Acisp
Acisp
Synopsis:
Although limit state design blends both serviceability and strength limit
states, most engineers tend to be less confident in serviceability limit states than in
strength limit states, especially when deflections of reinforced concrete slabs are
considered. A major source of this lack of confidence is the existence of many
uncertain variables in calculating slab deflections such as concrete properties
(e.g., modulus of elasticity, modulus of rupture), creep coefficient, curing regime
and duration, and the existence of construction loads. The absence of any
reliability coefficients in deflection calculations or deflection limits gives the
impression that engineers are expected to evaluate the exact deflection that will
take place on site.
1
ACI member Mahmoud M. Reda Taha, Ph.D., P.Eng. is a structural engineer at
Stantec Consulting Ltd., Calgary, Canada. He is a member of ACI Committee 440
(Fibre Reinforced Polymers-FRP) and 548 (Polymers in Concrete). His research
interests include reinforced and prestressed concrete structures, computer
applications in structural design, fracture mechanics, and use of FRP in structures.
INTRODUCTION
3
M
I e = I cr + (I g − I cr ) cr ≤ I g [1]
Ma
fr I g
M cr = [2]
yt
fr is the modulus of rupture of concrete; and yt is the distance from the centroidal
axis of gross section to the extreme fibre in tension. When Eq. [1] is used, design
codes allow the designer to estimate the additional long-term deflection resulting
from creep and shrinkage by multiplying the immediate deflection due to
sustained load by a simple multiplier, ζ, given by
S
ζ = [3]
1 + 50 ρ ′
2
where S is a time-dependent factor equal to 2.0, 1.4, 1.2, and 1.0, respectively, for
5 years or more, 12 months, 6 months, and 3 months; ρ′ (= As′ / bd) is the ratio of
compression reinforcement at mid-span for a simple span; As′ is the area of
compression reinforcement; b is the width of compression face of member; and d
is the distance from the extreme compression fibre to the centroid of tension
reinforcement.
Sherif and Dilger (4), Ghali and Azarnejad (5), and Gilbert (6) criticized
the use of Eq. [1] showing that if the actual moment Ma in a concrete slab is close
to the cracking capacity of the slab, Mcr, this equation would indicate that the
value of Ie is close to Ig when in reality the slab has already cracked. While
neglecting the tension stiffening effect in a cracked concrete section would have a
negligible effect on the estimated deflection in cases where the total moment Ma
is three to four times larger than Mcr, it would have a significant effect if Ma is
between one and two times Mcr (5).
The mean curvature method is a more accurate and more general method
for predicting deflections of concrete members. It has been adopted by the CEB-
FIP Model Code 90 (MC-90) (8). In this method, the deflection of a member can
be determined from the values of the curvature Ψ at a number of sections (9). For
example, the deflection at mid-span of a simply-supported, one-way slab using
three sections along the span with parabolic variation of Ψ between them is
determined by double integration of the curvature along the span (9). The mid-
span deflection, ∆Mid, can be given by the geometrical relationship
L2
∆Mid =
96
(
Ψ Left + 10Ψ Mid + Ψ Right ) [4]
where
Mi
Ψi = [5]
E c I ei
where L is the span length; M is the bending moment; and Ec is the modulus of
elasticity of concrete. Increasing the number of sections will increase accuracy.
Equations to estimate the deflection based on the calculated curvature at various
sections along the member with different end conditions are given by Ghali and
Favre (9). The mean curvature method accounts for the variation in cross-
3
sectional properties (i.e., the loss of stiffness) that inevitably occurs with time in
concrete members due to cracking.
4
parameters) have wide scatter in their values. Concrete properties affect the
predicted deflection both directly by affecting the structural stiffness of the
element, and indirectly by defining the way moment distribution and
redistribution takes place in the structural system. The focus of this study is on
examining the effect of the uncertainties of concrete properties on the accuracy of
the predicted deflections.
Modulus of Rupture
Equations [6] to [11] give the models to predict the modulus of rupture of
concrete in MPa as per ACI 318-99 metric and U.S. customary units editions (2),
CSA A23.3-M94 (3), Standards New Zealand NZS 3101 (19), Standards
Australia AS 3600-1994 (20), and the Ontario Highway Bridge Design Code
(OHBDC) (21), respectively. Figure 1 shows the variation in the modulus of
rupture of concrete with respect to its compressive strength for the different
design codes. It is worth mentioning that CSA A23.3-M94 (3) requires the
reduction of the value of fr to half of the value given by Eq. [8] when deflection of
5
two-way slabs is computed. This reduction is considered to account for the effect
of cracking due to restrained shrinkage of two-way slabs based on the work of
Thompson and Scanlon (17).
CEB-FIP MC-90 (8) and the new Canadian Highway Bridge Design Code
(CHBDC) (22) are the only codes to abandon the modulus of rupture as a criterion
for cracking of concrete. While CEB-FIP MC-90 (8) uses the mean tensile
strength to represent the limit of tensile strength of concrete, CHBDC (22) adopts
a new term called the cracking strength to represent the stress level at which
concrete cracks. The use of a reduced value to represent cracking is intended to
reflect the uncertainty in concrete cracking strength because of its significant
variation with shrinkage and thermal strains developed in the section (23).
Equations [12] and [13] give the concrete tensile / cracking strength adopted by
CEB-FIP MC-90 (8) and CHBDC (22), respectively for use in estimating
deflections.
2/3
f'
f t = 1.4 c CEB-FIP MC-90 (8) [12]
10
It has been suggested that most of the models developed to predict the
modulus of rupture of concrete based on its compressive strength are inaccurate
(24,25). Raphael (25) was the first to propose abandoning the square root models
for predicting the modulus of rupture of concrete based on its compressive
strength. He proposed the following expression
6
( )
f r = 0.7 f c'
2/3
[14]
Recent attempts to predict the concrete splitting tensile strength and its
modulus of rupture using the compressive strength have been reported by
Oluokun (26), and Légeron and Paultre (27). They revealed similar trends
presented in Eqs. [15] and [16], respectively. It is worth mentioning that the only
code that does not use the square root model is CEB-FIP-MC90 (8).
( )
f t = 0.214 f c' 0.69
[15]
( )
f r = 0.5 f c' 2/3
[16]
As it has been established that these two models have much lower
coefficients of variation than those models using the square root of fc′, it seems
that the insistence on using the square root of fc′ to predict the tensile strength
generally, and the modulus of rupture specifically hinders the attempts to develop
more accurate prediction models. Figure 2 compares these two models for
predicting the modulus of rupture of concrete using Eqs. [15] and [16]. The
variation between the predicted modulus of rupture from these equations is still in
the range of ± 15 percent.
Modulus of Elasticity
7
E c = γ 1.5 ( 0.043 f c' ) AS 3600-1994 (20) [21]
1/3
f'
E c = 21500 κ c CEB-FIP MC-90 (8) [23]
10
where γ in Eq. [21] is the concrete density (in kg/m3), and κ in Eq. [23] is a
parameter that equals 0.85 to account for the initial plasticity of concrete when
elastic analysis of the structure is to be performed. Figure 3 shows the variation in
the modulus of elasticity of concrete with respect to its compressive strength for
the different design codes. For concrete strengths from 20 MPa to 80 MPa, there
exists a difference in the predicted modulus of elasticity in the range of 4.3
percent, 7.6 percent, and an average of 10 percent between ACI 318-99 metric
(2), and CSA A23.3-M94 (3), AS 3600-1994 (20), and CEB–FIP MC-90 (8),
respectively. It is worth mentioning that the difference between the two Canadian
standards, CSA A23.3-M94 (3) and OHBDC (21) is ± 10 percent.
f cr = f r − f sh [24]
where fcr is the cracking strength of concrete. The stress fsh can be evaluated using
the following expression suggested by Gilbert (6).
3.5 ρ (1 + 0.8 φ c ) E s
f sh = E a ε sh [25]
E c + 3 ρ (1 + 0.8 φ c ) E s
where
8
As
ρ= [26]
bd
Ec
Ea = [27]
1 + 0.8 φ c
ε sh '
Ψ sc = Γ 1 − As [28]
h As
The factor Γ is taken as 0.7 for uncracked and 1.2 for cracked sections,
respectively; and h is the overall depth of the member.
φ
Ψ t = Ψ i 1 + c [29]
α
Based on this review, it can be concluded that the value of the modulus of
rupture specified in most design codes is a best guess rather than a fixed material
property. In addition, it can be seen that there is a large variation in the predicted
value of the modulus of elasticity. Because both creep and shrinkage
9
fundamentally affect the concrete stiffness and cracking capacity, it is possible to
consider the errors in φc and εsh to be incorporated in the errors assumed in the
modulus of elasticity and the modulus of rupture, respectively. Therefore, even
with the best construction procedures, the estimated modulus of rupture or
modulus of elasticity using code equations would probably have an error of ± 30
percent.
1
f r −real = f [30]
1.09 + 0.0034 f ' r
c
The best curing regime for HPC is seven days of wet curing followed by
dry curing. Due to tight construction schedules, these requirements are hardly met
in most situations, despite the widely known adverse effects on the performance
of HPC when inadequate curing is practiced (33,34). HPC is more susceptible to
shrinkage cracks than NSC. In addition, shrinkage cracks of HPC are more
harmful to the modulus of rupture than they are in NSC. Raphael (25), and
Légeron and Paultre (27) showed how the moisture gradient due to drying induces
10
tensile stresses at the concrete surface causing microcracks to develop, and
therefore reducing the modulus of rupture of NSC when it is improperly cured.
Although this behaviour applies to HPC, it only constitutes part of the picture.
The tensile stresses due to drying at the surfaces of HPC members are expected to
be higher than in NSC members due to the significant increase of the cementitious
material content (e.g., Portland cement, slag, silica fume, and/or fly ash). In
addition, the increased autogenous shrinkage of HPC is expected to cause
homogeneous tensile shrinkage strains in the whole member mass if shrinkage is
restrained. Thus, several microcracks can develop in the whole HPC member if
not properly cured rather than at the skin only as in the case of NSC.
HPC has a higher elastic modulus and a more brittle post-peak behaviour
than NSC. HPC also responds to elastic stresses in a different manner than NSC.
The weak aggregate/cement paste transition zone in NSC allows the cement paste
to dominate the response to stresses within the elastic range. In HPC, the
transition zone is much stronger than in NSC, and failure usually occurs within
the aggregate particles. Therefore, more stress transfer to the aggregate takes
place within the elastic range, and the aggregate shares a relatively large portion
of the elastic response (33). Baalabaki et al. (36) proved the possibility of
producing two HPC mixes with the same compressive strength and very different
moduli of elasticity. Thus, when HPC is used, the above models are not valid, and
their prediction of the modulus of elasticity is not accurate.
Research work revealed that HPC would have higher total shrinkage
strains and lower creep coefficients than NSC due to its low water/cementitious
materials ratio (33). Therefore, a reduction of curing periods, as discussed earlier,
would significantly affect the shrinkage and creep performance of HPC. Dilger
and Wang (37) showed that creep and shrinkage models for NSC cannot be used
for HPC. This is because of the different proportions between basic and drying
creep, and autogenous and drying shrinkage in HPC compared to NSC. Few
design codes have adopted models for predicting creep and shrinkage of HPC
(e.g., The French Code) (AFREM) (38).
11
APPLICATION OF THE THEORY OF ERROR PROPAGATION
TO DEFLECTION COMPUTATION
z = f(x, y) [31]
2 2
∂z ∂z ∂z ∂z
σ z2 = σ x2 +σy
2
+ 2 COV(x, y) [32]
∂
x ∂y ∂x ∂y
The first two terms in Eq. [32] represent the contribution of standard
deviations of x and y to the standard deviation of the function z. The latter term
represents the contribution of the correlation between x and y to the standard
deviation of the function z. The term COV(x,y) is the covariance of the variables x
and y. The covariance of any two variables x and y is a function of their standard
12
deviations σx and σy, and of the coefficient of correlation between the two
variables, ρ(x,y), as shown in Eq. [33]. The same can be applied to any number of
correlated and/or non-correlated variables. Applying the principles of error
propagation to Eq. [5], the standard deviation of the curvature, σΨ, can be
estimated using the following expression.
2 2 2
∂Ψ ∂Ψ ∂Ψ
σΨ2 = σ I2e + σ E2 2
+ σ M +
∂I e ∂Ec ∂M s
c s
∂Ψ ∂Ψ
2 COV(I e , Ec ) +
∂I e ∂Ec [34]
∂Ψ ∂Ψ
2 COV(Ec , M s ) +
∂Ec ∂M s
∂Ψ ∂Ψ
2 COV(I e , M s )
∂I e ∂M s
Ms is the service moment. Note that Ms in the above equation and all of its
correlated values will vanish since loading conditions are assumed to have no
errors as discussed earlier. Therefore,
2 2
∂Ψ ∂Ψ ∂Ψ ∂Ψ
σΨ2 = σ I2e + σ E2 + 2 COV(I e , Ec ) [35]
∂I e ∂Ec ∂I e ∂Ec
c
where
∂Ψ −Ms
= [36]
∂I e Ec I e 2
and
∂Ψ −M
= 2 s [37]
∂Ec E c I e
2 2
∂I e ∂I e ∂I ∂I e
σ I2e = σ E2 c + σ 2f + 2 e COV(E c , f r ) [38]
∂E c ∂f r ∂E c ∂f r
r
13
parameters discussed above can also be evaluated numerically if the values of
these parameters are predicted using the appropriate prediction models.
Ψ f = Ψt +Ψ a −Ψ d [39]
2 2 2
∂Ψ f ∂Ψ f ∂Ψ f
σΨ2 f = σΨ2 t
∂Ψ
+ σΨ2 a
∂Ψ + σΨ d
2
∂Ψ
+
t a d [40]
∂Ψ f ∂Ψ f
2
COV(Ψ t ,Ψ a ) +
∂Ψ t ∂Ψ a
∂Ψ f ∂Ψ f
2
∂Ψ COV(Ψ a ,Ψ d ) +
∂Ψ a d
∂Ψ f ∂Ψ f
2
COV(Ψ t ,Ψ d )
∂Ψ t ∂Ψ d
Consequently, it can be seen that the effect of repeating the process three
times to calculate the final curvature deflection, Ψf, is simply to accumulate
further error in the final curvature value.
Knowing the final curvature and the curvatures due to shrinkage at the end
supports, the deflection at mid-span can be evaluated as follows
L2
∆=
96
(
Ψ Left + 10Ψ f + Ψ Right ) [41]
For a simply-supported slab, ΨLeft and ΨRight are induced by shrinkage alone. As
Eq. [28] shows, shrinkage-induced curvature is not a function of the concrete
properties. Therefore, the first and last derivatives of Eq. [41] will vanish, and the
14
standard deviation of the deflection, σ∆, will be directly proportional to the
standard deviation of the final mid-span curvature, σΨf. Thus,
2 2 2
∂∆ ∂∆ ∂∆
σ ∆2 = σΨ2 Left + σΨ2 + σΨ2 [42]
∂Ψ Left f ∂Ψ f Right ∂Ψ Right
L2
σ ∆ = σ ψ f [43]
9.6
σ f r = RE f r ⋅ f r [44]
σ E c = RE Ec ⋅ Ec [45]
The relative error in the parameters REfr and REEc was assumed to range
from 0 to 30 percent. When the effect of the level of error in the modulus of
rupture was examined, a constant error in the modulus of elasticity of 5 percent
15
was assumed. Also when the effect of the change in the error in the modulus of
elasticity was examined, a constant error in the modulus of rupture of 5 percent
was assumed. Using these relative errors, the standard deviation of the modulus of
rupture and the standard deviation of the modulus of elasticity were assumed, and
then the standard deviation of the estimated deflection, σ∆, was evaluated. The
relative error in the estimated deflection, RE∆, can be predicted using Eq. [46] in
which ∆ is the final deflection including the error.
σ∆
RE ∆ = [46]
∆
Using the above procedure, a parametric study was carried out to examine
the effect of a number of parameters on the error in the estimated deflection.
These parameters included the concrete compressive strength, fc′, the
mathematical model for predicting the modulus of rupture, fr, the mathematical
model for predicting the modulus of elasticity, Ec, the value of the shrinkage
strain, εsh, the value of the creep coefficient, φc, and the sustained load ratio, q.
Case Study
Using the ACI 318-99 U.S. customary units model (2), and for a concrete
compressive strength of 40 MPa, the modulus of rupture and the modulus of
elasticity will be 3.9 MPa and 29,915 MPa, respectively. The gross moment of
inertia is 2.8×108 mm4, while the cracked moment of inertia is 4.5 ×107 mm4. The
age-adjusted modulus of elasticity is 11,506 MPa.
Using this data, the shrinkage stress can be calculated as (Eq. [25])
This shrinkage will reduce the cracking strength of concrete to 2.1 MPa (Eq. [24])
and, consequently, the cracking moment capacity to 7.9 kN.m. Due to the
16
omission of the compression reinforcement, the shrinkage will induce a curvature,
Ψsc, at both ends of the slab. This curvature is calculated as follows (Eq. [28])
600 × 10 −6
Ψ sc = 0.7 1 − 0 = 2.8 × 10 −6 mm −1 [48]
150 667
The effective moments of inertia and the curvatures due to the different
cases of loadings are summarized in Table 1. It is shown that the immediate
deflection of the slab due to live load is 22.7 mm, and the estimated total
deflection, ∆, including long-term deflection, is 25.1 mm. To evaluate the error in
the estimated final deflection, Eqs. [34] to [43] would be applied. The coefficients
of correlation between the different parameters incorporated in the covariance
calculations for this case study were obtained by determining the values of these
parameters as predicted using the ACI 318-99 model (2), and by considering
concrete strengths in the range of 20 to 110 MPa.
Upon examining the two figures, it can be seen that a similar level of error
in fr and Ec will not yield the same level of error in the estimated deflection. For
example, for an average error of 15 percent in fr and an assumed error of 5 percent
in Ec (Figure 4), the average error in the estimated deflection would be about 26
percent. For an average error of 15 percent in Ec and an assumed error of 5
percent in fr (Figure 5), the average error in the estimated deflection would be
about 20 percent. The relative error in the modulus of rupture, REfr, has a more
pronounced effect on the relative error in the estimated deflection, RE∆, than the
relative error in the modulus of elasticity, REEc. This is because the contribution
of the cracking moment and, consequently, the modulus of rupture to the effective
17
moment of inertia is much higher than the contribution of the modulus of
elasticity.
The effect of shrinkage strain was also examined. Error analysis was
repeated for the same case study and for a 40-MPa concrete compressive strength.
The error in fr ranged from 0 to 30 percent, while the error in Ec was assumed to
be constant at 5 percent. The level of shrinkage strains ranged from 200 to 1000
microstrains. Figure 6 shows the effect of changing the shrinkage strain on REEc
due to the error in fr. It can be noticed that the higher the shrinkage strain, the
lower the final cracking strength, fcr, (as in Eq. [24]) and consequently the lower
the contribution of fr to both the deflection and its error.
2
σΨ = J C J T [49]
where J is the matrix of derivatives and C is the matrix of variances given by Eqs.
[50] and [51], respectively.
∂Ψ ∂Ψ ∂Ψ ∂Ψ ∂Ψ
J = [50]
∂M ∂E c ∂f cr ∂ε sh ∂φc
and
18
σ M2 σMσE σMσ f σM σε σ M σφ
c cr sh c
σ E2c
σE σ f
c cr
σ E σε
c sh
σ E σφ
c c
C= σ 2f σ f σε σ f σφ
cr cr sh cr c
SYMMETRY σ ε2 sh
σε σφ
sh c
σ φ2
c
[51]
SUMMARY
ACKNOWLEDGEMENTS
The authors would like to thank Mr. Pak Wong, P.Eng. and Mr. Paul
Langohr, P.Eng., Stantec Consulting Ltd., and Dr. Naser El-Sheimy, the
University of Calgary for their thoughtful comments and constructive criticism.
19
NOTATION
20
z variable
α a term that accounts for the effect of cracking and the braking action
of the reinforcement
∆ final deflection at mid-span
∆Mid mid-span deflection
εsh shrinkage strain
φc creep coefficient
γ concrete density (in kg/m3)
Γ factor used in calculating shrinkage induced curvature
κ a parameter that equals 0.85 to account for the initial plasticity of
concrete when elastic analysis of the structure is to be performed
ρ ratio of tension reinforcement
ρ′ ratio of compression reinforcement
ρ(x,y) coefficient of correlation between two variables, x and y
σ standard deviation
Ψ curvature
Ψa curvature due to dead and live loads
Ψd curvature due to dead load
Ψf final curvature at any section of an element
Ψi initial curvature due to a sustained service moment
ΨL curvature due to live load
ΨLeft curvature at the left end of a simply-supported slab
ΨMid curvature at mid-span of a simply-supported slab
ΨRight curvature at the right end of a simply-supported slab
Ψsc shrinkage induced curvature
Ψt long-term curvature at any time t due to a sustained service moment
ζ a multiplier used to estimate the additional long-term deflection due
to creep and shrinkage
21
REFERENCES
8 CEP-FIP Model Code 90, Model Code for Concrete Structures, Comité
Euro-International du Béton (CEB) - Fédération Internationale de la Précontrainte
(FIP), Thomas Telford Ltd., London, UK, 1993.
11 Ghali, A., Elbadry, M., and Megally, S., Two-Year Deflections of the
Confederation Bridge, Canadian Journal of Civil Engineering, Vol. 27, No. 6,
December 2000, pp. 1139-1149.
22
12 Fling, R.S., Practical Considerations in Computing Deflection of
Reinforced Concrete, Designing Concrete Structures for Serviceability and
Safety, ACI SP-133, 1992, pp. 69-91.
23
25 Raphael, J.M., Tensile Strength of Concrete, ACI Journal, Vo. 81, No. 2,
March-April 1984, pp. 158-165.
28 Gardner, N.J., and Zhao, J.W., Creep and Shrinkage Revisited, ACI
Materials Journal, Vol. 90, No. 3, May-June 1993, pp. 236-246.
36 Baalabaki, W., Benmokrane, B., Chaallal, O., and Aïtcin, P.C., Influence
of Coarse Aggregate on Elastic Properties of High-Performance Concrete, ACI
Materials Journal, Vol. 88, No. 5, September-October 1991, pp. 499-503.
37 Dilger, W.H., and Wang, C., Shrinkage and Creep of High Performance
Concrete (HPC) – A Critical Review, Adam Neville Symposium on Concrete
Technology, Las Vegas, NV, June 1995, pp. 59-84.
24
38 Le Roy, R., de Larrard, F., and Pons, G., The AFREM Code Type Model
for Creep and Shrinkage of High Performance Concrete, Fourth International
Symposium on the Utilization of High Strength/High Performance Concrete,
Proceedings−Vol. 1, Paris, France, May 1996, pp. 387-396.
25
SUGGESTED KEYWORDS
26
LIST OF TABLES
27
TABLES
28
LIST OF FIGURES
Figure 4 Effect of concrete strength on the error in the estimated deflection due
to the error in the modulus of rupture (REEc = 5%)
Figure 5 Effect of concrete strength on the error in the estimated deflection due
to the error in the modulus of elasticity (REfr = 5%)
Figure 6 Effect of shrinkage strains on the error in the estimated deflection due
to the error in the modulus of rupture (REEc = 5%)
29
FIGURES
6.5
6.0
5.5
Modulus of rupture, MPa
5.0
4.5
4.0
3.5
3.0
2.5
2.0
10 20 30 40 50 60 70 80 90
Compressive strength, MPa
ACI metric ACI U.S. CSA NZS AS OHBDC
30
10
9
8
Modulus of rupture, MPa
7
6
5
4
3
2
1
10 20 30 40 50 60 70 80 90
Compressive strength, MPa
Oluokun (1991) Legeron and Paultre (2000)
31
50
45
Modulus of Elasticity, GPa
40
35
30
25
20
15
10 20 30 40 50 60 70 80 90
Compressive strength, MPa
ACI metric ACI U.S. CSA NZS
AS OHBDC CEB-FIP
32
50%
45%
Relative error in deflection 40%
35%
30%
25%
20%
15%
10%
5%
0%
0% 5% 10% 15% 20% 25% 30%
Relative error in modulus of rupture
20 MPa 30 MPa 40 MPa 50 MPa
Figure 4 Effect of concrete strength on the error in the estimated deflection due
to the error in the modulus of rupture (REEc = 5%)
33
30%
25%
Relative error in deflection
20%
15%
10%
5%
0%
0% 5% 10% 15% 20% 25% 30%
Relative error in modulus of elasticity
20 MPa 30 MPa 40 MPa 50 MPa
Figure 5 Effect of concrete strength on the error in the estimated deflection due
to the error in the modulus of elasticity (REfr = 5%)
34
50%
45%
Relative error in deflection 40%
35%
30%
25%
20%
15%
10%
5%
0%
0% 5% 10% 15% 20% 25% 30%
Relative error in modulus of rupture
200 400 600 800 1000
Figure 6 Effect of shrinkage strains on the error in the estimated deflection due
to the error in the modulus of rupture (REEc = 5%)
35