ACI STRUCTURAL JOURNAL TECHNICAL PAPER

Title No. 118-S41

Rational Approach for Computing Long-Term Deflection of

Reinforced Concrete
by Peter H. Bischoff and Lluis Torres

The ACI 318 Building Code computes time-dependent deflections

from creep and shrinkage as the product of immediate deflection
from sustained loads and a deflection multiplier that depends on
the duration of loading and amount of compression reinforcement.
An alternative methodology is proposed that extends the approach
developed for an effective moment of inertia Ie, recently adopted by
ACI 318-19 for computing immediate deflection to include time-de-
pendent deflections from creep under sustained load and shrinkage.
Separate components are developed for creep and shrinkage defor-
mation based on a rational tension stiffening model that accounts
for the reduced member stiffness from creep, shrinkage warping
of the uncracked and cracked sections, and loss of stiffness from
a lower cracking moment that results as tensile stresses develop
in the concrete from restraint to shrinkage and cause additional
cracking. Comparison is also made with the Eurocode 2 approach.
Validation with long-term test data shows close agreement within a
few percent for the proposed approach and Eurocode 2 approach.
ACI 318-14 underestimates deflection by 27% and ACI 318-19
overestimates deflection by 23% for the test results evaluated in Fig. 1—Short-term response of flexural member.
this study.
effective moment of inertia Ie is extended to include time-de-
Keywords: beams; creep; deflection; effective moment of inertia; long- pendent deflections from creep under sustained loading and
term; shrinkage; slabs. shrinkage. The proposed approach computes deformation
from creep, which is load-dependent, separately from that
INTRODUCTION caused by shrinkage. Deflection from creep arises from a
Deflection of flexural members under service load is reduced elastic modulus for concrete that includes imme-
computed by integrating curvature or using formulas for diate deflection. Deflection from shrinkage is caused by
elastic deflection that include an elastic concrete modulus shrinkage warping of a partially cracked section in addition
Ec and effective moment of inertia Ie. ACI 318 computes to a loss of stiffness from additional cracking through the
immediate deflection using an elastic deflection equa- use of a lower cracking moment that results from the devel-
tion, and additional time-dependent deflection from creep opment of restraint stress in the concrete as the concrete
and shrinkage by multiplying the immediate deflection shrinks over time. Comparison with test data demonstrates
value from sustained loads with a long-term multiplier the proposed approach can provide reasonable estimates of
that depends on the duration of loading and amount of long-term deflection with the correct material properties.
compression reinforcement. Another well-known approach
is to compute long-term deflection using an effective elastic SHORT-TERM DEFLECTION MODEL
modulus for concrete with a reduced value to account for ACI 318-19 uses an effective moment of inertia Ie devel-
creep under sustained loads plus a separate deflection value oped by Bischoff (2005, 2020) where a tension stiffening
from shrinkage warping (CEN 2004; fib 2013, 2019). moment ΔMts is added onto the cracked EcIcr response of the
This paper presents a rational approach for computing flexural member as shown in Fig. 1. Conversely, the tension
long-term deflection that builds on the principles used to stiffening curvature Δϕts can be subtracted from the EcIcr
develop an effective moment of inertia adopted by ACI response. Both approaches give the same result. Using the
318-19 (ACI Committee 318 2019) for immediate deflec- former approach for a defined service load moment Ma
tion of nonprestressed members. Comparison is made with
existing approaches and validated with experimental data.
ACI Structural Journal, V. 118, No. 2, March 2021.
MS No. S-2020-131, doi: 10.14359/51728192, received April 5, 2020, and reviewed
RESEARCH SIGNIFICANCE under Institute publication policies. Copyright © 2021, American Concrete Institute.
The approach adopted by ACI 318-19 for computing All rights reserved, including the making of copies unless permission is obtained from
the copyright proprietors. Pertinent discussion including author’s closure, if any, will
immediate deflection of nonprestressed concrete using an be published ten months from this journal’s date if the discussion is received within
four months of the paper’s print publication.

ACI Structural Journal/March 2021 215

 Ma = EcIcrϕa + ΔMts (1) K depends on the type of load and support conditions (ACI
Committee 435 2020). For example, the value of K for a
with ΔMts = βtsΔMcr; the tension stiffening moment at member with a uniformly distributed load w equals 1.0
cracking ΔMcr = ηMcr; and η ≅ 1 – Icr/Ig, where the moment for a simply supported member, 2.4 for a cantilever, and
of inertia of the uncracked transformed section Itr is approx- 1.2 – 0.2Mo/Ma for a continuous span with Mo = wL2/8. Ie is
imated with the gross moment of inertia Ig of the concrete computed using Ma at the critical section and accounts for
section neglecting the reinforcement. Icr is the moment of nonlinear behavior after the member cracks.
inertia of the cracked transformed section, Ec is the elastic
modulus of the concrete, and βts is a tension stiffening factor EXISTING ACI APPROACH FOR LONG-TERM
with an upper limit equal to 1 at cracking that gradually DEFLECTION
drops down to 0 at high loads relative to Mcr. For the cracking The total deflection Δtotal over the long term is comprised
moment Mcr = frIg/yt, fr is the rupture modulus of concrete, of the computed short-term deflection value Δi plus the addi-
and yt is the distance from the centroid of the uncracked tional time-dependent deflection Δt from the effects of creep
(gross) section to the tension face. under sustained loads and shrinkage warping. This gives
Rearranging Eq. (1) in terms of the curvature ϕa and Δtotal = Δi + Δt. ACI 318-19 computes additional time-depen-
setting this equal to Ma/EcIe gives dent deflection by multiplying the short-term (immediate)
deflection value from sustained loads (including the portion
Ma M (1 − ∆M ts / M a ) Ma  M cr  of live load that is sustained) with a single multiplier λΔ = ξ/
φa = = a = 1 − βts η M  (2) (1 + 50ρ′) that includes the combined effects of creep plus
Ec I e Ec I cr Ec I cr a shrinkage
leading to
Δt = [Δi,D + Δi,L(sus)]λΔ (5a)
I cr I cr
Ie = = (3) where Δi,D equals the immediate deflection from dead load
M  I  M   I 
2

1 − βts  cr  1 − cr  1 −  cr  1 − cr  and Δi,L(sus) equals the immediate deflection from the portion
 Ma   Ig   Ma   Ig  of live load sustained. The ACI time-dependent factor ξ
assuming βts = Mcr/Ma after cracking. Ie equals Ig before depends on the duration of sustained loading and is equal
cracking. Restraint to shrinkage from the internal reinforce- to 1.0, 1.2, 1.4, and 2.0 for loads sustained over a dura-
ment and supports causes tensile stresses to develop in the tion of 3, 6, 12, and 60 months or more, respectively (ACI
concrete, leading to a reduction in the external moment Committee 318 2019). The compression reinforcement ratio
needed to crack the member and additional time depen- ρ′ = As′/bd acts to reduce time-dependent deflections, where
dent cracking that reduces tension stiffening. Scanlon and As′ equals the area of compression reinforcement, b is the
Bischoff (2008) recommend using 2/3Mcr for the reduced member width, and the effective depth d is taken from the
cracking moment to give the expression for Ie adopted by compression face of the section to the centroid of the tension
ACI 318-19 (Bischoff 2020). reinforcement area As.
An alternative, more detailed approach permitted by ACI
I cr 318-19 is to compute the additional time-dependent deflec-
Ie  2 (4) tions from creep and shrinkage separately (ACI Committee
2  435 1966, 2020; Branson 1965, 1977), where
 M cr   I cr 
1  3  1  
 M a   Ig  Δt = Δcp + Δsh (5b)
 
when Ma > 2/3Mcr. Ie equals Ig when Ma ≤ 2/3Mcr. A reduced The deflection Δcp caused by creep of the concrete under
cracking moment of 0.8Mcr is recommended for FRP rein- sustained stress is given by
forced concrete because the FRP reinforcement provides
less restraint to shrinkage than steel reinforcement (Bischoff Δcp = krCtΔi (6)
and Gross 2011). ACI 318-14 and earlier editions have used
Branson’s (1965) equation for Ie at full Mcr with Ie = (Mcr/ where kr = 0.85/(1 + 50ρ′); Ct is the creep coefficient (related
Ma)3Ig + [1 – (Mcr/Ma)3]Icr ≤ Ig. to strains); and Δi is the immediate deflection from sustained
loads. The reduction factor kr is needed because the rela-
Computing deflection with Ie tive increase in curvature and corresponding increase in
Short-term (immediate) deflection Δi under service loads deflection is less than the relative increase in strain when
is calculated using either a linear elastic deflection equation the neutral axis depth shifts towards the tension face as the
such as Δi = K(5/48)MaL2/EcIe or by integrating curvature compressive strain increases over time from creep (Branson
M/EcIe along the member span L where Ie depends on the 1977). kr also accounts for restraint of creep strains from the
moment M at each location. For the deflection equation, Ma is presence of compression reinforcement. Time dependent
the service load moment at the critical section (located at the expressions for Ct that can range in complexity are readily
support for a cantilever and at midspan for simply supported available and depend on several factors that include the age
and continuous members), and the elastic deflection factor of concrete at loading, duration of loading, curing conditions

216 ACI Structural Journal/March 2021

that account for the relative humidity and age when drying
starts, size and volume-surface ratio of the member, cement
type, and compressive strength (ACI Committee 209 1992,
2008; ACI Committee 435 2020; fib 2013). A convenient
and simple expression relating the creep coefficient to the
ACI time-dependent factor ξ is given by Ct = (ξ/2.0) × Cult.
In the absence of more detailed information, ACI Committee
435 (1978) recommends using a value of 1.6 for the ultimate
creep coefficient Cult while ACI Committee 209 (1992, 2008)
recommends a value of 2.35 under standard conditions.
Deflection Δsh caused by warping that occurs when short-
ening from concrete shrinkage is non-uniform through the
depth of the member is obtained by integrating the shrinkage
curvature ϕsh along the member span, making sure to account
for secondary moments that develop when the member is
restrained at the supports. A general equation for shrinkage
deflection is given by

Δsh = kshϕshL2 (7a)

where the factor ksh depends on the support conditions and

distribution of ϕsh along the member span L (0.125 for a
simply supported member, 0.090 for a member with one end
continuous, 0.065 for a member continuous at both ends, and
0.5 for a cantilevered member). The shrinkage curvature ϕsh
due to restraint from the internal reinforcement is based on a
modification of Miller’s (1958) empirical method developed
by Branson (1965, 1977)

Ash ε sh
φ sh = (7b)
h
with Ash = 0.7(ρ – ρ′)1/3[(ρ – ρ′)/ρ]1/2 for (ρ – ρ′) ≤ 3% and Ash
Fig. 2—(a) Long-term response of flexural member
= 1 for (ρ – ρ′) > 3%. For this case, the tension reinforcement
(including short-term response); and (b) zoomed-in long-
ratio ρ = As/bd and compression reinforcement ratio ρ′ = As′/
term response of flexural member (initial part).
bd are expressed as percentages, and h is the member thick-
ness. Time-dependent expressions for the free shrinkage on the reinforcing ratios ρ and ρ′, ratios d/h and d′/h, concrete
strain εsh are readily available and depend on factors like compressive strength fc′, which affects the elastic concrete
those affecting creep (ACI Committee 209 1992, 2008; modulus, and creep coefficient Ct. ϕsh,cr/ϕsh,tr is greater for
ACI Committee 435 2020; fib 2013). A convenient expres- lower values of ρ and Ct and for higher values of fc′. The
sion relating the shrinkage strain to the ACI time-dependent relative difference between ϕsh,cr and ϕsh,tr is not affected by
factor ξ is given by εsh = (ξ/2.0) × εsh,ult. ACI Committee the shrinkage strain, although shrinkage does reduce the
435 (1978) recommends a value of 400 × 10–6 for the ulti- cracking moment. Figure 2(b) provides a zoomed-in view of
mate shrinkage strain εsh,ult while ACI Committee 209 (1992, the initial part of the long-term response.
2008) suggests using a value of 780 × 10–6 under standard For long-term calculations, an effective elastic modulus
conditions when more detailed information is not available. of concrete E̅c = Ec/(1 + Ct) that includes the immediate
and creep components of strain (Gilbert 1988) is used to
PROPOSED LONG-TERM DEFLECTION MODEL compute the long-term moments of inertia of the uncracked
The short-term deflection model for Ie is extended to transformed section I̅ tr and cracked transformed section I̅ cr.
include time-dependent deflections from creep and shrinkage The long-term (effective) modular ratio n̅ = Es/E̅c = n(1 + Ct)
based on an approach described by Bischoff and Darabi is used to transform the reinforcement in the uncracked and
(2012). This approach gives the total deflection Δtotal which cracked sections, where n = Es/Ec is the short-term modular
includes short-term deflection. ratio and Es is the elastic modulus of the reinforcement.
From Fig. 2(a), creep reduces the stiffness of the long-term
uncracked response E̅cI̅ tr which is offset by a curvature ϕsh,tr Model development
caused by shrinkage. Similarly, the long-term response of Referring to Fig. 2(a), the tension stiffening moment ΔMts
the fully cracked section has a reduced stiffness E̅cI̅ cr arising is added onto the long-term cracked response E̅cI̅ cr of the
from creep and is offset by a shrinkage curvature ϕsh,cr that is flexural member (similar to the approach used for short-term
typically between 1 to 6 times greater than ϕsh,tr depending deflection) which is offset by ϕsh,cr

ACI Structural Journal/March 2021 217

 Ma = E̅cI̅ cr(ϕa – ϕsh,cr) + ΔMts (8)

with ΔMts = βtsΔMcr; ΔMcr = η̅(Mcr′ + ΔMsh); and η̅ = 1 – I̅ cr/

I̅ tr. I̅ tr can be approximated with Ig for design but is kept as
I̅ tr for the purposes of this paper as it is more exact. Gilbert
(2003) suggests that the use of Ig is unnecessarily conserva-
tive in deflection calculations. Mcr′ is the reduced cracking
moment caused by the development of tensile stresses in
the concrete from restraint to shrinkage by the internal rein-
forcement. A reduction in tensile strength under sustained
load can also lead to a decrease in the cracking moment.
The offset moment from shrinkage ΔMsh = (E̅cI̅ cr/η̅)Δϕsh with
Δϕsh = ϕsh,cr – ϕsh,tr (refer to Fig. 2(a) and (b) for a visual
description of these terms). Hence, ΔMcr = η̅Mcr′ + E̅cI̅ crΔϕsh.
Rearranging Eq. (8) in terms of the curvature ϕa gives

M a (1 − ∆M ts / M a ) Fig. 3—Effect of short-term plus sustained loading combined

φa = + φ sh , cr (9a)
Ec I cr with deformation from shrinkage.
plus creep behavior under sustained loads and shrinkage
Ma  M cr  deformation as follows
a  1  ts    ts sh ,tr  1  ts  sh , cr (9b)
Ec I cr  Ma 
I cr
Compare Eq. (9a) and (9b) with Eq. (2) for short-term Ie  2 (11a)
deformation. Adopting a generalized expression for the 2 
 M cr   I cr 
curvature at Ma 1  3  1  
 M a   I tr 
Ma  
φa = + φ sh , e (10a)
Ec I e 2   2 
 M cr    3 M cr 
leads to a time-dependent (long-term) value for the effec- sh , e  3  sh ,tr  1     sh , cr (11b)
tive moment of inertia I̅ e used to compute curvature from the  M a    M a  
effects of creep under sustained loading that includes short-      
term deformation I̅ tr can be approximated with Ig for design as noted earlier.
Hence, before cracking when Ma ≤ 2/3Mcr, I̅ e = Ig and ϕa =
I cr Ma/E̅cIg + ϕsh,tr. After cracking when Ma > 2/3Mcr, ϕa = Ma/
Ie = (10b)
M′  I  E̅cI̅ e + ϕsh,e using Eq. (11a) for I̅ e and Eq. (11b) for ϕsh,e.
1 − βts cr 1 − cr 
Ma  I tr 
Offset curvatures caused by shrinkage
and an effective shrinkage curvature ϕsh,e The offset curvatures ϕsh,tr and ϕsh,cr arising from shrinkage
are obtained using the “equivalent tensile force method”
ϕsh,e = βtsϕsh,tr + (1 – βts)ϕsh,cr (10c) (Branson 1977) for both the uncracked and fully cracked
sections as depicted in Fig. 4 and are similar to those
I̅ e depends on the level of cracking in the member through computed with the fib Model Code 2010 (fib 2013). In this
the Mcr′/Ma ratio and βts factor, while ϕsh,e depends on the approach, an equivalent tensile shrinkage force Tsh = EsAsεsh
level of cracking through the βts factor only. Figure 3 demon- located at the centroid of the tension reinforcing steel is
strates how deformation from short-term plus sustained multiplied by the eccentricity e of the force relative to the
loading (Ma/E̅cI̅ e) is combined with the deformation value centroid of the long-term uncracked transformed section or
from shrinkage warping of the partially cracked section long-term cracked transformed section to give an equivalent
(ϕsh,e) which represents a weighted average of deformation shrinkage moment Tshe. Separate tensile forces are obtained
caused by shrinkage warping of the uncracked and cracked for the tension steel and compression steel with each having
sections. their own eccentricity. This approach adopts the same prin-
For long-term behavior, the tension stiffening factor βts ciple as using a fictious decompression force introduced by
= (Mcr′ + ΔMsh)/(Ma + ΔMsh). Simplifying this factor by Nilson (1976) for evaluating prestressed concrete.
ignoring ΔMsh gives βts ≈ Mcr′/Ma, which provides a conser- Dividing the shrinkage moment Tshe by the effective
vative estimate of deflection. With this substitution and concrete modulus E̅c and long-term moment of inertia
assuming a reduced cracking moment Mcr′ = 2/3Mcr for (depending on whether the section is uncracked or fully
steel-reinforced concrete, the proposed model for computing cracked) gives the respective offset curvatures. Warping
total deflection gives separate components for immediate deflection occurs because of nonuniform shortening through

218 ACI Structural Journal/March 2021

 Fig. 5—Long-term member response comparison.
of curvature along the member length from sustained load
is different from that of shrinkage, particularly when elastic
deflection equations are used to compute deflection. Short-
term deflection plus the effects of creep from sustained
loading are computed with the long-term effective moment of
inertia I̅ e using Eq. (11a), while deformation from shrinkage
Fig. 4—Equivalent tension force model used to compute is based on an effective shrinkage curvature ϕsh,e defined by
shrinkage curvature for section with tension reinforcement As. Eq. (11b) that also depends on the level of loading relative to
the depth of the section. For a member with tension rein- the cracking load as seen in Fig. 3. Branson (1965, 1977) and
forcement only others (Walkup et al. 2019) have typically based shrinkage
deformation calculations on the uncracked flexural member
nAs etr nAs eg only.
φ sh ,tr = ε sh ≈ ε sh (12a) For loads less than the defined cracking load moment,
I tr Ig
the effective shrinkage curvature equals the shrinkage
curvature ϕsh,tr of the uncracked section. At higher loads,
nAs ecr the shrinkage curvature equals a weighted average of the
φ sh , cr = ε sh (12b)
I cr uncracked and cracked shrinkage curvature values that grad-
ually approaches the shrinkage curvature ϕsh,cr of the fully
The eccentricities e̅ tr and e̅ cr of the equivalent tension
cracked section as the applied load becomes increasingly
force are taken relative to the centroids of the uncracked and
greater than the cracking load (Fig. 3). Hence, deformation
cracked long-term transformed sections respectively. For a
from both creep and shrinkage are dependent on the extent
more exact approach an age-adjusted effective modulus of
of flexural cracking which is directly linked to the magni-
concrete equal to Ec/(1 + χCt) that includes an aging coef-
tude of service load moment relative to the reduced cracking
ficient χ (assumed equal to 0.8) can be used to account for
moment value.
the gradual development of restraint stress in the concrete as
Figure 5 compares the proposed approach using both
shrinkage increases with time (Gilbert 1988).
the exact tension stiffening factor βts = (Mcr′ + ΔMsh)/(Ma +
The actual E̅cI̅ cr response is initially nonlinear because
ΔMsh) and approximate tension stiffening factor βts ≈ Mcr′/
of the axial load from the tensile shrinkage force Tsh acting
Ma with Mcr′ = 2/3Mcr. Comparison is made with the Euro-
on the cracked section, but converges quickly to the slope
code 2 approach (CEN 2004; fib 2013), which is similar to
of the shifted E̅cI̅ cr response as shown in Fig. 2(b). This
the proposed approach but slightly more conservative, as
behavior is similar but opposite to a prestressed member
discussed in the Appendix. The ACI 318 equations for fr
with a compressive axial force from the prestressing acting
(used to compute Mcr) and Ec are used for each comparison
on the cross section (Bischoff et al. 2018). Shrinkage can
including for Eurocode 2. For the typical example shown in
be considered as negative prestressing which decreases the
Fig. 5, approximating the tension stiffening factor βts with
cracking moment, causes an initial downwards deflection in
Mcr′/Ma increases deformation by approximately 4% at full
the uncracked member, and a shift in the E̅cI̅ cr response (rela-
service load (in this example, the service load moment Ma
tive to the E̅cI̅ tr response) to the right and downwards.
= 2.4Mcr), and by another 3% with the approximation made
in Eurocode 2. Differences between these approaches are
Discussion
not great and are affected primarily by the reinforcing ratio,
Figure 3 demonstrates how the proposed approach for
free shrinkage strain value and creep coefficient. Differ-
computing long-term deformation treats the time-dependent
ences related to the specific equations used by ACI 318
effects from creep and shrinkage separately. Separating the
and Eurocode 2 for the cracking stress (which affects Mcr)
deflection components is advantageous when the variation

ACI Structural Journal/March 2021 219

and Ec are not considered (refer to the Appendix for more slabs were simply supported with a 3.5 m (11.5 ft) span.
details about these differences). The values adopted for Ct The beams were loaded with two sustained point loads
and εsh are highly variable and play an important role also in applied at the third points, while the slabs were loaded with
affecting computed values of deformation. Deflection calcu- a superimposed load that was uniformly distributed. Load
lations at the design stage can easily vary by ±30% or more, from the member self-weight is included in the analysis.
and should be intended for control of deflection only unless The beams had a square cross-section 250 mm (9.8 in.) wide
specific details of the construction process, loading history, by either 333 or 348 mm (13.1 or 13.7 in.) high (depending
and material properties (including shrinkage and creep) can on the concrete cover) and reinforcing ratio of 0.53% and
be well controlled. 0.8%. The slab strips had a cross section 400 mm (15.75
in.) wide by 160 mm (6.3 in.) thick and reinforcing ratio of
Deflection under sustained loading 0.42%, 0.63%, and 0.84%. Hence, all members were lightly
Long-term deflection under sustained loading accounts for reinforced. Each pair of identical beams was subjected to
preloading from construction loads by computing the long- a sustained load corresponding to either 26 to 30% or 42
term effective moment of inertia I̅ e and effective shrinkage to 44% of the ultimate moment capacity at midspan (1.2
curvature ϕsh,e using the full dead MD plus live ML service to 1.5Mcr and 1.8 to 2.5Mcr), while each pair of identical
load moment. In other words, I̅ e and ϕsh,e are computed using slabs was loaded at 32 to 39% and 44 to 50% of the ulti-
Eq. (11a) and (11b), respectively, with Ma = MD + ML. Long- mate moment capacity at midspan (1.15 to 1.7Mcr and 1.5
term deflection from the sustained load is computed using to 2.3Mcr).
the elastic deflection equation Δ = K(5/48)MsusL2/E̅cI̅ e and The beams and slabs were moist-cured for 14 days and
added to the computed deflection from shrinkage using Eq. then loaded with a constant sustained load for a further 380
(7a) with ϕsh,e being substituted for ϕsh to give Δsh = kshϕsh,eL2. days. Temperature was not controlled and varied from 11 to
The sustained load moment Msus = MD + ML,sus. 24°C (52 to 75°F) during the period of testing. Humidity was
not reported and was likely not controlled either. Material
EVALUATION AND COMPARISON WITH TEST tests gave an average compressive strength fc of 18.3 MPa
RESULTS (2650 psi) and concrete modulus Ec of 22.82 GPa (3310 ksi)
The presented methodology is validated by comparison at the time of loading (14 days). Gilbert and Nejadi (2004)
with long-term tests from Jaccoud and Favre (1982) and used a computed value of fr = 0.6√fc (2.6 MPa or 375 psi) to
Gilbert and Nejadi (2004). Jaccoud and Favre tested seven compute the cracking moment Mcr for these tests because
simply supported slab strips with a span of 3.1 m (10.2 ft) of difficulties with matching the observed cracking moment
at five different levels of sustained load corresponding to with computed values based on the measured rupture
a midspan moment between 20 and 60% of the ultimate modulus fr at the time of loading. Efforts were made to
capacity. The magnitude of load relative to the cracking minimize shrinkage before loading by keeping the concrete
moment (0.71Mcr to 2.15Mcr) affects the level of cracking moist cured up to the time of loading at 14 days. The creep
in the member and therefore the tension stiffening effect. coefficient Ct had a measured value of 1.71 after 380 days
The slab strips were subjected to four-point loading (near of loading and relatively high shrinkage strain εsh of 825 ×
the third points) in addition to the member self-weight. 10–6. The ACI 318 time-dependent factor ξ equals 1.44 for
Specimens were identical and had a cross-section 750 mm these long-term deflection calculations. The elastic modulus
(29.5 in.) wide by 160 mm (6.3 in.) thick, tension reinforce- of steel Es is assumed to equal 200 GPa (29,000 ksi).
ment ratio of 0.58%, and compression reinforcement ratio of Computed values of long-term deflection (Δcalc) are
0.06%. The specimens were moist cured for the first 7 days compared with experimental values (Δexp) in Fig. 6 and
prior to being exposed in a laboratory-controlled environ- 7, showing lines for perfect prediction (Δcalc/Δexp = 1) and
ment at 21 ± 1°C (70 ± 1.8°F) and 60 ± 5% relative humidity. ±30% deviation (Δcalc/Δexp = 1.3 and 0.7). Statistical details
Load was applied at 28 days and sustained for either 365 or of the accuracy for all approaches evaluated are provided
510 days, depending on the specimen tested. in Table 1 for each study and for the two studies combined,
Average values of measured concrete properties at the age with mean values of the prediction ratio (Δcalc/Δexp), standard
of loading are provided for cylinder strength fc (30.6 MPa deviation (St. Dev.) and coefficient of variation (COV).
[4450 psi]), modulus of elasticity Ec (29.5 GPa [4280 ksi]), Comparison in Fig. 6 for the proposed approach and
and tensile strength fr (2.8 MPa [406 psi]). The concrete Eurocode 2 approach shows that both provide close agree-
creep coefficient Ct had an average measured value of 2.31 at ment mostly within a few percent of the measured deflec-
365 days and 2.39 at 510 days, while the measured shrinkage tion values. Both models provide a consistent prediction of
strain εsh was 268 × 10–6 at 365 days and 320 × 10–6 at 510 deformation with a reasonably low coefficient of variation
days. The ACI 318 time-dependent factor ξ used to compute for these tests. However, it is important to realize the signifi-
long-term deflection equals 1.4 for the specimens with a load cance of Mcr on computed values of deflection. For example,
duration of 365 days (12 months) and 1.55 at 510 days (17 using the measured rupture modulus fr at the time of loading
months) of loading. The steel reinforcement had an elastic (3.7 MPa or 537 psi) to compute Mcr for Gilbert and Nejadi’s
modulus Es of 205 GPa (29,730 ksi). Deflection calculations (2004) tests leads to an unconservative prediction of deflec-
are based on the measured material properties. tion with the proposed approach giving Δcalc/Δexp = 0.84 on
Gilbert and Nejadi (2004) carried out their long-term average (compared to Δcalc/Δexp = 0.98 from Table 1) because
tests on six beams and six slab strips. Both the beams and of Mcr having too high a value.

220 ACI Structural Journal/March 2021

 Table 1—Statistical comparison of calculated and
experimental deflections
Δcalc/Δexp
Standard
Approach Test series Mean deviation COV
A 0.97 0.143 14.7%
Proposed B 0.98 0.058 5.8%
A+B 0.98 0.094 9.6%
A 0.98 0.159 16.2%
Eurocode 2 B 1.05 0.066 6.3%
A+B 1.02 0.110 10.8%
A 1.34 0.375 27.9%

Fig. 6—Long-term deflection comparison with tests using ACI 318-19 B 1.17 0.079 6.8%
proposed approach and Eurocode 2. A+B 1.23 0.241 19.6%
A 0.68 0.278 41.1%
ACI 318-14 B 0.76 0.229 30.0%
A+B 0.73 0.244 33.4%

Note: Test Series A: Jaccoud and Favre (1982); Test Series B: Gilbert and Nejadi
(2004).

CONCLUSIONS
A rational approach is presented for computing long-term
deflections of reinforced (nonprestressed) concrete. The
proposed approach is based on the same principles used
to formulate an effective moment of inertia Ie adopted by
ACI 318-19 for computing short-term deflection. Separate
components are developed for creep and shrinkage deflec-
tion using a coherent model for tension stiffening that
encompasses both phenomena as follows:
Fig. 7—Long-term deflection comparison with tests using • Creep causes a reduction of stiffness in the concrete
ACI 318-19 and ACI 318-14. with time that is reflected by a decrease in stiffness of
Figure 7 provides a comparison of test data with ACI the load-deformation response for the member in both
318-14 using Branson’s (1965) equation for Ie at full Mcr the uncracked and cracked states. This leads to a long-
and with ACI 318-19 using Bischoff’s (2005) equation for term effective moment of inertia I̅ e used to compute
Ie at 2/3Mcr. ACI 318-14 underpredicts deflection by 27% deformation that includes both the short-term (imme-
on average for both sets of tests while ACI 318-19 over- diate) deformation and long-term deformation from
predicts deflection by 23% on average. ACI 318-19 gives creep under sustained loads.
a more flexible response and hence more deflection than • Shrinkage causes a shift in the member response modi-
ACI 318-14 because of the reduced cracking moment and fied by creep (in the direction of increasing deforma-
problems associated with Branson’s equation for Ie giving tions), in addition to a reduction of the effective cracking
too stiff a response for lightly reinforced members (Bischoff moment because of tensile stresses in the concrete that
2020). The accuracy of the two ACI 318 approaches is obvi- originate from restraint to shrinkage by the internal rein-
ously also dependent on the long-term deflection multiplier. forcement and supports. The reduced cracking moment
Using a single multiplier for the combined effects of creep decreases tension stiffening, which affects both creep
and shrinkage has been criticized in the past as the multi- and shrinkage deformation.
plier value is influenced by numerous factors that include Current ACI 318 provisions for time-dependent deflec-
the creep and shrinkage characteristics of the concrete, tions apply a single multiplier to the immediate deflection
curing and environmental conditions, age of concrete at first caused by sustained load. This approach is simple and easy
loading, level of sustained load, and nρ, which is a func- to apply but does not account for the mechanics of sectional
tion of the reinforcing ratio and elastic modulus of the rein- behavior nor the many factors affecting long-term deforma-
forcement (Gilbert 1999; Gilbert and Nejadi 2004; Marí et tion such as environmental conditions, the many types of
al. 2010; Torres at al. 2012). Others have commented on the current concretes (compressive strength, type of cement,
influence of the Ig/Icr ratio, controlled primarily by nρ, on the dosage, and so on), age at loading and level of loading,
deflection multiplier (Walkup et al. 2019). reinforcement ratios, and relationship between the moduli
of elasticity of the reinforcement and concrete. While
creep deflections depend on the sustained load for which

ACI Structural Journal/March 2021 221

a simple multiplier might be applicable, deformation from ACI Committee 435, 1978, “Proposed Revisions by Committee 435
to ACI Building Code and Commentary Provisions on Deflections,” ACI
shrinkage depends only partly on the level of applied load as Journal Proceedings, V. 75, No. 6, June, pp. 229-238.
demonstrated with the proposed model. Deformation from ACI Committee 435, 2020, “Deflection of Nonprestressed Concrete
shrinkage is also influenced by the amount of shrinkage Structures (ACI 435R-20),” American Concrete Institute (ACI), Farm-
ington Hills, MI, 56 pp.
and amounts of reinforcement in relation to the specimen Balázs, G. L.; Bisch, P.; Borosnyói, A.; Burdet, O.; Burns, C.; Ceroni, F.;
cross-section details. Cervenka, V.; Chiorino, M. A.; Debernardi, P.; Eckfeldt, L.; El-Badry, M.;
Comparison of the proposed approach with the ACI 318 and Fehling, E.; Foster, S. J.; Ghali, A.; Gribniak, V.; Guiglia, M.; Kaklauskas,
G.; Lark, R. J.; Lenkei, P.; Lorrain, M.; Marí, A.; Ozbolt, J.; Pecce, M.;
Eurocode 2 approaches is made and evaluated with a limited Pérez Caldentey, A.; Taliano, M.; Tkalcic, D.; Torrenti, J. M.; Torres, L.;
set of long-term deflection data. The proposed approach and Toutlemonde, F.; Ueda, T.; Vitek, J. L.; and Vráblík, L., 2013, “Design for
method used by Eurocode 2 provide close agreement within SLS According to fib Model Code 2010,” Structural Concrete, V. 14, No. 2,
pp. 99-123. doi: 10.1002/suco.201200060
a few percent of the test results. ACI 318-14 underpredicts Bischoff, P. H., 2005, “Revaluation of Deflection Prediction for Concrete
the defection results by 27% on average while ACI 318-19 Beams Reinforced with Steel and Fiber Reinforced Polymer Bars,” Journal
overpredicts the same set of data by 23% on average. Recall of Structural Engineering, ASCE, V. 131, No. 5, pp. 752-767. doi: 10.1061/
(ASCE)0733-9445(2005)131:5(752)
that ACI 318-14 uses Branson’s (1965) expression for Ie at Bischoff, P. H., 2020, “Comparison of Existing Approaches for
full Mcr to compute immediate deflection which is multiplied Computing Deflection of Reinforced Concrete,” ACI Structural Journal, V.
by the deflection multiplier to get the time-dependent deflec- 117, No. 1, Jan., pp. 231-240. doi: 10.14359/51718072
Bischoff, P. H., and Darabi, M., 2012, “Unified Approach for Computing
tion, while ACI 318-19 uses Bischoff’s (2005) expression for Deflection of Steel and FRP-Reinforced Concrete,” Andy Scanlon Sympo-
Ie at 2/3Mcr with the same long-term deflection multiplier. sium on Serviceability and Safety of Concrete Structures: From Research
The proposed approach requires the creep coefficient Ct and to Practice, SP-284, P. H. Bischoff and E. Musselman, eds., American
Concrete Institute, Farmington Hills, MI.
shrinkage strain εsh to compute long-term deflection. Use Bischoff, P. H., and Gross, S. P., 2011, “Design Approach for Calcu-
of a correct cracking moment that accounts for shrinkage lating Deflection of FRP Reinforced Concrete,” Journal of Composites
restraint has a significant effect on the accuracy of computed for Construction, ASCE, V. 15, No. 4, pp. 490-499. doi: 10.1061/(ASCE)
CC.1943-5614.0000195
values for both short- and long-term deflection that cannot Bischoff, P. H.; Naito, C. J.; and Ingaglio, J. P., 2018, “Immediate Deflec-
be overemphasized. It is recommended the alternative tion of Partially Prestressed Concrete Flexural Members,” ACI Structural
detailed approach permitted by ACI 318-19 be replaced with Journal, V. 115, No. 6, Nov., pp. 1638-1693. doi: 10.14359/51702381
Branson, D. E., 1965, Instantaneous and Time-Dependent Deflections
the proposed approach and be used as the basis to develop a of Simple and Continuous Reinforced Concrete Beams, HPR Report No.
set of more rational long-term deflection multipliers. 7, Part 1, Alabama Highway Department, Bureau of Public Roads, Mont-
gomery, AL., 78 pp.
Branson, D. E., 1977, Deformation of Concrete Structures, McGraw-
AUTHOR BIOS Hill, New York, NY, 546 pp.
Peter H. Bischoff, FACI, is a Professor in the Civil Engineering Depart-
CEN, 2004, “Design of Concrete Structures - Part 1-1: General Rules
ment at the University of New Brunswick, Fredericton, NB, Canada. He
and Rules for Buildings (Eurocode 2), European Standard BS EN 1992-1-
is a member of ACI Committees 224, Cracking; 360, Design of Slabs
1,” European Committee for Standardization, Brussels, Belgium.
on Ground; 435, Deflection of Concrete Building Structures; and 440,
Comité Euro-International du Béton (CEB), 1993, CEB-FIP Model Code
Fiber-Reinforced Polymer Reinforcement; and Joint ACI-ASCE Committee
1990 (MC-90), Thomas Telford, London, UK, 437 pp.
408, Bond and Development of Reinforcement. He received the ACI Design
fib, 2013, “fib Model Code for Concrete Structures 2010,” International
Award in 2014 and ASCE T.Y. Lin Award in 2020. His research interests
Federation for Structural Concrete, Lausanne, Switzerland.
include serviceability behavior of concrete structures and design with
fib, 2019, Serviceability Limit State of Concrete Structures, fib Bulletin
fiber-reinforced concrete.
92, International Federation for Structural Concrete, Lausanne, Switzer-
land, 209 pp.
ACI member Lluis Torres is a Professor in the Mechanical Engineering and
Gilbert, R. I., 1988, Time Effects in Concrete Structures, Elsevier,
Industrial Construction Department at the University of Girona, Girona,
Amsterdam, the Netherlands, 321 pp.
Spain. He is a member of ACI Committees 224, Cracking; 435, Deflection
Gilbert, R. I., 1999, “Deflection Calculation for Reinforced Concrete
of Concrete Building Structures; and 440, Fiber-Reinforced Polymer Rein-
Structures—Why We Sometimes Get It Wrong,” ACI Structural Journal, V.
forcement. His research interests include serviceability behavior of concrete
96, No. 6, Nov.-Dec., pp. 1027-1032.
structures and fiber-reinforced polymer-reinforced concrete structures.
Gilbert, R. I., 2003, “Deflection by Simplified Calculation in AS3600-
2001- On the Determination of fcs,” Australian Journal of Structural Engi-
ACKNOWLEDGMENTS neering, V. 5, No. 1, pp. 61-71. doi: 10.1080/13287982.2003.11464928
The work presented in this paper was initiated while the first author was Gilbert, R. I., and Nejadi, S., 2004, “An Experimental Study of Flex-
on a sabbatical visit with the AMADE Group (Analysis and Advanced ural Cracking in Reinforced Concrete Members under Sustained Loads,”
Materials for Structural Design), Department of Mechanical Engineering UNICIV Report No. R-435, The University of New South Wales, Sydney,
and Industrial Construction, Universitat de Girona, Girona, Spain. Australia, 61 pp.
Jaccoud, J. P., and Favre, R., 1982, “Flèche des Structures en Béton
Armé: Vérification Expérimentale d’une Méthode de Calcul (Deflection of
REFERENCES Reinforced Concrete Structures: Experimental Verification of a Computing
ACI Committee 209, 1992, “Prediction of Creep, Shrinkage, and Method),” Annales de l’Institut Technique du Bâtiment et des Travaux
Temperature Effects in Concrete Structures (ACI 209R-92 [Reapproved Publics, Paris, No. 406, Série Béton, V. 208, pp. 23-66. (in French)
2008]),” American Concrete Institute, Farmington Hills, MI, 47 pp. Marí, A. R.; Bairán, J. M.; and Duarte, N., 2010, “Long-Term Deflec-
ACI Committee 209, 2008, “Guide for Modeling and Calculating tions in Cracked Reinforced Concrete Flexural Members,” Engineering
Shrinkage and Creep in Hardened Concrete (ACI 209.2R-08),” American Structures, V. 32, No. 3, pp. 829-842. doi: 10.1016/j.engstruct.2009.12.009
Concrete Institute, Farmington Hills, MI, 46 pp. Miller, A. L., 1958, “Warping of Reinforced Concrete Due to Shrinkage,”
ACI Committee 318, 2014, “Building Code Requirements for Struc- ACI Journal Proceedings, V. 54, No. 5, May, pp. 939-950.
tural Concrete (ACI 318-14) and Commentary (ACI 318R-14),” American Nilson, A. H., 1976, “Flexural Stresses after Cracking in Partially
Concrete Institute, Farmington Hills, MI, 520 pp. Prestressed Beams,” PCI Journal, V. 21, No. 4, pp. 72-81. doi: 10.15554/
ACI Committee 318, 2019, “Building Code Requirements for Struc- pcij.07011976.72.81
tural Concrete (ACI 318-19) and Commentary (ACI 318R-19),” American Scanlon, A., and Bischoff, P. H., 2008, “Shrinkage Restraint and Loading
Concrete Institute, Farmington Hills, MI, 624 pp. History Effects on Deflection of Flexural Members,” ACI Structural
ACI Committee 435, 1966, “Deflections of Reinforced Concrete Flex- Journal, V. 105, No. 4, July-Aug., pp. 498-506.
ural Members (ACI 435.2R-66),” ACI Journal Proceedings, V. 63, No. 6, Torres, L.; Miàs, C.; Turon, A.; and Baena, M., 2012, “A Rational
June, pp. 637-674. Method to Predict Long-Term Deflections of FRP Reinforced Concrete

222 ACI Structural Journal/March 2021

Members,” Engineering Structures, V. 40, pp. 230-239. doi: 10.1016/j. Adopting the third option for Δϕcr defined by Eq. (A2c)
engstruct.2012.02.021
Walkup, S.; Musselman, E.; and Gross, S., 2019, “Effect of Service Load
leads to a different expression for ϕa
Levels on Long-Term Deflection Multiplier,” ACI Structural Journal, V.
116, No. 2, Mar., pp. 89-100. doi: 10.14359/51711137   M ′ + ∆M sh    M a 
φ a = βts  cr   + φ sh ,tr 
APPENDIX   M a + ∆M sh    Ec I tr 
Tension stiffening curvature approach for long-   M ′ + ∆M sh    M a 
term deformation (including Eurocode 2 approach) + 1 − βts  cr    E I + φ sh , cr  (A3c)
  M + ∆M sh 
Subtracting the tension stiffening curvature Δϕts from the a c cr

long-term E̅cI̅ cr response shown in Fig. 2(a) gives which gives the same computed value of curvature as Eq.
(A3a) and (A3b) when the same tension stiffening factor βts
Ma and reduced cracking moment Mcr′ are used. The ΔMsh term
φa = + φ sh , cr − ∆φts (A1)
Ec I cr which results from shrinkage is an added complication of
using Eq. (A3c), and the relative proportions for deforma-
where Δϕts = βtsΔϕcr and the maximum tension stiffening tion caused by load and shrinkage are not the same as those
curvature at cracking Δϕcr is defined as in Eq. (A2a) and (A2b) because of the ΔMsh term.
Equation (A3c) forms the basis of the Eurocode 2 (CEN
 M′ M′ 
(
∆φcr =  cr − cr  + φ sh , cr − φ sh ,tr
 Ec I cr Ec I tr 
) (A2a) 2004) and fib Model Code 2010 (fib 2013) expressions for
computing long-term deformation. However, this requires
setting ΔMsh = 0, which is another approximation leading to
with (Mcr′/E̅cI̅ cr – Mcr′/E̅cI̅ tr) = ηMcr′/E̅cI̅ cr and η = 1 – I̅ cr/I̅ tr.
a more conservative estimate of deflection as demonstrated
Alternatively, Δϕcr can be expressed by factoring the ratio
in Fig. 5. Assuming βts = Mcr′/Ma and Mcr′ = 0.5M cr =
Mcr′/Ma out of the first pair of terms in Eq. (A2a)
0.71Mcr (compare with Mcr′ = 2/3Mcr for the proposed
approach), along with setting ΔMsh = 0, gives
M cr′  M a M 
∆φcr =  (
− a  + φ sh , cr − φ sh ,tr (A2b)
M a  Ec I cr Ec I tr 
)

2
       Ma
2

φ a = 0.5 M cr Ma M cr
or as a third option  M   E I + φ sh ,tr  + 1 − 0.5  M   + φ sh , cr 
a c tr 
 a   Ec I cr 

 M ′ + ∆M sh   M a   (A4)
M 
∆φcr =  cr   (
− a  + φ sh , cr − φ sh ,tr 
 M a + ∆M sh   Ec I cr Ec I tr 
)
 Equation (A4) is identical to the Eurocode 2 and fib Model
(A2c) Code 2010 equations for long-term deformation under
sustained loading
This third option is based on similar triangles with the
vertex located at the idealized origin positioned at ΔMsh as M M
φ a = (1 − ζ) +ζ + (1 − ζ) φ sh ,tr + ζφ sh ,cr (A5)
shown in Fig. 2(a) and (b), with Δϕcr/(Mcr′ + ΔMsh) = [(Ma/ Ec I tr Ec I cr
E̅cI̅ cr + ϕsh,cr) – (Ma/E̅cI̅ tr + ϕsh,tr)]/(Ma + ΔMsh).
where ζ = 1 – β(Mcr/M)2 with β = 0.5 for a sustained load.
Substituting Eq. (A2a) into Eq. (A1) with Δϕts = βtsΔϕcr
The concept of a reduced cracking moment equivalent to
leads to
√βMcr was introduced in the CEB-FIP Model Code 1990
(CEB 1993) and applies in Eurocode 2 when M > Mcr as
Ma  M cr′  shown in Fig. 5.
φa =
Ec I cr 1 − βts η M  + βts φ sh ,tr + (1 − βts ) φ sh , cr Similar to the proposed approach in this paper, the
a
shrinkage curvatures are defined using ϕsh = n̅ (S/I)εsh for the
(A3a)
uncracked and fully cracked sections, where S represents the
first moment of the reinforcing steel area about the centroid
which is the same as Eq. (9b) and forms the basis of the
of uncracked and cracked sections (Ase̅ tr and Ase̅ cr for the
proposed model for long-term deflection assuming βts ≈ Mcr′/
tension reinforcement), and I is the moment inertia of the
Ma and Mcr′ = 2/3Mcr.
uncracked and cracked sections (I̅ tr and I̅ cr) based on an
Substituting the alternative expression for Δϕcr defined by
effective elastic modulus of concrete E̅c. Refer to Balázs et
Eq. (A2b) into Eq. (A1) leads to a slightly different expres-
al. (2013) for a more detailed explanation.
sion for ϕa given by Eq. (A3b) which is identical to Eq. (A3a)
Specific differences between the proposed ACI approach
except for the first two of four terms in Eq. (A3b) which
and Eurocode 2 approach for long-term deformation are
represent a rearranged form of the first term in Eq. (A3a).
summarized in Table A1. In general, the tensile strength of
concrete used to compute the cracking moment is between
  M ′   M    M ′   M 
φ a = βts  cr    a  + 1 − βts  cr    a  10 to 25% greater using the ACI equation for fr (except
  M a    Ec I tr    M a    Ec I cr  for slabs and shallow beams), while the elastic modulus
 (A3b) of concrete is between 10 to 30% less with the ACI equa-
+βts φ sh ,tr + (1 − βts ) φ sh , cr
tion for Ec depending on the concrete strength. Using an

ACI Structural Journal/March 2021 223

Table A1—Differences between ACI proposed approach and Eurocode 2 approach
Proposed ACI Approach Eurocode 2 (EC2) Approach

Reduced cracking moment Mcr′ 2/3Mcr 0.71Mcr*

 M cr′    M′   M cr′ 
2
  M′ 2
ϕsh,e φ sh ,tr + 1 −  cr   φ sh ,cr φ + 1 −  cr   φ sh ,cr
 M a   M a 
sh tr
  Ma     M a  
,

Flexure fr or axial fctm tensile stress fr = 0.62√fc′ in MPa fctm = 0.3(fc′)2/3 in MPa
used to compute Mcr† fr = 7.5√fc′ in psi fctm = 1.58(fc′)2/3 in psi

 f ′+ 8 
0.3

4730√fc′ in MPa 22, 000  c in MPa

Ec  10 
57,000√fc′ in psi
359,300(fc′ + 1160)0.3 in psi
*
EC2 uses a factor β = 0.5 for sustained and repeated loading (for Ma > Mcr) causing a shift in the deformation response equivalent to using a reduced cracking moment value √βMcr
= 0.71Mcr when β = 0.5.
†
EC2 permits the flexural tensile strength fctm,fl = [1.6 – h/1000]fctm ≥ fctm to compute Mcr in the absence of axial tensile stresses such as those caused by shrinkage or thermal effects
(h is in mm with 1 in. = 25.4 mm).

increased value of the cracking moment compared to Euro- curvature ϕsh,e arise from the approximation made by setting
code 2 decreases the computed value of deformation but ΔMsh = 0 in the Eurocode 2 approach. Figure 5 shows the
this is counteracted with using a lower value of elastic effect this approximation has on the difference between
modulus of concrete which increases deformation. Differ- deformation values computed with the two approaches.
ences between the expressions for the effective shrinkage

224 ACI Structural Journal/March 2021