than 87% of calcium
carbonate, by 8% of different silicates and for the remaining 5% by alkali,
aluminium, iron, gypsum, etc. The grain size distribution of the soil shows that it
is composed of clay (8%), silt (72%) and sand (20%), a very fine material.
5.3.4.2 Stones
Some compressive tests were carried out on cylindrical samples of the two stones,
limestone and travertine; their texture is shown in Fig. 5.10. The tests performed on
the limestone show that its strength when saturated at constant mass (11.56 N/mm2)
drops dramatically with respect to the strength measured when dried at constant
mass (17.98 N/mm2). The compressive strength of the travertine is very low and can
vary from 4 to 6 N/mm2 or more.
Table 5.1: Chemical analysis and bulk density of pillar P1A mortars.
Pillar P1A Pillar P1A
13CNMP3 (%) Filling of cracks (%)
SiO2 4.27 2.32
Al 0.70 0.77
Fe2O3 1.05 0.28
CaO 50.22 41.05
MgO 0.41 0.14
Na2O 0.47 0.43
K2O 0.98 1.06
SO3 0.64 39.02
Loss on ignition 41.13 14.83
CO2 40.50 11.92
Ins. Res. 4.20 1.73
Soluble Silica 0.66 0.30
Cl 0.025 0.028
Bulk density 1313.00 kg/m3 1489.00 kg/m3
100
90
80
70
60
(%)
50
40
30
20 CNMABS1A
10 CNMN51
0
0.01 0.10 1.00 10.00 100.00
(mm)
Figure 5.9: Grain size distribution.
It should be stressed that the overall strength of the pillars, or better of the external
regular leaf, was lower due to the influence of the mortar joints. Tests carried out at
the Politecnico showed that the strength of a masonry reproducing the external
leaves drops by 4045% when mortar joints are present [7].
Furthermore, it should also be taken into account the reduction due to dimen-
sional ratio when dealing with the real structure.
Figure 5.10: Calcarenite (Stone of Noto) and travertine.
In order to know the response of the two stones to the elastic waves, the ultra-
sonic velocity was determined by transmission on stone blocks. The two materials
show very different behaviour. In fact, in the case of the limestone (calcarenite) the
values are almost constant, between 2912 and 3157 m/s with an average of 3068 m/s.
The values of the travertine are more scattered, with a measured velocity between
1325 and 3548 m/s, and an average of 1823 m/s. The scattering of the data is due
to the presence of large voids, randomly distributed in the material, and confirms
the results of the mechanical tests.
5.3.4.3 Injectability tests

Injectability tests proposed in [8] were carried out in laboratory on materials
sampled from the internal part of the pillars and walls, and from the collapsed pillars
of the cathedral [4]. Grout injection was controlled directly on-site as well [4, 9].
Finally it was decided that injection could not help in improving the behaviour
of the damaged pillars since it could not penetrate or strengthen the very thick but
weak mortar joints.
5.3.5 On-site tests
5.3.5.1 Flat-Jack tests

A single flat-jack test was carried out on the pillar P1E in order to know the state
of stress simply due to the dead load of the pillar itself and a value of 0.85 N/mm2
was found at a height of 3.00 m. Taking into account the missed weight of arches,
vaults and dome in the collapse, it is easy to make the hypothesis that the pillars
must have been under a non-negligible state of stress.
Double flat-jack tests were also carried out on pillars P1E and P1A and on the
external walls of the cathedral.
2.0

CNJ1d
1.5
Elastic modulus:1760[N/mm2]
Stress [N/mm2]
1.0 CNJ2d
Elastic modulus:190[N/mm2] Local stress
0.5
l v
0.0
-6.0 -4.0 -2.0 0.0 2.0 4.0 6.0

Strain [m/mm]
Figure 5.11: Double flat-jack test carried out on the external and internal part of
P1A.
A double flat-jack test was also carried out on the inner part of pillar P1A in
order to check the behaviour of its weakest part. Figure 5.11 shows the difference
between the external leaf (CNJ1D) and the core (CNJ2D), which had a much
higher deformability and lower strength.
5.3.5.2 Application of sonic pulse velocity test to pillars

As a confirmation of the state of damage, and also a calibration of the procedure,
sonic pulse velocity tests were carried out on the remains of the collapsed pil-
lars, as well [9]. It is well known that ultrasonic frequencies can not be used on
rubble walls due to the high attenuation caused by joints, voids and homogeneities.
Nevertheless, travertine and calcarenite being so different the ultrasonic velocities
measured in laboratory on single blocks were very useful.
Figure 5.12 localises as an example the test position on the pillar P1E and in the
correspondent collapsed pillar called PE.
Measurements were taken at different heights. It was impossible to position
equal levels for all the pillars due to the presence of safety scaffolding. Neverthe-
less it was clear that the material of the external blocks was changing from the base
(calcarenite) to the top of the pillars (travertine).
The measurements were also carried out on some parts of the external walls as
a comparison. Figure 5.13 shows the average values found for pillars on the leftand
for the external pillar and wall called M1B.
Low velocity values were systematically recorded for all the tested pillars of the
cathedral from about 1.001.50 m onwards, i.e. above the limestone base. The
values reported in Fig. 5.14 are average values of the measurements carried out in
the two orthogonal directions. The pillar P1B shows the lowest values of the sonic
velocities recorded at each level compared to the other P1i pillars. The pillar state
Figure 5.12: Geometry of the pillars P1E and PE and localisation of the sonic
tests.
Sonic velocity
[m/sec]
1800
1400
1200
1000
800
600
400
Figure 5.13: Vertical distribution of the Figure 5.14: Sonic velocity

sonic velocity measured on the pillars distribution in the P1A section
and the walls. at 90 cm.
is in fact characterised by very serious damage, as evident from the crack pattern
in Fig. 5.7.
5.3.6 Design decisions
The accurate and detailed survey carried out by a multidisciplinary team was very
helpful for the designers who had to take many difficult decisions. The crack pat-
tern survey revealed large vertical cracks already present and filled with gypsum
mortars in the sixties when the timber roof of the cathedral was replaced by a
concrete roof. These damages indicate, together with the laboratory results that
Figure 5.15: State of damage observed after the removal of the plaster.
the material used for the construction was very weak and damaged by long-term
effects; the collapse perhaps would have taken place later if the earthquake had
not occurred.
Figure 5.15 shows, as a confirmation, the state of damage of one of the pillars as
observed after the complete removal of the plaster.
The pillars on the left could not be preserved due to the high state of damage
caused by the weak technique of construction and the weak materials; therefore it
was decided to demolish and rebuild them, together with the collapsed ones using
better materials and technique of construction, i.e. hydraulic mortars obtained with
hydrated lime and pozzolana, calcarenite stones avoiding travertine and good con-
nections between the external leaf of the pillars and the core.
The soil and foundation seem to be acceptable everywhere; the only exception
can be made for the foundation of the central nave pillars, which can eventually
follow a new conception.
5.3.7 The dismantling of the remaining pillars
The substitution of the left pillars takes place in alternate order demolishing one
pillar at a time and reconstructing it. Before this operation, the vaults of the left
nave are supported by a stiff steel structure (Fig. 5.16).
Figure 5.16: Steel structure supporting the vaults.
Figure 5.17: Detail of fractures.
The dismantling of every single pillar is carried out in successive steps,

demolishing stone by stone every single course. The use of the local travertine is
confirmed, as well as the serious state of damage and the lack of connection
between the external stone leaf and the core.
The stones showed passing through cracks or deep cracks; when lifted by the
workers, the blocks often broke, revealing the internal large voids.
Furthermore the fractures in the external surface could be observed also in the
internal rubble, even if less readable for the high inhomogeneity of the masonry.
They in fact can follow the boundary between the mortar and the pebbles, but also
go through every single stone (Fig. 5.17).
Figure 5.18 shows as an example of the sequence of demolition of one course of
pillar P1B. It can be seen that the vertical and horizontal joints are so weak that the
dismantling can be carried out by hands.
Figure 5.18: Example of a sequence of demolition of one course of pillar P1B.
Furthermore the inside of the stones, which are practically all broken or fis-
sured, is full of voids and very weak. This explains why the pillar did not bear the
state of stress for a long time. Creep phenomena have certainly developed during
the life of the pillars, lowering their strength.
Moreover, the internal filling of the pillars showed very low damage compared
to the external leaves.
This situation could be recently explained by some compressive tests carried out
on three-leaf stimulating model of the section of the Noto pillars [7]. From these
tests it is clear that the stress under compression is transferred even at low values
to the external leaves when the internal one, free from stresses, is detached from
the bearing one and does not suffer damages even at peak stress level.
References
[1] Binda, L., Baronio, G., Gavarini, C., De Benedictis, R. & Tringali, S.,
Investigation on materials and structures for the reconstruction of the par-
tially collapsed Cathedral of Noto (Sicily). Proc. STREMAH 99, Dresden,
Germany, pp. 323332, 1999.
[2] Tringali, S., De Benedictis, R., La Rosa, R., Russo, C., Bramante, A.,
Gavarini, C., Valente, G., Ceradini, V., Tocci, C., Tobriner, S., Maugeri,
M., Binda, L. & Baronio G., The reconstruction of the Cathedral of Noto.
Proc. Int. Symp. on Earthquake Resistant Engineering Structures (ERES II),
Catania, pp. 499510, 1999.
[3] De Benedictis, R., Tringali, S., Gavarini, C., Binda, L. & Baronio G.,
Methodology applied to the removal of the ruins and to the survey of the
remains after the collapse of the Noto Cathedral in Sicily. Proc. STREMAH
99, Dresden, Germany, pp. 529538, 1999.
[4] Binda, L., Baronio, G., Tiraboschi, C. & Tedeschi, C., Experimental research
for the choice of adequate materials for the reconstruction of the Cathedral
of Noto. Construction Building Materials, Special Issue, 17(8), pp. 629639,
2003.
[5] Baronio, G., Binda, L., Tedeschi, C. & Tiraboschi C., Characterization of the
materials used in the construction of the Noto Cathedral. Construction Build-
ing Materials, Special Issue, 17, pp. 557571, 2003.
[6] Baronio, G. & Binda, L., Experimental approach to a procedure for the in-
vestigation of historic mortars. Proc. 9th Int. Brick/Block Masonry Conf.,
Berlin, pp. 13971405, 1991.
[7] Binda, L., Anzani, A. & Fontana, A., Mechanical behaviour of multiple-
leaf stone masonry: experimental research. 10th International Conference
Structural Faults and Repair, London, 13 July 2003, Engineering Technical
Press, Edinburgh, Keynote Lecture, 2003.
[8] Binda, L., Modena, C. & Baronio, G., Strengthening of masonries by injec-
tion technique. Proc. 6th NaMC, Vol. I, Philadelphia, pp. 114, 1993.
[9] Binda, L., Saisi, A. & Tiraboschi, C., Application of sonic tests to the
diagnosis of damage and repaired structures. NDT&E Int. Journal, 34(2),
pp. 123138, 2001.
CHAPTER 6
Monitoring of long-term damage in long-span

masonry constructions
P. Roca1, G. Martnez1, F. Casarin2, C. Modena2, P.P. Rossi3,

I. Rodrguez4 & A. Garay4
1Universitat Politcnica de Catalunya, Barcelona, Spain.
2Universit degli Studi di Padova, Italy.
3R. teknos Srl, Bergamo, Italy.
4Labein, Bilbao, Spain.
6.1 Introduction
Monitoring is a key activity in the study of ancient structures, providing reliable
insight into its present condition and the significance and progress of damage.
Monitoring can contribute to identifying and evaluating existing damage and help
determine which active physical phenomena are involved in its generation. How-
ever, which monitoring procedures and strategies are to be considered depends
highly on the type of processes experienced by the structure. Four different
phenomena large deformation, tensile damage, compressive damage and large-
scale fragmentation are considered and discussed with regard to monitoring pos-
sibilities for measuring and characterizing them. Not only the structural effects
(rotations, displacements, crack openings, etc.), but also the actions experienced
by the construction (wind, earthquake, thermal cycles, soil settlements, etc.) are to
be measured together over a sufficiently long period comprising several years. The
use of a detailed numerical model may, in most cases, allow accurate physical and
quantitative interpretation of the information obtained.
6.2 Monitoring and long-term damage

Large deformation and damage are observed in almost all ancient masonry con-
structions. Structures and particularly masonry structures are not fully inert
and static objects, but living entities which experience active processes through-
out their entire life span. Such processes, related to construction effects, mate-
rial decay, environmental actions and extraordinary actions (such as earthquakes)
manifest in deformation and damage which develop in the long term.
Monitoring can provide a certain degree of insight into the condition of the
structure and the possible presence of active processes associated with incremental
damage. However, the results of a monitoring programme are only fully compre-
hensible when analysed in the light of the historical nature of the construction or,
in other words, when its results are interpreted as an evidence of processes which
act and evolve over a deep-time or historical scale.
Monitoring can be understood as the attempt to open a small window in the
domain of time, over a response that develops over centuries or millennia. The
challenge, thus, is to develop possible hypotheses or conclusions on the condition
of the structure and the phenomena acting upon it, based on just a small, almost
infinitesimal, patch or picture of the variation of the structural response in the time
domain (Fig. 6.1).
Deformation and damage develop as a superimposition of different phenomena,
some of which act persistently, some cyclically or periodically, others occurring
only on isolated occasions. The effects of these different actions result in the final
response of the structure. Obtaining realistic conclusions through a monitoring
programme requires the ability to unravel the registered information into its differ-
ent components, so that they can be related to different actions or phenomena. In
particular, the information obtained can include an assortment of reversible (cyclic)
components mixed with the long-term accumulation of irreversible components.
Obtaining meaningful hints related to long-term damage requires the ability to
amplitude
Damage
Safety
DT
DT Long term
Figure 6.1: Monitoring as a window over historical time.

Long-Term Damage in Long-Span Masonry Constructions 127
(a)
t
(b)
(c)
t
(d)
Figure 6.2: Breakdown of a captured response into (a) cyclic, (b) circumscribed,
(c) monotonic and (d) components.
distinguish the unidirectional, accumulative trends from the totality of data regis-
tered during the monitored period (Fig. 6.2).
Characterizing long-term damage is a challenging task due to the slowness of
the processes involved and the fact that they may be masked by more apparent,
short-term variations caused by present environmental actions. Both the viability
of characterizing long-term damage by monitoring and the possible strategies that
can be used to attain it are discussed in the following paragraphs.
6.3 Role of monitoring in the study of ancient constructions

Monitoring provides quantitative information on the response of the structure across
a short, recent period of time. Monitoring may specifically allow recognition of
incremental processes over a term reasonable for engineering purposes, and thus
provide information useful for the study and restoration of ancient constructions.
Both structural analysis and monitoring deal with quantities and thus allow
direct comparison. Monitoring results can be used in combination with a numeri-
cal model, provided that not only the parameters associated with the response
(deformations, displacements, rotations, vibrations, etc.) are measured, but also
those characterizing actions (environmental thermal effects, ground motion, etc.).
The role of monitoring in the study of an ancient construction is better under-
stood in the light of the application of scientific methodology based on a multidis-
ciplinary approach. Experts involved in the study of historical structures of the
architectural heritage base their research on a combined set of activities, including
historical investigation, inspection and structural modelling. Monitoring constitutes
HISTORY PRESENT MONITORING STRUCTURAL

CONDITION MODELLING
EMPIRICAL EVIDENCE HYPOTHESES
Figure 6.3: Activities involved in the study of an ancient construction and their role
in the application of the scientific method. The structural model is the
receptacle of the hypothesis to be validated by the empirical evidence
provided by historical research, inspection and monitoring.
a fourth complementary activity intimately related to the rest (Fig. 6.3); used
alongside the other activities, monitoring contributes to the successful application
of the scientific method in the study of ancient constructions.
Applying the scientific method first requires adopting a set of hypotheses and,
second, the use of available empirical evidence to prove them. Some of the activi-
ties mentioned (specifically, structural modelling) are related to the first stage of
the process, namely, the adoption of hypotheses. The structural model is the recep-
tacle of the hypotheses on the physical and mechanical nature of the construction.
History, inspection and monitoring are activities intended to provide the empirical
evidence needed to validate these hypotheses or to correct or improve them to a
satisfactory extent.
More specifically, monitoring produces quantitative measurements which allow
comparison to the numerical predictions of the structural model. System identifi-
cation can be undertaken to adjust the material properties or morphological fea-
tures of the model. An updated or calibrated model, with enhanced predictive
capacity, is thus obtained.
6.4 Monitoring: methodology and requirements

6.4.1 Technology
Nowadays a variety of sensors and associated equipment is available for measur-

ing structural movements (including absolute or relative displacements, rotations,
settlements or accelerations) and environmental parameters such as internal or
external temperatures, humidity or wind force and direction. Electronic devices
and sophisticated digitizers provide reliable automatic data collection systems and
fast and remote recovery of large amounts of data.
Removable mechanical extensometers or electrical crack gauges are commonly

used to measure relative movements between crack surfaces. Not only is it the
opening of cracks which is meaningful; attention must also be paid to the relative
movements tangent to the crack faces; three different extensometers, placed
perpendicularly, can be used to measure possible movements between crack
faces.
Absolute horizontal movements of vertical members can be registered by means
of direct pendulums with measuring systems based on telecoordinometers. Rela-
tive horizontal movements can be measured more inexpensively and easily by
means of long-base extensometers. Rotations of either vertical or horizontal
elements can be measured by fixed or removable clinometers.
Differential settlements can be measured by levemetric vessels where the level
of the liquid is registered by an electrical transducer. Settlement gauges and pie-
zometers are used to analyse the deformation of the soil foundation in relation to
the water table variations.
Air temperature and the temperature gradient across the wall thickness can be
measured by thermal-gauges fitted inside small diameter boreholes drilled in the
walls.
A monitoring programme must be laid-out in accordance with a precise defini-
tion of its objectives and scope. The design of a monitoring system must bear in
mind conditions related to the environment (protection, accessibility, etc.), the
necessary accuracy (of instruments and also of the entire system), system reliability
(possibility of self-diagnosis, redundancy, etc.), flexibility (easy substitution and
recalibration of sensors) and the maintenance needs [1].
More information of the technological alternatives and their application to
specific studies can be found in [14].
6.4.2 Distinction between dynamic and static monitoring
The monitoring system must be adequately designed to satisfy its intended pur-
pose. Rather than universal, all-purpose systems, more specialized systems aimed
at specific targets may be more efficient and also less expensive.
In particular, a distinction can be made between static monitoring, aimed at the
continuous measurement of gradual, slow-varying parameters over a long period,
and dynamic monitoring, aimed at the intensive measurement of sudden variations
caused by isolated and short-lived actions (such as micro-tremors or hurricanes),
over a brief interval of time (Fig. 6.4).
Static monitoring requires the regular measurement of small variations over
lengthy periods comprising several years. In principle, there is no need to register
measurements at a very high frequency. A few measurements per minute, or even
per hour, may be enough to characterize the variations caused by daily climatic
cycles or other periodical or gradual effects.
Dynamic monitoring is intended to characterize the dynamic or seismic response
of the building. It can be carried out by means of dynamic tests measuring the
motion of the building caused by forced or natural vibration.
Figure 6.4: Continuous static monitoring, circumscribed (threshold) dynamic

monitoring and continuous dynamic monitoring.
Another possibility is to install a fixed system capable of self-activating and

capturing the motion of the structure at every occurrence of a micro-tremor or any
other significant vibration source above a certain threshold.
Dynamic monitoring requires the ability to capture a very dense amount of
information during a very short interval. Thousands of readings per minute (for
instance, 200 readings per second) may be needed to adequately characterize the
oscillation of the structure caused by an external source of vibration, and to later
carry out the signal processing leading to the determination of significant dynamic
properties such as the shapes of the vibration modes, frequencies and damping.
High sensitivity sensors are needed when measuring natural vibrations caused by
traffic, wind or micro-tremors. Fixed dynamic monitoring may provide valuable
information specifically related to the response of the structure during micro-tremors
or even significant earthquakes. Long-term variations of damage are also better
measured by means of a fixed system left active over a long period. Depending on
the chosen threshold, a fixed dynamic system may require significant data storage
capacity (or alternatively, frequent information transfer from data-loggers to other
storage media). Meaningful information has been recorded using this type of
systems for several ancient towers. In the case of the Torrazzo (civic tower) of
Cremona, the monitoring system allowed the detection of wind-forced oscillations
due to vortex shedding excitation [5].
The continuous capture of dynamic motion over long periods, covering several
months or even years, is also possible thanks to more recent technological devel-
opments concerning dynamic data acquisition. Modern portable instruments,
equipped with large storage capacity (tens or hundreds of Gb), allow the capture
of continuous and dense information over long periods of time without having to
set up an activating threshold. This last possibility, used in the dynamic monitoring
of Mallorca Cathedral (Section 6.7.1), is particularly useful when the amplitude of
the seismic motion expectable in the short term is very low and similar in magni-
tude to the effects of wind or traffic; in these cases, it may be difficult or even
impossible to set a threshold that would allow the specific capture of seismic
motion. Linking the equipment to GPS time, by means of a GPS antenna, allows
the information collected to be synchronized with seismic events registered at seis-
mic stations. This specifically enables information related to meaningful seismic
episodes to be extracted from the entire volume of data registered over a long
period.
System identification on data recorded during low-intensity earthquakes has
been successfully used to characterize the dynamic. response and the effect of soil
structure interaction in the case of Hagia Sophia in Istanbul [6]. Non-linear behav-
iour was identified even at very low response levels; this non-linearity can be
related to existing damage and might become more evident for large intensities.
Dynamic monitoring provides the only way to experimentally measure param-
eters related to the global structural behaviour of the historical construction. How-
ever, its real contribution to a clear understanding of structural damage propagation
is strongly limited due to several causes. The parameters related to the dynamic
response of the structure behave always in the non-linear range (at least those of
interest for damage detection) and are highly sensitive to the local or global mate-
rial properties and the support conditions. Furthermore, the dynamic response of
the structure may be highly influenced by the soilstructure interaction. No theo-
retical or numerical tools are yet available to simulate such effects in an accurate
way, and thus to assist in the interpretation of the influence and variation of such
parameters. However, dynamic monitoring can be very useful in carrying out
model calibration or sensitivity analyses, especially when combined with comple-
mentary information on other experimentally measured static parameters (local
Young modulus, local stresses, via flat-jack measurements).
6.4.3 Requirements
Besides the technical challenges posed by the need to acquire information reliably,
technicians also face the need to interpret results adequately. In order to ensure
the correct interpretation of measurements, a series of requirements should be
considered:
1. Before or while undertaking a monitoring programme, detailed characteriza-

tion of the building is needed. Historical investigation and geometrical and
morphological surveys are needed to allow correct interpretation of the monitoring
output. Monitoring will normally be accompanied (or preceded) by characteriza-
tion based on non-destructive or quasi non-destructive tests aimed at determining
the internal morphology of the structural members and the mechanical properties
of the materials. Damage patterns (particularly major cracks) must also be recog-
nized and carefully documented. The foundation (soil and structure) must be
characterized carefully since it may significantly influence the motion and

deformations to be monitored.
2. Specific and actually monitorable targets should be selected, adequately related
to the physical phenomena to be identified or analysed. In particular, long-term
damage will require the monitoring of effects such as crack opening and defor-
mation. Characterizing different phenomena (soil settlements, deformations
caused by thermal cycles, mechanical deformations caused by loads, etc.)
requires specific monitoring strategies.
3. Actions affecting the construction are to be monitored in combination with its
structural response. This normally requires monitoring of climatic environmental
parameters (temperature and humidity), wind parameters (speed and direction),
seismic ground motion and soil settlements, among others. The obvious aim is
to correlate causes (actions) and effects (structural response). Furthermore, the
varying effect on the structure of different actions can only be extracted accu-
rately from the overall response if the actions themselves have been accurately
measured and characterized.
4. Even if climatic actions are not the target they will have to be characterized,
since their impact on the structure is normally very prominent and may alter or
even mask deformations caused by other possible effects, such as those specifi-
cally linked to long-term damage. In this case, characterizing climatic actions
is necessary in order to determine and cancel out the climatic component in the
monitoring output. In order to characterize incremental, long-term processes,
monitoring must be designed to allow a clear distinction between the reversible
or cyclic components of the parameters measured, on the one hand, and their
irreversible, cumulative components on the other.
5. An accurate numerical model must be available to interpret the results and cor-
relate the causes identified (measurements related to actions) with their effects
(deformations or displacements measured at different critical points of the
building) in light of hypotheses on the configuration and condition of the struc-
ture. Characterizing the action in the time domain will later allow its numerical
simulation and comparison between the numerical prediction and the actual
response measured. An identification process can then be carried out by ade-
quately modifying such hypotheses until a satisfactory coincidence is attained
between the numerical predictions and the measurements.
6. Monitoring must be carried out over a period long enough to cover the entire
duration of the cyclic actions at work; since annual variations in temperature
must be considered in all cases, the minimum acceptable period is a complete
year. Additional years will be of value to confirm the tendencies observed and
appraise their possible long-term evolution. In fact, a period of four years is a
more reasonable minimum, since it allows confirmation of tendencies and de-
tection of anomalous, local (in time) measurements produced by extraordinary
actions or alterations in the monitoring system itself.
7. In order to provide meaningful information with regard to the monitoring target,
critical points of the structure must be selected. Prior numerical simulation may
help determine the optimal configuration and location points. Normally, the
most adequate points to install sensors will be those showing comparatively

large or even maximum displacement.
8. The global nature of structural response must be taken into account while de-
signing the monitoring strategy or when interpreting its results. Monitoring
points should not be chosen individually or based on local considerations, but
according to a strategy involving meaningful structural parts or even the entire
structure.
9. Moreover, the monitoring system must be designed to allow redundant mea-
surement of related effects, allowing results to be interpreted more consistently
and soundly. For instance, displacements or rotations of a faade experiencing
a gradual out-of-plumb can be measured in combination with related crack
openings experienced at the junction of the faade with other walls.
6.5 Measuring damage and deformation related to

historical or long-term processes
6.5.1 Monitoring and long-term damage
Long-term processes may cause damage to manifest in different ways. Some of the
most significant structural disorders observed are (1) overall large deformations,
(2) cracks in elements subject to tension, (3) cracking or crushing in elements sub-
jected to compression and (4) large cracks causing separation between different
structural components or significant portions of the building (fragmentation). The
following paragraphs consider these types of disorders and how monitoring might
contribute to their characterization.
6.5.2 Structural deformation
Large deformations affecting piers, buttresses and other structural components

are commonly observed in ancient constructions (Fig. 6.5). In most buildings,
deformation has been determined by many different actions occurring both dur-
ing the construction process and the later historical life of the building. Numeri-
cal approaches may normally provide fair qualitative simulations of the deformed
shape, but they fail at predicting, even roughly, the absolute values of deformations
and displacements. In many cases, the real deformation is one or more orders of
magnitude superior to those predicted by instantaneous or short-term numerical
analysis; this is so even when non-linear material or geometrical effects, or even a
conventional treatment of primary creep, are considered.
Important effects related to deformation can be ascribed to construction. On the
one hand, construction of historical structures spanned large periods of time which,
in turn, included lengthy stages during which the structure was subject to provisional
support conditions; during these intermediate phases, the structure was forced to
develop resisting mechanisms not entirely consistent with its final arrangement and
design. Significant deformation could be expected during these phases, due to the
.
Figure 6.5: Kk Ayasofya Mosque in Istanbul. Example of large deforma-
tion of an ancient building. Condition of the real construction and
predicted deformed shape (m) [7].
more deformable or mobile character of the incomplete substructures and provi-

sional wooden, iron or masonry buttressing systems.
Even after structure completion, creep will tend to partially amplify the defor-
mation acquired at intermediate construction phases, causing both a significant
increase of deformation and possible stress redistributions, eventually leading to
cracking due to internal deformation incompatibility.
Actions occurring after the construction process may have also contributed very
significantly to the continuous increase of deformation. Extraordinary actions such
as large earthquakes may produce important lesions and irreversible deformations.
Low-intensity earthquakes or repeated occurrences of hurricane-force winds may
act cumulatively to cause ever-increasing damage and deformation. The individual
effect of daily or annual thermal cycles is minimal; however, a certain, irreversible
increment of deformation may take place after each cycle, thus contributing to a
significant increase in overall deformation over very long periods of time. It must
be noted that the effects of cyclic actions do not dissipate with time, but may
increase and accelerate as the construction becomes increasingly damaged.
Damage affecting the construction, which in normal circumstances always
increases due both to the aforementioned and other possible causes, will, in turn,
enlarge the sensitivity of the construction towards a variety of actions. This situa-
tion contributes to constantly increasing (never-mitigating) stiffness reduction and
long-term deformation, or even to an acceleration of long-term deformation, which,
in the worst cases, can lead to the collapse of the construction. Since gravity is the
most persistent action, it is not strange that such constant increases of flexibility
and deformation may manifest as a monotonic, non-asymptotic amplification of

the initial deformed shape due to dead load. Generally speaking, all historical
actions may contribute towards amplifying the initial dead load deformation to a
lesser or greater extent. Owing to this, the progress of structure deformation must
generally be registered, regardless of the type of damage or physical phenomenon
to be characterized. However, the progress of overall deformation of the structure
alone can hardly provide enough information to determine the cause or phenom-
enon generating the progress of damage. Other possible parameters or effects
(such as crack opening) must also be measured simultaneously for identification of
the involved phenomenon to be more conclusive.
6.5.3 Tensile damage in arches and vaults
Tensile damage in arches and vaults will manifest in cracking and deformation. As
is well known, the tensile strength of masonry is almost negligible, meaning that
a certain amount of cracking may easily develop in members subject to tension
effects or eccentric loading. In fact, masonry may initially show significant tensile
strength. The presence of a certain amount of tensile strength may help under-
stand the stability for a limited period of time of some intermediate construction
configurations. However, tensile strength is easily lost in the medium or the long
term due to a variety of effects (settlements, vibrations, deformation cycles, etc.)
causing micro-cracking or cracking in the mortar and the stone and the separation
of both along their interface. Viable ancient buildings were designed in such a
way that equilibrium did not require any tensile strength at their final construction
configuration.
This type of cracking, which is not normally too meaningful, is not necessarily
linked to long-term damaging processes caused by the decay of the material itself.
Severe tensile effects such as severe cracks or large deformation are likely to
be caused by other indirect effects appearing at the buttressing elements or at the
foundation. For instance, differential soil settlements, or the decay of the material
in piers, buttresses or footings, may produce a loss in the capacity of such elements
to properly counter-balance the thrust of arches or vaults, which in turn will expe-
rience cracks and openings between voussoirs associated with plastic hinges
(Fig. 6.6). Monitoring these cracks, however, will be of use for characterizing the
mobility of the structure and the overall progress of damage. Due to the percepti-
ble deformation of the elements and measurable opening of the cracks, this type of
effects can be easily monitored.
6.5.4 Damage of compressed members
The authors have observed frequent vertical cracking in the piers of long-span
buildings. Vertical cracks and related lesions in piers are particularly frequent in
Gothic cathedrals and churches. In many cases, severe cracking, spalling or mate-
rial bursting has appeared in spite of the moderate average compressive stress
caused by gravity load (Fig. 6.7a).
Figure 6.6: Opening between voussoirs in a transverse arch (Tarazona Cathedral).
As is well known, cracks parallel to the direction of applied compression may

appear in materials such as concrete or stone, even for stresses significantly lower
than the compressive strength. In tests performed at the Laboratory of Structural
Technology of the Universitat Politcnica de Catalunya on specimens made of
stacked sandstone blocks with 1 cm mortar joints [8], longitudinal cracks appeared
for an applied stress ranging from 30 to 60% of the compression strength of the
specimen. As expected, specimens made of larger units showed a greater tendency
to crack under moderate compression. The lowest ratios between cracking stresses
and compression strength (30%) were obtained for specimens made of large blocks
(40 20 20 cm), while the largest ratios (60%) were obtained for specimens with
the smallest units (20 10 10 cm, placed horizontally in both cases).
In order to understand the actual existing damage, long-term phenomena leading
to progressive deterioration during historical periods must be accounted for. As
observed apropos of the study of recent collapses [9, 10], the long-term effect of
creep under constant stress may induce significant, cumulative damage in rock-
like materials. As mentioned by Binda et al., damage accumulation (eventually
leading to collapse) may occur for stress values significantly lower than the normal
strength obtained by standard monotonic compression tests. The same authors
found that such phenomena could start at 4050% of normal strength value.
Actions other than dead load may also contribute to long-term damage and couple
synergistically with the effect of creep.
As previously mentioned, the construction process (and the construction tech-
niques used) may induce mid-term or long-term effects. Aspects such as the con-
struction sequence, the duration of the construction, or the use and removal of
scaffoldings and other auxiliary elements, may influence the later behaviour of the
overall building and even cause deferred lesions or other structural disorders.
(a)
(b) (c)
Figure 6.7: Deterioration of piers in Tarazona Cathedral, Spain. (a) Longitudinal

cracks and bursting of compressed material at the springing of diago-
nal arches (b) Artificial thinning of piers. (c) Coupled mechanical and
chemical deterioration.
A specific aspect could be the use of small wooden wedges as a device to keep
stone units in position while the mortar had not yet hardened, which in some cases
(for instance, in Mallorca Cathedral, Spain) produced high stress concentrations
leading to a deterioration of the external face of the stone (Fig. 6.8a) and causing
cracks. Past human actions may be very significant in causing additional damage.
In some cases, large operations undertaken during the life of the construction for
purposes unrelated to the structure may cause significant alterations in the geom-
etry of the piers or other structural members (such as the thinning of piers in
Tarazona Cathedral, Spain, to make space for a wooden choir, Fig. 6.7b); in other
(a) (b)
(c) (d)
Figure 6.8: Effects related to intense compression in the piers of Mallorca

Cathedral. (a) Deterioration due to wooden construction wedges. (b)
Crack caused after the insertion of a steel rod into a joint. (c) Cracks
caused by loss of mortar in the joint. (d) Loose wedge along a corner.
cases, small, apparently inoffensive actions, may reveal to be potentially damaging

after some time (such as the insertion of iron or wooden devices in compressed
members for ornamental or liturgical purposes, Fig. 6.8b). On the other hand, lack of
maintenance or inadequate historical repairs may also contribute to an accelerated
deterioration of the building. Initially minor construction or material defects such
as the loss of a portion of mortar, in Fig. 6.8c may cause cracks to initiate in
zones subjected to significant compression.
The repeated occurrence of extraordinary actions, such as earthquakes or
hurricane-force winds, even if moderate in intensity, also contributes with irreversible,
cumulative effects. Such actions may cause dramatic increases in the eccentricity
of the applied vertical forces at the base of the piers; in turn, this will cause addi-
tional vertical cracking or other effects associated with increased maximum
compression stresses.
Stone or mortar decay due to chemical attack (as observed at the base of the
piers of Tarazona Cathedral, Fig. 6.7c), may couple synergistically with mechani-
cal effects related to sustained compression forces and thus cause accelerated
deterioration.
Cracks in piers or compressed members may be also caused when large blocks
accommodate to deformation or irregularities in mortar joints. In this case, such
cracks are not necessarily related to actual long-term and cumulative damaging
processes. Normally, this type of crack will affect individual blocks and will not
propagate causing long discontinuities along the height of the member. Monitoring
will allow distinguishing between an already stabilized effect, caused by this type
of accommodation, or a more severe, destabilized phenomenon compromising the
safety of the construction.
Conversely, a truly concerning effect is the development of separated wedges in
compressed zones, which in turn produce a reduction of the available resisting
section of the members (Fig. 6.8d). This effect usually starts in less-confined parts,
in the corners of a rectangular or polygonal pier for example.
Measuring variations due to long-term damage in compressed members is
intrinsically difficult, due to the very small magnitude of associated movements or
crack openings. In fact, such movements may be far smaller than those caused by
climatic actions, and thus remain masked. Measuring this type of damage may
require very long monitoring periods (much longer than the 4-year minimum previ-
ously mentioned) and highly sensitive equipment. A sufficiently long monitoring
period may be the only possibility of telling them apart from the entire measure-
ments. Furthermore, certain forms of damage (such as cracks and material losses)
may develop suddenly, with almost no previous indication, making anticipated
symptomatic detection through monitoring very difficult.
6.5.5 Fragmentation
Another common type of damage, not necessarily related to long-term decay, con-
sists of the division (or fragmentation) of the structure into large structural parts
or substructures. Similarly, division owing to large cracks affecting the entire con-
tact between structural members or parts of large members is not uncommon in
historical constructions. The cause for this type of response can be found in dif-
ferent phenomena (soil settlements, thermal contraction or dilation, construction
effects, etc.). In many cases, the safety of the structure is barely affected because
the resulting structural parts are self-stable.
Soil settlements are a very frequent cause for fragmentation. They can induce
the generation of new cracks, enlargement of existing ones or opening of construc-
tion joints. The process may stabilize whenever the divided structure becomes
cinematically compatible with the settlements.
(a) (b)
Figure 6.9: Fragmentation. (a) Separation between the membrane and the trans-
verse arch of a cross-vault in Mallorca Cathedral. (b) Crack developed
from a partial construction joint in Girona Cathedral.
Similar effects may be caused by cyclic winter contractions of the structure. In

this case cracks or separation planes act as a contraction joint. In Mallorca Cathe-
dral, wide discontinuities can be observed between some of the transverse arches
and the vaults of the nave (Fig. 6.9a). Unlike soil settlements, which tend to miti-
gate with time as the soil consolidates, thermal cycles act continuously and may
cause indefinite cumulative damage.
Cracks or separation planes, acting as expansion or settlement joints, may easily
develop starting at weak planes generated during construction itself. In some
ancient buildings, construction joints were finished by interlocking stone units
without filling the joints with mortar (Fig. 6.9b).
6.6 Structural modelling and monitoring

Structural modelling is of extreme interest to enhancing the possibilities and
understanding provided by monitoring. In fact, monitoring fulfils its overall poten-
tial when used in combination with a model of the structural response. Structural
analysis is useful both when designing a monitoring system and when later inter-
preting the information collected.
On the one hand, a prior structural analysis may contribute to better define sig-
nificant aspects of the monitoring system. Simulation using a numerical model can
help lay out adequate and truly informative monitoring, for instance by casting
light on the most significant variables to be measured, the expectable ranges of

variation (which are meaningful for selecting the type of sensors) or the best loca-
tion for the sensors.
On the other hand, and more importantly, a numerical model will help interpret
monitoring results by comparing them with the predictions of a numerical analy-
sis. This comparison may allow, among other aspects, a more quantitative (or
absolute) understanding of the variations measured by sensors. For this purpose, it
is essential, as mentioned in Section 6.4.3, not only to characterize the structural
response but also the main actions that have affected the structure during the period
monitored. Wind (force and direction), temperature, humidity and accelerations
caused by micro-tremors may be included among the effects which have acted on
the structure and generated meaningful measurements. The numerical model can
be used to produce a prediction of the response of the structure when subjected to
one or more of the actions measured. Comparison with the response actually mea-
sured will provide criteria for identifying some of the structural or material proper-
ties (in particular, stiffness of materials or structural members) and for improved
calibration of the model.
In the case of studies addressing long-term phenomena (such as long-term dam-
age), the numerical model needs to have special capabilities in order to allow the
simulation of historical processes. Ideally, the numerical model should be capable
of simulating most of the present or historical actions that have affected the con-
struction; it should also allow sequential analysis in order to simulate the construc-
tion process and the possible structural alterations or repairs that followed. Specific
constitutive equations for long-term creep of masonry or stone-like materials, such
as the one proposed by Papa and Taliercio [9], are of utmost importance for the
purpose here referred. As shown in other chapters of the present book, significant
efforts are presently being made for the experimental characterization and numerical
modelling of long-term creep-induced damage.
6.7 Case studies

6.7.1 Dynamic monitoring of Mallorca Cathedral
The main space of Mallorca Cathedral, built between 1350 and 1601, consists of a
three-nave building, 77 m long and 35 m wide, comprising seven bays with central
and lateral vaults spanning 19.9 m and 8.75 m respectively (Fig. 6.10). The height
of the central vaults at crown reaches 44 m. The central vaults are sustained on
octagonal piers with circumscribed diameters of 1.6 and 1.7 m. The remarkable
slenderness of the piers, rising up 22.7 m to the springings of the lateral vaults,
contributes towards generating the impression of an immense, diaphanous inner
space. The construction has reached our days in a satisfactory state of conserva-
tion. However, some structural disorders can be observed, including (1) significant
deformations affecting the piers; (2) vertical cracks at the base of some of the
piers, sometimes forming surface wedges partially expelled from the core of the
Figure 6.10: Interior of Mallorca Cathedral.
pier (Fig. 6.8d); (3) wide cracks developed throughout the contact lines between
transverse arches and vaults (Fig. 6.9a) and (4) significant deformations affecting
the flying arches.
A detail of the structure, intended to determine the causes and significance of
the mentioned lesions and to provide criteria for its future conservation, is now
being carried out. The study, funded by the Spanish Ministry of Culture, includes
historical research, detailed geometric surveillance, non-destructive testing and
structural analysis. Mallorca Cathedral is also one of the case studies considered
for the EUIndia cooperation project for improving the Seismic Resistance of Cul-
tural Buildings (contract ALA/95/23/2003/077-122) funded by the EC. As part of
the activities envisaged within this last project, a monitoring system aimed at char-
acterizing dynamic response has been also implemented.
The system consists of a 24-bit resolution dynamic acquisitor connected to two
triaxial accelerometers, one of which has been installed on top of a vault of the
central nave. The acquisitors clock is disciplined to GPS time by means of a GPS
antenna (Fig. 6.11). Among the different strategies mentioned in Section 6.4.2,
continuous dynamic measurement has been preferred to allow the capture of low-
intensity oscillations. The system, with a sensitivity of 10-6g, allows continuous
acquisition at 100 sps and 600-plus days of storage. Meaningful seismic episodes
are detected thanks to information provided by the nearest seismological station,
ETO 8 in Mallorca. Information corresponding to any interval measured in GPS
Figure 6.11: Triaxial accelerometer (left) and system with GPS antenna used for
dynamic monitoring (right) placed over the transverse arches.
time can be then easily extracted from the entire data. The system was installed
during April 2005.
Preliminary results, which provide some hints as to the dynamic response of the
structure, are already available. The effect of the Northern Chile earthquake of
13 June 2005, produced measurable effects on the building. Figure 6.12 shows the
variation of accelerations in the time domain and the resulting spectral distribution
for two different time intervals (windows) retrieved, for both the ETO 8 seismic
station in Mallorca and the accelerometer placed above one of the cathedral vaults.
Note that, due to the great distance from the epicentre, the building was mostly
subjected to low-frequency oscillations. In spite of this, the building experienced a
certain excitation and its fundamental vibration mode can be clearly distinguished
in the peak corresponding to 1.68 Hz in the spectral diagrams. Both windows pro-
duced similar results. The acceleration referred is that produced in the longitudinal
direction of the building, which showed the largest amplitude due to the fragmenta-
tion caused by the cracks between transverse arches and vaults (Fig. 6.9a).
The mentioned frequency matches the measure previously obtained by means
of a dynamic test carried out at the level of the vaults, in April 2005, during which
traffic and wind vibrations were measured. In turn, another dynamic test, based on
Nakamuras (or H/V relationship) technique [11] and executed at the ground sur-
face, yielded a soil natural frequency of 2.0 Hz. The corresponding peak can also
be recognized in Fig. 6.12d (right). It must be noted that the building is founded
over strata composed of quaternary sediments to a depth of about 80 m over the
bed rock and that certain concern exists over the possible dynamic amplification
effects that could be caused by such strata.
The analysis of additional information generated in the occasion of future earth-
quakes may provide larger and more accurate information on the dynamic response
of the building. This information will be used, among other purposes, to calibrate
a detailed numerical model of the building.
(a) Northern Chile Earthquake 13/06/2005 ETOS station EW Northern Chile Earthquake 13/06/2005 ETOS station
record (window 1) EW record (window 2)
6.00E-06
5.00E-06 1.50E-06
4.00E-06
Acceleration (m/sec2)
1.00E-06
3.00E-06
2.00E-06 5.00E-07
1.00E-06 0.00E+00
0.00E+00
-1.00E-06 -5.00E-07
-2.00E-06 -1.00E-06
-3.00E-06
-4.00E-06 -1.50E-06
637 687 737 787 1850 1900 1950 2000
Time (sec) Time (sec)
(b) Northern Chile Earthquake 13/06/2005 Mallorca Northern Chile Earthquake 13/06/2005 Mallorca
cathedral station EW (window 1) cathedral station EW (window 2)
2.50E-03 3.00E-03
2.00E-03
2.00E-03
1.50E-03
1.00E-03
1.00E-03
5.00E-04
0.00E+00 0.00E+00
-5.00E-04
-1.00E-03 -1.00E-03
-1.50E-03 -2.00E-03
-2.00E-03
-2.50E-03 -3.00E-03
637 687 737 787 1850 1900 1950 2000
Time (sec) Time (sec)
(c) Fourier Spectra comparative (window 1) Fourier Spectra comparative (window 2)

1.00E-03
1.00E-03
1.00E-04
Fourier Amplitude
Fourier Amplitude
1.00E-04
1.00E-05 1.00E-05
1.00E-06 1.00E-06
1.00E-07 1.00E-07
1.00E-08 1.00E-08
1.00E-02 1.00E-01 1.00E-00 1.00E-01 1.00E 1.00E-02 1.00E-01 1.00E-00 1.00E-01 1.00E-02
Frequency (Hz) Frequency (Hz)
Mallorca cathedral station ETOS station Mallorca cathedral station ETOS station
(d) Transfer Function (window 1) Transfer Function (window 2)

1.00E+08 1.00E+04
1.00E+07
1.00E+06 1.00E+03
Amplitude
Amplitude
1.00E+05
1.00E+04 1.00E+02
1.00E+03
1.00E+02 1.00E+01
1.00E+01
1.00E+00 1.00E+00
1.00E-02 1.00E-01 1.00E+00 1.00E+01 1.00E+02 1.00E-02 1.00E-01 1.00E+00 1.00E+01 1.00E+02
Frequency (Hz) Frequency (Hz)
Figure 6.12: Effect of a long-distance earthquake on Mallorca Cathedral: (a) ac-

celerogram captured in nearby seismic station and (b) at the vaults of
the structure; (c) corresponding spectral diagrams and (d) transference
functions. Left and right: windows corresponding to two different time
intervals.
6.7.2 S. Maria Assunta Cathedral, Reggio Emilia, Italy
The S. Maria Assunta Cathedral in Reggio Emilia, Italy, is one of the most impor-
tant religious buildings of the city. Built on a previous roman construction around
857 AD, it underwent several style remodellings during the thirteenth, fifteenth,
sixteenth and seventeenth centuries. The Cathedral presents a Latin cross plan
with three naves and transept; the crypt is placed below the presbytery. The build-
ing ends in three semicircular apses, the central longer than the two lateral ones
(Figs 6.13 and 6.14).
A static monitoring system was positioned to evaluate a settlement phenomenon
affecting one of the pillars sustaining the massive cupola at the crossing between
the main nave and the transept. Such settlement, bringing visible deformations on
the horizontal structures connected to the pillar (see Fig. 6.15), is a historical
phenomenon, in the sense that it was noticed since long time, but it seemed to
worsen in the last period, so it was decided to control its possible evolution.
Figure 6.13: View of the Cathedrals faade.

Figure 6.14: Plan, longitudinal and transverse cross-section.
Figure 6.15: Slope in the access steps to the aisle apse due to the sinking of the
pillar, indication of the affected pillar.
The acquisition system is composed of 10 long base cable extensometers,

13 electric extensometers, 2 multi-base extensometers equipped with 3 measure-
ment bases each, 2 measurement panels equipped with switch and thermometer
(Fig. 6.16). The acquisition is semi-automatic, meaning that the system is able to
store the data in the panels memory, and at determined time intervals it is neces-
sary to recover the data on-site.
Figure 6.16: View of a long-base extensometer, of an electric extensometer placed

across a fissure and of the extensometers positioning plan, at the
crypts level.
The system was implemented in a way to have redundant information, from

the measurement obtained at specific reference points. In fact, the settlement of
the pillar brings deformations also on the surrounding structural elements and
determines visible crack patterns, evidently related to the observed phenomenon.
Aiming at defining further settlements of the pillar, the different types of exten-
someters positioned are able to detect the relative displacements of the vertical
structures, to measure variations in the openings on the main fissures, to evaluate
the settlements of the foundations of the crypts pillars with respect to fixed
points of the underlying soil, at the depths of 5, 10 and 15 m. The temperature,
fundamental parameter for the comprehension of seasonal non-monotone behav-
iours, is also monitored.
A year and a half have passed since the system has been active, hence at least
the seasonal effects can be appreciated. Figure 6.17 shows the crack mouths rela-
tive displacements, plotted versus the observation date. From the visualized data,
30.00
0.20
28.00
0.10
26.00
EL1
EL2
RELATIVE DISPLACEMENT (mm)
0.00 24.00 EL3

EL4
.00
,30
,40
.10
.50
.40
.30
.38
.40
,00
,00
.30
.05
.00
.30
,30
TEMPERATURE (C)
16
10
15
15
10
15
15
09
16
15
10
12
09
10
15
11
22.00 EL5
/04
/04
/04
/04
/04
/04
/04
/04
/04
/04
/05
/05
/05
/05
/05
/05
-0.10
/03
/04
/05
/06
/07
/08
/09
/10
/11
/12
/03
/04
/05
/06
/07
/01
EL6
10
09
12
09
07
06
07
05
09
13
08
12
17
22
29
19
EL7
20.00
-0.20 EL8
EL9
18.00 EL10
-0.30 EL11
16.00 EL12
EL13
-0.40 TEMPERATURE
14.00
-0.50
12.00
-0.60 10.00
DATE
Figure 6.17: Crack mouth opening, extensometers EL1 . . . EL13.
it is possible to notice that the seasonal effect is fully visible (contraction when the
temperature is higher), and that the relative displacements are contained within
narrow variations (the tenth part of a millimetre/half mm). There still are some
sensors that indicate some trends that are not directly related to the temperature;
longer observation periods are needed in order to define if such singular relative
displacement tendencies are of importance for the studied phenomenon.
6.7.3 Vitoria Cathedral
The Cathedral of Santa Maria, also known as the Old Cathedral or Vitorias Cathe-
dral, is located on the highest part of the city of Vitoria. The construction of this
interesting monument of Gothic Style began in the thirteenth century and finished
in the fifteenth. The building, declared Historic-Artistic Monument, has the form
of a pronounced Latin cross with three naves in the main hall space and a head wall
with ambulatory and radial chapels (Fig. 6.18). Throughout its life, many interven-
tions have taken place in the monument in order to solve different structural prob-
lems. The first intervention happened just a few years after the construction was
finished. The last intervention stands among the major ones carried out through
the ages and took place in 1967. Twenty-six years after that, the temple had to be
closed to the general public owing to the appearance of worrying signs of ancient
structural problems reactivating.
Figure 6.18: Interior of Vitoria Cathedral.
At that time, people in charge of the conservation of the Cathedral opted deci-
sively for embarking on a long study towards the integral understanding of the
building before starting new structural interventions. They conceived the Cathe-
dral Refurbishment Programme as a living project open to the public. The restora-
tion of Vitoria Cathedral has deserved the Europa Nostra 2002 Prize, the highest
award given by the European Union for the conservation and enhancement of the
citys cultural heritage.
The structural studies have demonstrated that the important historical deforma-
tions of the cathedral are due to insufficient capacity of the piers to adequately
resist the thrust of the vaults. The original structure had no flying arches and all the
buttressing capacity depended on the piers.
The monitoring of the Cathedral is one of the most important activities in the
process of the study and restoration of the structure. In the beginning it provided
knowledge on the structural behaviour and its stability, and nowadays it allows
detecting states of risk during the restoration works. Likewise, monitoring is used
to gauge the repercussion of the consolidation actions on the whole structure.
The monitoring began in November 1992 and, at the moment, is intended to

continue indefinitely. Since 1998, the Labein Technological Centre is responsible
for it. In order to improve monitoring, the equipment installed was complemented
in 1999 with new sensors.
The monitoring equipment consists of temperature and relative humidity sen-
sors (5), crackmeters (16), clinometers (6), extensometers (load cells and chains,
10) and convergence meters (9). Sensors are located in strategic points to detect
the largest and more significant movements. Readings of all sensors are taken
automatically every twelve hours and once in a week a computer collects all
(a)
40,00
35,00
30,00
TEMPERATURE C
25,00
20,00
15,00
10,00
5,00
0,00
-5,00
-10,00
-15,00
16/03/00
08/05/00
01/07/00
23/08/00
15/10/00
07/12/00
30/01/01
24/03/01
16/05/01
08/07/01
31/08/01
23/10/01
15/12/01
06/02/02
01/04/02
24/05/02
16/07/02
07/09/02
31/10/02
23/12/02
14/02/03
08/04/03
01/06/03
24/07/03
15/09/03
07/11/03
31/12/03
22/02/04
15/04/04
07/06/04
31/07/04
22/09/04
14/11/04
06/01/05
01/03/05
23/04/05
15/06/05
07/08/05
30/09/05
22/11/05
DATE
T1 T2 T3 T4
(b)
1178,000
936,000
694,000
452,000
LOAD (Kg.)
210,000
-32,000
-274,000
-516,000
-758,000
-1000,000
16/03/00
08/05/00
01/07/00
23/08/00
15/10/00
07/12/00
30/01/01
24/03/01
16/05/01
08/07/01
31/08/01
23/10/01
15/12/01
06/02/02
01/04/02
24/05/02
16/07/02
07/09/02
31/10/02
23/12/02
14/02/03
08/04/03
01/06/03
24/07/03
15/09/03
07/11/03
31/12/03
22/02/04
15/04/04
07/06/04
31/07/04
22/09/04
14/11/04
07/01/05
01/03/05
23/04/05
15/06/05
08/08/05
30/09/05
22/11/05
DATE
PE 1 PE 2 PE 3 PE 4
Figure 6.19: (a) Fluctuation of temperature and (b) movements registered in exten-
someters from 16 March 2000.
readings by telephone from the headquarters of Labein. During special periods expe-
riencing important operations, such as movements of the strengthening structures in
the main nave of the temple, readings of all the sensors and the interpretation are
done daily.
The control carried out to date has given very useful information about the evo-
lution of the movements and damage manifested in the structure. The active move-
ments have been perfectly identified, and it is well known that these movements
are cyclical in time and related with the changes of temperature and humidity
(Fig. 6.19). The main active movement detected is the opening of the central nave
and the transept caused by the expansion of the materials of the vaults as a result
of the rise in temperature.
The evolution of movements obtained at the same time of two consecutive years
shows that the structure seems to be now stable. However, certain changes mea-
sured during three years since 2002 indicate that the monitoring system is detecting
the effect of the present use of the Cathedral and associated infrastructures, as well
as the consolidation works that are being carried out.
6.8 Conclusions
In the framework of a global study of an ancient structure, monitoring contrib-
utes with quantitative information related to the current response of the structure
subjected to a combination of ordinary actions (dead load, thermal environmental
actions, micro-tremors, wind, traffic, etc.). This information may provide significant
evidence as to the type and progression of damage affecting the structure. However,
characterizing long-term damage by means of monitoring is a challenging task that
requires specific strategies involving lengthy observation periods.
Most damage manifestations including large deformation, cracking under
tension or compression and crushing or bursting under compression are caused
by the combined effect of a variety of physical processes, comprising the long-
term creep of the materials, cyclic environmental actions or repeated extraordi-
nary actions. They are also influenced by historical facts related to construction,
utilization and maintenance. The measurements obtained (crack openings, dis-
placements, accelerations, etc.) will include all these effects bundled into a single
response; interpreting the results requires the mixed responses to be broken down
into cyclic and reversible processes and monotonic (or cumulative) ones. A fur-
ther step consists of breaking measurements into the components associated with
different effects acting on the structure (wind, temperature, earthquakes, traffic,
etc.). The latter specifically requires monitoring of the actions themselves by
means of appropriate equipment (thermometers, hygrometers, anemometers,
seismometers, etc.).
Accurate numerical simulations of structural response, by means of detailed
structural models, are also needed, both for the results to be interpreted and for
reliable conclusions to be reached regarding the condition of the building and the
actions or processes contributing to its decay or damage.
Acknowledgements
The assistance of the Chapter of Mallorca Cathedral, Studio Mauro Severi
Architetto, Studio Associato di Ingegneria Gasparini, Reggio Emilia, Fundacin
Catedral de Sta. Mara de Vitoria and the EC project ALA/95/ 23/2003/077-122 in
the development of the studies presented is gratefully acknowledged.
References
[1] Rossi, P.P., The importance of monitoring for structural analysis of monu-
mental buildings. Inspection and Monitoring of the Architectural Heritage,
Ferrari Editrice: Bergamo, pp. 253265, 1997.
[2] Rossi, P.P. & Rossi, C., Surveillance and monitoring of ancient structures.
Recent developments. Structural Analysis of Historical Constructions II.
International Center for Numerical Methods in Engineering: Barcelona,
pp. 163178, 1998.
[3] Rossi, C. & Rossi, P.P., A low cost procedure for quick monitoring of monu-
ments and buildings. On site Control and Evaluation of Masonry Structures,
eds. L. Binda & R.C. de Vekey, RILEM publications S.A.R.L: Bagneux,
pp. 8194, 2001.
[4] Modena, C., Advanced monitoring systems. On site Control and Evaluation
of Masonry Structures, eds. L. Binda & R.C. de Vekey, RILEM publications
S.A.R.L.: Bagneux, pp. 95104, 2001.
[5] Binda, L., Falco, M., Poggi, C., Zasso, A., Mirabella Roberti, G., Corradi, R.
& Tongini Folli, R., Static and dynamic studies of the Torazzo in Cremona
(Italy), the highest masonry bell tower in Europe. Bridging Large Spans from
Antiquity to Present, Sanayi-i-Nefise Vakfi: Istanbul, pp. 100110, 2000.
[6] akmak, A.S., Natsis, M.N. & Mullen, C.L., Foundation effect on the
dynamics of Hagia Sophia. Structural Studies of Historical Buildings IV,
eds. C.A. Brebbia & B. Leftheris, Computational Mechanics Publications:
Boston and Southampton, pp. 310, 1995.
[7] Massanas, M., Roca, P., Cervera, M. & Arun, G., Structural analysis of
Kk Ayasofya Mosque in stanbul. Structural Analysis of Historical
Constructions IV, Balkema: Amsterdam, pp. 679686, 2004.
[8] Hernndez, P., Ensayo en laboratorio de prismas de obra de fbrica en
piedra. Universitat Politcnica de Catalunya (UPC): Barcelona, 1998.
[9] Papa, P. & Taliercio A., Prediction of the evolution of damage in ancient
masonry towers. Bridging Large Spans from Antiquity to Present, Sanayi-i-
Nefise Vakfi: Istanbul, pp. 135144, 2000.
[10] Binda, L., Saisi, A., Messina, S. & Tringali, S., Mechanical damage due to
long term behaviour of multiple leaf pillars in Sicilian Churches. Historical
Constructions, University of Minho: Guimaraes, pp. 707718, 2001.
[11] Nakamura, Y., A method for dynamic characteristics estimation of subsur-
face using microtremor on the ground surface. Report of Railway Technical
Research Institute (RTRI), 30(1), pp. 2533, 1989.
CHAPTER 7
Modelling of the long-term behaviour of

historical masonry towers
A. Taliercio & E. Papa

Milan, Italy.
7.1 Introduction
In March 1989 the Civic Tower in Pavia, a Middle-Ages masonry building nearly
58 m high, suddenly collapsed, killing four people and destroying part of the
adjacent Cathedral, along with some of the buildings overlooking the surround-
ing square [1]. No special event, to which the collapse might be attributed, was
monitored in the preceding months. Also, no evidence of significant soil settle-
ments was found. The results of a number of accelerated creep tests performed
on samples extracted from the debris, along with finite element (FE) analyses of
the tower, indicated that a possible origin for the collapse was the cumulation
of creep-induced damage in time (see also Section 1.2). Indeed, under sustained
loading some of the specimens failed at stress levels (see Section 2.2) comparable
to the maximum values yielded by the numerical analyses. This remark was the
starting point for an extensive research programme aiming at (1) characterizing
the creep behaviour of ancient masonry; (2) developing theoretical models suit-
able to the description of creep evolution and creep-induced damage; and, finally,
(3) assessing the safety of ancient masonry buildings subjected to heavy persistent
loads through nonlinear structural analyses.
In this chapter, a theoretical model developed by the authors is first described (Sec-
tion 7.2); the procedure employed to identify the model parameters from results of
accelerated creep tests is also outlined in Section 7.2.3. Then, the numerical results of
structural analyses of two masonry towers are presented (Section 7.3), with particular
emphasis on the description of the damage evolution in time, up to the predicted creep
time to failure for the building. Finally, the obtained results are critically reviewed, and
future improvements for the theoretical model are pointed out (Section 7.4).
7.2 A continuum damage model for masonry creep

In this section, a model that is capable of accounting for the material damage
under increasing stresses, the damage-induced creep acceleration, and the
damage-induced anisotropy is presented. The model is addressed to brittle
materials, such as concrete or rubble masonry, which are assumed to be isotropic
at their virgin (undamaged) state. More details on the model can be found in
[2]. Throughout the section, compressive stresses and strains are assumed to be
positive.
7.2.1 Unidimensional viscoelastic model with damage
Starting point for the development of the proposed model is the so-called Burgers
rheological model, which consists of a Kelvin element in series with a Maxwell
element (Fig. 7.1). Both elements are composed by a spring and a dashpot: in the
first one, the two components are in parallel, whereas the Maxwell element has the
two components in series.
The spring constants are denoted by EK, EM and the relaxation times of the
dashpots by tK, tM; the superscripts K and M stand for Kelvin and Maxwell,
respectively. The spring of the Maxwell element accounts for the instantaneous
(elastic) response of the material; the Kelvin element describes primary creep,
whereas the Maxwell dashpot simulates the steady-state (secondary) creep. A
frictional (or Bingham) element is inserted between the spring and the dashpot of
the Maxwell element to prevent the activation of secondary creep at low stress
levels (say, lower than s0 in uniaxial conditions).
The classical Burgers model cannot describe the decay in stiffness experienced
by a material element subjected to stresses increasing beyond the elasticity limit,
nor allow for the tertiary stage that precedes creep failure. These effects can be
reproduced by accommodating some damage variables in the model (see similar
proposals for concrete by Bazant and Chern [3], or Cervera et al. [4]). Figure 7.1 shows
primary creep
secondary and tertiary creep
(Kelvin) EK
(Maxwell-Bingham)
s0
E M (1D) t M (1D)
elastic behaviour
tK (with damage)
Figure 7.1: Modified Burgers rheological model.

LONG-TERM BEHAVIOUR OF HISTORICAL MASONRY TOWERS 155
the rheological model, as modified by the authors. The damage variable, D, reduces
the elastic modulus of the Maxwell spring, thus accounting for the decrease in
stiffness induced by an increase in stress. D affects also the relaxation time of the
Maxwell dashpot, making it possible to capture the tertiary creep phase and creep
failure. The value of the damage variable ranges between 0 (virgin state) and 1
(complete failure); this variable is supposed to start increasing as a suitable thresh-
old for the elastic strain is attained. Primary creep is supposed to be unaffected by
damage.
Assuming infinitesimal deformations, the total strain e(t) of the Burger model is
the sum of the strains in the Maxwell and Kelvin elements. The stressstrain (rate)
law for the Kelvin element reads
K K K K
s = E e + h e& . (7.1a)
The strain rate in the Maxwell element due to any stress increment can be conve-
niently expressed as the sum of a delayed term:
M s
e& = M 2
H (s s0 ) (7.1b)
h (1 D )
and an instantaneous term, which is the rate of the elastic strain:
el s
e = M
. (7.1c)
E (1 D )
In eqns (7.1ac), ha = taEa, a = M or K, and H is the Heaviside function.

The form of the damage evolution law has now to be specified. It will be assumed
that the variable driving the damage process is a damage force, Y, which, accord-
ing to thermodynamical rationales, is given by
1 M el 2
Y = E (e )
2
(see e.g. Lemaitre and Chaboche [5]). . .

. VThe. sdamage rate will be assumed to be the sum of two contributions: D = D +
s
D . D dominates under increasing stress. It can be obtained by differentiating the

following expression, which conforms to a proposal of La Borderie et al. [6] for
concrete:
S 1
D = 1 , (7.2)
M BH
1 + AH (Y Y0 H ) / E
where Y0H is a threshold value for the damage force, AH and BH are material parame-
ters, and are Macauley brackets. Different parameters define the damage evolution
laws, depending on whether strains are compressive (H = C) or tensile (H = T).
.
DV dominates under sustained stresses: the following expression, borrowed from
rock mechanics (see Chan et al. [7]), is proposed for this term:
D& = a1H (Y / E ) 2 H D ( log D) 3 H ,

V M a a
(7.3)
where aiH (i = 1, 2, 3; H = C or T) are additional material parameters.

. . .
Eventually, it is possible to integrate the strain rate e = eK + eM to obtain the
damaged creep law at constant stress s, with the boundary condition e(0) =
s/[E (1 D(0))].
M
A detailed parametric study of the model is presented in [8]. For the sake of
illustration, here only some creep curves computed according to the proposed
model are shown in Fig. 7.2 for a given set of material parameters, at different
values s values. Provided that creep failure is reached, these curves exhibit the
classical S-shape corresponding to the sequence of primary, secondary and tertiary
creep. Also note that the time to failure decreases as the stress intensity increases,
in agreement with experiments.
Under any stress history, a numerical procedure has usually to be employed to
integrate the strain response of any material element. If the element undergoes a
stress increment si = si si1 during any (small) time interval ti = ti ti1, the
relevant strain increment ei can be approximately computed assuming the stress to
vary linearly during the time step and no increase in damage to occur. This gives:
si in
ei = + ei . (7.4)
Ei
10 7 MPa
8
6 MPa
transversal strain (*1000) axial strain
4 4 MPa
2 3 MPa
0
3 MPa
-2
-4
-6
7 MPa 6 MPa
-8
4 MPa
-10
0 200000 400000 600000
time (sec)
Figure 7.2: Simulation of uniaxial creep tests at different stresses.

The first term on the RHS of eqn (7.4) is the (damaged, elastic) strain increment
due to the stress increment, where
1 1 H (si s0 ) ti 1 tK ti
= 1+ M + K 1 t 1 exp t K .
E (1 Di 1 ) 2 t (1 Di 1 ) E
M
Ei i
(7.5)
The second term is the (inelastic) strain increment due to the stress acting on the
material element at the beginning of the step, and is given by
in si 1 H (si s0 ) ti ti K
ei = + 1 exp
t K i 1
M 2
e , (7.6)
h (1 Di 1 )
k is given by the recursive formula
where ei1
si 1t K
t t K
1 exp Ki 1 + exp Ki 1 ei 2 .
K
ei 1 = (7.7)
ti 1 E
K
t t
This explicit forward Euler integration scheme proposed here is not unconditionally
stable, so that sufficiently small time steps have to be taken. This condition also ensures
the reliability of the assumption that no increase in damage occurs during ti.
7.2.2 Three-dimensional viscoelastic model with damage
When the proposed constitutive model is extended to 3D solids, it is assumed that

the damage variable, D, is a second-order symmetric tensor; its eigenvalues will be
denoted by Da (a = I, II, III). Each principal direction of this tensor (xa) is assumed
to be aligned with the normal (na) to a plane microcrack that forms and grows at any
point in the solid. Because of the choice made for the damage variable these direc-
tions are mutually perpendicular, so that the damage-induced local material anisot-
ropy is, in the more general case, orthotropy. This is consistent with the assertion
made by Kachanov [9], stating that even for high crack densities with interacting
cracks, the effective elastic properties remain orthotropic with good accuracy.
The damage process driving variable in 3D is supposed to be a non-dimensional
damage force, y = eeleel, which is basically an equivalent strain measure. This is
a quite common assumption in damage models, where equivalent stresses [7] or
equivalent strains [10] are often employed as damage indicators. The first damage
direction, xI, is supposed to activate as the maximum eigenvalue of y exceeds a
threshold value in tension (y0T) or compression (y0C), according to the sign of the
relevant principal strain. From now onwards, the material element is no longer
isotropic: this means that, in general, stress tensor and strain tensor are not collinear
in the continuation of the load history. An additional damage direction, xII, can acti-
vate in the plane orthogonal to xI as the maximum direct component of the damage
force tensor, i.e. yaa = na(yna), attains either one of the threshold values in tension
or compression. The third possible damage direction, xIII, is then automatically iden-
tified, having to be orthogonal to both xI and xII. Note that at any point in the solid
the activated damage directions are fixed throughout the subsequent load history, so
that the proposed model can be classified as a non-rotating, smeared crack model.
As for the damage evolution law, here it was chosen to employ eqns (7.2) and
(7.3) for each one of the principal values of the damage tensor, which is supposed
to be related to yaa (a = I, II, III).
The time integration scheme presented in the preceding paragraph will be now
generalized to the 3D case. In view of the implementation of this scheme into a FE
code, from now onwards a matrix notation will be employed. The six independent
stress and strain components in any Cartesian reference frame (x1, x2, x3) will be
gathered into two arrays:
s = {s11, s22, s33, s12, s23, s31}T; e = {e11, e22, e33, 2e12, 2e23, 2e31}T,
where the classical Voigts notation for strains was adopted.
Let any stress increment si be applied to a material element during any time inter-
val ti. By generalizing eqn (7.4), the relevant strain increment can be expressed as:
ei = Ci si + eiin, (7.8)
where Ci is the compliance matrix at time ti1. Similar to eqn (7.5), this matrix can
be expressed as the sum of three matrices:
Ci = Ciel + CiM + CiK. (7.9)
Owing to the orthotropic nature of the damaged material, it is expedient to give
the expressions of the first two matrices in the reference frame of the principal
directions of damage (xI, xII, xIII):
1 n n
0 0 0
y yI,II yI,III
I,I
n
0
1
0 0
yII,II yII,III

1
0 0 0
el 1 yIII,III
C i = M ,
E 2(1 + n )
0 0
yI,II
2(1 + n )
symm. 0
yII,III
2(1 + n )

yI,III
t =t
i 1
(7.10)
1 n n
0 0 0
y2 2
yI,II
2
yI,III
I,I
n
0
1
0 0
2
yII,II
2
yII,III

1
0 0 0
M t 2
yIII,III
C i = i
M ,
2h 2(1 + n )
0 0

2
yI,II
2(1 + n )
symm. 2
0
yII,III
2(1 + n )
2
yI,III
t =t
i 1
(7.11)
with yjk = [(1Dj)(1Dk)]1/2, j, k = I, II, III. These matrices can be rotated in any
^
Cartesian reference frame using the classical transformation rule Ci = T Ci TT,
where the transformation matrix T depends on the direction cosines of the angles
defining the orientation of (x1, x2, x3) to (xI, xII, xIII) (see e.g. [11]).
Note that damage affects the off-diagonal terms in the compliance matrices in
eqns (7.10) and (7.11) through geometric averages of the relevant components:
this is in accordance with proposals found in the literature for ductile [12] and
brittle materials [13, 14]. The Heaviside function that affects the term ti /tM in
eqn (7.5) was omitted for brevity; it will be implicitly assumed that secondary
creep activates along any direction when a certain threshold (in terms of damage
forces) is exceeded.
The third matrix in eqn (7.9) is unaffected by damage and can be expressed, in
any reference frame, as CiK = CKs (1li), where CKs is the flexibility matrix for
an isotropic material with Youngs modulus EK and Poissons ratio n and
t
K
t
li = 1 exp Ki . (7.12)
ti t
Finally, the inelastic strain increment eiin in eqn (7.8) can be obtained by
generalizing eqn (7.6):
t
ei
in
= 1 exp Ki e iK1 +2 CiM s i 1 ,
t (7.13)
where
K Ks ti 1 K
e i 1 = li 1C s i 1 + exp e .
t K i 2
(7.14)
7.2.3 Identification of the model parameters and comparisons

with experimental results
In this section the procedure to identify the model parameters is briefly outlined,
referring to tests on masonry samples, subjected to uniaxial compressive stresses
either monotonic or constant in time.
On the whole, 17 constitutive parameters are required to characterize the mechan-
ical response of masonry according to the proposed model. Two parameters (EM,n)
are associated with the elastic behaviour of the virgin (undamaged) material, which
is supposed to be isotropic; three more (EK, tK and tM) define the viscoelastic behav-
iour in the absence of damage. The damage effects that characterize the material
behaviour under monotonic or sustained stress are separately taken into account by
introducing six different parameters for each kind of damage contribution (namely,
AH, BH, y0H for the static damage contribution, eqn (7.2); a1H, a2H, a3H for the viscous
damage contribution, eqn (7.3)). The threshold stress s0 above which secondary
creep activates is supposed to correspond to the damage threshold y0H.
The model parameters have to be identified both in tension (H = T) and in com-
pression (H = C). As direct tension tests are not easy to perform, the model param-
eters defining the tensile behaviour of the material can be approximately estimated
according to the evolution of the transversal strains during simple compression
tests. In the continuation, attention is mainly focused on the identification of the
parameters defining the damage evolution in compression.
The elastic modulus EM can be evaluated by best fitting the first sensibly linear
part of the axial stressstrain plots relative to monotonic tests (usually between 30
and 60% of the peak stress; segment 01 in Fig. 7.3) and by averaging the com-
puted experimental slopes. The elasticity limit furnishes the threshold values for
the (normalized) damage forces (y0C). The model parameters that define the static
damage evolution law in compression (AC, BC) can be derived by best fitting the
nonlinear part of the stressstrain plot in uniaxial compression (123 in Fig. 7.3).
In particular, AC mainly affects the strength value, whereas BC is related to the
shape of the softening branch [6].
The parameters associated with the viscoelastic behaviour of the material (EK,
K
t ) are determined according to the primary creep phase in creep tests at relatively
low stresses, so that the strain evolution in time is unaffected by damage.
Starting from the time evolution of the axial strain at constant stress above the
elasticity limit, s0, it is possible to estimate the relaxation time tM of the Maxwell
dashpot of the rheological model in Fig. 7.1, which is related to the secondary (or
steady-state) creep rate, dein/dt (phase 12 in Fig. 7.4).
The parameters defining the viscous damage law, eqn (7.3), could be evaluated
by computing the decrease in elastic stiffness during the reloading phases between
50 5
2 2
40 4 3
1 1
strain (*1000)
stress (MPa) 2
30 3
3 3
1
20 2
10 1
0 0
0
0 0
-15 -10 -5 0 5 10 0 50000 100000 150000 200000 250000
transversal strain (*1000) axial strain time (sec)
Figure 7.3: Stressstrain plot Figure 7.4: Axial strain versus time
relative to a monotonic test on a for a uniaxial creep test on a concrete
concrete specimen. specimen.
two subsequent steps at constant stress. The relevant increase in damage, divided
by the time spent between the two reloading phases, gives an estimate of the
average damage rate, which can be used to identify a1C, a2C and a3C. If unloading
reloading cycles are not performed, a1C, a2C, a3C can be derived by best fitting the
tertiary phase of the creep curves (23 in Fig. 7.4) for tests performed at higher
stress levels. In particular, the creep time to failure of the material at different
stress levels defines the values of a1C and a3C; a2C controls the rate of the viscous
damage (see also [8]).
As true creep tests at constant stress are extremely time-demanding, it may be
expedient to identify the model parameters according to the results of pseudo- (or
accelerated) creep tests, of the type described in Chapter 2. Figures 7.5 and 7.6 show
the results of two tests on rubble masonry samples, taken from the ruins of the Civic
Tower of Pavia, together with their numerical simulation according to the proposed
model; the viscous parameters were basically identified by best fitting the axial strain
creep curve. The experimental plot of Fig. 7.5 was obtained testing a 200 mm 200
mm 350 mm prism, coming from the external layers of the masonry constituting
the tower (Section 2.2.6), whereas the data in Fig. 7.6 refer to a test carried out on a
300 mm 300 mm 510 mm prism taken from the internal part of the walls (Section
2.2.4). Tests were performed in controlled thermo-hygrometric conditions (20C
and 50% RH) using a hydraulic machine capable of applying a maximum constant
load of 1000 kN. Further details on the characteristics of the tested sample and the
experimental procedure are reported in Section 2.2.
In the test to which Fig. 7.5 refers, the compressive stress applied to the specimen
was progressively increased at a rate of 2.5 103 MPa/s; at the end of each step
of 0.25 MPa, the load was stopped and kept constant for about three hours. On the
contrary, Fig. 7.6 refers to a test characterized by phases at constant stress of a
transversal strain (*1000) axial strain (*1000)
transversal strain (*1000) axial strain (*1000)

4
experimental test experimental test
6 numerical simulation numerical simulation
2
3
0 0
-3
-2
-6
-9 -4
0 50000 100000 150000 200000 250000 0 200 400 600 800 1000
time (sec) time (days)
Figure 7.5: Straintime plots for a Figure 7.6: Straintime plots for a
uniaxial test on a rubble masonry uniaxial test on a rubble masonry
sample with load steps of a duration sample with load steps of variable
of 3 hours each; experimental data and duration; experimental data and
numerical simulation. numerical simulation.
duration that ranges between 15 and 200 days. Accordingly, the global duration of
the former test is equal to less than three days, whereas the duration of the latter
one is of about three years.
In Table 7.1 the values of the model parameters used in the numerical simulations
are reported. It can be noted that the two tested materials are similar as far as the
damage evolution laws (both static and viscous) are concerned. The two materials
essentially differ in terms of creep behaviour: indeed, the relaxation times (tK and tM)
employed in the simulation of the two tests differ of some orders of magnitude.
Upon the whole, the agreement between test data and theoretical predictions
shown in the two figures is satisfactory. The main difference lies in the prediction
of the transversal strains when failure is approached, which are underestimated by
the model. According to Herrmann and Kestin [15], this problem could be circum-
vented by introducing additional inelastic (permanent) strains related to the dam-
age variables: this was successfully done, for example, by Papa and Taliercio [16],
with reference to uniaxial creep tests in a previously developed model; a similar
extension will be considered in future developments of the present research.
The model was also used to simulate creep tests performed on masonry samples
taken from the crypt of Monza Cathedral (see Section 2.3). All the (prismatic)
specimens were subjected to uniaxial compression applied by subsequent steps of
0.25 MPa, at a rate of 1.25 103 MPa/s. The stress is then kept constant for one
hour and half, before the following stress increment is applied. Table 7.2 reports
the specimen sizes and the main results of three of the tests performed (see also
Table 2.9). Readers are referred to Section 2.3 for a detailed description of the
specimens and the experimental results.
Table 7.1: Tests on samples taken from Pavia Civic Tower: model parameters used
in the numerical simulations shown in Figs 7.5 and 7.6.
Model parameter Test of Fig. 7.5 Test of Fig. 7.6
EM (MPa) 3500
tM 690000 (s) 1228 (days)
EK (MPa) 30000 3400
tK 7500 (s) 5 (days)
n 0.2 0.25
AC 14.3 106
BC 1.2
AT 47.4 106
BT 1.05
y0C 5.7 107
y0T 5.7 109
a1C 2.2 (s1) 2.2 (days1)
a2C 0.45
a3C 1.45
a1T 7.7 (s1) 7.7 (days1)
a2T 0.35
a3T 1.13
Table 7.2: Samples taken from the crypt of Monza Cathedral: main experimental
results.
Failure stress Time to Failure strain
Test Size (mm) (MPa) failure (s) (m/mm)
CDM I p4 200 200 314 4.25 55736 8.87
CDM II p4 200 200 314 2.75 21974 7.57
CDM II p11 200 200 316 5.30 109772 4.98
Some of the model parameters (EM, EK, n, y0C, y0T, tM) were identified according to
the previously outlined procedure. The remaining parameters were initially given
the same values as the samples extracted from the ruins of Pavia Civic Tower. In
this way, the axial strain versus time plot identified by triangles in Fig. 7.7 was
obtained.
To increase the predicted time to failure, three numerical tests were made, either
by decreasing AC (from 800 (EM)1.2 to 380 (EM)1.2), or by increasing a2C (from
0.45 to 0.55); in the latter case, it was found that better results are obtained if, at
the same time, a2T is reduced (from 0.35 to 0.25). The results of these tests are
shown in Fig. 7.7 (plots identified by circles and diamonds). Decreasing AC slack-
ens the evolution of damage in accelerated creep tests between two subsequent
load steps. Changing a2C and a2T makes the transversal viscous damage variable to
6 experimental tests
AC=380
axial strain (*1000)
a2C=0.55a2T=0.25
AC=800
4
a2C=0.55a2T=0.25
AC=380
a2C=0.45a2T=0.35
2
0
0 20000 40000 60000 80000 100000
time (sec)
Figure 7.7: Comparison between the results of accelerated creep tests on samples
taken from the crypt of the Monza Cathedral and the predictions of the
theoretical model.
evolve more quickly than the axial one, so that the phenomenon of dilatancy when
approaching failure can be captured (which is not the case with the original choice
for the parameter values).
For further analyses, it was decided to take AC = 380 (EM)1.2 = 56.3 106, a2C = 0.55,
a2T = 0.25: these values approximately give the same time to failure as the experi-
mental test of intermediate duration. The final set of identified values is reported
in the third column of Table 7.3 and compared with the homologous set of values
for Pavia Tower (second column).
In Fig. 7.8, the predicted axial, transversal and volumetric strains are plotted ver-
sus time for a material element subjected to a sustained compressive stress of 1.23
MPa (which is the average vertical stress computed at the base of the bell-tower of
Monza Cathedral). Note that the volumetric strain turns from positive (compaction)
to negative (dilatancy) when creep failure is approached: this is a feature peculiar to
the failure of brittle materials, which can be captured thanks to the anisotropic nature
of the damage model employed, allowing for a faster increase in transversal strain
than in axial strain as the creep-induced damage process evolves.
Finally, a problem that rose in the structural analyses presented in the follow-
ing Section had to be tackled. Using the value for tM identified according to
accelerated creep tests led to the prediction of unrealistically short times to fail-
ure of the analysed structures. Thus, in the applications presented hereafter, it
was decided to set tM equal to a very high value (1000 years), so as to make the
estimate for the time to failure of the structures unaffected by the choice made
for the value of this parameter. In Fig. 7.9 the numerical results obtained with
Table 7.3: Comparison between the values of the model parameters for specimens
taken from Pavia Tower and Monza Cathedral.
Value
Model
parameter Pavia Monza
EM (MPa) 3500 2800
tM (s) 690000 110000
EK (MPa) 30000 20000
tK (s) 7500
n 0.2 0.15
AC 14.3 106 56.3 106
BC 1.2 1.5
AT 47.4 106 37.5 106
BT 1.05
y0C 5.7 107 3.6 109
y0T 5.7 109 3.6 1011
a1C (s1) 2.2 8
a2C 0.45 0.55
a3C 1.45 1.13
a1T (s1) 7.7
a2T 0.35 0.25
a3T 1.13
4 vertical strain
horizontal strain
3
volumetric strain
2
1
strain (*1000)
-1
-2
-3
-4
0 500 1000 1500 2000 2500
time (years)
Figure 7.8: Vertical, horizontal and volumetric strain versus time for a material
element subjected to a uniaxial compressive stress of 1.23 MPa.
0
8
transversal strain (*1000)

experimental tests
6 -2
axial strain (*1000)
numerical simulation
(110000 sec)
numerical simulation -4
4
(1000 years)
experimental tests
2 -6 (110000 sec)
(1000 years)
0 -8
0 20000 40000 60000 80000 100000 0 20000 40000 60000 80000 100000
time (sec) time (sec)
(a) (b)
Figure 7.9: Accelerated creep tests on samples taken from the crypt of Monza
Cathedral: experimental results and theoretical predictions with differ-
ent values for the relaxation time of Maxwells spring: (a) axial strain
versus time; (b) transversal strain versus time.
two values for tM in the simulation of the tests performed on samples taken from
Monza Cathedral are shown: note that, although the secondary creep rate is in
general underestimated, the time to failure of the material is not very sensitive to
the choice made for tM.
7.3 Structural analyses of two masonry towers

The theoretical model described in Section 7.2 was implemented into a commercial
FE code, suitable for nonlinear structural analyses and endowed with a user-oriented
interface (ABAQUS). This code was used to predict the time behaviour of two
massive Middle-Ages masonry towers in Italy: the Civic Tower of Pavia (Section
1.2) collapsed in March 1989, and the bell-tower of the Cathedral of Monza
(Section 1.4.2), which has recently been restored.
7.3.1 The Civic Tower of Pavia
More details on the research programme carried out on this building can be found
in Sections 1.2.2, 1.2.3, 2.2.2 and in [17]. The lower part of the tower (22 m high)
was discretized using three-dimensional FEs of different shapes (see Fig. 7.10).
The total number of d.o.f.s of the FE model is about 11000. The loads acting on
the model are the dead weight of the discretized part (unit weight: g = 18 kN/m3)
and the weight of the upper part of the tower, including the belfry, which was
assumed to be evenly distributed at the top of the discretized part, with an intensity
q = 675 kN/m2. These loads are supposed to grow quickly (in about ten years),
up to their final value; afterwards, they remain constant throughout the rest of the
analysis. The model parameters employed in the analysis are listed in the second
column of Table 7.3 (except for tM = 1000 years).
q = 675 kN/m2
g = 18 kN/m3
22 m
(a) (b)
Figure 7.10: (a) Civic Tower of Pavia; (b) FE model of the lower part.
Figure 7.11 shows the evolution in time of the vertical displacement of a corner at
the top of the FE model. It is important to note that failure is reached in about 400 years:
this value matches the experimental collapse time (1989 AD) reasonably well, if com-
puted from the end of sixteenth century when the construction of the belfry ended.
The time evolution of damage is represented in Fig. 7.12 in the form of contour
plots of the trace of the damage tensor. Note that damage is negligible at the end of
construction (Fig. 7.12a). At this time, it was found that the maximum vertical (com-
pressive) stress is of the order of 1.8 MPa at the corner of the base of the tower located
near the entrance: although this stress is compatible with the short-term strength of
the masonry forming the tower, it was found to be able to bring the material to failure
if acting for a long time (see Section 2.2). Damage evolves slowly for about 200 years
(Fig. 7.12b and c); then, it starts growing quickly leading the tower to failure for loss
of bearing capacity in the region near the most stressed base corner (Fig. 7.12d).
The numerically identified failure mechanism of the tower, that is, the deformed
shape at incipient failure, is shown in Fig. 7.12e: the tower seems to collapse by
crushing the corner near the entrance. It is interesting to note that this failure
mechanism agrees both with the survey of the tower ruins (see Section 1.2) and
verbal evidences of some witnesses.
7.3.2 The bell-tower of Monza Cathedral
The second masonry structure analysed is the bell tower of Monza Cathedral
(Fig. 7.13a). This tower is presently under restoration, as it showed evidence of
diffused cracking (Fig. 7.13b): more details on the research programme carried
-7
-6
displacement (cm) -5
-4
-3
-2
-1
0
0 5 10 15 20 25 30 35 40 45 50
time (years *10)
Figure 7.11: FE analysis of Pavia tower: vertical displacement of a node at the top of
the model versus time.
(a) (b)
(c) (d) (e)
Figure 7.12: FE analysis of Pavia Tower: damage contour plots different times: (a)
end of the construction; (b) 100 years; (c) 210 years; (d) numerical
collapse (417 years); (e) failure mode.
3
1
(a) (b) (c)
Figure 7.13: Cathedral of Monza and bell-tower; (b) crack pattern before restora-
tion; (c) FE mesh of the lower part of the bell-tower.
out on this building can be found in Section 1.4.2 and in [18]. Only the lower
part of the tower was discretized by FEs, for a height of 35 m (see Fig. 7.13c).
The FE model employed to analyse the Monza tower consists of isoparametric
eight-noded 3D FEs; the total number of d.o.f.s of the mesh is around 18000.
In the analysis, the tower was supposed to be loaded only by its self-weight,
with a unit weight of the material equal to 18 kN/m3. The upper part of the tower,
which was not discretized by FEs, enters the analysis only as an evenly distributed
load of 50.7 kN/m2 at the top of the FE model: this load was estimated according
to the volume of the upper part and the belfry (which is about 1500 m3) and the
area of the cross-section of the tower at 35 m (which is 53.54 m2).
The values of the parameters employed in the analysis are listed in Table 7.3,
third column, except for tM = 1000 years.
In Fig. 7.14 the results of the FE analysis are presented in the form of contour
plots of the maximum eigenvalue of the damage tensor at different times. t = 0
corresponds to the end of the construction phase of the tower: it can be seen that
damage is nearly negligible, meaning that the tower would be deemed to be safe
according to a short-term analysis. A maximum vertical stress of the order of 1.23
MPa was computed at the base of the tower, which is nearly twice the average
vertical stress computed as the ratio of the tower weight to the base area.
Figure 7.14c refers to the estimated damage distribution at the end of the current
century (around 2100 AD): it is interesting to note that damage evolves in time
mainly in the vicinity of the openings located in the western and eastern sides of
the tower, thus matching the monitored experimental cracks (see Fig. 7.13b). The
diffusion of damage is relatively slow up to 300 years, whereas it accelerates after
this time (corresponding to the end of nineteenth century) leading the tower to
(a) (b)
(c) (d) (e)
Figure 7.14: Contour plots of one of the principal components of the viscous dam-
age tensor at (a) 150 years, (b) 300 years, (c) 500 years and (d) 650
years (estimated failure); (e) deformed FE mesh at failure.
0.8 1.6
0.4 u1
1.2
displacement (mm)
displacement (mm)
u3
0 u1
u3 0.8
-0.4
0.4
-0.8
-1.2 0
0 200 400 600 800 0 200 400 600 800
time (years) time (years)
Figure 7.15: Horizontal displacement components versus time for two nodes of
the FE mesh located at a height of about 9 m.
failure after about 650 years. Fig. 7.14e shows the deformed FE model at the pre-
dicted time to failure: note that the barreling of the lower part is matched by the
spalling of the corners of the tower observed in reality (Section 1.4.2).
To emphasize the potentially catastrophic increase in damage with time, in
Fig. 7.15 the horizontal displacement components of a couple of nodes located at
the level of the most damaged zone are plotted versus time; the sign of the displace-
ments is immaterial, as it depends on the choice made for the reference frame (see
Fig. 7.13c). Note how in these plots the three stages typical of the creep behaviour
of materials experiencing damage effects are visible, namely primary, secondary
(or steady-state) and tertiary creep; the latter phase precedes structural failure.
Note that the predicted time to failure of the tower is much shorter than the time
to failure of a material element subjected to a uniaxial compressive stress of 1.23
MPa (see Fig. 7.8): this is attributable to the stress concentrations originated by the
diffusion of damage throughout the tower.
7.4 Remarks and future perspectives

The numerical model developed to analyse damage effects induced by heavy per-
sistent loads in existing masonry buildings is potentially an effective tool to assess
the safety of these structures. Both the predicted time to failure of Pavia Civic
Tower and the predicted crack pattern in the bell-tower of Monza Cathedral were
found to match experimental evidences fairly well.
In particular, the predicted failure mechanisms for the analysed structures may
give indications regarding the effectiveness of possible strengthening techniques.
Pavia Civic Tower was found to collapse for a sort of rotation about one of the
corners, which got apparently crushed whereas the rest of the building did not
significantly participate in the failure mechanism (Fig. 7.12e). Thus, hooping the
tower would not likely make it safer against collapse, as this technique would not
hinder the predicted mode of failure.
On the contrary, hooping Monza bell-tower might actually contrast barrelling

at the base of the building, the deformation mode that precedes failure according
to numerical analyses (see Fig. 7.14e): this restoration technique is currently being
employed to strengthen the tower (see Section 8.3.3, or [19]).
Regarding the evaluation of the model parameters, the relaxation time of the
Maxwells dashpot, tM, was found to play a crucial role in the prediction of the
residual life of the analysed structures. An open problem is how this parameter can
be realistically estimated according to accelerated creep tests rather than true creep
tests. This is strictly related to the detection of the actual exhaustion of the primary
creep phase (related to tK) in tests of a given, finite duration. This problem should
be addressed in the continuation of this research.
On the side of the accuracy of the FE model, there is a point that should have to
be tackled in future researches, regarding the mesh-sensitivity of the numerical
solution. It is well known that, in the presence of a strain-softening constitutive
law, strains localize in narrower bands as the FE mesh is refined. A non-local
version of the presented damage model will be formulated, according to proposals
of other researchers (see e.g. Saetta et al. [20]), to overcome this limitation of the
present model and make numerical predictions unaffected by the mesh size.
Other problems that will be tackled in continuation of the present research to
improve the numerical model are (1) the distinction between cracks activated in
tension or compression and (2) the so-called unilateral effect. Indeed, in the pres-
ent version the model makes no distinction between damage directions associated
to tensile strains and to compressive strains, assuming that both are aligned with
the direct strain which induces the damage process. A more realistic modelization
of the damage process should account for the experimentally observed inclination
of cracks with respect to the compressive strains by which they are induced. The
unilateral effect consists in a recovery in stiffness when an opened crack heals, for
instance, as a consequence of a stress reversal from tension to compression. This
effect could be described by introducing two independent second-order damage
tensors, say DC and DT, that separately allow for damage induced by compressive
and tensile strains, rather than a single damage tensor that evolves differently,
according to the sign of the principal strains.
References
[1] Binda, L., Gatti, G., Mangano, G., Poggi, C. & Sacchi-Landriani, G., The
collapse of the civic tower of Pavia: a survey of the materials and structure.
[2] Papa, E. & Taliercio, A., A damage model for brittle materials under
non-proportional monotonic and sustained stresses. Int. J. Numer. Analyt.
Methods in Geomechanics, 29, pp. 287310, 2005.
[3] Baant, Z.P. & Chern, J.C., Strain softening with creep and exponential
algorithm. J. Engng. Mech., ASCE, 111, pp. 391415, 1985.
[4] Cervera, M., Oliver, S., Oller, S. & Galindo, M., Pathological behaviour of
large concrete dams analysed via isotropic damage models. Proc. 2nd Int.
Conf. on Computer Aided Analysis and Design of Concrete Structures, Zell

am See (A), pp. 633644, 1990.
[5] Lemaitre, J. & Chaboche, J.L., Mechanics of Solid Materials. Cambridge
Univ. Press: Cambridge (UK), 1990.
[6] La Borderie, C., Berthaud, Y. & Pijaudier-Cabot G., Crack closure effect in
continuum damage mechanics: numerical implementation. Proc. 2nd Int.
Conf. on Computer Aided Analysis and Design of Concrete Structures, Zell
am See (A), pp. 975986, 1990.
[7] Chan, K.S., Bodner, S.R., Fossum, A.F. & Munson D.E., A constitutive
model for inelastic flow and damage evolution in solids under triaxial com-
pression. Mechanics of Materials, 14, pp. 114, 1992.
[8] Papa, E., Taliercio, A. & Binda, L., Creep failure of ancient masonry: exper-
imental investigation and numerical modelling. Structural Studies, Repairs
and Maintenance of Historical Buildings VII, ed. C.A. Brebbia, WIT Press:
Southampton (UK), pp. 285294, 2001.
[9] Kachanov, L.M., Elastic solids with many cracks: a simple method of anal-
ysis. Int. J. Solids Structures, 23, pp. 2343, 1987.
[10] Mazars, J., A description of micro- and macro-scale damage for concrete
structures. Engng. Frac. Mech., 25, pp. 729737, 1986.
[11] Lekhnitskii, S.G., Theory of Elasticity of an Anisotropic Body. Holden-Day:
San Francisco (USA), 1963.
[12] Cordebois, J.P. & Sidoroff F., Anisotropic damage in elasticity and plastic-
ity. J. Mc. Th. Appl. (Special issue), pp. 4560 (in French), 1985.
[13] Ortiz, M., A constitutive theory for the inelastic behaviour of concrete.
Mech. of Materials, 4, pp. 6793, 1985.
[14] Suaris, W., Ouyang, Ch. & Fernando, V.M., Damage model for cyclic load-
ing of concrete. J. Engng. Mech., ASCE, 116(5), pp. 10201035, 1990.
[15] Herrmann, G. & Kestin, J., On the thermodynamic foundations of a damage
theory in elastic solids. Strain Localization and Size Effects due to Damage and
Cracking, eds J. Mazars & Z.P. Bazant, Elsevier: London, pp. 228232, 1998.
[16] Papa, E. & Taliercio, A., Anisotropic damage model for the multiaxial static
and fatigue behaviour of plain concrete. Engng Frac. Mech., 55(2), pp.
163179, 1996.
[17] Papa, E. & Taliercio, A., Creep modelling of masonry historic towers.
Structural Studies, Repairs and Maintenance of Heritage Architecture VIII,
ed. C.A. Brebbia, WIT Press: Southampton, pp. 131140, 2003.
[18] Papa, E., Taliercio, A. & Binda, L., Safety assessment of ancient masonry
towers. Proc. 2nd Int. Congress Studies in Ancient Structures, eds G. Arun
& N. Sekin, Istanbul, 1, pp. 345354, 2001.
cathedral bell-tower. Construction and Building Materials, Special Issue,
Elsevier Science Ltd., 16(7), pp. 385395, 2002.
[20] Saetta, A., Scotta, R. & Vitaliani, R., Mechanical behaviour of concrete
under physical-chemical attacks. J. Engng. Mech., ASCE, 124(10), pp.
11001109, 1998.
CHAPTER 8
Repair techniques and long-term

damage of massive structures
C. Modena & M.R. Valluzzi

Dipartimento di Costruzioni e Trasporti, University of Padua, Italy.
8.1 Introduction
Brick-masonry load bearing elements of heavy historic structures such as towers,
heavily loaded pillars, and large masonry walls (i.e. curtains) frequently exhibit
very typical mechanical deterioration phenomena like (a) formation of vertical or
sub-vertical, thin but very diffused cracks and (b) more or less local detachment of
the outer leaf in multiple leaf walls [1, 2].
Such a particular crack pattern is often not attributable to common causes of
damage like seismic events, foundation settlements, instantaneous increase of
external loads (e.g. for added storey or building changes) or to chemical, physical
and mechanical degradation of the basic materials. On the contrary, it is due to the
prevalent effect of the dead load and to the connected time dependent phenomena,
often combined with cyclic loads, like wind actions, thermal and hygroscopic
excursions, or bell ring oscillations (in bell-towers) [3, 4]. Because of the large
weight of this sort of structures, wind and temperature loads do not cause substan-
tial increase of stresses (shear and flexure) at the bottom. On the contrary, they can
act in combination with the previously mentioned condition and contribute to
worsening the crack situation in the heaviest portions of the structure with their
cyclic trend. In fact, close to the ultimate strength of the material, alternate cycles of
loading and unloading can lead to fatigue damage, which can increase the failure
hazard of the structure.
As it is well known, cracks appear very thin until the failure, which happens
suddenly, without any apparent warning (such as large cracks or spalling), even in
close proximity of the collapse.
In the past, such problems were disregarded in comparison with other more
evident critical conditions (large cracks, out-of-plumb, relevant deformations, etc.),
because the presence of the above-mentioned cracking pattern was considered as

a steady state of the structure. The sudden collapse of some of them induced
researchers to deepen the study in order to find specific solutions of repair and
strengthening.
Experimental investigations performed in collaboration between the University
of Padova and the Polytechnic of Milan [5, 6] demonstrated that the creep damage
can be effectively counteracted by the bed joint reinforcement technique (known
later also as structural repointing). It requires the insertion of minimal reinforce-
ment of the superficial layer of the masonry, and a correct execution and selection
of compatible materials.
This chapter describes the most significant experimental researches that contrib-
uted to validate the above-mentioned technique for specific application on real
case studies. Stainless steel bars were first adopted, then FRP (fibre reinforced
polymer) bars and thin plates have been more recently considered; the use of dif-
ferent types of embedding materials (lime-based or additivated mortars, resins) is
also discussed. Finally, some applications of the proposed technique on masonry
structures are briefly illustrated [7, 8].
8.2 The bed reinforcement technique

The reinforcement of the mortar bed joint in brick masonry structures has been
recently considered for the strengthening and repair of massive brick masonry
structures like towers, pillars and heavy loaded walls.
The technique is performed by the insertion of small diameter reinforcing ele-
ments (stainless steel or FRP bars or plates) in the mortar bed joints previously
excavated (up to about 68 cm) and then filled by a repointing material (compati-
ble mortars should generally be used, also possibly with additives for particular
requirements). The intervention can be executed at one or both sides of the wall,
depending of the in situ conditions (accessibility, thickness, type of masonry, etc.).
Transversal ties, inserted into drilled holes successively sealed, can improve the
confining action of the bars, both in the longitudinal and in the transversal direc-
tions, especially in the case of multi-leaf masonry, where out-of-plane phenomena,
as the detachment of the external leaf can occur (Fig. 8.1). Such a technique is
particularly suitable to brick walls having regular courses of mortar. Moreover, the
technique can be successfully applied when the control of cracking due to different
settlements and thermal and moisture movements is required.
It exploits the bond actions among bars and mortar to counteract the dilation
caused by the compression loads. It is clearly a surface intervention but a combina-
tion of this technique with rebuilding (where bricks are particularly damaged) and/
or injections (especially for multi-leaf masonry walls with internal core having a
high percentage of voids), can improve the strength of the damaged compressed
wall and consequently reduce the cracking.
The application of the technique involves many aspects as follows: (1) the pre-
liminary preparation of the joints, (2) the choice of the reinforcement in terms of
Repair Techniques for Massive Structures 177
(a) (b) (c)
Figure 8.1: Examples of different techniques in different types of brick masonry

walls: (a) single-leaf bearing wall; (b) multi-leaf wall with external
bearing walls; (c) multi-leaf wall with external veneer leaf.
material (steel bars or plates, or FRP laminates) and in terms of spacing and amount
to obtain the best performances, (3) the type of repointing material (mortar or
resins) and (4) the aesthetic of the wall face.
The intervention is characterized by the following operative phases:
1. Possible removal of plaster or finishing from the surface, to check the masonry
condition.
2. Cutting of the bed mortar joints by using suitable (very common) tools; the
recesses should be at least 10 mm high and 5070 mm deep, so that the rein-
forcement can be inserted and the remaining mortar in the masonry can bear the
applied loads.
3. Accurate inspection of the masonry: it should be appropriate to inject some
large voids or replace some bricks.
4. Removal of powder or rubble through compressed air, or water, or particular
solvents, depending on both the existent and the repointing materials. In par-
ticular, water can be used if a mortar is adopted for repointing, such that exces-
sive absorption from the mortar to the bricks can be avoided, whereas the
groove should be kept dry in case of the synthetic mortars (additivated with
resins).
5. Placing of a first layer of repointing material, which should be accurately com-
pacted; a proper embedding material should be used. Mortars are usually made
of hydraulic lime for a better compatibility (chemical, physical and mechanical)
with the existent ones, and can contain special additives (e.g. with expansive
properties to compensate the shrinkage during the hydration phase). Synthetic
resins (epoxy, acrylic or polyester) may be used for particular cases, for example,
when the achievement of the strength is needed in a shorter time.
6. Placing of the reinforcing material: steel bars or plates (stainless, in general) or
FRP laminates can be used. To increase the friction between reinforcement and
Figure 8.2: Detail of embedding phase in strengthened joint.
mortar rough surfaces should be preferred, as reinforced steel bars. To the same
purpose, steel bars or plates can be previously sandblasted in order to clean
their surface and to improve the adhesion with the repointing material. More-
over, the placing of spacers can be appropriate to separate the reinforcement
from the surface of the bricks. Finally, the use of more bars with smaller diam-
eter rather than a single bar with larger diameter should be preferred. Moreover,
due to the small dimension of the bed joints (usually around 1015 mm) only
reduced sizes of reinforcement (46 mm in diameter for rods) can be inserted.
7. A second layer of embedding material has to be applied over the bars to cover
them accurately; further bars or other reinforcement type can be put in if
necessary.
8. A final layer of repointing material should be placed in the last 1520 mm avail-
able, to seal the horizontal joints and for aesthetic and homogeneity purpose;
special sands or pigments can be used to obtain particular effects.
The proposed technique does not show particular difficulty of application; some care
is required in some operative phases (cutting of the bed joint, cleaning, repointing,
etc.) but it can be performed quite easily and quickly (Figs 8.2 and 8.3).
8.3 The experimental campaigns

First experimental investigations [57, 9] with applications of low diameter (5 or
6 mm) stainless steel reinforced bars showed the efficiency of the intervention in
reducing the transverse dilation of walls.
Later on, investigations were focused on verifying the possible application of FRP
materials. Monotonic, cyclic and creep simulating tests were performed first on pan-
els strengthened with circular section carbon bars (CFRP) (5 mm in diameter) [10]
Figure 8.3: Arrangement of bar and tie into an excavated joint before final sealing
of hole.
and then, more recently, with thin rectangular section CFRP strips (1.5 5 mm)
(Fig. 8.21) [9, 11, 12].
The advantages of the use of carbon fibres instead of steel reinforcement are
mainly related to their complete corrosion immunity, but many aspects still need
to be deeply investigated. Despite their high strength, FRPs are very brittle; thus,
even though the tensile stresses able to provoke rupture are not practically achiev-
able in the moderate stress conditions detectable in the masonry structures, FRPs
inductility to bending and folding (e.g. for anchorage) does not allow low-cost
solutions. Moreover, they are sensitive to high temperatures [13] and constitutive
laws able to describe the interface behaviour among reinforcementmortarbrick
are so far not comprehensively known in masonry [8].
More advantages may be gained with the use of thin strips in comparison to
circular bars, due to their higher adaptability to the unevenness of the joints and
better behaviour against splitting failures [11], which may enable more superficial
and less obtrusive interventions, but further research both at global and local level
are required for the proper validation of this reinforcement system.
Finally, the correct application of the techniques requires that compatible
embedding mortars should be properly selected to avoid further undesirable prob-
lems to the original masonry.
8.3.1 Laboratory tests on the use of stainless steel bars
Experimental work carried out on 19992001 by the Polytechnic of Milan and the
University of Padua on brick panels having dimensions of about 25 50 110 cm
primarily showed that the bars are able to contribute in bearing the tensile stresses
acting into the bricks and, consequently, to reduce the dilation of the wall.
A preliminary finite element model (FEM) study calibrated on the basis of the
materials characterization contributed to quantify the stresses borne by the bars
and the bricks and to optimize the geometrical distribution of the bars into the
panels (Fig. 8.4). This can provide information for the choice of the reinforcement
in the design phase. By using two 6 mm diameter bars every three joints, the
migration of the tensile stresses from the bricks to the reinforcing material was
estimated to be approximately 40% of the total. The same analysis demonstrated
that higher reduction of the tension (over 50%) can be obtained by using smaller
bars (e.g. one bar of 5 mm in diameter, see Fig. 8.5) in every bed joint, but such
condition can be inapplicable in practice (more damaging of the masonry during
execution and consequent higher costs of intervention). On the contrary, higher
diameter bars, always total amount of reinforcement being equal (e.g. 3 bars of
8 mm in diameter every 9 joints), can lead to a reduction of the tensile stresses only
by 20%.
Therefore, in the experimental campaign, two stainless steel bars of 6 mm diam-
eter (26) were used for every three bed joints of mortars.
Different configurations (either both sides or just one side reinforced) and dif-
ferent types of repointing material (hydraulic mortar or resins) were considered.
Mechanical characteristics (compressive strength and elastic modulus) of basic
materials were as follows: fc = 5.4 MPa and E = 3890 MPa for the hydraulic mor-
tar, fc = 5090 MPa for resin type 1, and E = 69 GPa and fc = 5565 MPa for resin
type 2. Two series of laboratory tests were performed (see Table 8.1 and Fig. 8.6):
(1) monotonic compressive tests on six masonry prisms repaired at one side, after
Figure 8.4: Reduction of principal tensile stress in bricks (the model concerns the
upper-left portion of a panel) before (left) and after (right) insertion of
steel bars every three joints (stresses are in MPa).
(a) 38 (b) 26 (c) 15
Figure 8.5: Optimization of bed joint reinforcement in brick masonry panel (the
same amount of steel is distributed): the support of a numerical analy-
sis showed a reduction of tensile stresses in bricks of 20, 40 and 50%,
for the cases (a) (38), (b) (26) and (c) (15), respectively. The
positioning of two bars every three joints optimizes structural perfor-
mance with minimal obtrusiveness.
damaging caused by previous compressive tests and, (2) creep simulating tests on
six prisms strengthened (i.e. without any previous damage) on both sides. Such
latter actions can be performed by progressive increments of compressive loads,
kept constant for a fixed time (e.g. three hours) up to collapse, which occurs for
excessive deformations (tertiary creep phase).
As regards the monotonic series of tests, the repaired panels reached only about
50% of the original strength and were characterized by a stiffness ranging between
43% for the consolidated sides and 29% for the non-consolidated side compared
to the original stiffness (see Fig. 8.7). Due to the limited dimension of the samples
in comparison with a whole existing wall, the damage caused by the cutting of the
joints in the post-compression phase was probably in most cases excessive. Any-
way, it is important to notice that the main results have been obtained in terms of
reduction of the dilation and of the vertical cracking of the masonry. In particular, all
the prisms showed a reduced cracks pattern in the repaired sides, whereas in the non-
repaired side the damage increased both in number and in the depth of fissures.
As for the accelerated creep series of tests, the scheme of loading was able to
show a crack pattern very similar to the typical observations in the real structures,
with lower transversal deformation obtained for the strengthened walls, especially
in the main sides (see Fig. 8.8).
Figure 8.9 shows the vertical and the horizontal deformation versus stress
diagrams of the creep tests. It can be noticed that in the proposed configuration
182
Table 8.1: Experimental programme: stainless steel bars reinforcement (19992001).
Loading Reinforcing Reinforcing
Series Sample test Intervention material case Repointing material
0-SS M01 Creep
(unreinforced) M03 Creep
1-SS (one MU6H1 Monotonic Repair 26 A Hydraulic lime mortar
side reinforced)
MU6H2 Repair 26 A Hydraulic lime mortar
MU6H3 Repair 26 B Resin type 1
MU6H4 Repair 26 C Resin type 2
MU6H5 Repair 26 B Resin type 1
MU6H6 Strengthening 26 A Hydraulic lime mortar Hydrated
2-SS (two sides M02 Creep Strengthening 26 E lime and pozzolana mortar
reinforced)
M04 Monotonic Strengthening 26 D Hydraulic lime mortar with resin

M05 Monotonic Strengthening 26 E Hydrated lime and pozzolana
mortar
M06 Creep Strengthening 26 D Hydraulic lime mortar with resin
original mortar original mortar

A A
Case D Case E
A A
hydraulic lime mortar hydraulic lime and
with resin pozzolana
Figure 8.6: Panels configuration investigated during experimental campaign.
(Fig. 8.10) the reinforcement does not have great influence on the strength and the
vertical deformations, whereas it has a great role in the control of the horizontal
deformations. In particular, the reinforcement was able to reach the tertiary creep
conditions at deformations around 70% of the original case; moreover, in such
ultimate phase, the prisms reinforced by technique D, showed an increment of
more than 25% in strength. Reduction of lateral deformation was around 3739%
(Fig. 8.11), even though the increase in strengthening is between 5 and 28%.
8.3.2 Laboratory tests on the use of CFRP bars and thin strips
As for CFRP rods, laboratory tests were performed both at local and global level,
to characterize the mechanical properties of the basic materials and their mutual
7.00
MU6H1
volumetric
6.00 v - BCD sides (repaired)
l - BCD sides (repaired)
v - A side (plain)
l - A side (plain)
5.00
before-repair curves
stress [MPa]
4.00 volLVDT l, v
3.00
2.00
1.00
0.00
-40.0 -35.0 -30.0 -25.0 -20.0 -15.0 -10.0 -5.0 0.0 5.0 10.0
strain [m/mm]
Figure 8.7: Typical stressstrain diagram before and after repair for first series
of panels (Case A).
Side A Side B Side A Side B

(a) (b)
Figure 8.8: Crack pattern detected in second series of panels: (a) unreinforced
prism MO1; (b) MO6 (Case D).
(a)
5.0
Case D (M06)
4.0
Plain panel (M01)
3.0
[MPa]
Case E (M02)
2.0
1.0
0.0
1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0
v [10-3]
(b)
5.0
Case D (M06)
Case E (M02)
4.0
3.0 Unreinforced prism (M01)

[MPa]
2.0
1.0
0.0
0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
l sides AB [10-3]
Figure 8.9: Creep test results: (a) vertical and (b) horizontal deformations of
panels.
interaction, and of full-scale masonry panels. Pull-out and sliding tests on triplets
were performed to characterize respectively the bar and the brick interfaces with
the mortar and the embedding materials.
Solid clay bricks having dimensions 50 120 250 mm were used. Their aver-
age compressive strength is around 16.55 MPa, whereas the flexural and the ten-
sile strength (from Brazilian test) are 4.75 and 2.30 MPa, respectively. The parent
mortar was a premixed product based on a natural hydraulic lime binder (1.35 and
4.80 MPa for flexural and compressive strength respectively, and 4.500 MPa for
elastic modulus, after 28 days of curing). As repointing materials an epoxy resin
(having 40.10 and 81.50 MPa for flexural and compressive strength) and a pre-
mixed fibre-reinforced hydrated lime and pozzolana mortar (having 3.80 and 16.70
MPa for flexural and compressive strength and 16.000 MPa) were adopted. Pul-
truded CFRP bars 5 mm in diameter, twisted and sand-coated in order to improve
Figure 8.10: General scheme of reinforcement technique with steel bars.
Figure 8.11: Maximum transversal deformation measured during creep tests car-
ried out on steel reinforced panels [12].
their bond properties, were characterized by high tensile strength (2300 MPa), a
modulus of elasticity of 130 GPa, and an ultimate strain of 1.8%.
Results at local level demonstrated a higher sliding strength developed at the
interface between resin and brick (4.60 MPa against the 3.40 MPa for the mortar
case), but related to a more brittle failure mechanism, characterized by a sudden
detachment of the elements, which involved the surface layers of the bricks also.
Cohesion and friction coefficients were 0.88 and 40.45, and 1.34 and 47.37 for
the fibre reinforced mortar and the resin, respectively. To check the bond between
the bars and the embedding materials pull-off tests were performed. Under a trans-
verse compressive stress equal to the maximum value utilized in the triplet tests
(3.00 MPa) resins were able to develop bond strength higher than twice that of
mortar (20.60 MPa against 9.45 MPa) but, similar to the triplet test results, the
failure was more brittle and involved also the splitting of the bricks.
At overall level, ten panels of 110 52 25 cm were subjected to strengthening
by CFRP bars and subsequent uniaxial compression, under monotonic or cyclic
loads. Cycles were executed at load levels corresponding approximately to 1/2 or
1/3 of the monotonic strength (for 2-cycle and 3-cycle cases respectively). To
make up for the current unavailability of foldable FRP bars or of special corner
reinforcing elements, the bars were anchored in the short sides of the samples by
using sleeves composed by a metal tube filled by resin laid on a square plate,
whose reliability was tested by a preliminary experimental phase [10]. The whole
experimental programme is given in Table 8.2.
Results confirmed that the best results are obtained for symmetric reinforce-
ment (both sides) and for the use of lime-based mortar (Fig. 8.12). The maximum
increase in strength was around 40% (2S.M panels, embedded with lime-based
mortar), with a slight increase in the modulus of elasticity too. A great reduction
Table 8.2: Experimental programme: CFRP bars reinforcement (2003).

Loading Reinforcing Repointing
Series Sample test Intervention material material
0-CFRPb UR.1 Monotonic
(unreinforced) UR.2 2 cycles
1-CFRPb 1S.M.1 Monotonic Strengthening 15 Fibre-reinforced
(one side hydrated
reinforced) lime &
pozzolana
mortar
1S.M.2 2 cycles Strengthening 15 Fibre-reinforced
hydrated
lime &
pozzolana
mortar
1S.R.1 2 cycles Strengthening 15 Resin type 3
2-CFRPb 2S.M.1 2 cycles Strengthening 15 Fibre-reinforced
(two sides hydrated
reinforced) lime &
pozzolana
mortar
2S.M.2 3 cycles Strengthening 15 Fibre-reinforced
hydrated
lime &
pozzolana
mortar
10.00
Unreinforced (UR.2) Stress (MPa)
Mortar (1S.M.1)
8.00 Resin (1S.R.1)
6.00
4.00
2.00
Strain ()
0.00
-8.00 -6.00 -4.00 -2.00 0.00 2.00 4.00 6.00 8.00
10.00
Stress (MPa)
8.00
6.00
Unreinforced (UR.2)
Mortar (2S.M.1)
4.00 Resin (2S.R.1)
2.00
Strain ()
0.00
-8.00 -6.00 -4.00 -2.00 0.00 2.00 4.00 6.00 8.00
Figure 8.12: Stressstrain diagrams for 1-CFRP series (a, one side strengthening)
and for 2-CFRP series (b, two sides strengthening).
of the crack pattern was detected at the end of the tests on the reinforced sides of
the strengthened panels in comparison with the plain ones (decrease of the Poisson
ratio of about 50%) (Figs 8.13 and 8.14). Panels repointed with epoxy resins also
gave good results, but their behaviour is characterized by a brittle failure mecha-
nism and by the presence of higher opening of cracks in the longitudinal side at the
ultimate load (Fig. 8.12).
The strength values obtained on the tested panels in both the first cracking and
the ultimate phases are depicted in Fig. 8.15. Cyclic tests performed in two or three
steps showed that after the first loading phase (up to 30% of the stress peak) the
increment in strength is still possible, due to the elastic properties of the materials
and the possible settling of the panel. On the contrary, after the second cycle (up
to 60% of the stress peak), the internal damage is progressively increasing, thus
the retrieval of the previous strength is not possible anymore. Compared to the
strengthened masonry, first cracks appears very close to the ultimate stress in the
Figure 8.13: Crack pattern of the 1S.R.1 panel (side A is the strengthened one).
Figure 8.14: Details of one-side (left) and two-sides (right) strengthened panels
after failure.
unstrengthened masonry (Fig. 8.16); this confirms that the capacity to slowdown
the failure process by the progressive plasticization of the masonry is particularly
significant for symmetric configurations of strengthening.
In comparison to rebars, CFRP thin strips (having a rectangular section and
thickness around 1.5 mm), have a higher potential, possibly due to more superficial
placements and higher flexibility (Figs 8.17 and 8.18). Pultruded CFRP strips are
10.00
8.00
6.00
s (MPa)
4.00
2.00
0.00
UR.1 UR.2 1S.M. 1S.M. 1S.R.1 1S.R.2 2S.M. 2S.M. 2S.R.1 2S.R.2
First crack 5.96 6.15 6.67 6.94 6.19 5.32 8.02 7.36 7.67 6.11
Failure 6.17 6.68 7.25 7.42 6.99 7.48 9.67 8.18 8.79 8.43
Figure 8.15: Compression strength values detected on panels.
Figure 8.16: Detail of the cracking of the UR.2 panel.
externally sanded to improve adhesion with embedding materials; measured ulti-

mate tensile strength was 1334 MPa at a corresponding strain of 1.8%, with a
Youngs modulus of 73.20 GPa. Behaviour at interface and overall levels was
extensively investigated experimentally.
A first step of the research was carried out at the University of Padua, including the
selection and characterization of the materials, the CFRP positioning configuration
and a series of monotonic tests on wall samples [10, 14]. Clay bricks had a compres-
sive and flexural strength of 17.24 and 6.40 MPa, respectively. Ordinary hydraulic
lime mortar used for the laying of the bed joints had compressive and flexural strength
after 28 days of curing of 10.32 and 0.63 MPa, respectively. To exploit the high per-
formances of the FRP a high-strength hydraulic lime mortar was selected for the
repointing phase; it had corresponding values of 15.61 and 0.83 MPa.
The preliminary monotonic tests carried out on these panels showed the weak-
ness of the corner layout of the reinforcement, which led to premature failure
Figure 8.17: Configuration of a plain panel subjected to laboratory test and detail
of reinforcement with thin CFRP strips.
Figure 8.18: Experimental study of the interface behaviour of CFRP strips and
numerical simulation showing the stress migration along the anchor-
ing length and the influence of the rectangular section in limiting the
splitting of the mortar, due to the elliptical stress distribution around
the strip [11].
mainly concentrated at the top of the panels, as strips were not overlapped at the
corners (Fig. 8.19). This was due to the unavailability of special anchorage devices,
as L-shaped strips or bars, which may counteract such a brittle mechanism.
The second step of the research, carried out at the Polytechnic of Milan, was
focused on the evaluation of the creep behaviour of the reinforced masonry [8, 12].
Monotonic and creep tests were carried out on plain and strengthened samples,
with CFRP strip inserted at every horizontal joint on the four sides (basic configu-
ration, type A) or with CFRP sheets applied around the corners and the lateral
sides of the specimen (configuration B) (Fig. 8.20). This was to prevent the prema-
ture failure at corners, as described above.
The experimental programme executed on panels in different configurations is
reported in Table 8.3.
As for monotonic tests, as expected, ultimate axial load of the panels were
quite similar, for both reference and reinforced panels. After cracking a relevant
Figure 8.19: Failure mode of unreinforced and reinforced panels.
Figure 8.20: Configuration B for creep tests: CFRP thin strips at every joint and
CFRP sheets around corners.
Table 8.3: Experimental programme: CFRP thin strips reinforcement (200305).
Sample Reinforcing
Series label Loading test Intervention configuration Repair case Repointing material
0-CFRPs UR1 Monotonic Repair Each joint
(unreinforced) in all sides
UR2 Monotonic
UR3C Creep
1-CFRPs 1SAS1 Monotonic Strengthening Strips at each joint on sides High-strength
Repair Techniques for Massive Structures

(one side A,B,D + sleeves on A hydrated lime mortar
reinforced)
1SA1 Monotonic Repair Strips at each joint on sides Each joint High-strength
A,B,D on side C hydrated lime mortar
2-CFRPs 2SA1 Monotonic Strengthening Strips at each joint on sides High-strength
(two sides A,B,C,D hydrated lime mortar
reinforced) 2SA2 Monotonic Strengthening Strips at each joint on sides High-strength
A,B,C,D + FRP sheets hydrated lime mortar
2SB1 Monotonic Strengthening Strips at each joint on sides High-strength
B,D and every second joint hydrated lime mortar
on sides A,C
2SA3C Creep Strengthening Strips at each joint on sides High-strength
A,B,C,D hydrated lime mortar
2SA4C Creep Strengthening Strips at each joint on sides High-strength
A,B,C,D + FRP sheets hydrated lime mortar
193
improvement in the reduction of the horizontal deformation is obtained with both

configurations of strengthening (A = each joint, B = every second joint, for mono-
tonic tests). Both strengthened specimens (2SB1 and 2SA2), at a stress level equal
to the UR2 ultimate stress, showed an average horizontal strain more than five
times lower than the reference panel (Fig. 8.21a). The strip insertion in every sec-
ond joint was particularly effective, as the maximum horizontal strain reached by
configuration B, corresponding to 13 MPa of compression stress, was almost the
same as the one of configuration A. The only difference among the two configura-
tions performances consists in higher ultimate stress and strain reached by specimen
2SA2, which was obviously due to the damage concentration located across the non-
strengthened bed joints. As for the effect of repair after preliminary compression,
unexpectedly, both repaired specimens revealed a noticeable performance,
as the first cracking stress was reached again with a reasonable horizontal strain.
(a) 16.0
14.0
12.0
10.0
Stress [MPa]
8.0
6.0
UR.2 AV
4.0
2S.B.1 AV
2.0
2S.A.2 AV
0.0
-5.0 -4.0 -3.0 -2.0 -1.0 0.0 1.0 2.0 3.0 4.0 5.0
-2.0
strain [mm/m]
(b) 16.0
14.0
12.0
10.0
Stress [MPa]
8.0
6.0
UR.2 AV
4.0 1S.A.1 AV*
2.0 UR.2.R AV
1S.A.1.R AV
0.0
-5.0 -4.0 -3.0 -2.0 -1.0 0.0 1.0 2.0 3.0 4.0 5.0
-2.0
strain [mm/m] * instrumentation removed at first cracking
Figure 8.21: Axial stress versus strain curves of plain and strengthened specimens
(a) before and (b) after repairing (AV means strain is average of sides
A and C; horizontal strain negative values).
This means that cracks which occurred after the first test could not freely propagate,
as would happen, even at negligible axial stress levels, when loading failed plain
masonry. Moreover, the repaired specimens were able to bear load up to the former
cracking stress (around 9 MPa). At that point, due to the presence of the FRP strips,
new cracks could not appear in the central areas of the main sides, but appeared
close to the corners, where anchoring of the strips is lower. Panel 1SA1, after first
damage and subsequent repairing (1SA1R) showed a coefficient of Poisson passing
from 0.37 to 1.19 in the repaired face (side C) (Fig. 8.21b). This consistency dem-
onstrates that CFRP repointing is an effective technique to limit lateral dilation of
masonry members in both damaged and undamaged conditions.
Also for creep tests, as expected, the strength is not affected by the reinforce-
ment application; in particular, panels strengthened by FRP, despite different fibre
configuration and materials that have been used, showed a lower lateral dilation
(Fig. 8.22) than unreinforced ones.
(a)
(b)
Figure 8.22: (a) Peak stresses and (b) maximum horizontal deformations mea-
sured for creep tests (in test 2SA4C detachment of sheet at corners
affected the experimental measures of deformation) [12].
A typical diagram which resulted from creep simulating test is shown in

Fig. 8.23. The prevalent failure mechanisms are similar to the failure detected
during the monotonic tests, and reinforced panel appears cracked with wider
expulsions and delamination of the external parts of the bricks (Fig. 8.24). In the
16.0
14.0
12.0
Stress [N/mm2]
10.0
8.0
6.0
4.0
2.0 v average between the plates (103)
0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0
0.0
5000
10000
Time [sec]
15000
20000
25000
30000
Figure 8.23: Creep simulation on masonry prisms strengthened with FRP strips
(Polytechnic of Milan) [12]
Figure 8.24: Crack pattern of the panel 2SA3C.

case of application of the CFRP reinforcement at the corners, the detachment of

the adhesion system from the masonry at the end of the test was observed.
8.4 Case studies

The experimental validation of the performances of the bed reinforcement tech-
nique allowed proposing it for the consolidation of several historic constructions
and monuments at risk in Italy, in order to counteract the dilation under high com-
pressive stresses by improving the material toughness. It is worth to remark that,
in real cases, more damage typologies can coexist, thus the proper selection and
combination of different intervention techniques should be considered. In many
cases, both injections and limited rebuilding (the traditional scuci-cuci) may be
used, to reduce the stress concentration and to replace the most damaged resistant
parts, respectively.
Those techniques act locally on improving the mechanical behaviour of the
material for the rehabilitation of the proper load bearing capacity of the structure.
Despite a number of researches and efforts devoted to FRP materials, until now
real applications have been performed only with stainless steel bars in combina-
tion with lime-based repointing mortars, as they can assure more guarantees of
effectiveness and durability, even in exceptional conditions not easy to predict.
In the following, some representative case studies of towers and structural
components of churches in Italy are described [6, 8, 9].
8.4.1 The bell-tower of the Basilica of S. Giustina in Padua
The bell-tower of S. Giustina Basilica in Padua is a three-leaf masonry structure 70

m tall, built during the thirteenth century up to 40 m and raised up to the current
height in the seventeenth century (Fig. 8.25). This event was responsible for the
overloading conditions of the lowest, and poorer, part of the structure, character-
ized by a serious crack pattern, mainly concentrated at the corners. Some spalling
in the internal bricks of the bell-tower was also noticed, possibly due to a worse
quality of the structural elements used on the inside.
There are signs of rebuilding of at least one whole corner performed in the past
(also other similar interventions are visible in other towers still standing) which
proves the effectiveness of that solution in preventing the structure from collapse.
The point under consideration is why this solution works, notwithstanding the fact
that during the operation the local overstress is to increase substantially. The dam-
age in fact involves basically in its first phase the external leaf, where masonry is
stiffer, and due to the availability of a large resisting area on the usual thick sec-
tions (which allowed one to execute the intervention for safety considerations), the
structure is able to accept during time a stress re-distribution both for damage
and rebuilding operations.
In the most deteriorated portions (corner and basement) diffused interven-
tion of injections, partial rebuilding and bed reinforcement technique [5] were
(a) (b)
(c)
Figure 8.25: S. Giustina bell-tower (Padova, Italy): (a) general view, (b) typical
cracking due to creep, (c) cracking and spalling of brick elements.
executed (Fig. 8.26). In particular, local rebuilding (scuci-cuci) was extensively

used during the works aimed at improving the safety conditions of the tower, espe-
cially in the lower, less visible parts of the construction, combined with injections,
that contribute by reducing stress concentrations, whereas in the more visible parts
steel injections are combined with the introduction of small diameter bars into the
bed joints [15].
(a) (b) (c)
Figure 8.26: S. Giustina bell-tower interventions: (a) application of injections, (b)

partial rebuilding (scuci-cuci) and (c) phases of the bed joint reinforce-
ment (insertion of bar and situation before injecting the pin hole).
8.4.2 The pillars of S. Sofia church in Padua
Similar conditions have been found in some pillars in the S. Sofia church, also
located in Padua. Several cracks have been detected and, some traditional repair
techniques as steel ties (see Fig. 8.27), injections and rebuilding have been applied
in several parts of the structure.
8.4.3 The bell-tower of S. Giovanni Battista Cathedral in Monza (Milan)
It is the bell-tower of the Cathedral of Monza, a sixteenth-century building made

of solid brick masonry walls, which show passing-through large vertical poten-
tially dangerous cracks on some particularly weak portions of the western and
eastern sides [9]. They are slowly but continuously opening as given by a monitor-
ing roughly active since 1927. Wide cracks are also present in the corners of the
tower up top 30 m (Fig. 8.28a). Furthermore, a damaged zone at a height of 11
to 25 m with a multitude of very thin and diffused vertical cracks is present (Figs
8.28b,c and 8.29).
Design of intervention was mainly aimed at providing an overall confining
action of masonry walls, limiting the dilation of the material [16]. In the specific
case, due to the large diffusion of the crack pattern (11 9.5 m) and to its exten-
sion in the wall depth (45 cm over 140 cm), to reconstruct a wall or the use of floor
tie roads in order to reduce the lateral deformation of the structure was inapplica-
ble. Therefore, repair and retrofitting of masonry was extensively performed by
grout injection, to re-establish homogeneity, uniformity of strength and continuity
of masonry walls, and by bed joint reinforcement technique by insertion into a
groove (68 cm deep) of one or two small diameter reinforcing bars connected to
the inner leaf by transversal short pins (Fig. 8.29). Both excavated joints and drilled
holes for pins insertion are successively sealed by mortar and grout, respectively
(Fig. 8.30).
(a) (b)
(c)
Figure 8.27: S. Sofia church pillars (Padova): (a) general view of the church,
(b) provisional measures on a cracked pillar, (c) application of the
technique.
(a) (b)
(c)
Figure 8.28: S. Giovanni Battista bell-tower: (a) detachment of the corner; (b and
c) thin cracks on the wall [6, 15].
8.5 Final remarks

The counteraction of creep damage in massive structures is a topical subject, as
many historical constructions, mainly towers or large columns and walls, are in
hazardous conditions. The bed joint reinforcement technique revealed its potential
in preserving those structures from collapse, as it has been widely experimen-
tally investigated with small diameter reinforced steel bars applications and use
Figure 8.29: Crack pattern detected in the western and eastern sides of the bell-
tower of Monza and scheme of the combination of the interventions
in the mostly damaged portions.
Figure 8.30: On-site insertion of reinforced bars into the joints [6].
of compatible embedding materials. The possible introduction of FRP materials,

especially for small rectangular section bars, seems to be promising in similar con-
texts, even if in some peculiar aspects it needs deeper investigation (local behav-
iour at interface levels, anchorage systems, sensitivity to high temperatures, use of
resins with low compatibility with original materials, etc.).
The technique is pretty easy to be performed in-situ and does not require par-
ticular skills or special tools. The proper selection of strengthening material is a
crucial point of the intervention, especially for the possible application on existing
historic masonry buildings. The use of high-strength products as embedding mate-
rial can be inappropriate, especially in the case of epoxy resins, due to the more
brittle behaviour both at local (interfaces with bricks and with the reinforcing bar)
and global level. Finally, symmetric interventions are recommended, in order to
optimize the global performances of the strengthened walls.
References
[1] Binda, L. & Anzani A., The time-dependent behaviour of masonry prisms:
an interpretation. The Masonry Society Journal, 11(2), pp. 1734, 1993.
[2] Anzani A., Binda, L. & Mirabella Roberti, G., The effect of heavy persis-
tent actions into the behaviour of ancient masonry. Materials and Structures,
33(228), pp. 251261, 2000.
[3] Binda, L., Anzani, A. & Gioda, G., An analysis of the time-dependent
behaviour of masonry walls. 9th International Brick/Block Masonry
Conference, Berlin, 1991, Vol. 2, pp. 10581067, 1991.
[5] Binda, L., Modena, C., Valluzzi, M.R. & Zago, R., Mechanical effects of
bed joint steel reinforcement in historic brick masonry structures. Struc-
tural Faults + Repair 99, 8th International Conference and Exhibition,
London, England, 1315 July 1999 (CD-ROM).
[6] Binda, L., Modena, C., Saisi, A., Tongini Folli, R. & Valluzzi, M.R., Bed
joints structural repointing of historic masonry structures. 9th Canadian
Masonry Symposium Spanning the Centuries, Fredericton, New Brunswick,
Canada, 46 June 2001 (CD-ROM).
[7] Valluzzi, M.R., Binda, L. & Modena, C., Mechanical behavior of historic
masonry structures strengthened by bed joints structural repointing. Con-
struction and Building Materials, 19(1), pp. 6373, 2004.
[8] Valluzzi, M.R., Tinazzi, D. & Modena, C., Strengthening of masonry
structures under compressive loads by FRP strips: local-global mechanical
behaviour. Science and Engineering of Composite Materials, Special Issue,
12(3), pp. 203218, 2005.
cathedral bell-tower. Construction and Building Materials, Special Issue,

16(7), pp. 385395, 2002.
[10] Valluzzi, M.R., Disar, M. & Modena, C., Bed joints reinforcement of
masonry panels with CFRP bars. Proc. of the International Conference
on Composites in Construction, Rende (CS), Italy, September 2003, eds
D. Bruno, G. Spadea & N. Swamy, pp. 427432, 2003.
[11] Valluzzi, M.R., Tinazzi, D., Garbin, E. & Modena, C., FEM modelling
of CFRP strips bond behaviour for bed joints reinforcement technique.
Proc. of STRUMAS VI, Computer Methods in Structural Masonry 6,
September 2003, Rome, Italy, eds T.G. Hughes & G.N. Pande, Computers
& Geotechnics Ltd: Swansea, pp. 149155, 2003.
[12] Saisi, A., Valluzzi, M.R., Binda, L. & Modena, C., Creep behavior of brick
masonry panels strengthened by the bed joints reinforcement technique us-
ing CFRP thin strips. Proc. of SAHC2004: IV Int. Seminar on Structural
Analysis of Historical Constructions Possibilities of Experimental and
Numerical Techniques, Padova, Italy, November 2004, pp. 837846, 2004.
[13] Garbin, E., Moro, L., Valluzzi, M.R. & Modena, C., Influence of high tem-
perature on the bed joint reinforcement of brick masonry by CFRP bars.
Proc. of CCC2005 Composites in Constructions International Conference,
Lyon, France, July 2005 (CD-ROM).
[14] Tinazzi, D., Valluzzi, M.R., Bianculli, N., Lucchin, F., Modena, C. &
Gottardo, R., FRP strengthening and repairing of masonry under compres-
sive load. Proc. 10th International Conference on Structural Faults and
Repair, London, 13 July 2003, (CD-ROM), Engineering Technical Press:
Edinburgh, 2003.
[15] Valluzzi, M.R., Casarin, F., Garbin, E., da Porto, F. & Modena, C., Long-term
damage on masonry towers: case studies and intervention strategies. 11th
International Conference on Fracture, Turin (Italy), 2025 March 2005
(CD-ROM).
[16] Binda, L., Modena, C. & Valluzzi, M.R., Il restauro del campanile del Duomo
di Monza: scelte di progetto e tecniche dintervento. Arkos Scienza e
Restauro dellArchitettura, 1, Genn./Marzo 2003, pp. 4453, 2003.
CHAPTER 9
Simple checks to prevent the collapse of heavy

historical structures and residual life prevision
through a probabilistic model
L. Binda, A. Anzani & E. Garavaglia

Department of Structural Engineering, Politecnico di Milano, Milan, Italy.
9.1 Introduction
A survey on approximately 60 Italian ancient towers, carried out to collect infor-
mation on the most common symptoms of structural decay, particularly on the
crack patterns, is described in this chapter. A simple procedure, aimed at achieving
a pre-diagnosis of the tower structures and at identifying the cases where more
detailed investigations are to be designed for defining the safety of the structure, is
proposed, which local institutions might adopt.
The vulnerability assessment of historic buildings towards the effects of persis-
tent loading and the achievement of a reliable lifetime estimate on a deterministic
basis is often very complex. Therefore, an attempt has been made at solving the
problem through a probabilistic approach.
9.2 The safety of ancient towers

Masonry towers are largely diffused in Italy and Europe and constitute an important
part of our heritage which requires suitable protection. The problem of their safety,
which became particularly evident after the collapses of monumental buildings
mentioned in previous chapters [1, 2], has to be tackled by local public institutions,
which not always can afford very costly diagnostic techniques.
On the other hand, before large investments are made for monitoring and/or
designing a repair intervention, a preliminary screening should be carried out to
qualify the buildings depending on the estimated degree of damage. For this purpose,
some help might come if guidelines were provided by local authorities, so that a
classification of the state of the towers based on direct observation could be rou-
tinely performed.
As an example, the results of an investigation carried out on some Italian towers
are presented and a simple procedure to analyse masonry towers is tentatively
proposed.
9.2.1 A survey on Italian cases
A survey of about 60 ancient Italian towers has been carried out, to collect infor-
mation on the most common symptoms of structural decay, particularly on the
crack patterns. The analysed towers present various degrees of damage from an
almost complete integrity to a very serious situation; therefore a classification has
been necessary to categorize some classes with similar characteristics [3].
Different types of towers have been observed, which are qualified by different
functions and show specific features that are typical of their group: family-towers,
bell-towers, boundary wall-towers, house-towers. A data base has been built with
charts and forms, where the information collected for any single tower has been
recorded.
In particular, geometric data and notes on the material and texture characteris-
tics of the masonry have been gathered. An accurate photographic survey has been
performed so as to report on the main cracks in the building facades, indicating,
when possible, the ones which cross the wall thickness. Historical information on
the buildings has been collected to know their load history and any particular event
which could have modified it.
In some cases, experimental data on the state of stress were available because
some towers are being monitored by private or public research organizations. In
the other cases, when the geometry was known, the vertical stress at ground level
was estimated assuming a density of 18 and 20 kN/m3, respectively, for brick and
stone masonry.
To get information on possible instability risks, the critical load has been calculated
as a percentage of the acting dead load, assuming a Young modulus of 200 N/mm2
according to the values previously detected on several masonries by use of flat
jacks. Two calculations have been done assuming two limit conditions depending
on the assumed mutual constraints between orthogonal walls: on the one hand
perfect continuity has been assumed, on the other hand any single wall has been
considered as isolated. Of course this hypothesis is far more pessimistic but it is
also on the safe side.
9.2.2 Comments on the observed crack patterns
Different kinds of crack patterns have been observed on the examined towers which
might be due to different causes. Of course only a direct experimental investiga-
tion with a complete non-destructive and/or destructive evaluation could provide
Checks to Prevent the Collapse of Heavy Historical Structures 207
a satisfying diagnosis but such very costly survey can not always be afforded.
Nevertheless, some interesting considerations can be made based on the available
information, which have more easily been obtained, and a comparison between
similar cases can be done given also some experience which has been gained by
previous investigations.
Examples of two kinds of damages will be presented trying to connect, when pos-
sible, the observed cracks to the experimental knowledge on long-term behaviour of
masonry.
In Figs 9.19.4 two leaning towers are shown: the Garisenda family-tower
(Bologna, twelfth century) and the S. Maria del Pero bell-tower (Monastier
[Treviso] tenth century), both made by brick masonry. The first one is 47.5 m high,
with a base of 8 8 m and three-leaf walls 2.35 m thick at ground level. The sec-
ond one is 45 m high, with a base of 12.9 12.9 m and solid walls 3.5 m thick at
ground level.
In both cases all four fronts of the tower are badly cracked: it is interesting to
note that the sloping face is the most damaged one, with a concentration of flaws
at the base level and at mid-height in the case of the Garisenda and a multitude of
vertical cracks in the case of S. Maria del Pero. The walls orthogonal to the direc-
tion of the slope are also cracked vertically, as if the wall had split into two parts
which sheared mutually: the crack path always follows a preferential way, which
runs along a line of openings.
Figure 9.1: Garisenda family-tower (Bologna, twelfth century), north, west, south
and east fronts.
Figure 9.2: Bell-tower of S. Maria del Pero (Monastier, tenth century).
In Figs 9.59.8 two towers damaged by compression effects are shown: the bell-
tower of the Duomo of Monza (sixteenth century) [4] (see also Chapter 1) and the
Coronata family-tower (Bologna, thirteenth century) both of them made again by
brick masonry. The first one is 78.5 m high, with a base of 9.8 10.3 m and solid
walls 2.1 m thick at ground level and at present is subjected to a reinforcing inter-
vention. The second one is 61 m high, with a base of 9 9 m and three-leaf walls
2.8 m thick at ground level.
In both cases vertical cracks are visible. In the case of Monza cracks positioned
at the corner indicate that the wall thickness is possibly split. Many cracks are also
present in the lower part for a height of about 30 m; the most damaged face is the
western one, where cracks crossing the wall thickness rise up to about 15 m also
cutting the bricks. In the case of the Coronata Tower, the most damaged face is the
southern one where a concentration of cracks is visible just above the stone
basement and a single long crack runs up for about 12 m along a line of open-
ings. In general, vertical cracks are always present whenever a compressive stress
concentration takes place, the path of which is greatly influenced by the masonry
texture.
Figure 9.3: Garisenda family-tower Figure 9.4: Bell-tower of S. Maria del

(Bologna, twelfth century): detail of Pero (Monastier, tenth century): detail
a crack. of a crack.
9.2.3 Elaboration of the collected data
After the calculations described above, a comparison between the obtained values
and those corresponding to the collapsed Tower of Pavia has been made.
In Fig. 9.9 the distribution of average vertical stress at ground level is shown,
where it appears that in most cases the stress level varies between 0.25 and
0.75 MPa, but in some cases it is higher than that of Pavia.
In Fig. 9.10 the vertical stress is shown as a function of the tower height. Of
course a linear relationship appears, with house-towers and boundary wall-towers
showing the lowest values of the considered parameters. The stress value of the
Tower of Pavia is shown again.
In Fig. 9.11 the average stress value is related to the ratio between the tower
height and the wall thickness where again, despite some scatter in the data, a linear
relationship can be observed.
In Fig. 9.12 the critical loads, calculated according to the two assumptions pre-
viously mentioned, are plotted as a percentage of the dead load versus the tower
heights. Again the levels corresponding to the Tower of Pavia are indicated. It is
interesting to notice that in the case of Pavia very low values of this ratio have been
Figure 9.5: Bell-tower of the Duomo Figure 9.6: Coronata family-tower

of Monza (sixteenth century): west (Bologna, thirteenth century): east
and east fronts. and south fronts.
obtained, even in the case of the more optimistic assumption when a value not
much greater than 1 has been obtained.
As far as the other towers are concerned, it has to be observed that unfortunately
for many of them (the tallest ones) the analysed ratio lies below the Pavia level and
also below 1. Of course this is not equal to a statement of a critical situation for
these towers, but certainly it might be enough to recommend that a deeper evalua-
tion be carried out.
9.3 A probabilistic model for the assessment

of historic buildings
The results of the creep and pseudo-creep tests described in Chapter 2 have been
interpreted by means of a probabilistic model, based on the definition of a random
Figure 9.7: Bell-tower of the Duomo Figure 9.8: Coronata family-tower

of Monza (sixteenth century): detail (Bologna, thirteenth century): detail on
on a crack. a crack.
50
Pavia Tower
4 Bell-Towers
40
Frequency [%]
Family-Towers
3
30 Boundary Wall-Towers
v [MPa]
House-Towers
20 2
10 1
Pavia Tower
0
-0.5 0.0 0.5 1.0 1.5 2.0 2.5 0
0 20 40 60 80 100 120
v [MPa] H [m]
Figure 9.9: Bar chart of the average Figure 9.10: Average stress value at
vertical stress at ground level. ground level versus tower height.
variable as a significant index of vulnerability, and on the solution of the clas-

sic problem of reliability in stochastic conditions. A comparison between verti-
cal and horizontal strain-rate is put forward and the application of the proposed
procedure to the Tower of Monza is attempted. Aim of the research is to provide a
4 1E+2
Bell-Towers
Family-Towers
3 1E+1
v [MPa]
Boundary wall-Towers c) Pavia Tower
Pcr /P
House-Towers
2 1
Pavia Tower Pavia Tower
1 0.1
i)
0 1E-2
0 10 20 30 40 50 0 40 80 120
H/S H [m]
Figure 9.11: Average stress value at Figure 9.12: Critical load over
ground level versus the ratio between acting dead load versus tower height:
tower height and the wall thickness. c represents collaborating walls and
i represents isolated walls.
mathematical model able to predict possible failures of heavy masonry structures

due to long-term damage, allowing preventive repair interventions.
By experimental evidence, the strain evolution connected with a given stress
history of a viscous material such as a historic masonry can be described through
. .
the parameters ev and eh, respectively, defined as the vertical and horizontal strain-
rate taken on the linear branches of the strain versus time diagrams shown by the
specimen at the stress level s, remaining constant for a certain time interval.
For each s the high randomness connected with the changing of strain-rate, due
.
to the high non-homogeneity of the masonry here tested, brings to consider e as a
random variable with a certain distribution of values (Fig. 9.13).
The diagrams, referred to the horizontal strain components, show an initial
pseudo-constant branch with low values of the strain-rate, followed by a clearly
increasing part that ends with the collapse [5].
Following this way, the deformation process can be interpreted as a stochastic
.
process of the random variable e. The strain-rate also depends on the stress level s
corresponding to which the deformation is recorded. Therefore, for each stress
.
level s the strain-rate e (measured in e/s) can be modelled with a probability den-
. .
sity function (PDF) fE (e, s) that results to be dependent on the stress s and on the
.
strain-rate e. The experimental measurements are taken at discrete stress values
. . .
s*, the PDF fE. (e, s) becomes fE. (e, s*) and is dependent on the strain-rate e and
on s* which is constant at every step. Therefore the modelling of the strain-rate
. .
behaviour depends only on the random variable e. In order to model fE. (e, s*), at
every stress level s* a family of theoretical distributions has to be chosen. No
doubt that, in the choice of a distribution modelling a given phenomenon has to be
connected not only to the physical aspects of the phenomenon itself but also to the
characteristics of the distribution function in its tail, where often no experimental
data can be collected. This last aspect of the matter can be investigated by analysing
.
the behaviour of the immediate occurrence rate function fE. (e, s*) connected with
the chosen distribution function. On this subject more details are described in [6]
and [7].
0.005
0.0045
0.004
0.0035
h [( h 10 3 )/sec]
0.003
f . ( , )
.
0.0025

_.
h = 0.002
0.002
.
0.0015 _.
h = 0.001
0.001 _.
h = 0.0005
0.0005
0
1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 2.8 3
* v [N/mm2]
Figure 9.13: Interpolation of the horizontal strain-rate versus applied stress and the
.
modelling of fE. (e, s*).
The physical knowledge of the phenomenon has shown a relationship between

the secondary creep strain-rate and the residual life of the material [8, 9]. In view
of the preservation of historic buildings from major damage or even failure, it
would be very convenient to indicate a critical value of the strain-rate under which
the residual life of the building is conveniently greater than the required service
.
life. Here a conventional value of e may be assumed as a critical value indicating
a safety limit. Consequently, for a given stress level s* the probability to record
the critical strain-rate connected with the secondary creep safety limit increases
.
if the strain-rate e increases. Therefore, it seems correct to assume that, at a given
stress level s* the higher is the strain-rate, the higher
. . is the
. probability
. that the
secondary creep strain-rate falls in the interval {E < e E + dE}. The assumed
hypothesis, as a satisfied (but not unique) physical interpretation of the decay
.
process, leads to model the strain-rate e at the stress level s* with a Weibull dis-
tribution (Fig. 9.13).
This family of distributions presents an. immediate occurrence rate function.
.
fE. (e, s*) which increases if the value of E increases and tends to if E ;
this fact seems to respect the physical interpretation of the strain-rate behaviour
previously commented.
For a correct probabilistic modelling of the process, the parameters involved in
the chosen PDF of the considered random variable need to be estimated. For a
.
probabilistic modelling of e, the maximum likelihood method, applicable through
a computer code, is preferable to the least squares method.
It is furthermore interesting to evaluate the probability for the system of reach-
.
ing or exceeding a given strain-rate level e over a stress history. This probability
.
can be seen as the shadowed area above e as shown in Fig. 9.14 and [5, 810]. This
. .
area can be calculated by using the survive function E. (e, s*) = 1 FE. (e, s*)
.
where FE. (e, s*) is the cumulative distribution function of the density chosen. The
.
calculation of E. (e, s*) is possible with the use of any kind of computer code for
.
numerical integration. For different strain-rate levels e h, the survive function has
been evaluated for all stress levels s*. The calculated values allow the plotting of
an experimental fragility curve connected to each chosen strain-rate levels (see
Fig. 9.15, Section 9.4.1 and [7, 10]). Since the experimentally measured strain-
rates only refer to a discrete number of stress levels, it would be quite convenient
to have a suitable tool capable to predict, though in probabilistic terms, the system
behaviour at any stress level. Following this approach the deterioration process
can be treated as a reliability problem [11, 12]. Indeed the reliability R(t) concerns
the performance of a system over time and it is defined as the probability that the

system does not fail during the time t. Here this definition is extended and R(s)
is assumed as the probability that a system exceeds a given significant strain-rate
.
e with a stress s. The random variable that is used to quantify reliability is which
0.005
0.0045
0.004
0.0035
h [( h 10 3 )/sec]
0.003
f . ( , )
.
0.0025

0.002
.
0.0015
0.001 _.
h = 0.0005
0.0005
0
1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 2.8 3
* v [N/mm2]
.
Figure 9.14: Exceedance probability to cross the threshold e .
.
is just the stress to exceed strain-rate e . Thus, from this point of view, the reliabil-
ity function is given by [11, 13]
R (s ) = Pr( > s ) = 1 F (s ), (9.1)

where F (s) is the distribution function for and represents the theoretical model-
ling of the experimental fragility curves.
In order to model the experimental fragility curves and to evaluate F (s), a
Weibull distribution has been chosen [5, 10, 13].
In fact, this distribution seems to be a good interpretation of the physical phe-
nomenon: the larger the stress level, the higher the probability that a critical strain-
.
rate e , connected with the creep phenomenon, will happen for s value included in

the next ( + d) interval. Therefore, distributions with the function (s) increas-
ing with s and tending to as s are needed. Also this time the Weibull dis-
tributions satisfy this requirement.
The fitting of the experimental fragility curves with the Weibull distribution has
been made using an optimization procedure for the parameters estimation through
a computer code involving the least squares method. Since in this case the param-
eters estimation is made on cumulative distributions the least squares method is
preferable to the maximum likelihood method.
9.4 Fragility curves from the experimental data

.
9.4.1 Fragility curve e versus s applied to creep tests
Referring to the creep tests described in Chapter 2 and considering the strain ver-
sus time evolution, the interpolation of the strain-rate versus stress and the model-
.
ling of fE.(e, s*) have been plotted as shown in Fig. 9.13. On it, as possible critical
.
strain-rate values, three different e h have been identified connected to the initiation
of the secondary creep phase [14]. In Fig. 9.15 the experimental and theoreti-
cal fragility curves connected with horizontal strain-rate thresholds are reported.
They describe the probability to exceed the critical thresholds as a function of the
reached stress level sv.
If applied to real cases, this type of prediction allows to evaluate, for instance,
the results of a monitoring campaign on a massive historic building subjected to
persistent load and to judge whether the creep strain indicates a critical condition
in term of safety assessment (see Section 9.5).
Of course, the precocious recognition of a critical state will allow designing a
strengthening intervention to prevent total or partial failure of the construction.
9.4.2 Comparison between vertical and horizontal strain-rate
Referring to a previous research that analysed the experimental results obtained in

the vertical direction [5], possible critical values of the vertical strain-rate, connected
1 F (v)
0.9

0.8 h = 0.0005
0.7
0.6
h = 0.001
0.5
0.4
0.3
h = 0.002
0.2
0.1
0
1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 2.8 3
v [N/mm2]
Figure 9.15: Horizontal strain-rate: experimental () and theoretical () fra-

gility curves.
to the initiation of the secondary creep may also be indicated. Therefore, to com-
pare the experimental results obtained in the horizontal and vertical directions,
the same vertical and horizontal strain-rates thresholds have been chosen. In
Fig. 9.16 the experimental and theoretical fragility curves connected with the
previous thresholds are reported for the vertical strain-rates.
In Tables 9.1 and 9.2 a comparison between vertical and horizontal strain-rate in
term of exceedance probability is reported.
Comparing these results it can be observed that always the exceedance of the
chosen threshold strain-rate is performed at a lower stress level in the case of
horizontal strain. This is quite in agreement with the dilatant behaviour of ancient
masonry when approaching failure, as shown by Fig. 2.13a, where the horizontal
strain appears to be higher and developing at a higher rate than the vertical ones.
This is also confirmed by Fig. 2.15b where the crack pattern of a prism at the end
of the test is shown: having loaded the specimen vertically in compression, the
cracks follow a mainly vertical path, therefore giving an apparent horizontal dila-
tion. However, from a probabilistic point of view caution must be offered to tails
of distribution where usually not much data are present and where the prediction
becomes critical.
.
9.4.3 Fragility curve e versus s applied to pseudo-creep tests
Also in the case of pseudo-creep tests as those shown in Fig. 2.17, possible criti-
cal strain-rate values, connected to the initiation of the secondary creep phase,
1 F (v)

0.9
0.8
0.7
0.6
0.5
_.
0.4 v = 0.001
0.3
_. _.
0.2 v = 0.0005 v = 0.002
0.1
0
1 1.2 1.4 1.6 1.8 2 2.2 2.4
v [N/mm2]
Figure 9.16: Vertical strain-rate: experimental () and theoretical () fragility

curves.
.
Table 9.1: Probability to exceed ev = 0.0005 for different sv.
Exceedance probability
.
of ev (%) sv (N/mm2)
10 1.40
63 1.81
90 2.00
.
Table 9.2: Probability to exceed eh = 0.0005 for different sv.
.
of eh (%) sv (N/mm2)
10 1.38
63 1.75
90 1.90
were found. In Fig. 9.17 the experimental and theoretical fragility curves con-
.
nected with the threshold e h = 1.00e 005 are reported.
Table 9.3 shows that for sv = 4.22 N/mm2 63% of population (samples) could
be already failed.
The same observation made in the previous paragraph can apply here.
1 F (v)
0.9
0.8
0.7
0.6 _.
h = 1.00e-005
0.5
0.4
0.3
0.2
0.1
0
0 1 2 3 4 5 6 7 8 9 10
v [N/mm2]
Figure 9.17: Experimental () and theoretical () fragility curves.
.
Table 9.3: Probability to exceed eh for different sv.
.
Exceedance probability of eh(%) sv (N/mm2)
10 2.37
63 4.22
90 5.25
9.4.4 Comparison between vertical and horizontal strain-rate
In Fig. 9.18 the experimental and theoretical fragility curves connected with the
.
threshold e h = 1.25e 005 are reported for both vertical (Fig. 9.18a) and horizontal
(Fig. 9.18b) strain-rates. Comparing these results with those reported in Tables 9.4
and 9.5, a tendency similar to that shown by the creep test is confirmed, but here
some additional comments have to be made. It can be observed that in 10% of the
cases the exceedance of the chosen threshold strain-rate is performed at a lower
stress level in the case of vertical strain.
On the contrary, in 63 and 90% of the cases the exceedance of the chosen thresh-
old strain-rate is performed at a lower stress level in the case of horizontal strain,
exactly like in the creep test. In fact, at low stress values the material response is
still viscoelastic; therefore, the higher strain is shown in the loading direction.
This was not evident in the case of creep tests because the load history of the
different prisms was not the same and less regular data were available, therefore
(a) (b)
1 F (v) _. 1 F (v)
0.9 v = 1.25e-005 0.9
0.8 0.8 _.
h = 1.25e-005
0.7 0.7
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10
v [N/mm2] v [N/mm2]
Figure 9.18: Experimental (, ) and theoretical () fragility curves: (a) vertical

strain; (b) horizontal strain.
.
Table 9.4: Probability to exceed ev = 1.25e 005 for different sv.
.
of ev(%) sv (N/mm2)
10 2.35
63 4.30
90 5.40
.
Table 9.5: Probability to exceed eh = 1.25e 005 for different sv.
.
of eh(%) sv (N/mm2)
10 3.00
63 4.23
90 4.75
on a statistical basis the phenomenon was not evident. At higher stress level,
approaching failure, the tendency is inverted: the material response turns into vis-
coplasticity with visible crack appearance corresponding to dilatant behaviour, as
shown by Fig. 2.17. Again, this is confirmed also by the crack patterns (Fig. 2.19).
9.5 Application to the bell-tower of Monza

The proposed probabilistic approach has been applied for the first time to a real
case trying to evaluate the results of the monitoring of a massive historic building
subjected to persistent load. The bell-tower of Monza, a sixteenth-century struc-

ture built in solid brickwork masonry, had suffered major and diffused cracks due
to high compression (Fig. 9.5).
After the constitution of a Technical Committee in 1976 the building, together
with the Cathedral, was subjected to a systematic control, setting up 31 fixed bases,
7 of which were on the Tower corresponding to the major cracks. The bases had a
length of about 400 mm, and the measurements were taken, starting in January
1978, every month during the first three years and every three months subsequently.
The instrument used was a millesimal deformometer. After the recorded increase
of the cracks aperture on the tower and an anomalous geometry recorded in the
Cathedral, in 1992 a new committee was constituted by Politecnico di Milano which
installed a static control system that included the continuation of the geometric
evaluation of the cracks. Each of the previous bases was substituted by a couple of
new bases placed above and below the other ones, having a length of 200 mm, the
readings being taken at the same periodicity. Considering the data collected until
1999 (Fig. 9.19), the influence of thermal variation on the crack opening can
clearly be observed.
Nevertheless, tracing a regression line between the data, a neat increase of the
aperture in time is visible; in the case of the base shown in Fig. 9.19 a rate of 6.48
m/year can be measured until 1986 and a higher rate of 24.94 m/year can be
measured subsequently, clearly indicating a worsening of the Tower static condi-
tions. After the static survey, a consolidation intervention on the Tower was
required, which is still in progress.
In order to compare the rate of crack opening of the Tower with the strain-rate
measured in the laboratory, the monitoring readings were divided by the base
length and the rate in m/mm over seconds was calculated.
400
300
200
aperture variation [m]
100
-100
-200
6.48 m/year
-300 24.94 m/year
-400
19 8
19 9
19 0
19 1
19 2
19 3
84
19 5
19 6
19 7
19 8
19 9
19 0
19 1
19 2
19 3
19 4
19 5
19 6
19 7
19 8
20 9
00
7
7
8
8
8
8
8
8
8
8
8
9
9
9
9
9
9
9
9
9
9
19
19
years
Figure 9.19: Monitoring of the bell-tower of Monza: opening variation of a crack

versus time.
1 F (v)
0.9
_.
= 1e-9
0.8
0.7
0.6
0.5 _.
= 2e-9
0.4
0.3
0.2
0.1
0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
[N/mm2]
Figure 9.20: Experimental (, ) and theoretical () fragility curves.
Relating this calculated rate with the values of the vertical stress, locally mea-
sured by flat-jack tests at the same height of the crack monitoring, the fragility
curves shown in Fig. 9.20 were built for the Tower cracks for two different thresh-
.
olds e In this case the strain-rate recorded is lower than the strain-rate obtained by
creep and pseudo-creep laboratory tests.
Here the modelling appears a hazard, but on the basis of the results obtained by
laboratory tests it is possible to suppose the same distribution (Weibull distribu-
tion) for modelling the probability of exceedance of the thresholds chosen.
Although two data are not sufficient to investigate the Weibull shape, the results
obtained look like an interesting example of possible application of the procedure
to real cases, where more suitable data were available.
9.6 Conclusions
An accurate visual survey based on experimental and analytical experience has
been shown to give important information on the state of damage of massive build-
ings like towers. Typical major or diffused cracks were found capable of giving
preliminary information on the state of damage of these buildings. The collapse of
a badly damaged structure happens generally very rapidly, but it may be avoided
if the damage symptoms as cracks, deformations, etc. are carefully taken into
account as early as possible. To this purpose the following procedure is suggested

as a preliminary investigation, which public institutions may adopt:
Geometrical survey and preparation of drawings of the building using simple

tools like photography and the subsequent restitution by software for image
elaboration;
Representation of the crack pattern visually detected over the geometrical
restitution; measure of the wall thickness.
Interpretation of the crack pattern and recognition of its causes: tilting, effect of
dead load, others;
Rough calculation, based on geometrical data, density of the material and the
assumption of homogeneous and elastic material, of the maximum stress value;
Rough calculation of the critical load value under the same hypothesis.
These preliminary results can indicate the possible need of a more detailed inves-
tigation and/or of periodically repeated surveys.
A probabilistic model has also been applied to the study of the long-term behaviour
of masonry specimens subjected in laboratory to creep and pseudo-creep. The chosen
model seems to appropriately interpret the experimental results also capturing the
passage between viscoelastic and viscoplastic behaviour. An attempt of applying the
procedure to the results of long-term monitoring of displacements, particularly of
crack opening, was carried out. The research will continue with the aim of providing
a tool for preventing the masonry failure under particular states of stress.
Acknowledgements
Architects E. Bertocchi and D. Trussardi are gratefully acknowledged for creating
the database on Italian towers. Dr. G. Cardani is gratefully acknowledged for the
data elaboration. Thanks are given also to ISMES for providing information on the
towers. The research was carried out with the support of MIUR Cofin 2000.
References
[2] Binda, L., Anzani, A. & Mirabella Roberti, G., The failure of ancient towers:
problems for their safety assessment. Conf. on Composite Construct.
Conventional and Innovative, IABSE, Insbruck, pp. 699704, 1997.
[3] Binda, L. & Anzani, A., The safety of ancient masonry towers: a survey on
the effects of heavy dead loads. Int. Conf. on Studies in Ancient Structures,
Istanbul, Turkey, pp. 207216, 1997.
[4] Modena, C., Valluzzi, M.R., Tongini Folli, R. & Binda L., Design choices
Cathedral Bell-Tower. Construction & Building Materials, 16(7), pp. 385

395, 2002.
[5] Anzani, A., Binda, L. & Garavaglia, E., The vulnerability of ancient build-
ings under permanent loading: a probabilistic approach. Proc. of 2nd Int.
Symp. ILCDES 2003, Integrated Life-Cycle Design of Materials and Struc-
tures, Kuopio, Finland, 13 December 2003, ISSN 0356-9403, ISBN 951-
758-436-9 Helsinki, Finland, UE., I, pp. 263268, 2003.
[6] Garavaglia, E., Lubelli, B. & Binda, L., Service life modelling of stone and
bricks masonry walls subject to salt decay. Proc. of Integrated Life-Cycle
Design of Materials and Structures, ed. A. Sarja, RILEM/CIB/ISO, Pro 14,
Technical Research Center of Finland (VTT); Helsinki, 1, pp. 367371,
2000.
[7] Binda, L., Garavaglia, E. & Molina, C., Physical and mathematical model-
ling of masonry deterioration due to salt crystallisation. Proc. of 8th Int.
Conf. on Durability of Building Materials and Components, eds M.A.
Lacasse & D.J. Vanier, NRC-CNRC: Ottawa, I, pp. 527537, 1999.
[8] Taliercio, A.L.F. & Gobbi, E., Experimental investigation on the triaxial
fatigue behaviour of plain concrete. Magazine of Concrete Research,
48(176), pp. 157172, 1996.
[9] Anzani, A., Binda, L. & Mirabella Roberti, G., The effect of heavy persis-
tent actions into the behaviour of ancient masonry. Materials and Struc-
tures, 33, pp. 251261, 2000.
[10] Garavaglia, E., Lubelli, B. & Binda, L., Two different stochastic approaches
modeling the deterioration process of masonry wall over time. Materials
and Structures, 35, pp. 246256, 2002.
[11] Evans, D.H., Probability and Its Applications for Engineers, Marcel Dekker,
Inc.: New York, NJ, USA, 1992.
[12] Melchers, R.E., Structural Reliability Analysis and Prediction, Ellis Horwood
Ltd: Chichester, West Sussex, England, 1987.
[13] Bekker, P.C.F., Durability testing of masonry: statistical models and methods.
[14] Garavaglia, E., Anzani, A. & Binda, L., Probabilistic model for the assess-
ment of historic buildings under permanent loading. J. of Materials in Civil
Engineering, ASCE, USA, 18(6), pp. 858867, 2006.
Conclusions
This book is probably the first one completely dedicated to the study of long-term
behaviour of massive historic buildings. Although the creep behaviour of materials
such as rocks, concrete and steel has been well known for a long time, very little
research has been carried out on the behaviour of masonry. The fact that historic
masonry walls are usually characterized by very thick sections, to better diffuse the
dead load and hence reduce the stresses due to compression, has caused a lack of
interest in the long-term behaviour of masonry structures.
Collapses, such as the one of the S.Marco bell-tower in Venice and others that
occured previously, were never interpreted as caused by long-term progressive
damage, but rather by high stress or some other cause never connected to time.
Only after the collapse of the Civic Tower in Pavia, and the following experimental
research, was the phenomenon of continuous damage due to heavy loads taken into
account (Chapter 1).
The importance of persistent loads in the damage of historic masonry has been
studied experimentally and their effects on the mechanical properties of the material
have been shown. Constant load step (pseudo-creep) tests turned out to be a suitable
procedure for analysing creep behaviour, having the advantage of being carried
out more easily than long-term tests. Primary, secondary and tertiary creep phases
have been clearly observed, together with their relationship with the stress level, a
damage development being associated with an increase in the stress level.
Considering the last load step for each specimen tested with pseudo-creep tests,
the secondary creep rate, which is the strain rate during the phase of stable damage
growth, has been calculated before collapse and then related to the duration of the
last load step, which can be regarded as the residual life of the material. A strong
correlation exists between creep time to failure and secondary creep rate which,
accordingly, can be used as a reliable parameter to predict the residual life of a
material element subjected to a given sustained stress. In view of preserving the
historical heritage, it would be useful to define similar relationships to evaluate, for
instance, the results of a monitoring campaign on a massive historic building
subjected to persistent load, to judge whether the creep rate indicates a critical
condition in terms of safety assessment. Of course, the precocious recognition of a
critical state will allow one to design a strengthening intervention to prevent total or
partial failure of the construction.
226 LEARNING FROM FAILURE
The action of cyclic and persistent loads has proved to cause severe damage to the
mechanical properties of ancient masonry. The fatigue life of masonry under uni-axial
cyclic compression was also related to the secondary creep strain rate, which is the
strain rate during the phase of stable cyclic damage growth (Chapter 2).
Concerning the possibility of carrying out creep tests in the long-term, it can be
concluded that creep tests on ancient masonry prisms should be performed by
applying the load in successive steps, at a given time interval, starting from a low
stress level. In this way, a thorough description of the viscous behaviour of the
material can be obtained. Creep tests in which the load is applied in a single step are
inadequate in the case of ancient masonry due to the high scatter in the mechanical
properties and to the small number of specimens usually available.
The time period between successive load steps should be sufficiently long to
extinguish primary creep. In fact, the evolution of different stress levels of the strain
rate associated with secondary creep can only be evaluated in such a way. From the
results obtained on regular masonry prisms tested, a minimum time period under
sustained loading of 70 to 80 days should be adopted (Chapter 3). Finally, it should be
stressed that secondary creep was found to initiate at 60 to 70% of the compressive
strength. In the case of soft-stone masonry the secondary creep can initiate even at
40% of their compressive strength (Chapter 5).
Much more complicated is the modelling of the phenomenon, even if the approach
by fracture mechanics is suggested by some authors. Authors such as N. Shrive
et al. suggest that the step-by-step in time analysis is a powerful tool to investigate
the change of stresses due to creep over a large time range. Stresses in a material
can rise and fall due to the effects of creep (or fall and then rise). Although the
effective modulus method using the final creep coefficient can accurately estimate
the final stresses in the components of a composite material (e.g. masonry) due to
creep, the method may not predict intermediate peak stresses accurately: one needs
to know when they will occur (Chapter 4).
The numerical model developed by E. Papa and A. Taliercio to analyzse damage
effects induced by heavy persistent loads in existing masonry buildings is to be
considered as an effective tool to assess the safety of these structures. Both the
predicted time to failure of Pavia Civic Tower and the predicted crack pattern in the
bell-tower of Monza Cathedral were found to match experimental evidences fairly
well. In particular, the predicted failure mechanisms for the analyzsed structures
may give indications regarding the effectiveness of possible strengthening techniques.
On the accuracy of the finite element (FE) model, there is a point that will have
to be tackled in future researches, regarding the mesh-sensitivity of the numerical
solution. It is well known that, in the presence of a strain-softening constitutive law,
strains localize in narrower bands as the FE mesh is refined. A non-local version
of the presented damage model will be formulated, according to proposals of other
researchers, to overcome this limitation of the present model and make numerical
predictions unaffected by the mesh size.
Other problems that will be tackled in the continuation of the research to improve
the numerical model are (1) the distinction between cracks activated in tension or
compression and (2) the so-called unilateral effect (see Chapter 7).
LEARNING FROM FAILURE 227
A probabilistic model has also been applied by E. Garavaglia, A. Anzani and
L. Binda to the study of the long-term behaviour of masonry specimens subjected in
laboratory to creep and pseudo-creep. The chosen model seems to appropriately
interpret the experimental results also capturing the passage between visco-elastic
and visco-plastic behaviour. An attempt at applying the procedure to the results of
long-term monitoring of displacements, particularly of crack opening, was carried
out. The research will continue with the aim of providing a tool for preventing
masonry failure under particular states of stress (Chapter 9).
One of the most difficult tasks is to understand when the damaged structure can
reach such a level of risk so that a deep investigation or intervention is needed. An
accurate visual survey based on experimental and analytical experience has been
shown to give important information on the state of damage of massive buildings
like towers. Typical major or diffused cracks were shown to be capable of giving
preliminary information on the state of damage of these buildings. The collapse of
a badly damaged structure happens generally very rapidly, but it may be avoided if
the damage symptoms such as cracks, deformations, etc. are carefully taken into
account as early as possible. To this purpose the following procedure is suggested
as a preliminary investigation which public institutions may adopt:
Geometrical survey and preparation of drawings of the building using
simple tools like photography and the subsequent restitution by software
for image elaboration;
Representation of the crack pattern visually detected over the geometrical
restitution; measure of the wall thickness.
Interpretation of the crack pattern and recognition of its causes: tilting,
effect of dead load, etc.;
Rough calculation, based on geometrical data, density of the material and
the assumption of homogeneous and elastic material, of the maximum stress
value;
Rough calculation of the critical load value under the same hypothesis.
These preliminary results can indicate the possible need for a more detailed
investigation and/or for periodically repeated surveys.
In this case an exhaustive investigation on site and in laboratory as the one
described in Chapter 1 should be carried out, after which a decision can be made as
to start a monitoring of the structure (Chapter 6), or in case of short-term risk, to
provide an intervention design.
In case of monitoring, the measurements obtained (crack openings,
displacements, accelerations etc.) will include all the effects caused by static, cyclic,
dynamic loads, soil settlements, etc. bundled into a single response; interpreting the
results requires the mixed responses to be broken down into cyclic and reversible
processes and monotonic (or cumulative) ones. A further step consists of breaking
measurements down into the components associated with different effects acting on
the structure (wind, temperature, earthquakes, traffic, etc.). The latter specifically
requires monitoring of the actions themselves by means of appropriate equipment
(thermometers, hygrometers, anemometers, seismometers etc.).
228 LEARNING FROM FAILURE
The techniques to be applied in the case of long-term or fatigue damages are

well described in Chapter 8. No heavy interventions are needed; the aim is to confine
the dilation of the masonry; therefore steel tie rods at different heights can tackle
instability of the walls. Bed joint reinforcements in the most damaged areas have
high potential in preserving those parts from collapse. The technique is pretty easy
to be applied on site and does not require special skill or special tools and it is not
so far destructive. The proper selection of strengthening materials is a crucial point
of intervention.
The research on long- term behaviour of masonry structures will still continue,
particularly on the experimental and modelling procedures. Nevertheless, it is
worthwhile to inform the professionals and Cultural Heritage curators of a
phenomenon which can affect in a serious way, and particularly in seismic areas,
the life of our massive structures.
...for scientists by scientists
Structural Studies, Repairs and Maintenance of

Heritage Architecture X
Edited by: C.A. BREBBIA, Wessex Institute of Technology, UK
The importance of architectural heritage for the historical identity of a region, town
or nation is now widely recognized throughout the world. In order to take care of our
heritage we need to look beyond borders and continents to benefit from the
experience of others and to gain a better understanding of our cultural background.
Featuring contributions from the Tenth International Conference on Structural
Studies, Repairs and Maintenance of Heritage Architecture, this book covers a broad
spectrum of topics including: Heritage Architecture and Historical Aspects; Regional
Architecture; Structural Issues; Seismic Vulnerability Analysis of Historic Sites;
Maintenance; Seismic Behaviour and Vibrations; Surveying and Monitoring;
Material Characterization; Material Problems; Protection and Prevention; Simulation
Modeling; Environmental Damage; Assessment and Retrofitting; Preservation and
Prevention; Historical Dockyards, Shipyards and Buildings; Underwater Heritage;
Surveying Techniques; Rivers, Lakes, and Canals Heritage; Site Protection, Oral
Traditions and Stories.
WIT Transactions on The Built Environment, Vol 95
ISBN: 978-1-84564-085-9 2007 736pp 238.00/US$475.00/357.00
WITPress
Ashurst Lodge, Ashurst,
Southampton,
SO40 7AA, UK.
Tel: 44 (0) 238 029 3223
Fax: 44 (0) 238 029 2853
E-Mail: witpress@witpress.com
The Great Structures in Architecture

From Antiquity to Baroque
F.P. ESCRIG, Universidad de Sevilla, Spain
Starting in antiquity and finishing in the Baroque, this book provides a complete
analysis of significant works of architecture from a structural viewpoint. A
distinguished architect and academic, the authors highly illustrated exploration will
allow readers to better understand the monuments, get closer to them and to explore
whether they should be conserved or modified.
Contents: Stones Resting on Empty Space; The Invention of the Dome; The Hanging
Dome; The Ribbed Dome; A Planified Revenge Under the Shadow of Brunelleschi;
The Century of the Great Architects; The Omnipresent Sinan; Even Further;
Scenographical Architecture of the 18th Century; The Virtual Architecture of the
Renaissance and the Baroque.
Series: Advances in Architecture, Vol 22
ISBN: 1-84564-039-X 2006 272pp 95.00/US$170.00/142.50
The Revival of Dresden

Edited by: W. JGER, Technical University of Dresden, Germany and
C.A. BREBBIA, Wessex Institute of Technology, UK
In 1945 the ancient City of Dresden was destroyed by massive bombardments and
much of its rich architectural heritage appeared to have been obliterated forever.
Over the last half-century, however, Dresden has been lovingly reconstructed with the
active collaboration of its citizens. This process, now culminating in the rebuilding of
the Frauenkirche (the Church of Our Lady) is documented in this unique book.
Partial Contents: THE REVIVAL OF THE CITY: The Contribution of
Preservationists to the Reconstruction of the Semper Opera House; Restoration of the
Castle in Dresden; The Reconstruction of Taschenberg Palace; The Conservation of
the Neustadt District as Part of the Cultural Cityscape. THE FRAUENKIRCHE:
The Citizens Initiative to Promote the Rebuilding; A Construction of Stone and Iron
Structural Concept for Reconstruction of the Dresden Frauenkirche; Structural
Proof-Checking Using a Complete 3D FE-Model.
ISBN: 1-85312-787-6 2000 272pp 97.00/US$159.00/145.50
Environmental Deterioration of Materials

Edited by: A. MONCMANOVA, Slovak Technical University, Slovakia
This book deals with the fundamental principles underlying the environmental
degradation of widely used and economically important construction materials.
The invited contributions cover aspects such as the deterioration mechanisms of
materials and metal corrosion, environmental pollutants, micro- and macro-climatic
factors affecting degradation, the economic impact of damaging processes, and
fundamental protection techniques for buildings, industrial and agricultural facilities,
monuments, and culturally important objects. Basic details of ISO standards relating
to the classification of atmospheric corrosivity and low corrosivity of indoor
atmospheres are also included.
Designed for use by materials, corrosion, civil and environmental engineers,
designers, architects and restoration staff, this book will also be a useful tool for
managers from different industrial sectors and auditors of environmental management
systems. It will also be a suitable complementary course book for university students
in all of the above disciplines.
ISBN: 978-1-84564-032-3 2007 336pp 98.00/US$188.00/147.00
The Conservation and Structural Restoration of

Architectural Heritage
G. CROCI, University of Rome La Sapienza, Italy
The book should be seen and known about by all engineers and architects
who are developing their work in the field.
THE STRUCTURAL ENGINEER
...instructive and fascinating.... The excitement and challenges of preserving and
stabilizing historic buildings is captured by this very readable book.
JOURNAL OF ARCHITECTURAL CONSERVATION
Designed for use by all professionals involved or interested in the preservation of
monuments, the purpose of this book is to contribute to the development of new
approaches in the area.
Many of the examples examined, including the Colosseum, the Tower of Pisa and the
Pyramid of Chephren, are the result of work carried out during Giorgio Crocis
distinguished career.
Featuring numerous black and white photographs and illustrations by the author, the
text is divided into two main sections entitled The Scientific Approach to the Study
of Architectural Heritage and Structural Analysis of Masonry Buildings.
ISBN: 1-85312-482-6 1998 272pp 148.00/US$237.00/222.00
Structures Under Shock and Impact IX

Edited by: N. JONES, The University of Liverpool, UK and C.A. BREBBIA,
Wessex Institute of Technology, UK
This book contains the papers presented at the Ninth International Conference on
Structures Under Shock and Impact.
The shock and impact behaviour of structures is a challenging area, not only because
of the obvious time-dependent aspects, but also because of the difficulties in
specifying the external dynamic loading characteristics for structural designs and
hazard assessments and in obtaining the dynamic properties of materials. Thus, it is
important to recognise and utilise fully the contributions and understanding emerging
from theoretical, numerical and experimental studies on structures, as well as
investigations into the material properties under dynamic loading conditions.
Featured topics include: Impact and Blast Loading Characteristics; Material
Response to High Rate Loading; Missile Penetration and Explosion; Protection of
Structures from Blast Tools; Behaviour of Structural Concrete; Structural Behaviour
of Composites; Interaction between Computational and Experimental Results; Energy
Absorbing Issues; Structural Crashworthiness; Structural Serviceability Under Impact
Loading; Seismic Engineering Applications.
WIT Transactions on The Built Environment, Vol 87
ISBN: 1-84564-175-2 2006 592pp 190.00/US$335.00/285.00
Computational Mechanics for Heritage Structures

B. LEFTHERIS, Technical University of Crete, Greece, M.E. STAVROULAKI,
Technical University of Crete, Greece, A.C. SAPOUNAKI, Greece and
G.E. STAVROULAKIS, University of Ioannina, Greece
This book deals with applications of advanced computational-mechanics techniques
for structural analysis, strength rehabilitation and aseismic design of monuments,
historical buildings and related structures.
The authors have extensive experience working with complicated structural analysis
problems in civil and mechanical engineering in Europe and North America and have
worked together with architects, archaeologists and students of engineering.
The book is divided into five chapters under the following headings: Architectural
Form and Structural System; Static and Dynamic Analysis; Computational
Techniques; Case Studies of Selected Heritage Structures; Restoration Modeling
and Analysis.
Series: High Performance Structures and Materials, Vol 9
ISBN: 1-84564-034-9 2006 288pp+CD-ROM 130.00/US$234.00/195.00

Luigia Binda-Learning From Failure - Long-Term Behaviour of Heavy Masonry Structures - WIT Press (2008) PDF

Загружено:

Сведения о документе

Оригинальное название

Авторское право

Доступные форматы

Поделиться этим документом

Поделиться или встроить документ

Параметры публикации

Этот документ был вам полезен?

Это неприемлемый материал?

Авторское право:

Доступные форматы

Luigia Binda-Learning From Failure - Long-Term Behaviour of Heavy Masonry Structures - WIT Press (2008) PDF

Загружено:

Авторское право:

Доступные форматы

Learning from Failure

Long-term Behaviour of Heavy

For USA, Canada and Mexico

British Library Cataloguing-in-Publication Data

A Catalogue record for this book is available

Library of Congress Catalog Card Number: 2007922340

The texts of the papers in this volume were set

WIT Press 2008

Printed in Great Britain by Athenaeum Press Ltd.

Preface ............................................................................................................ xiii

Failures due to long-term behaviour of

L. Binda, A. Anzani & A. Saisi

1.2 The collapse of the Civic Tower of Pavia:

1.2.1 Description and historic evolution of the tower

1.2.2 First experimental results and interpretation of the failure causes

1.2.2.1 Structure and morphology of the walls

Figure 1.3: Photogrammetric survey of the remains of the tower.

Figure 1.4: Cross-section of the wall of the Civic Tower of Pavia.

Figure 1.6: View of the complete

1.2.2.2 Geotechnical investigation

1.2.2.3 Physical, chemical and mechanical tests on the components

1.2.2.4 Compression tests on masonry prisms

Figure 1.7: se curves of Figure 1.8: se curve obtained for a cyclic

1.2.3 Long-term tests

1.2.3.1 Fatigue tests

Figure 1.9: se curve obtained during Figure 1.10: se and et curves

1.2.3.2 Constant load tests

1.3 Long-term behaviour of masonry structures

1.3.1 Deformation during mortar hardening

Figure 1.14: Pillar of the Cathedral of Salisbury [11].

Figure 1.16: Church of St. Maria la

Figure 1.17: Dependency of strain on the loading rate.

Figure 1.18: First, secondary and tertiary creep.

The influence of time on the mechanical behaviour of masonry structures under

1.4 Collapse and damage of towers due to long-term

1.4.2 The bell-tower of Monza Cathedral and the Torrazzo of Cremona

The bell-tower of the Cathedral of Monza is a sixteenth-century building made of

at a height of 30 m, together with a damaged zone at a height of 1125 m with a

1.5 The role of investigation on the interpretation of the

1.5.1 The bell-tower of the Cathedral of Monza

The bell-tower of the Cathedral of Monza, is a masonry structure 70 m high, with

first floor at 11 m. No damages were reported in other known calamities, such as

Figure 1.24: Single flat-jack tests of Monza.

(a) 3.00 (b) 3.00

1.5.2 The Torrazzo of Cremona

2.50 inner walls (rooms)

Figure 1.26: Single flat-jack tests of Cremona.

(a) 4.00 (b) 4.00

Figure 1.27: Stressstrain plot at (a) 7.2 m and (b) 19 m height.

1.6 Comparison between the two towers

[12] Taupin, J.L., Rflexions sur la cathedrale Saint-Pierre de Beauvais.

Experimental researches into long-term

A. Anzani1, L. Binda1 & G. Mirabella Roberti2

thermo-hygrometric conditions and especially designed testing apparatus, a more

failure limit curve

viscosity limit curve

Figure 2.1: Pseudo-creep testing: simulation of monotonic test at an equivalent

2.2 Tests on the masonry of the Civic Tower of Pavia

(a) (b) (c)

Figure 2.3: Civic Tower of Pavia before failure.

(a) (b) (c)

2.2.1 Characterization by sonic tests

2.2.2 Monotonic tests on prisms of different dimensions