Академический Документы
Профессиональный Документы
Культура Документы
Written discussion
D. A. Ponniah, BSc, MEng, PhD, MBA, CEng, MICE, and D. J. Prentice, BEng, PhD closes 29 May
1998
j As a result of recent European Community 2. Recent studies6-8 have attempted to
directives, it is necessary to reassess a large evaluate the significance of the interaction
number of arch bridges in the UK to ascertain between the arch and the fill. It has been shown
their load-carrying capacity. Current methods through both experimental and theoretical work
of assessment can be conservative, and that the fill significantly affects the capacity of
improvements to these are urgently required. the arch. These studies conclude that when the
The effect of the overlying fill on the load- effect of the fill is taken into account the capacity
carrying capabilities of an arch bridge is an of the arch is significantly greater than that
important aspect that is still not yet fully predicted by the methods mentioned above.
understood. This paper addresses some of Depending on the soil type, depth of fill and the
these issues through the presentation of the configuration of the arch, the load-carrying
results from a series of load tests on a double- capacity can be enhanced by up to 40%.9
span model arch bridge. The model arch was However, only a relatively small proportion of
constructed in timber with a dry granular fill, this work has been specifically concerned with
and approximately 90 tests were carried out to multi-span bridges,10,11 particularly when
investigate the effects of load position and fill compared to the percentage in existence. In
depth. A limited number of tests were also view of the limited research on these structures,
undertaken to quantify parameters such as the this investigation was initiated to provide a
density of fill, the presence of structural greater understanding of the interaction
backing and the effect of the secondary arch. A between the arch barrels and the surrounding
regression analysis of the results was carried fill.
out and an expression for the failure load of 3. The research can be carried out by a
arch bridges has been derived. This is number of means, such as theoretical methods,
validated against full-scale-test data from a field tests and model tests. While theoretical
number of previous arch-bridge tests. methods can be used to investigate many
variables quickly and cheaply, such analyses are
Keywords: brickwork & masonry; bridges; limited in use because of the idealizations
models (physical) required. Alternatively, field tests are real and
provide realistic data, but are enormously D. A. Ponniah,
expensive and relevant only to a limited number Senior Lecturer,
Introduction of parameters. Model tests incorporating much Department of
Civil &
In recent years, a considerable amount of effort of the real structure can be used to investigate a
Environmental
has gone into research concerning the load- wide range of parameters within a reasonable
Engineering,
carrying capacity of arch bridges.1¹3 The reason cost and time-scale. While this paper describes a University of
for this upsurge in interest can be largely programme of small-scale testing, it is only Edinburgh
attributed to recent European Community through a combination of these techniques that
directives which state that bridges must be the full picture can emerge.
capable of taking an increased vehicle weight of 4. The aims of the investigation were:
40 tonne by the year 1999. However, the majority
(a) to examine the effects of various para-
of arch bridges in the UK were constructed
meters, such as the fill depth and load
between the 17th and 19th centuries,4 and were
position, on the failure load of a double-span
‘designed’ to carry loads much less than those
arch
on current road networks. As a result, these
(b) to compile a databank of relevant informa-
bridges are required to be reassessed to
tion to allow a regression analysis to be
ascertain their present-day load-carrying
undertaken on the failure-load results.
capacity. The common methods of assessment,
D. J. Prentice,
such as the MEXE5 and the mechanism
Research Fellow,
method,1 have been demonstrated to be Experimental methods Department of
unreliable when compared to load tests on full- Civil &
scale bridges,4 and this has been largely Description of the model Environmental
attributed to the lack of understanding of the 5. A semi-circular double span arch was Engineering,
contribution made by the fill material constructed to the dimensions shown in Fig. 1. University of
surrounding the arch. The proportions were initially selected to Edinburgh
81
Downloaded by [ University of Sussex] on [11/09/16]. Copyright © ICE Publishing, all rights reserved.
PONNIAH AND PRENTICE
Load Counterbalance
platen
Weights
added in Solid timber
increments abutment
82
Downloaded by [ University of Sussex] on [11/09/16]. Copyright © ICE Publishing, all rights reserved.
LOAD-CARRYING
CAPACITY OF MASONRY
ARCH BRIDGES
11. As detailed in the previous section, an Table 1. Details of the test programme
alternative fill was also used for a series of six
Parameter investigated Fill depth (corresponding load positions): mm
tests to quantify the effect of sand grading on the
overall failure load and collapse mechanism. Main study 5, 10, 22·5, 35, 50, 75 (1–10)
This sand was a mixture of equal quantities of Single-span arch 5 (3); 10 (2); 22·5, 35 (6); 50, 75 (4)
the sand described above and a finer dry silica Stiff ‘wedge’ backing 5 (3); 10 (2); 22·5, 35 (6); 50, 75 (4)
sand. A comparison of the particle-size Loose ‘rubble’ backing 5 (3); 10 (2); 22·5, 35 (6); 50, 75 (4)
distributions is shown in Fig. 4 with the Alternative fill 5 (3); 10 (2); 22·5, 35 (6); 50, 75 (4)
alternative fill (type B) more widely graded than
the original fill used in majority of the tests
(type A). The uniformity coefficient for sand
type A is considerably smaller than that for the
alternative sand type B, with values of 1·59 and
8·2, respectively.
12. The alternative fill was placed and
compacted using the same method as that
described above. The average in situ density
using this material was measured at 1693 kg/m3
with a variation of 61%. This represents an
increase of 12% from the density of the original
sand, which is as expected due to the wider
grading of the material. A series of shear box
tests was carried out on samples of varying
densities in order to establish the relationship
with the angle of shearing resistance f. The Fig. 3. Load positions
corresponding value of f was 418, which is coefficient,14 which is used to establish the and system of reference
appropriate for a dense, rounded and well- correlation between two variables x and y, and is
graded sand13 as in this case, but is not signifi- given by
cantly higher than the original sand type A. P
ðx ¹ x̄Þð y ¹ ȳÞ
r¼ P P ð1Þ
Regression analysis Î ðx ¹ x̄Þ2 ð y ¹ ȳÞ2
13. For large sets of experimental data, Its value lies between ¹1 and þ1 and gives an
establishing relationships between the indication of the linearity of the relationship.
dependent and independent variables is often The closer the data are to a linear relationship,
desirable or even necessary, but can be the nearer the correlation coefficient is to ¹1 or
reasonably difficult. Regression analysis is a þ1. If there is almost no association between x
means of ascertaining whether such a and y, the coefficient will be close to zero.
relationship exists between a determinant and 14. Whilst correlation demonstrates how
its associated independent parameters. A much association there is between two
common method used in scientific work is the variables, linear regression analysis is used to
Pearson product moment correlation find the best equation which describes the
Downloaded by [ University of Sussex] on [11/09/16]. Copyright © ICE Publishing, all rights reserved.
PONNIAH AND PRENTICE
Fig. 6. Variation in
failure load with load
position and depth of fill
84
Downloaded by [ University of Sussex] on [11/09/16]. Copyright © ICE Publishing, all rights reserved.
LOAD-CARRYING
CAPACITY OF MASONRY
ARCH BRIDGES
Fig. 8. Comparison of
single to double span
failure loads
85
Downloaded by [ University of Sussex] on [11/09/16]. Copyright © ICE Publishing, all rights reserved.
PONNIAH AND PRENTICE
single-span arch is consistently greater than substantial structural bond between the spans.
that of the double-span one by an average of 20%. Two types of backing were used to model these
At the lower fill depths of 5 and 10 mm, there is two situations. The first was formed using a
not such a significant difference in the failure solid timber wedge up to half the height of the
loads for the two models. This verifies the point crown and cut to the precise dimensions of the
made above; that is, due to the reduced depth of model so that it was situated in direct contact
the overlying fill, the double-span tends to with the extrados of the two arches. The second
behave as a single arch. However, it can be was modelled using timber arch voussoirs,
concluded that at depths of fill greater than which were again placed up to half the height of
approximately 3% of the span, the failure load of the crown. The height of backing chosen was
a single span is significantly greater than that of not specific as there is some variation in the
a similarly proportioned double span. This is extent of haunching used in old bridges.11
due to the increase in the overall flexibility of a 25. Figure 9 shows the comparison of the
double span, which is clearly much greater than original failure loads with those from the
a single span and which may develop a failure addition of both the stiff and loose backing,
hinge mechanism earlier. which resulted in an increase in the failure load
21. The Department of Transport give at all depths of fill. The stiff wedge produces a
recommendations on multi-span arch bridge greater increase, with an average value of 20%
assessment in clause 6.26(i) of BD21/93 Part being recorded, whereas the loose voussoirs
3,16 and state that: result in an increase of 13%. Once again the
effect is more pronounced at greater depths of
Any individual span of the bridge may be assessed as a
fill. It can thus be concluded that increasing the
single span arch provided the adjacent intermediate
supports and spans are structurally adequate. overall lateral stability or stiffness of the arch
system through the inclusion of structural
This is qualified by stipulating in the next part of backing results in an enhanced failure load.
the clause that structural adequacy is achieved if 26. This effect is described in clause 6.28 of
no tension occurs at ultimate limit state (ULS) in BD21/93 Part 3,16 where it is stated that:
the adjacent span or supports to the span being
Such construction details [backing] have the effect of
loaded. However, clause 6.29 allows the
raising the line of the horizontal thrust on the adjacent
assessing engineer the freedom to disregard the
arch thereby reducing the likelihood of any tension
above recommendations if they are considered occurring at the top of that arch.
unnecessary, when assessing a structure with
short, stocky intermediate piers. In this case, it This is a recognition by the Department of
is permissible to assess each span of the bridge Transport that the addition of backing between
as a series of single-span arches. the arches can have an advantageous effect on
22. This is not consistent with the results the strength of the arch. The results presented
observed in the tests presented in this paper in this paper are thus consistent with this
which highlight the importance of the flexibility statement.
of the arch–fill–pier system on the estimated 27. The existence of backing or haunching
failure load. If the arch system is sufficiently in a multi-span system of arches is an important
rigid then it would be appropriate to assess the parameter that must be considered in any
multi-span bridge as a series of single spans. assessment. Unfortunately, it is often very
Alternatively, if the interaction extends to more difficult to ascertain the existence of this as
than one arch then there would be a reduction in arches are mostly assessed without the luxury
the failure load, which could be as much as 20%. of the removal of the fill to determine such
The assumption that a stock pier can simulate an details. However, it is clear that consideration
abutment is perhaps an oversimplification, should be given to the additional benefit in
because the interaction may occur through the terms of the increased strength that haunching
fill and not only through the pier and arch of this nature could provide. The addition of
barrels. structural backing between arches may also
prove to be a viable repair option to
Effect of ‘backing’ on the failure load understrength arches and could result in
23. Many multi-span arch bridges were reasonable cost savings when compared with
constructed with the addition of structural the cost of replacing the structure.
backing between adjacent arches to increase the
strength of the overall structure.11 Additional Effect of fill density on the failure load
tests were therefore undertaken in an attempt to 28. The fill material has been shown to be an
quantify the effect this has on the failure load important element of the arch system. When
and hinge mechanism of the model. considering the properties of the material the
24. Haunching of this nature was two most influential parameters are the friction
traditionally formed by either placing loose angle, which is a strength-related parameter,
rubble masonry blocks between the arches or and the density, which is related to the stiffness.
using mortar or concrete to construct a more As part of this investigation it was decided to
86
Downloaded by [ University of Sussex] on [11/09/16]. Copyright © ICE Publishing, all rights reserved.
LOAD-CARRYING
CAPACITY OF MASONRY
ARCH BRIDGES
identify the effect density has on the failure load, methods of assessment. In addition, the
without significantly altering the friction angle. replacement of the fill with a more dense
This was achieved by using a wider grading of material could also be viewed as a viable option
sand which increased the density from 1515 to for repair of an understrength arch.17
1693 kg/m3 . A number of tests were repeated
with the alternative sand, and the corresponding Derivation of regression equations
failure loads were considerably greater than for 30. Regression analysis was undertaken on
the original material (Fig. 10). As described in the test results to derive a relationship between
the previous sections, the effect is again more the failure load W and the parameters
pronounced at greater depths of fill. The higher investigated within this investigation. The
failure loads are therefore attributed to the effects of varying the arch span, the ring
increase in density of the fill, where an increase thickness and the span/rise ratio were
of 12% produces an average increase in failure ascertained from the results of a previous
load of 40%. similar experimental test programme by
29. The effect of the fill is already Fairfield.12 In this programme a total of 120 tests
incorporated in some methods of assessment, were carried out on a single-span semi-circular
but the density of the fill is only considered as an arch and a span/rise ratio of 4, with a span of
additional dead weight on the structure. The 700 mm and a barrel thickness of 35 mm. The
possible effects on the load-carrying capacity combined set of results covered the majority of
from the fill density are not catered for. the range of parameters likely to be experienced
However, further quantification of this in arch-bridge assessment.
relationship would be required before 31. From regression analysis the following
amendments could be recommended to current relationship was obtained:
Downloaded by [ University of Sussex] on [11/09/16]. Copyright © ICE Publishing, all rights reserved.
PONNIAH AND PRENTICE
W ¼ 2·45h ¹ 65·7ðxs Þ0·1 67·0na þ 45·6sr þ 6·71st 5·31, 10·69 and 4·65, respectively. The signifi-
ð4Þ cance of the parameter is evaluated by compari-
son with the critical value calculated as 2·13 for
where W is the failure load (N), h is the depth of 95% confidence limits and four degrees of free-
fill at the crown (mm), xs is the distance from the dom. The depth of fill h is clearly the most signi-
left springing to the load point (mm), na is the ficant parameter, with a value of 18·25, but all
number of arches, sr is the span/rise ratio and st the parameters have t-values greater than the
is the span/thickness ratio. critical value and are therefore also statistically
32. The values estimated from equation (4) significant.
versus the actual failure load are plotted in
Fig. 11, with the best-fit line drawn at 458. The
coefficient of determination R2 for this Application of regression equation to full scale
relationship was 79%, indicating a good field tests
correlation. The values of t indicating the 33. Over the last 10 years the Transport
significance of any particular parameter were Road Laboratory has undertaken a number of
estimated for h, xs , na , sr and st as 18·25, 10·3, tests to failure on redundant arch bridges.4 In
Table 2. Results from the regression analysis of the full-scale field tests
Bridge Span: Rise: Ring h: xs na sr st Width Actual Estimated Difference:
location m m thickness: mm of bridge: failure: failure: %
m m kN kN
Bridgemill 18·3 2·85 0·711 203 2·32 1 6·4 25·7 8·3 3100 1195 61·4
Bargower 10 5·18 0·558 1200 2·25 1 1·9 17·9 8·7 5600 4936 11·8
Preston 4·95 1·64 0·36 380 2·1 1 3 13·8 5·7 2110 1057 49·9
Prestwood 6·55 1·43 0·22 165 2·1 1 4·6 29·8 3·8 228 179 21·4
Torksey 4·9 1·15 0·343 246 2·04 1 4·3 14·3 7·8 1080 1047 3·1
Shinafoot 6·16 1·18 0·39 215 2·08 1 5·2 15·8 7 2524 909 64
Strathmashie 9·42 2·99 0·6 410 2·17 1 3·2 15·7 5·8 1325 1176 11·3
Barlae 8·53 1·69 0·45 295 2·15 1 5 19 9·8 2900 1656 42·9
Rothm road 8·9 2·3 0·50 710 2·2 3 3·9 18·4 6·5 3500 2140 38·8
88
Downloaded by [ University of Sussex] on [11/09/16]. Copyright © ICE Publishing, all rights reserved.
LOAD-CARRYING
CAPACITY OF MASONRY
ARCH BRIDGES
this section the results obtained from the load fill of 12% produced an average increase in
tests are compared with the estimates obtained failure load of 40%. Both this and/or the
using the regression equation derived from the addition of backing may be viable repair
series of small-scale tests. The bridges and their options to understrength arches.
characteristics are summarized in Table 2. Also (e) Regression analysis was carried out to
included in this table is the measured failure obtain an expression for the failure load of
load, the estimated failure load from the arch bridges, which can be used to obtain
regression equation and the percentage initial approximate estimates.
difference between the two loads. The estimated
loads have been amended to allow for both the Acknowledgements
length of the applied surface load and the width The authors would like to thank the EPSRC
of the arch under investigation, since the and the Highways Agency for funding this
solution to the equation is based on the research. In addition, the technical staff of the
dimensions of the timber model. Department of Civil and Environmental
34. As can be seen from Table 2, some Engineering at the University of Edinburgh and
reasonable correlation has been obtained. Messrs Scott Baker, Brian McCall, George
Although all the estimated failure loads are less McKinlay and Eric Hutchinson are gratefully
than the actual values, no clear trends are appar- acknowledged.
ent in the comparisons. It must be noted that the
regression equation was derived from model References
arch tests, of which the majority had a span/rise 1. HEYMAN J. The Masonry Arch. Horwood,
ratio of 2, without spandrel walls and without Chichester, 1982.
surfacing. When it is considered that the field 2. CHOO B. S., COUTIE M. G. and GONG N. G. Finite
tests were carried out on masonry arches with element analysis of masonry arch bridges using
span/rise ratios between 1·9 and 6·4, which tapered elements. Proceedings of the Institution of
encompassed a variety of spandrel walls and Civil Engineers, 1991, 91, 755–770.
3. SMITH F. W., HARVEY W. J. and VARDY A. E. Three-
black-tops, and containing undocumented fill
hinge analysis of masonry arches. The Structural
types, the comparisons are quite favourable. Engineer, 1990, 68, No. 11, 203–207.
35. In conclusion, in this section we have 4. PAGE J. Masonry Arch Bridges – State of the Art
presented the practical application of a Review. HMSO, London, 1993.
regression equation derived from the results of 5. MEXE. Military Load Classification of Civil Bridges
the small-scale timber tests. If it were possible to by Reconnaissance and Correlation Methods. Mili-
incorporate the more specific details of an arch tary Engineering Experimental Establishment,
bridge as described above, the estimates made Christchurch, 1963 (SOLOG study B38).
using the regression equation could be 6. PRENTICE D. J. and PONNIAH D. A. New techniques
enhanced. Further improvement of the in the elastic analysis of arch bridges using image
processing. Journal of the British Society for Strain
regression equation is also planned with a series
Measurement, 1996, 139–144.
of tests to identify the effects of the span/rise 7. FAIRFIELD C. A. and PONNIAH D. A. Model tests to
ratio and the fill density. Nevertheless, the determine the effect of fill on buried arches.
authors believe that this method provides the Proceedings of the Institution of Civil Engineers,
basis for a technique that will ultimately provide Structures and Buildings, 1994, 104.
the practising engineer with a straightforward 8. PONNIAH D. A., FAIRFIELD C. A. and PRENTICE D. J.
means of obtaining an initial estimate of the Fill stresses in a new brick arch bridge subject to
failure load for both single- and multi-span heavy axle-load tests. Proceedings of the Institution
arches. of Civil Engineers, Structures and Buildings, 1997,
122, 173–185.
9. PRENTICE D.J. An Appraisal of the Geotechnical
Conclusions Aspects of Multi-span Masonry Arch Bridges. PhD
36. The conclusions drawn from this study Thesis, University of Edinburgh, 1996.
are as follows: 10. MELBOURNE C., GILBERT M. and WAGSTAFF M. The
behaviour of multi-span masonry arch bridges.
(a) The lowest failure loads for a multi-span Proceedings of the 1st International Conference on
arch occur around the region of midspan of Arch Bridges, Bolton, 1995.
the loaded arch. 11. HARVEY W. J. and SMITH F. W. The behaviour and
(b) On average, a 20% increase in failure load assessment of multispan arches. The Structural
was recorded for the similarly proportioned Engineer, 1991, 69, No. 24, 411–417.
single-span arch than for the double-span 12. FAIRFIELD C. A. Soil–Structure Interaction in Arch
Bridges. PhD thesis, University of Edinburgh,
arch.
1994.
(c) The addition of a solid wedge simulating the
13. LAMBE T. W. and WHITMAN R. V. Soil Mechanics.
backing between the arches increased the Wiley, New York, 1969.
failure load by, on average, 20%. A similar 14. OWEN F. and JONES, R. Statistics. Pitman, London,
loose backing produced an average 1990.
increase of 13%. 15. RYAN B. F. and JOINER B. L. MINITAB Handbook.
(d) An increase in the density of the overlying Duxbury Press, Belmont, CA, 1994, 3rd edn.
89
Downloaded by [ University of Sussex] on [11/09/16]. Copyright © ICE Publishing, all rights reserved.
PONNIAH AND PRENTICE
16. DEPARTMENT OF TRANSPORT. The Assessment of 17. FAIRFIELD C. A. and PONNIAH D. A. A method of
Highway Bridges and Structures. Department of increasing arch bridge capacity economically.
Transport, London, 1993, Departmental Standard Proceedings of the Institution of Civil Engineers,
BD 21/93. Structures and Buildings, 1996, 116, 109–115.
90
Downloaded by [ University of Sussex] on [11/09/16]. Copyright © ICE Publishing, all rights reserved.