Engineering Structures 32 (2010) 361–372

Contents lists available at ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Effects of support settlement on continuous deep beams and STM modeling

Ning Zhang, Kang-Hai Tan ∗
School of Civil and Environmental Engineering, Nanyang Technological University, 50 Nanyang Avenue, Singapore 639798, Singapore

article info abstract

Article history: This paper investigates the effects of differential support settlement on the strength and behavior of
Received 29 March 2008 two-span Continuous Deep Beams (CDBs). An experimental program consisting of six specimens has
Received in revised form been carried out with spring supports to investigate the effects of differential settlement and the web
24 September 2009
reinforcement on the beam behavior. Test results including crack patterns, load–deflection curves, strains
Accepted 26 September 2009
Available online 17 October 2009
in steel, serviceability loads and ultimate loads are presented and discussed with the effects of support
stiffness and web reinforcement. It is shown that settlement of the middle support significantly affects
Keywords:
the serviceability load, crack pattern and failure mode. However, for ultimate strength, the beams exhibit
Continuous deep beams considerable tolerance in accommodating differential support settlement, particularly for those with web
Concrete reinforcement. A direct Strut-and-Tie Model (STM) is proposed for beams subjected to support settlement.
STM Good agreement is found between the test results and STM predictions on support reactions and ultimate
Support settlement loads. The method can be easily implemented through an excel spreadsheet, thus making it an attractive
Web reinforcement tool for engineers.
Experimental program © 2009 Elsevier Ltd. All rights reserved.

1. Introduction moments and support reactions in continuous deep beams were

Differential settlement is commonly encountered in structural
and geotechnical engineering if the loads acting on the structure
are unevenly distributed to the supports, or the foundations are
resting on grounds with non-homogeneous soil properties. Even
in laboratory experiments, it is difficult to completely eliminate
differential settlement. When the middle support of a two-span
continuous deep beam settles more than that at either ends, a sag-
ging moment is formed over the middle support. Consequently,
the middle reaction force is relieved somewhat and the end re-
action forces are increased. Leonhardt and Walther [1] tested two
reinforced continuous deep beams, both of which failed in shear
at their interior shear spans. The test results showed that the dis-
tributions of bending moments and support reactions in continu-
ous concrete deep beams were substantially different from those
in continuous shallow beams. Rogowsky et al. [2] reported sev-
enteen two-span concrete deep beams tested to failure with a
central point load at each span. It was found that beams with ver-
tical web reinforcement showed more ductility and failed at much
higher loads than their counterparts without web reinforcement.
Chemrouk [3] conducted an experimental program consisting of
12 continuous deep beams. Experimental results confirmed earlier
findings by Rogowsky et al. [2] that the distributions of bending
∗ Corresponding author. Tel.: +65 679 052 97; fax: +65 679 166 97.
E-mail address: ckhtan@ntu.edu.sg (K.-H. Tan).
fundamentally different from those predicted by ordinary beam
theory. The end reactions and the mid-span moments were greater
than the predicted values by ordinary beam theory. Ashour [4] did
a study on continuous deep beams and concluded that the shear
capacity is more influenced by vertical web reinforcement than
horizontal web reinforcement. Differential support settlement is
commonly believed to be critical in continuous deep beams, and
may significantly affect the distribution of reactions and aggravate
the crack widths, thereby lowering the ultimate strength [4–6].
However, detailed studies about continuous deep beams subjected
to controlled differential support settlement are scarce. The effects
of support settlement still remain vague and not fully understood.
Due to the scarcity of test data, specimens with or without web re-
inforcement are designed with a range of middle support stiffness
K , to simulate varying degrees of differential support settlement.
The test program is also aimed to provide test validation for the
proposed strut-and-tie model, which incorporates the influence of
support settlement. Two variables are introduced in the authors’
experimental program, namely, relative middle support stiffness
K and web reinforcement ratio. The term K quantifies the vertical
stiffness of the middle support relative to the end supports:
R R
K = = (1)
∆ ∆B − (∆A + ∆C )/2
where R = the reaction force of middle support; ∆ = the relative
settlement of middle support; ∆A , ∆B , ∆C = the respective abso-
lute settlement of support A, B and C . K equals infinity if there is
no differential support settlement.

0141-0296/$ – see front matter © 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.engstruct.2009.09.019
362 N. Zhang, K.-H. Tan / Engineering Structures 32 (2010) 361–372

Table 1
Details of the specimens.
Beam h bw a le b c1 c c2 c ρs d ρh e ρv e fy f fyw g fc0 Support
notations (mm) (mm) (mm) (mm) (mm) (%) (%) (%) (N/mm2 ) (N/mm2 ) (N/mm2 ) conditionsh
End Middle

TCDB-1-1 750 180 2000 63 63 1.86 0.00 0.00 514.9 525.7 35.6 Rigid Rigid
TCDB-1-2 750 180 2000 63 63 1.86 0.00 0.00 514.9 525.7 36.1 Spring A Spring B
TCDB-1-3 750 180 2000 63 63 1.86 0.00 0.00 514.9 525.7 38.0 Rigid Spring B
TCDB-2-1 750 180 2000 63 63 1.86 0.36 0.28 514.9 525.7 38.9 Rigid Rigid
TCDB-2-2 750 180 2000 63 63 1.86 0.36 0.28 514.9 525.7 39.3 Spring A Spring B
TCDB-2-3 750 180 2000 63 63 1.86 0.36 0.28 514.9 525.7 38.5 Rigid Spring B
a
bw = beam width.
b
le = Effective span (support to support distance).
c
c1 , c2 = the distances from beam top to the top steel centroid and from beam soffit to the bottom steel centroid, respectively.
d
ρs = reinforcement ratio of longitudinal steel (top and bottom).
e
ρh , ρv = reinforcement ratios of horizontal and vertical web steel, respectively.
f
fy = yield strength of main steel.
g
fyw = yield strength of web steel.
h
Details of Spring A and Spring B are referred to Fig. 2.

2. Experimental program
2.1. Specimen details
A total of six concrete deep beams of dimension 180 mm
(width) × 750 mm (height) × 4400 mm (length) are fabricated. The
beam notation is typically written as ‘TCDB-A-B’, where TCDB is
an acronym for ‘‘two-span continuous deep beam’’. Two variables
are used to characterize the specimens, namely: the web reinforce-
ment configuration ‘A’ and the support configuration ‘B’. There are
two types of web reinforcement configuration ‘A’ with ‘‘1’’ indicat-
ing no web steel and ‘‘2’’ both horizontal and vertical web rein-
forcement. The second character ‘B’ has three variations. The first
type is a deep beam tested on three rigid supports (coded 1). The
second is a deep beam tested on three elastic supports (coded 2),
and the last refers to one with two rigid end supports and one elas-
tic middle support (coded 3). The three different support arrange-
ments lead to three different degrees of settlement at the middle
support. As an example, specimen TCDB-2-3 is provided with both
horizontal and vertical web reinforcement, and rested on rigid end
supports and an elastic middle support. The details of the speci-
mens and their respective support conditions are summarized in
Table 1. The beam size and longitudinal reinforcement are kept
the same for all six specimens, among which, the last three speci-
mens are with web reinforcement, whereas the first three without.
Two types of reinforcement are used in the specimens, namely,
deformed high tensile steel (T bar) for top and bottom longitudi-
nal reinforcement and plain round mild steel (R bar) for web rein-
forcement. The reinforcement details of the specimens are shown
in Fig. 1. The top and bottom faces of beams are reinforced with
four longitudinal T20 bars, arranged in two layers throughout the
whole length. The bottom longitudinal bars have a 90◦ bend at
both ends so as to have sufficient anchorage length. Local reinforce-
ment cages are provided at locations of loading and support points
to prevent premature crushing. The concrete mix is designed to
have 28-day compressive cube strength of 40 MPa. Chippings of
10 mm size are used as coarse aggregates to prevent congestion be-
tween bottom layers of reinforcement. Three cubes and cylinders
for each beam are prepared and cured under the same condition as
the beams. They are tested on the same day as the beams (Table 1).
2.2. Instrumentation and test setup
Spring elements A and B (Fig. 2) are specially fabricated for
the experimental program, with nominal spring stiffnesses of 68
and 101 kN/mm, respectively, as experimentally verified by the
manufacturer. It should be noted that spring elements with any
higher stiffness are very costly and are not used in the program. A
typical experimental set-up on three elastic supports is shown in
Fig. 3. This set-up (adopted for specimens TCDB-1-2 and TCDB-2-2)
is to produce a smaller differential support settlement compared
to solely using spring element B under the middle support. Two
point loads are applied from the top at each mid-span by two
actuators, which include inbuilt load cells and attached swivel
heads. The computer-driven actuators are controlled by a console,
which can toggle between the load- or displacement-control mode.
The loading area on top of the beam is leveled off by casting a
thin layer of dental stone capping to ensure proper contact. A
steel plate of 200 mm in length is then placed over the capping
as the loading plate. To record reaction forces, 200-ton and 100-
ton load cells are laid under the beam soffit for middle and end
supports, respectively. Each end support consists of a concrete
block and spring element A with a roller support on top of it. The
middle support has spring element B at the bottom and a 200-ton
hydraulic jack over it for adjusting the support level. The swivel
head attached to the top of the jack serves as a pin bearing so that
the middle support is free to rotate but restrained from translation.
The lengths of support plates are 150 mm and 200 mm for end and
middle supports, respectively. Support settlement is monitored by
two 25 mm LVDTs on each edge of the support as shown in Fig. 3,
making a total of four LVDT measurements for each support. The
beam deflection of each span is monitored by two 100 mm LVDTs
located at the middle soffit. Uniaxial strain gages are installed at the
longitudinal and web reinforcement, as shown in Fig. 4, indicated
by ‘SG’ followed by a sequence number.
The set-ups for specimens with one elastic support (TCDB-1-3
and TCDB-2-3) and three rigid supports (TCDB-1-1 and TCDB-2-1)
are similar to those shown in Fig. 3, except that the appropriate
spring elements are replaced with rigid concrete blocks.
Before testing begins, a load Pi = 50 kN is applied to a speci-
men at each loading point, then the middle support is jacked up to
support the beam. The height of middle support is adjusted so that
the reaction forces are equal to the values obtained from the linear
elastic finite element analysis of the deep beam. This is done to en-
sure that the middle and end supports are strictly leveled at the be-
ginning of each test. This is permissible as at the uncracked stage,
the load–reaction response of a continuous deep beam conforms
to elastic finite element analysis [4]. Then the pre-applied load Pi
is reduced to almost zero. Load-control mode is first adopted at an
increment of 20 kN before any cracks are formed on the beam sur-
face. After the first flexural crack is detected, the load increment
is increased to 40 kN until the total load reaches around 75% of
predicted failure load from finite element analysis. Then both ac-
tuators are toggled to displacement-control mode to prevent the
specimen failing in a sudden manner.
 N. Zhang, K.-H. Tan / Engineering Structures 32 (2010) 361–372 363

Fig. 1. Dimensions of specimens and typical reinforcement layout — Top: TCDB-1 series beams, Bottom: TCDB-2 series beams.

Fig. 2. Details of the spring elements — Top: Spring A, Bottom: Spring B.

3. Experimental results and discussion
The specimens are tested under intended differential support
settlement by utilizing spring elements at the middle and end
supports. By using different spring elements, different values of
support stiffness K can be achieved. As the experimentally
recorded points in R-versus-∆ (Eq. (1)) plots do not fall exactly
onto a straight line, linear regression is used to obtain the slope
(support stiffness K ). It should be noted that the relative support
stiffness K includes both the spring stiffness and the beam stiffness.
The value of K for each beam (Table 2) is very sensitive to support
settlement due to high stiffness of deep beams. Under a reaction
force of 1000 kN, a slight difference in the support settlement of
0.2 mm can result in a change in the K value of about 50 kN/mm.
It is noted that even for specimens tested with the same support
configuration, there was still a slight difference in K values (less
than 5%). This is probably due to some uncertain factors that may
affect the support settlement, for example, deformation of contact
surfaces between steel plates over the support, and deformation of
a layer of dental stone cast at the beam soffit. Generally, specimens
tested with three rigid supports (TCDB-1-1, TCDB-2-1 (r)) have K
values exceeding 400 kN/mm, while K value of specimen TCDB-
2-1 is above 500 kN/mm because the number of contact surfaces
between the steel plates is minimized and a much thinner layer
of dental stone is cast. Thus, the middle support settlement of
TCDB-2-1 is relatively smaller than the counterparts. Specimens
with three elastic supports (TCDB-1-2, TCDB-2-2) have K values
around 200 kN/mm, while specimens with elastic middle support
and rigid end supports (TCDB-1-3, TCDB-2-3) have the lowest K
value around 90 kN/mm.
3.1. Crack patterns
The crack patterns of the failed specimens are shown in Fig. 5.
The load at which each crack was first observed is also indicated
in the figure. Those cracks believed to be the cause of failure are
marked in bold and the hatched areas indicate concrete crushing
or spalling. It can be seen that the crack patterns of beams with web
reinforcement (last four specimens) are more developed than their
counterparts without web reinforcement (first three specimens).
364 N. Zhang, K.-H. Tan / Engineering Structures 32 (2010) 361–372

Fig. 3. Test setup for specimens loaded on three elastic supports (TCDB-1-2 and TCDB-2-2).

Fig. 4. Locations of strain gages (with STM indicated in dashed line) — Top: TCDB-1 series beams, Bottom: TCDB-2 series beams.

Table 2
Summary of experimental results.
Beam Relative stiffness of fc0 (MPa) Ultimate load First positive First negative First diagonal Serviceability load
notation middle support K Pexp (kN) flexural crack load flexural crack load crack load Pser (kN)
(kN/mm) Ppcr (kN) Pncr (kN) Pdcr (kN)

TCDB-1-1 422 35.6 797 240 420 400 400

TCDB-1-2 194 36.1 637 210 400 330 350
TCDB-1-3 83 38.0 640 180 n/a 280 280
TCDB-2-1 436 38.9 1001 230 410 450 490
TCDB-2-2 200 39.3 1001 220 490 490 450
TCDB-2-3 88 38.5 957 190 n/a 480 400

The cracking loads, serviceability loads (corresponding to 0.3 mm
crack width), ultimate loads and failure modes of the specimens
are summarized in Table 2. The specimens remain elastic with no
visible cracks until flexural cracking takes place. The first positive
flexural crack load Ppcr (per actuator), corresponding to the sagging
movement at the middle span, is typically around 180–260 kN.
The first negative flexural crack load Pncr , corresponding to the
hogging movement over the middle support, is at the level of
400–490 kN. The presence of web reinforcement does not seem
to influence the magnitude of cracking loads (both positive and
negative). However, the settlement of middle support substantially
influences the beam cracking strength. As K decreases, the beam
experiences a larger settlement. The first positive flexural crack
occurs at an earlier stage whereas the first negative crack over
the middle support occurs at a later stage (or even does not occur
for TCDB-1-3 and TCDB-2-3 in Fig. 5). This is attributed to an
increased sagging moment at the mid-span and a reduced hogging
moment over the middle support. Particularly for beams with
web reinforcement, as K decreases from 436 to 88 kN/mm, Ppcr is
reduced by 27%; for beams without web reinforcement, a reduction
of 25% is recorded as K decreases from 422 to 83 kN/mm.
Typically, one long flexural crack of about 2/3 of the beam
depth occurs over the middle support, accompanied by several
smaller flexural cracks near the top reinforcement. Contrary to
the shallow beam theory, first flexural crack occurs at the mid-
span rather than over the middle support. This distinct behavior
of CDBs due to shear deformation has also been reported by other
researchers [4,6–8]. As stiffness K of middle support decreases, this
difference in behavior becomes even more evident. For specimens
TCDB-1-3 and TCDB-2-3, negative flexural cracks do not occur
throughout the test. This is attributed to large settlement of
middle support, which renders the concrete portion over middle
support in compression. Flexural crack widths never reach the
serviceability limit of 0.3 mm or cause eventual beam failure.
 N. Zhang, K.-H. Tan / Engineering Structures 32 (2010) 361–372 365

Fig. 5. Crack patterns of specimens at failure.

Generally, they are finely distributed in the vicinity of bottom
reinforcement, indicating mobilization of the bottom tie.
Diagonal cracks occur after negative flexural cracks have taken
place and widen quickly under increasing load. The first diagonal
crack load Pdcr ranges from 280–540 kN, or 0.45 to 0.55 of ultimate
load. It is significantly influenced by the support stiffness and
web reinforcement. The serviceability limit of cracking (0.3 mm)
is reached by diagonal cracks in the specimens. As diagonal crack
widths widen quickly, a few of them merge and develop into
critical diagonal cracks that finally lead to beam failure.
3.2. Load–deflection response
The load-versus-mid-span deflection curves of the failure spans
of specimens are plotted in Fig. 6. The settlements of end supports
have been excluded from the beam mid-span deflections. As can
be seen from the figure, TCDB-1-1 and TCDB-2-1 exhibit the high-
est beam stiffness whereas TCDB-1-3 and TCDB-2-3 the lowest.
It is clear that the support stiffness K has a significant effect on
the load–deflection response. As K decreases, the gradient of the
load–deflection curve becomes gentler. It is attributed to larger
366 N. Zhang, K.-H. Tan / Engineering Structures 32 (2010) 361–372
Fig. 6. Load–deflection curves for specimens.
settlement of middle support relative to end supports. However,
web reinforcement has little effect on the beam stiffness. Gen-
erally, for the same support conditions, there are no observable
differences in load–deflection curves between specimens with or
without web reinforcement. Notwithstanding that, web reinforce-
ment significantly increases the ultimate shear strength for all sup-
port conditions.
3.3. Support reactions
The load-versus-reaction-force responses are plotted in Fig. 7. It
is found from the figure that the distribution of support reactions
is significantly affected by support stiffness K . As K decreases, the
middle support settles more and the end supports take over more
loads, thus relieving the reaction of the middle support. Particu-
larly for beams with web reinforcement, 23% of the middle support
reaction is transferred to the end supports at the same load level
when K decreases from 436 kN/mm to 88 kN/mm. A similar re-
duction in middle support reaction is observed for beams without
web reinforcement. This suggests a considerable redistribution of
beam internal forces due to support settlement. However, the pres-
ence of web reinforcement has negligible influence on redistribu-
tion ability. This becomes obvious from Fig. 7a, in which specimens
with and without web reinforcement exhibit almost the same
load–reaction response. Asin [6] also reported similar findings for
the influence of web reinforcement from his experimental pro-
gram. At failure, the reaction force of the middle support in a two
span continuous shallow beam is about 15% greater than its deep
beam counterpart with the same span (Fig. 7a). This distinct behav-
ior of CDBs has been reported by other researchers before [2,4,6,8,
9] and is re-confirmed in the authors’ experiments. Further more,
this difference grows larger as support stiffness K decreases. Fig. 7b
and c show the effect of spring stiffness on the reaction force of the
middle support. Compared to the middle support reaction of spec-
imen TCDB-2-3 (K = 88 kN/mm), its shallow beam counterpart
is about 55% higher (Fig. 7c). This is attributed to higher hogging
moment over middle support in shallow continuous beams. The
reaction forces at failure of the specimens are included in Table 3.
3.4. Serviceability and ultimate beam strengths
The dimensionless ultimate load Pexp /(fc0 bw h) and service load
Pser /(fc0 bw h) of the beams are plotted against support stiffness K in
Fig. 8. Under the same support conditions, that is, with elastic or
rigid supports, beams with web reinforcement exhibit higher ser-
viceability loads (0.3 mm cracking) than beams with plain concrete
web. This is because the web reinforcement resists tensile stresses
in the cracked concrete and deters the widening of crack width.
However, web reinforcement plays a more important role in ulti-
mate strength. This increase in ultimate strength is more evident
when K decreases, and an enhancement as high as 48% is found in
specimen TCDB-2-3 compared to its counterpart without web re-
inforcement TCDB-1-3 (Fig. 8b). Test results show an upward trend
of the serviceability and ultimate strengths with an increase of K .
Serviceability load is significantly reduced as K decreases. How-
ever, it is interesting to note that for beams with web reinforce-
ment, the ultimate strength is relatively less sensitive to support
stiffness, compared to those without, as opposed to the significant
redistribution of beam reaction due to change of support stiffness.
This can be explained by a strut-and-tie approach. At the same load
level, as K decreases, the compression force in the interior diagonal
strut (Fig. 9) decreases whereas the compression force in exterior
diagonal strut (Fig. 9) increases, thus delaying the failure within
the interior shear span near the middle support. In this sense, the
beam capacity does not deteriorate significantly. The CDB exhibits
considerable tolerance in accommodating the differential support
settlement, particularly for those beams with web reinforcement.
In the case when the compression force in exterior diagonal strut
is increased to the extent that initiates failure earlier than interior
diagonal strut does, the failure zone would shift from the interior
shear span to the exterior. This is observed in specimens TCDB-1-
3 and TCDB-2-3 (Fig. 5). In the following, a strut-and-tie model is
developed to predict the shear strengths of tested CDBs.
4. Modeling CDBs with support settlement
A direct Strut-and-Tie Model (STM) has been systematically de-
veloped by several researchers for calculating the ultimate shear
strengths of single-span deep beams (SSDBs) with or without
web reinforcement [10], prestressing tendons [11] and web open-
ings [12]. Tan and Cheng [13] used the STM to investigate the size
effect on shear strength of deep beams by considering the effect
of strut geometry, web bar spacing and diameter. The STM had
been further developed (referred as Modified STM) to achieve bet-
ter prediction accuracy and consistency and was successfully ex-
tended to continuous deep beams (CDBs) with rigid supports [14],
size effect in deep beam shear strength [15], SSDBs under unsym-
metrical loadings [16], and SSDBs with prestressing tendons [17].
Although numerical methods such as the nonlinear finite strip
technique have been proposed to analyze continuous deep beams
and showed encouraging results [18], a rational mechanical model
is scarce.
A proposed strut-and-tie model (STM) for CDBs with a point
load at each mid-span is shown in Fig. 9. It can be idealized as a
statically indeterminate truss with a spring stiffness K under the
middle support B to consider the effect of differential settlement
between the middle and end supports (Fig. 10).
Assuming bilinear elastic–plastic material properties for con-
crete struts and steel bars, the reaction force of middle support is
solved and denoted as X = Γ · P, where Γ is a support reaction
factor. Its detailed derivations are included in Appendix. Hence, the
member forces of the model can be expressed as follows:
Γ
 
P
Fc1 = 1− · (2)
2 sin θs
Γ P
Fc2 = · (3)
2 sin θs
P
T1 = (Γ − 1) · (4)
tan θs
Γ
 
P
T2 = 1− · (5)
2 tan θs
 N. Zhang, K.-H. Tan / Engineering Structures 32 (2010) 361–372 367

a b

Fig. 7. Influence of experimental parameters on distribution of support reactions — (a): effect of web reinforcement; (b) & (c): effect of support stiffness.

Table 3
Load distribution at ultimate stage.
Beam notations Reaction forces (kN) Loads (kN)
Left support Right support Middle support Left actuator Right actuator

TCDB-1-1 324 325 946 797 798

TCDB-1-2 260 259 755 637 636
TCDB-1-3 318 318 643 640 639
TCDB-2-1 389 388 1223 1001 999
TCDB-2-2 418 419 1163 999 1001
TCDB-2-3 497 499 917 957 956

where Fc1 and Fc2 represent the respective forces in the exterior Kcri is termed the critical support stiffness. It can be determined by
and interior diagonal struts (Fig. 10a), T1 and T2 represent the re- equating the tie force T1 in Eq. (4) to zero:
spective tension forces in the top and bottom longitudinal rein-
forcement, P represents the point load acting on the beam, and θs is 4Ec Astr5 cos θs sin2 θs
Kcri = (6)
m + 2p cos3 θs − 1 le


the inclined angle of diagonal strut with respect to the horizontal.
There are two scenarios to be considered: (I) the middle support
where Ec is the modulus of concrete, Astr5 is the average cross-
stiffness K is larger than a threshold value Kcri ; although the middle
sectional area of interior strut (Fig. 9), le is the beam effective span,
support settles under loading, it is sufficiently rigid to effect a
and m and p are ratios of axial stiffness of truss members. Details
hogging moment over the middle support. Consequently, the top
of these terms are included in Appendix.
member of the truss is in tension and acts as a tie (Figs. 9a and 10a).
For calculating the maximum force P, Mohr’s linear interaction
(II) K is smaller than Kcri , so the middle support settles considerably
failure criterion is adopted at the interface between the strut and
due to its low stiffness, and a sagging moment is present over the
the tension–compression nodal zones:
middle support. Therefore, the top of the beam is in compression,
represented by a top horizontal concrete strut in the truss model f1 f2
(Figs. 9b and 10b). The threshold value of middle support stiffness
+ =1 (7)
ft fc0
368 N. Zhang, K.-H. Tan / Engineering Structures 32 (2010) 361–372

4As2 fy sin θs Asw fyw sin(θs + θw )

a ftB = + + fct .
Ac / sin θs Ac / sin θs
Nodal zone C
1
PnC = (10)
(3−2Γ ) sin 2θs Γ −2 cos2 θs +Γ cos2 θs
ftC Ac
+ 2fc0 Astr2 sin θs

4T2 max − 2T1a Asw fyw sin(θs + θw )

ftC = + + fct
Ac / sin2 θs Ac / sin θs
where Ac = bw dc is the effective cross-sectional area of a deep
beam (Fig. 9); θw is the angle between the web reinforcement
and the horizontal axis of beams at the intersection of the web
steel bar and the diagonal strut (Fig. 9). Astr1 , Astr2 and Astr3 are the
cross-sectional areas at the ends of the tapered concrete struts as
illustrated in Fig. 11, given by Eqs. (A.2)–(A.4) in Appendix; T1 max
and T2 max are the respective yield force of top and bottom steel
T T
reinforcement; T1a = T1 T2 max and T2a = T2 T1 max are the tension
2 1
force in the top steel at the yielding of bottom steel and tension
force in the bottom steel upon yielding of top steel, respectively;
T1a and T2a should not exceed the respective yield strength of top
b and bottom reinforcement.
The predicted ultimate strength will be the minimum among
Eqs. (8)–(10) for the respective nodal zone A, B and C, denoted as
Pn .
Pn = Min(PnA , PnB , PnC ). (11)
When K < Kcri (Case II), nodal zone A experiences a compre-
ssion–compression stress state, so the failure load is governed only
by the compression–tension nodal zones B and C (Fig. 9b). Thus, no
further consideration is given to nodal zone A. The depth of top
concrete strut l0d (Fig. 9b) is assumed to be equal to the effective
depth of upper nodal zone ld (Fig. 11). The support reaction factor
Γ can be similarly obtained from Eq. (A.9), except that the ratio n
is substituted by n0 as follows:
Ec Astr5
n0 = (12)
Ec Astr6 + Es As1
where Astr6 (= ld bw −As1 ) is the cross-sectional area of top concrete
strut. After obtaining the member forces of the STM, failure load
Fig. 8. Effects of support stiffness K on — (a): service load; (b): ultimate load. Pn can be predicted by applying Eq. (7) to nodal zones B and C in a
similar fashion as before and adopting the smaller value of the two.
where f1 and f2 are the principal tensile and compressive stresses
Pn = Min(PnB
0
, PnC
0
) (13)
at the nodal zone respectively; fc0 is the concrete cylinder stre-
0 0
ngth representing the maximum compressive strength in the f2 where PnB = PnB as in Eq. (9), and PnC is expressed as below:
direction, and ft represents the maximum tensile capacity in the
0 1
f1 direction. As the flexural failure of deep beams falls outside the PnC = (14)
(2−Γ ) sin 2θs Γ −2 cos2 θs +Γ cos2 θs
scope of current paper, tie yielding is not considered as a failure +
0 A
ftB c 2fc0 Astr2 sin θs
criterion in the suggested model.
When K > Kcri (Case I), three different tension–compression where ftC0 is the corresponding tensile capacity of nodal zone C:
nodal zones (A, B and C) are identified in the model as shown 4T2 max Asw fyw sin(θs0 + θw )
in Fig. 9a. Similar to the procedure of modeling CDBs with rigid ftC0 = + + fct .
supports [14], Mohr’s failure criterion is applied to these nodal Ac / sin 2
θ0
s
Ac / sin θs0
zones (Fig. 11) and the ultimate load determined from each nodal In summary, the steps for direct calculation of ultimate strength
zone can be obtained as follows: Pn of CDB are summarized as follows:
Nodal zone A Step 1 Determine θs from Eq. (A.7);
Step 2 Determine m, n, p, ξ from Eqs. (A.10)–(A.13);
1 Step 3 Determine Γ and Kcri from Eqs. (A.9) and (6), res-
PnA = (8)
(5Γ /2−3) sin 2θs Γ /2−Γ cos2 θs +cos2 θs pectively;
+ fc0 Astr3 sin θs
ftA Ac
Case I: K > Kcri
4T1 max − 2T2a Asw fyw sin(θs + θw ) Step 4 Determine PnA , PnB and PnC from Eqs. (9), (10) and (8), res-
ftA = + + fct .
Ac / sin θs 2
Ac / sin θs pectively;
Step 5 Determine the ultimate load Pn from Eq. (11).
Nodal zone B Case II: K < Kcri
0 0
Step 4 Determine PnB and PnC from Eqs. (9) and (14), res-
1
PnB = (2−Γ ) sin 2θ (9) pectively;
s (1−Γ /2) sin θs
ftB Ac
+ fc0 Astr1
Step 5 Determine the ultimate load Pn from Eq. (13).
 N. Zhang, K.-H. Tan / Engineering Structures 32 (2010) 361–372 369

(a) K > Kcri .

(b) K < Kcri .

Fig. 9. Two-span CDB under two-point loading.

Fig. 11. Nodal zones in the strut-and-tie model for CDB.

Fig. 10. Truss models for two-span CDB on elastic supports.

5. Model verification
In the extreme case when K = 0, the two-span CDB reduces to
single span and the support reaction factor Γ = 0. Correspond- Fig. 12 demonstrates the development of steel strains in
ingly, in the STM, the interior diagonal struts (Astr5 ) vanish and specimens. Figures on the left refer to the measured strains on
the tension–compression nodal zone C no longer exists. Thus, the TCDB-2-1 with rigid supports, while those on the right refer to
member forces (Eqs. (2)–(5)) are reduced to: TCDB-2-3 with elastic middle support. The yield strains of longi-
P tudinal and web reinforcement are εy = 2512µ and εyw = 2503µ,
Fc1 = respectively. The locations of strain gages can be found in Fig. 4. The
sin θs strain development of SG 7-9 in the bottom steel is comparable to
Fc2 =0 one another, validating the assumption of a bottom tie. SG 6 and
T1 = − cot θs P SG10, located at the respective nodal zones C and B (Fig. 9), show
T2 = cot θs P . substantial delay in their strain development (Fig. 12a, b). This can
be attributed to the interference of compression forces transferred
Hence, the proposed STM for CDBs is identical to the one proposed into the nodal zone from diagonal concrete struts. The strain de-
for simply-supported deep beams [14]. velopment of SG 1 and 2 at the top steel is very similar along the
370 N. Zhang, K.-H. Tan / Engineering Structures 32 (2010) 361–372

a b

c d

e f

Fig. 12. Development of steel strains in TCDB-2-1 and TCDB-2-3.

course of loading. For TCDB-2-1, they are in tension whereas for
TCDB-2-3, they are slightly in compression. Though SG 3 in TCDB-
2-1 shows some delay in development due to the disturbance of
compressive stresses near the loaded area, the differences among
the measured strains in SG3, SG1 and SG2 are not significant. Thus
the assumption of top reinforcement acting like a tie is still valid.
SG 5 is located outside the load point and is in essence under
compressive stress, indicating that the top tie member exists only
between the two load points. Web reinforcement within the inte-
rior shear span yielded prior to or at imminent failure of the beam
(Fig. 12e). This suggested the interior diagonal strut was effective.
For specimens tested on elastic middle support and rigid end sup-
ports (TCDB-1-3 and TCDB-2-3), the bottom reinforcement yielded
at the beam failure (Fig. 12b) and the top reinforcement was in
compression throughout the test (Fig. 12d), suggesting the action
of a top horizontal compression strut between the two loads. This
 N. Zhang, K.-H. Tan / Engineering Structures 32 (2010) 361–372
Table 4
Comparison between STM predictions and test results.
Beam notations fc0 (MPa) 2Pexp (kN) 2PSTM
TCDB-1-1 35.6 1595 1456
TCDB-1-2 36.1 1274 1455
TCDB-1-3 38.0 1279 1346
TCDB-2-1 38.9 2000 1803
TCDB-2-2 39.3 2000 1781
TCDB-2-3 38.5 1912 1643
Mean
SD
COV
a
Xexp = tested middle support reaction forces.
b
XSTM = predicted middle support reaction forces by STM.
is due to an increased sagging moment caused by large support set-
tlement. Web reinforcement within the interior shear span did not
yield (Fig. 12f) because the interior diagonal strut (Fig. 9b) carried
a smaller load compared to the exterior diagonal struts (Fig. 9b) on
rigid supports. The exterior shear span attained its shear capacity
prior to the interior shear span, and led to the failure of the beam
(Fig. 5c, g).
The model’s verification on failure loads is carried out using the
authors’ tested specimens in Table 1. The Kcri value of the speci-
mens is calculated from Eq. (6) to be 99 kN/mm. Thus, specimens
TCDB-1-3 (K = 83 kN/mm) and TCDB-2-3 (K = 88 kN/mm) can
be categorized under Case II where the top horizontal member is
in compression—concrete strut. This is evidenced by the crack pat-
terns of the two beams (TCDB-1-3 and TCDB-2-3) where no neg-
ative flexural cracks are observed until failure takes place (Fig. 5).
The stiffness K of the other specimens is larger than 99 kN/mm,
thus these beams belong to Case I, where the top member acts as
a tie.
The STM calculations for CDBs can be easily implemented using
a spreadsheet. For modeling efficiency, bilinear elastic–plastic
material properties are assumed. This simplification for solving
the statically-indeterminate truss (Fig. 9) is validated by the
good agreement of the ratio of STM-to-experimentally-measured
middle support reaction forces at failure (XSTM /Xexp in Table 4). It
suggests that the proposed STM correctly represents the internal
force distribution of continuous deep beams at ultimate stage. The
STM-versus-experiment ultimate strength ratios (PSTM /Pexp ) are
also summarized in Table 4. The STM for CDBs generally performs
well in predicting the ultimate strengths with an overall mean of
0.95 and a C.O.V of 0.110. According to the authors’ model, the
failure loads of all Case I specimens are governed by nodal zone
B, whereas Case II specimens fail at nodal zone A of exterior shear
span. This conforms to the experimental observations that Case I
beams fail at the interior shear span whereas Case II specimens at
the exterior (critical crack leading to failure is highlighted in Fig. 5).
The model also predicts a reduction trend in beam capacity as K
grows (Fig. 8b). The curves are comparable to experimental results.
It is interesting to note that when K is infinite, i.e. there is no rel-
ative settlement between supports, the proposed STM will reduce
to the CDB model comprising three perfectly rigid supports [14].
The latter was in good agreement with test results of a total of 54
CDB specimens taken from the literature [14].
6. Conclusions
This paper presents the results of an experimental program de-
vised to investigate the effects of differential support settlement on
the strength and behavior of CDBs. Major findings are summarized
below:
1. All the specimens failed in shear. Support settlement signifi-
cantly influences the cracking behavior and crack patterns of
371
Xexp a XSTM b PSTM /Pexp XSTM /Xexp
946 883 0.91 0.93
755 819 1.14 1.08
643 631 1.05 0.98
1223 1097 0.90 0.90
1163 1008 0.89 0.87
917 788 0.86 0.86
0.95 0.94
0.104 0.078
0.110 0.083
continuous deep beams. The crack patterns of beams with web
reinforcement are more extensively developed than those with-
out web reinforcement.
2. Support settlement significantly reduces the serviceability load
(corresponding to a 0.3 mm crack width) and beam stiffness;
however the beams exhibit considerable tolerance in accom-
modating the support settlement. The reduction in ultimate
strength is not significant, particularly for beams with web re-
inforcement. The failure zone shifts from the interior shear span
to the exterior as the middle support stiffness K falls below crit-
ical support stiffness Kcri (Kcri corresponds to the support stiff-
ness that just transforms the horizontal tie on the top face of
the beams above the middle support into a strut). Web rein-
forcement significantly increases the beam capacity, though it
has little effect on beam stiffness. Furthermore, web reinforce-
ment is effective on restraining the widening of diagonal cracks,
thus increasing the serviceability load.
3. For specimens with middle support stiffness K greater than the
critical support stiffness (i.e. K > Kcri ), it was observed that
neither the bottom nor the top longitudinal steel reinforce-
ment yielded during the test. Web reinforcement within the
interior shear span yielded prior to or at imminent failure of
the beam. This suggested that the interior diagonal concrete
strut was effective. For specimens with K less than the criti-
cal support stiffness (i.e. K < Kcri ), the bottom reinforcement
yielded at failure whereas the top reinforcement was in com-
pression throughout the test. This is due to sagging moment
caused by large middle support settlement. The exterior shear
span reached its shear capacity prior to the interior shear span
and led to the failure of the beam.
4. The distribution of support reactions is significantly affected by
the middle support stiffness K . As K decreases, the middle sup-
port attracts a smaller percentage of the total load while the end
supports take more, suggesting a considerable redistribution of
beam internal forces. The web reinforcement has a negligible
influence on the reaction distributions.
A direct STM is proposed to calculate the ultimate strength of
CDBs with support settlement. Mohr’s interactive failure criterion
is applied to the tension–compression nodal zones and no empir-
ical stress limit is required. The model predicts the critical middle
support stiffness Kcri , below which the top member of STM trans-
forms from a tension tie to a compression strut. This theoretical
derivation is backed by experimental observations in strain mea-
surements and crack patterns. Explicit expressions of failure loads
for the beams are derived. The proposed model is evaluated by
the authors’ test program consisting of six medium-strength con-
crete CDBs with K values ranging from 83 to 436 kN/mm and with
or without web reinforcement. It is shown that the model predic-
tions for support reactions and ultimate beam strengths are in good
agreement with test results. The simplicity in principle and its ease
in implementation show clearly that the proposed STM is a promis-
ing tool for engineers.
372 N. Zhang, K.-H. Tan / Engineering Structures 32 (2010) 361–372

Acknowledgements and bottom longitudinal steel bars to the beam top and beam soffit
respectively (Fig. 9).
The authors would like to express their gratitude to Nanyang From the Theorem of Minimum Complementary Potential:
Technological University for research grant RG9/99. The first
∂ Uc
author also acknowledges a graduate research scholarship from ∆B = = 0. (A.8)
Nanyang Technological University, Singapore. ∂X
From Eqs. (A.1) and (A.8), the expression for X can be obtained as
Appendix. Derivation of reaction force in STM follows:
2(m + 2n cos3 θs + 2p cos3 θs )
The stress–strain relationship of concrete under compression is X =Γ ·P = ·P
taken as being linear for simplicity: 1 + m + 4n cos3 θs + 2p cos3 θs + 2ξ sin2 θs
(↑) (A.9)
σc = Ec εc (σc ≤ fc0 )
where m, n, p are the ratios of the axial stiffness of truss members:
where Ec is the modulus of elasticity of concrete
p in compression
and can be taken from ACI code (Ec = 4730 fc0 , MPa) for normal Astr5
strength concrete. m= (A.10)
Astr4
The stress–strain relationship of steel in tension is given by:
Ec Astr5
σs = Es εs (σs ≤ fs ). n= (A.11)
Es As1
For simplicity, the strain hardening of the steel reinforcement and Ec Astr5
the softening of concrete are not considered in the model. p= (A.12)
Es As2
The truss members in Fig. 10 are internally determinate but ex-
ternally indeterminate to 1◦ . Releasing the middle support force X and ξ is the stiffness ratio between the interior diagonal strut and
and applying the Crotti–Engesser theorem, the total complemen- the support spring:
tary energy in the truss is:
2Ec Astr5 cos θs
X Z Z  ξ= . (A.13)
Uc = ε dσ dV Kle


V
2
For a perfectly rigid support, ξ = 0.
P − X2 le
X 2 le
2 sin2 θs cos θs 8 sin2 θs cos θs References
= +
Ec Astr4 Ec Astr5
 2 [1] Leonhardt F, Walther R. Deep beams. Deutscher Ausschuss Fur Stahlbeton
P − X2 le Bulletin178, 1966, Wilhelm Ernst and Sohn (Berlin), 1966, CIRIA English
(X −P )2 le Translation, Jan. 1970.
2 tan2 θs tan2 θs X2
+ + + (A.1) [2] Rogowsky DM, MacGregor JG, Ong SY. Tests of reinforced concrete deep beams.
Es As1 Es As2 2K ACI J 1986;83(4):614–23.
[3] Chemrouk M. Slender concrete deep beams: Behaviour, serviceability and
where the term X is the reaction force of the middle support B; le is strength. Ph.D. Thesis. Newcastle: The University of Newcastle upon Tyne;
the effective span measured between center-to-center of the sup- 1988.
ports; θs is the inclined angle of the diagonal strut; Ec and Es are [4] Ashour AF. Tests of reinforced concrete continuous deep beams. ACI Struct J
1997;94(1):3–12.
the elastic modulus of concrete under compression and the steel [5] Maimba PP. Effects of support continuity on the behaviour of slender
bar under tension, respectively. The terms Astr4 and Astr5 represent reinforced concrete deep beams. M.Eng. thesis. Newcastle: The University of
the average cross-sectional areas of the exterior and interior ta- Newcastle upon Tyne; 1989.
[6] Asin M. The behaviour of reinforced concrete continuous deep beams. Delft
pered diagonal struts, respectively (Fig. 10). They are expressed as University Press; 1999. p. 168.
follows: [7] Poh SP. Strength and serviceability of prestressed concrete deep beams. M.Eng.
thesis. Singapore: Nanyang Technological University; 1995.
Astr1 = bw lf cos θs + lb sin θs


(A.2) [8] Subedi NK. Reinforced concrete two-span continuous deep beams. Proc Instn
Civ Engrs Structs Bldgs 1998;128(1):12–25.
Astr2 = bw lf cos θs + lc sin θs


(A.3) [9] Kong FK. Reinforced concrete deep beams. Van Nostrand Reinhold; 1990. xvi,
p. 288.
Astr3 = bw (ld cos θs + la sin θs ) (A.4) [10] Tan KH, Tang CY, Tong K. A direct method for deep beams with web
reinforcement. Mag Concrete Res 2003;55(1):53–63.
Astr1 + Astr3 [11] Tan KH, Tong K, Tang CY. Direct strut-and-tie model for prestressed deep
Astr4 = (A.5) beams. J Struct Eng 2001;127(9):1076–84.
2 [12] Tan KH, Tong K, Tang CY. Consistent strut-and-tie modelling of deep beams
Astr2 + Astr3 with web openings. Mag Concrete Res 2003;55(1):65–75.
Astr5 = (A.6) [13] Tan KH, Cheng GH. Size effect on shear strength of deep beams: Investigating
2 with strut-and-tie model. J Struct Eng 2006;132(5):673–85.
2(h − c1 − c2 ) [14] Zhang N, Tan KH. Direct strut-and-tie model for single and continuous deep
θs = arctan (A.7) beams. Eng Struct 2007;29(11):2987–3001.
le [15] Zhang N, Tan KH. Size effect in RC deep beams: Experimental investigation and
STM verification. Eng Struct 2007;29(12):3241–54.
where [16] Zhang N, Tan KH, Leong CL. Single-span deep beams subjected to unsymmet-
bw is the beam width; la , lb and lc are the widths of the support rical loads. J Struct Eng 2009;135(3):239–52.
and load bearing plates (Fig. 9); lf and ld are the respective effective [17] Wang GL, Meng SP. Modified strut-and-tie model for prestressed concrete
deep beams. Eng Struct 2008;30(12):3489–96.
depths of the bottom and upper nodal zone, lf = 2c2 , ld = 2c1 [18] Lin ZH, Raoof M. Application of nonlinear finite strip technique to concrete
(Fig. 11); c1 and c2 are the distances from the centroid of the top deep beams. Eng Struct 1995;17(10):725–36.