ACI STRUCTURAL JOURNAL TECHNICAL PAPER

Title No. 114-S106

ACI 318-14 Criteria for Computing Instantaneous Deflections

by Caitlin Mancuso and F. Michael Bartlett

This paper investigates current procedures to compute instan-

taneous deflections of reinforced concrete beams. Deflections
predicted using the Branson Equation with the cracking moment
computed using the full modulus of rupture, in accordance with
ACI 318-14, when compared to deflections of 65 simply supported
and continuous beams tested by others, are increasingly uncon-
servative and variable as the flexural reinforcement ratio reduces
below 1%. Using the Branson Equation with the cracking moment
computed using one-half the modulus of rupture, or the Bischoff
Equation with the cracking moment computed using two-thirds
the modulus of rupture give consistent and conservative deflection
predictions, whether the beam is idealized as a single element or as
a number of discretized elements. The Bischoff Equation can accu- Fig. 1—Variation of Ie/Ig with reinforcement ratio (after
rately compute the effective moment of inertia for members ideal- CAC 2016).
ized using discretized elements if the exponent in its denominator
is increased from 2 to 3. The cracked rigidity EcIcr of concretes that
uncracked regions as springs in parallel, when in fact they are
are at last 1 day old can be accurately estimated using conven- in series (Bischoff 2007). Therefore, the Branson Equation
tional assumptions and calculation procedures. For discretized overestimates the tension-stiffening effect and may so under-
element idealizations, the use of 20 equal-length elements per span estimate deflections, especially for lightly reinforced members
is sufficiently accurate for practical use. (CAC 2016). Bischoff therefore also proposed the following
equation, the “Bischoff Equation,” based on the correct
Keywords: Bischoff Equation; Branson Equation; early-age concrete prop- mechanical model, for members idealized as single elements
erties; effective moment of inertia; moment-curvature analysis.

I cr
INTRODUCTION Ie = (2)
Flexural cracking occurs in a reinforced concrete member   I   M 2
1 − 1 − cr   cr  
at discrete locations where the applied moment exceeds   Ig   Ma  

the cracking moment. Deflections are therefore computed
using an effective moment of inertia, Ie, that accounts for For Mcr/Ma > 1, the numerator of Eq. (2) can be negative
the “tension stiffening effect” (for example, Wight [2015]) and, in this case, Ie is taken as Ig. It is unclear whether the
of the uncracked regions between cracks. Branson (1965) Bischoff Equation requires modification for use when the
proposed the following equation for the effective moment member is idealized as discrete elements.
of inertia that has been adopted in the 1971 and subsequent Flexural members are subjected to additional tensile
editions of the ACI 318 Code stresses due primarily to restraint of concrete shrinkage. CSA
Standard A23.3-14, “Design of Concrete Structures,” (CSA
3
  M 3 2014) permits use of the Branson Equation to compute Ie but
M 
I e =  cr  I g + 1 −  cr   I cr (1) requires for this calculation that the cracking moment, Mcr,
 Ma    M a   be computed using a reduced modulus of rupture, 0.5fr, for
beams and one-way and two-way slabs. ACI 318-14 permits
where Ig is the gross moment of inertia of the uncracked the use of the Branson Equation with the cracking moment
section; Icr is the moment of inertia of the cracked section; computed using the full modulus of rupture, fr. To account
Mcr is the cracking moment; and Ma is the applied moment for the effect of restrained concrete shrinkage, Scanlon and
due to service loads. Equation (1), the “Branson Equation,” Bischoff (2008) recommend that Mcr be calculated using
provides an average effective moment of inertia intended to 0.67fr for use with the Bischoff Equation.
be used when the member is idealized as a single element for Figure 1 (adapted from CAC 2016) shows the variation
analysis. If the member is idealized as a number of discret- of Ie/Ig with the flexural reinforcement ratio ρ for simply
ized elements, Branson recommended computing effective supported beams as computed using the Branson and
moments of inertia for each element, based on the maximum
applied moment within the element, using a revised version ACI Structural Journal, V. 114, No. 5, September-October 2017.
MS No. S-2016-390, doi: 10.14359/51689726, was received November 7, 2016, and
of Eq. (1) with the exponent increased from 3 to 4. reviewed under Institute publication policies. Copyright © 2017, American Concrete
The Branson Equation is based on an incorrect mechan- Institute. All rights reserved, including the making of copies unless permission is
obtained from the copyright proprietors. Pertinent discussion including author’s
ical model that represents the rigidities of the cracked and closure, if any, will 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/September-October 2017 1299

Fig. 2—Stress-strain relationships for young-aged concretes: (a) 9.7 hours; and (b) 1 day.
Bischoff Equations, Eq. (1) and (2), respectively, using the 2. Determine the optimal element length in a discretized
full and reduced moduli of rupture. The figure is derived by: analysis for design office use that balances the desire for
computing the ultimate moment capacity for the given ρ; accuracy with the demands of computational resources.
assuming the service dead and live loads have equal magni- 3. Determine necessary modifications for the Bischoff
tudes; and computing the associated Ma for use in Eq. (1) Equation to allow its application to members idealized using
and (2). For ρ greater than 0.8%, which is typical of beams, discretized elements.
the four approaches yield virtually identical results: each 4. Conduct a sensitivity analysis of the deflections obtained
requires Ie to approach Icr. For ρ between 0.3 and 0.7%, using the Branson and Bischoff Equations for single-element
which is more typical of one- and two-way slabs, the Ie value or multiple-element idealizations using the full or reduced
computed using the Branson Equation, Eq. (1), with the full moduli of rupture.
fr is markedly greater than that computed using the Bischoff 5. Investigate the accuracy of deflections computed using
Equation, Eq. (2), with the full fr. For ρ less than 0.8%, the Ie the Branson and Bischoff Equations for single- or discret-
computed using the Branson Equation with 0.5fr is markedly ized-element idealizations with the full or reduced moduli of
less than that computed using the Branson Equation with the rupture using experimental data reported by others.
full fr, and roughly equivalent to that computed using the
Bischoff Equation with 2/3fr. Thus, the provisions of ACI RESEARCH SIGNIFICANCE
318-14 (ACI Committee 318 2014) for lightly reinforced The research reported in this paper demonstrates that
flexural members are quite unconservative with respect to the provisions in ACI 318-14 for computing instantaneous
these other procedures. deflections are unconservative for lightly reinforced one-
The following two additional questions arise concerning and two-way slabs. It proposes revisions to procedures using
the use of either the Branson or the Bischoff Equation for the Branson or Bischoff Equations, when the member is
computing instantaneous deflections: idealized as a single element or as discretized elements, that
1. Construction schedules may require load to be applied are justified by comparison to test data reported by others. It
to concrete members at very young ages and so impact the provides guidance concerning the accuracy of the cracked
fundamental stress-strain response assumed for concrete flexural rigidity, EcIcr, computed using conventional assump-
in compression and the relationship between the elastic tions and calculation procedures for young-aged concretes.
modulus and the compressive strength. Therefore, there is
a need to quantify any differences between the compressive IMPACT OF STRESS-STRAIN RESPONSE OF
stress-strain responses of young-aged and mature concretes. YOUNG-AGED CONCRETES
It is also necessary to verify that conventional idealizations Experimental stress-strain data obtained by others (Khan
and simplifications conventionally assumed for flexural 1995; Jin et al. 2005) indicate that the stress-strain rela-
analysis apply to young-aged concretes, and in particular tionships for very young concrete can differ from those for
that the ascending portion of the moment-curvature response 28-day-old concrete as conventionally idealized using the
is accurately approximated by the cracked flexural rigidity, Todeschini (1964), Modified Hognestad (1951), and other
EcIcr, computed using conventional methods. relationships (for example, Wight [2015]). Khan (1995)
2. There are no guidelines available to determine a suit- investigated the compressive stress-strain responses of 30,
able element length when idealizing a single member as a 70, and 100 MPa (4400, 10,000, and 14,500 psi) strength
number of discretized elements. A shorter element length concretes at ages from 72 hours to 91 days, testing a total
will increase the accuracy of the results, but at the cost of of approximately 300 cylinder specimens. Typical stress-
greater computational demand. strain responses observed at concrete ages of 9.7 hours and
The objectives of the research reported in this paper are 1 day are shown in Fig. 2(a) and (b), respectively, with the
therefore as follows: Todeschini and Modified Hognestad stress-strain relation-
1. Investigate the impact of the stress-strain response for ships corresponding to the maximum observed stress. For
young-aged concrete on the cracked flexural rigidity, EcIcr, concrete that is 9.7 hours old, both the Todeschini and Modi-
computed using conventional methods. fied Hognestad relationships overestimate the stress in the

1300 ACI Structural Journal/September-October 2017

Fig. 3—Moment-curvature response for young-aged concretes, ρ = 0.5%: (a) 9.7 hours; and (b) 14.5 hours.
ascending portion of the curve and erroneously predict The 14.5-hour old concrete, Fig. 3(b), has sufficient
significant strain softening. When the concrete is 1 day compressive strength to ensure steel yield at ultimate, and
old, (Fig. 2(b)), these shortcomings tend to disappear, as the the ultimate nominal moment computed using the conven-
observed response is initially stiffer and exhibits strain soft- tional stress block idealization corresponds well to that
ening after the maximum stress is achieved. Jin et al. (2005) computed from the observed stress-strain data. The four
investigated the compressive stress-strain responses of high- computed moment-curvature relationships are in close
and normal-strength concretes at ages between 12 hours and agreement, indicating that calculations based on the conven-
28 days and made similar observations. tional simplifying assumptions are valid and idealizing the
Moment-curvature analyses were performed to determine compressive stress-strain response using the Todeschini
the impact of the young-age concrete compression stress- or Modified Hognestad relationships is appropriate at this
strain relationship on the flexural response for cross sections concrete age.
with a range of reinforcement ratios. The applicability of The flexural rigidity, EcIcr , of the cracked member is a
simplifying approximations conventionally adopted for flex- necessary quantity for deflection calculations. ACI 318-14
ural analysis were investigated, specifically: linear-elastic permits the secant elastic modulus of normalweight concrete
response before cracking; linear-elastic cracked response to be computed as
before yield of the reinforcement; and idealization using the
concrete compressive stress block at ultimate. Figures 3(a) Ec = 57, 000 f c′ (3)
and (b) show the moment-curvature responses for a lightly
reinforced section, ρ = 0.5%, computed by idealizing the where fc′ and Ec are in psi. Elastic secant moduli, for a range
concrete compression stress-strain relationships using the of stresses from zero to 40% of the compressive strength,
conventional simplifying assumptions, the Todeschini and were computed from the stress-strain data reported by Khan
Modified Hognestad relationships, and the data reported by (1995) and Jin et al. (2005). At very young ages, Eq. (3) over-
Khan (1995) for concrete ages of 9.7 and 14.5 hours, respec- estimates the elastic modulus by up to 30% (Mancuso 2016).
tively. For the 9.7-hour old concrete, Fig. 3(a), analysis using Figures 4(a) and (b) show the moment-curvature relation-
the conventional simplifying assumptions indicates that the ship calculated for concrete ages of 9.7 hours and 1 day,
extreme-fiber concrete compression stress at the cracking respectively, using the Khan (1995) stress-strain data and the
moment is 66% of the concrete compressive strength f0. At flexural rigidity, EcIcr. For the latter, Ec is computed using
greater curvatures, the assumption of a linear-elastic stress- Eq. (3) and Icr is computed for a rectangular cross section as
strain relationship for concrete in compression is even less
justifiable, and so is not shown. Estimating the yield moment b(kd )3
+ nAs ( d − kd ) (4)
2
assuming linear-elastic cracked section analysis is incorrect I cr =
3
because the associated extreme-fiber concrete compression
stress greatly exceeds f0. At ultimate, failure is initiated by where b and d are the width of the section and the effective
the concrete crushing in compression: the nominal capacity depth of the reinforcing steel; n is the modular ratio Es/Ec; As
computed using Whitney’s Equation (1937) for compres- is the reinforcing steel area; and
sion-initiated failure is shown and is conservative with
respect to the moment-curvature response computed using k= (nρ)2 + 2nρ − nρ (5)
the stress-strain relationship observed experimentally by
Khan (1995). The moment-curvature response predicted In both cases, the computed flexural rigidities correlate
using the Todeschini stress-strain relationship markedly well to the moment-curvature relationship at lower moment
underestimates the ultimate capacity predicted using the values. This may seem surprising given that Eq. (3) from
experimental data, as it incorrectly captures significant strain ACI 318-14 tends to overestimate the observed elastic secant
softening as shown in Fig. 2(a). The response predicted modulus for very young concretes. As shown in Table 1,
using the Modified Hognestad relationship provides a better however, the greater Ec computed using Eq. (3) reduces the
approximation because the implied strain softening is less.

ACI Structural Journal/September-October 2017 1301

Fig. 4—Flexural rigidity-based moment-curvature relationships (Khan).

Table 1—Flexural rigidity of young-age concretes

9.7-hour-old concrete 1-day-old concrete
Ec, MPa n Icr, 10 mm
6 4
EcIcr, kN.m2
Ec, MPa n Icr, 106 mm4 EcIcr, kN.m2
Khan data 5000 40.0 3830 19,100 14,900 13.4 1770 26,300
Conventional methodology 6520 30.7 3220 21,000 15,300 13.1 1730 26,500

modular ratio, n, and so causes the cracked moment of inertia, Figure 5(b) shows the variation of Ie/Ig computed using
Icr, to be lower. The resulting product, EcIcr, computed using different idealizations for the case with ρ of 0.5% and wL/wD
Eq. (3), (4), and (5), is only 9.6% greater than that obtained of 0.5, which typically shows the greatest variation between
from the moment-curvature relationship derived using the the results obtained using the different idealizations. On
Khan (1995) stress-strain data. the left side of the figure, the Ie/Ig values for the constant-
length 50-element-per-half-span idealization more accu-
Mesh sensitivity analysis rately simulate the uncracked region of the member than the
When a member is idealized using discretized elements, coarser constant-length five-element-per-half-span ideal-
choosing the optimal element length is necessary to balance ization. Similarly, the comparison for the variable-length
the accuracy of the results with the computational require- element idealizations on the right side of the figure again
ments. Using larger numbers of shorter elements yields more indicates that the shorter element lengths—six elements in
accurate results, but at a higher computational cost, so the the negative moment region and 12 in the positive moment
sensitivity of the discretized analysis result to the element regions—yield more accurate results than the longer element
length was investigated. lengths—two elements in the negative moment region and
The case of a three-span continuous member was explored three in the positive moment region.
first. The member was designed for the steel reinforce- Figure 5(c) shows the computed curvatures, Ψ = M/EcIe,
ment at the negative moment regions over the supports to normalized by the maximum negative curvature at the
have reinforcement ratios ρ– of 0.5%, 1%, and 1.5%. Live- support. On the left side, the five-element idealization yields
to-dead load ratios wL/wD of 0.5, 1.0, and 1.5 were then smaller curvatures in the positive moment region and larger
assumed to design the steel reinforcement in the positive curvatures in the negative moment region, compared to the
moment regions in the span and to determine service loads 50-element idealization. The total midspan deflection is
for computing deflections. For these nine different loading the summation of the areas under the left or right half of
and reinforcement ratio combinations, various idealizations the curvature diagram: the coarse five-element-per-half-
using constant-length or variable-length elements were span idealization causes the curvatures at the support and
investigated. In each case, deflections were computed, for midspan regions to be overestimated and underestimated,
the case of dead load on all spans and live load on the interior respectively, so the resulting computed midspan deflection
span only, using the Branson fourth-power equation because is less than that computed using the 50-element-per-half-
it has been previously verified for analysis using discretized span idealization. Similarly, the curvatures associated with
elements (Branson 1965). the coarser variable-length idealization result in a smaller
Figure 5(a) shows the bending moment distribution along midspan deflection as shown on the right side of the figure.
the interior span, normalized as a fraction of the maximum Figures 5(d) and (e) show the midspan deflection incre-
simple span moment Mo = (wD + wL)L2/8. The higher ments computed using the constant-length and variable-length
moment gradients in the negative support moment regions element idealizations, respectively. In each case, the total
suggest that smaller element lengths may be necessary in midspan deflection is equal to the summation of the deflection
these regions, so idealizations with variable-length elements increments on either side of midspan. The coarser constant-
were investigated. length and variable-length element idealizations, 5 and 2–_3+
per half span, respectively, cause the increments of deflec-

1302 ACI Structural Journal/September-October 2017

Fig. 5—Analysis of discretized three-span beam with various element lengths (Branson).
tion to be underestimated in the positive moment region and computed using the 50-element idealization is always
slightly overestimated in the negative moment region. The greater than those computed using the coarser idealiza-
associated computed midspan deflections are therefore less tions. Decreasing the number of elements will therefore
than those computed using the finer idealizations. cause the computed deflection to be increasingly uncon-
In Table 2, the maximum deflection computed using five, servative, which is not desirable. For the constant-length
seven, 10, and 20 equal-length elements per half span and element idealizations, this unconservative error is less than
four combinations of equal-length elements in the posi- 6% if at least five elements are using in each half-span. For
tive and negative moment regions are compared to those practical design office applications, use of an idealization
computed using the 50-element idealization. All the differ- with 10 constant-length elements per half-span is probably
ences are positive, indicating that the midspan deflection satisfactory as the maximum unconservative error is 2.2%,

ACI Structural Journal/September-October 2017 1303

Table 2—Sensitivity of maximum deflection to element length
Percent difference with respect to 50 element result
wL/wD = 0.5 wL/wD = 1.0 wL/wD = 1.5
No. of elements ρ = 0.5% ρ = 1.0% ρ = 1.5% ρ = 0.5% ρ = 1.0% ρ = 1.5% ρ = 0.5% ρ = 1.0% ρ = 1.5%
50 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
20 0.2 0.2 0.2 0.2 0.1 0.1 0.2 0.1 0.2
10 2.2 0.6 1.3 1.0 0.7 0.3 0.6 0.5 0.8
7 4.3 2.2 1.2 2.2 0.5 1.9 1.7 1.3 0.8
5 5.7 3.8 1.6 3.9 2.1 3.0 2.8 2.1 3.7
6–_12+ 0.8 0.3 0.0 0.3 0.3 0.1 0.2 0.2 0.1
4–_12+ 1.6 0.1 0.6 0.5 0.1 0.3 0.3 0.2 0.2
6 _7
– +
1.4 0.4 0.6 0.4 0.4 0.4 0.3 0.4 0.4
2 _3
– +
5.2 3.6 2.2 3.6 1.9 1.8 3.8 1.4 1.8

particularly given the observation by ACI Committee 435 were determined from the factored moment resistances, and
(1995) that “the magnitude of actual deflections in concrete the service loads were back-calculated for the given live-to-
structural elements...can only be estimated within a range of dead load ratio. The maximum deflection was computed for
20-40 percent accuracy.” the case of dead load on both spans and live load on one span
The trend for the variable-length element idealizations only. To account for the unsymmetric loading and the stat-
shown in Table 2 is similar. The maximum deflection ical indeterminacy, the point of zero rotation (and maximum
computed using 6–_7+ idealization is more accurate than that deflection) was located. The adjacent element lengths were
computed using the 2–_3+ idealization. The variable-length then adjusted to create a node at this location to facilitate
2–_3+ idealization generally yields computed deflections computation of the maximum deflection.
that closely approximate those obtained using the constant- Figure 6 summarizes results for a two-span beam, for ρ–
length five-element idealization, indicating that any effect of of 1.0%, wL/wD of 0.5, and live load on the left span only,
the moment gradient near the supports is marginal. computed using the Bischoff third-power equation and Mcr
computed using 0.67fr. Figure 6(a) shows the variation of
COMPARISON OF SINGLE-ELEMENT AND normalized moment, which exceeds Mcr for a long length of
DISCRETIZED-ELEMENT IDEALIZATIONS the left span, a short length of the right span, and at the interior
Deflection calculations were carried out to verify the support. Figure 6(b) shows the variation of Ie/Ig and Icr/Ig, indi-
accuracy of the Branson fourth-power equation, determine cating that Ie approaches Icr in these cracked regions. Figure
an appropriate exponent in the Bischoff Equation for the 6(c) shows the variation of curvature, which is relatively
discretized-element idealization, and compare the results small in the uncracked regions where the applied moments
obtained using both equations with the full and reduced are small and the gross moment of inertia is effective.
moduli of rupture for both single-element and discret- Finally, Fig. 6(d) shows the final deformed shape.
ized-element idealizations. Preliminary analyses indicated Table 3 summarizes the maximum deflections computed
that using an exponent of 3 in the Bischoff Equation with using the Branson Equation with Mcr computed using 0.5fr
a discretized-element idealization would yield comparable and the Bischoff Equation with Mcr computed using 0.67fr
results to those obtained using an exponent of 2 with the for both the single-element and discretized-element ideal-
single-element idealization (Mancuso 2016). izations. The differences between the deflections computed
Simply supported single-span members, and two- and three- using the third- and fourth-power Branson Equations are
span members that are continuous at their interior supports slight, and between the second- and third-power Bischoff
were investigated. The procedure for the two-span case is Equations are also slight. The Branson Equation, whether
the most complex and will be presented herein. The other used for the single-element or discretized-element idealiza-
cases are reported in Mancuso (2016). The set of two-span tion, typically predicts slightly larger deflections than the
beams investigated included three flexural reinforcement Bischoff Equation for either idealization, although again
ratios in the negative moment region ρ– of 0.5%, 1%. and these differences are slight.
1.5%, and three live-to-dead load ratios wL/wD of 0.5, 1.0, and Table 4 shows the deflections corresponding to the
1.5. The member length was chosen so that the minimum various cases of the simply supported beams investigated.
thickness requirements in ACI 318-14 are not satisfied and Again, the differences between the third- and fourth-power
so deflection checks are required. For each beam, the design Branson Equations are slight, and between the second- and
strength ϕMn in the negative moment region was determined third-power Bischoff Equations are also slight. Increasing
for the target reinforcement ratio and, for the given live-to- the reinforcement ratio increases the member capacity and
dead load ratio, the necessary design strength in the posi- thus the applied moment but does not change the cracking
tive moment region was computed. Other section properties, moment. Thus, the ratio Mcr/Ma will decrease, the effective
including Mcr , Ig, and Icr, were calculated. The factored loads moment of inertia will decrease, and the midspan deflection

1304 ACI Structural Journal/September-October 2017

Fig. 6—Analysis of discretized two-span continuous beam (Bischoff).
will increase. As the wL/wD ratio increases, the total service tests reported by others that represent a wide range of flex-
load decreases and the deflection is reduced. ural reinforcement ratios, span-depth ratios, section geome-
Table 5 compares the maximum deflections computed tries, curing conditions, and applied load types. The database
using the two equations and two idealizations for the various consists of 42 simply supported and 11 two-span contin-
three-span cases. The maximum deflection of the center span uous members with rectangular cross sections, four simply
increases as wL/wD increases because the pattern live loading supported T-beams, and eight simply supported T-beams
increases the applied midspan positive moment at that loca- with cantilevers at one support. Details of the various spec-
tion. The greatest difference between the single-element imens investigated are summarized in Tables 6(a) and 6(b)
idealization and discretized-element idealization occurs for the simply supported and continuous beam specimens,
with the lowest values of ρ and wL/wD, 0.5% and 0.5, respec- respectively.
tively. These deflections are in the range of L/600 to L/1000, Figure 7 shows the impact of using the full and reduced
however, and so are unlikely to be critical. moduli of rupture with the Branson and Bischoff Equa-
tions to compute the deflections of two simply supported
ACCURACY OF VARIOUS DEFLECTION members reported by Branson (1965). Specimens SB-1
CALCULATION PROCEDURES and SB-3 were reinforced with one (ρ = 0.7%) or three
This paper has considered eight distinct procedures to (ρ = 2.1%) No. 3 bars, respectively, and had cross sections
compute instantaneous deflections using the Branson or of 4 x 5 in. (approximately 100 x 125 mm) and spans of
Bischoff Equations, for single-element or discretized-ele- 9 ft (approximately 2.75 m). They were loaded when
ment idealizations, using Mcr computed using the full or a the concrete was 28 days old by their self-weight and a
reduced value of fr. To assess the accuracy of these proce- uniformly distributed load consisting of uniformly spaced
dures, deflections were computed for 65 beam and slab iron bricks. The left sides of Fig. 7(a) and (b) show the

ACI Structural Journal/September-October 2017 1305

Table 3—Computed two-span continuous beam deflections
ρ– = 0.50%; ρ+ = 0.40%; ϕM–n = 49.7kN.m; ϕM+n = 40.7 kN.m; wf = 12.4 kN/m
Branson Bischoff
wL/wD Single element ∆MID Discretized ∆MID % difference Single element ∆MID Discretized ∆MID % difference
0.5 20.2 21.9 –7.6 20.3 21.3 –4.4
1 21.1 21.5 –2.0 21.1 20.7 2.3
1.5 21.4 21.2 1.2 21.5 20.4 5.7
ρ = 1.00%; ρ = 0.82%; ϕM n = 95.9 kN.m;
– + –
ϕM+n = 80.1 kN.m; wf = 24.0 kN/m
Branson Bischoff
wL/wD Single element ∆MID Discretized ∆MID % difference Single element ∆MID Discretized ∆MID % difference
0.5 28.4 28.8 –1.6 27.4 28.4 –3.4
1 29.7 30.2 –1.6 28.7 29.6 –3.2
1.5 30.4 30.5 –0.4 29.4 29.7 –1.3
ρ = 1.50%; ρ = 1.18%;
– +
ϕM–n = 111.2 kN.m; ϕM+n = 137.7 kN.m; wf = 34.0 kN/m
Branson Bischoff
wL/wD Single element ∆MID Discretized ∆MID % difference Single element ∆MID Discretized ∆MID % difference
0.5 31.0 31.2 –0.5 30.5 31.0 –1.7
1 32.5 32.7 –0.7 31.9 32.5 –1.8
1.5 33.3 33.5 –0.5 32.6 33.1 –1.3

Table 4—Computed simply supported beam deflections

ρ+ = 0.50%; ϕM+n = 50.0 kN.m; wf = 19.8 kN/m
Branson Bischoff
wL/wD Single element ∆MID Discretized ∆MID % difference Single element ∆MID Discretized ∆MID % difference
0.5 21.0 21.5 –2.1 20.3 20.9 –2.8
1 19.7 20.2 –2.3 19.1 19.6 –2.7
1.5 19.0 19.5 –2.5 18.4 18.9 –2.7
ρ = 1.00%;
+
ϕM+n = 95.9 kN.m; wf = 37.9 kN/m
Branson Bischoff
wL/wD Single element ∆MID Discretized ∆MID % difference Single element ∆MID Discretized ∆MID % difference
0.5 25.4 25.5 –0.1 24.9 25.2 –1.3
1 24.2 24.2 –0.1 23.7 24.0 –1.4
1.5 23.5 23.5 –0.1 23.0 23.3 –1.5
ρ = 1.50%;
+
ϕM+n = 137.7 kN.m; wf = 54.4 kN/m
Branson Bischoff
wL/wD Single element ∆MID Discretized ∆MID % difference Single element ∆MID Discretized ∆MID % difference
0.5 27.3 27.3 0.0 27.0 27.2 –0.6
1 26.0 26.0 0.0 25.7 25.9 –0.7
1.5 25.2 25.2 0.0 25.0 25.1 –0.7

normalized applied and cracking moments and normalized length, and so Ie equals Icr for most of the member length.
effective and cracked moments of inertia, respectively, for Thus, the computed deflection is quite sensitive to the
the lightly loaded member, SB-1. The right sides show the assumed modulus of rupture for the lightly reinforced beam
same quantities for the more heavily loaded member, SB-3. and relatively insensitive to the assumed modulus of rupture
For SB-1, the applied moment does not exceed the cracking for the heavily reinforced beam. For both beams, the effec-
moment computed using the full modulus of rupture; thus, tive moment of inertia computed using the Branson Equa-
Ie equals Ig in either single- or discretized-element ideal- tion with 0.5fr is slightly more severe than that computed
izations. On the other hand, for SB-3, the applied moment using the Bischoff Equation with 0.67fr.
exceeds the cracking moment, whether computed using the Figure 8(a) shows the variation of the test-to-predicted
full or reduced moduli of rupture, for most of the member ratios for the various deflection calculation procedures using

1306 ACI Structural Journal/September-October 2017

Table 5—Computed three-span continuous beam deflections
ρ– = 0.50%; ρ+ = 0.34%; ϕM–n = 50.0 kN.m; ϕM+n = 34.2 kN.m; wf = 15.2 kN/m
Branson Bischoff
wL/wD Single element ∆MID Discretized ∆MID % difference Single element ∆MID Discretized ∆MID % difference
0.5 6.9 5.6 18.8 6.6 4.4 33.5
1 10.7 10.2 5.2 11.2 9.6 14.0
1.5 12.9 13.0 –0.6 13.5 12.7 6.2
ρ = 1.00%; ρ = 0.66%; ϕM n = 95.9 kN.m;
– + –
ϕM+n = 65.1 kN.m; wf = 28.9 kN/m
Branson Bischoff
wL/wD Single element ∆MID Discretized ∆MID % difference Single element ∆MID Discretized ∆MID % difference
0.5 14.3 15.9 –11.7 13.6 15.1 –10.8
1 19.4 21.6 –11.3 18.6 20.7 –11.6
1.5 22.2 24.8 –11.4 21.3 23.9 –12.2
ρ =1.50%; ρ =1.00%; ϕM n=137.7kN.m;
– + –
ϕM+n= 95.9kN.m; wf = 42.1kN/m
Branson Bischoff
wL/wD Single element ∆MID Discretized ∆MID % difference Single element ∆MID Discretized ∆MID % difference
0.5 16.0 18.3 –14.1 15.5 17.8 –14.4
1 21.5 24.1 –12.1 20.9 23.6 –12.7
1.5 24.5 27.3 –11.4 23.9 26.8 –12.4

Table 6(a)—Simply supported beam tests by others

Study Comment Curing Applied loading No. of specimens ρ+ L/h
Point loads @ L/3 Four beam 0.53% 10 and 11
Gilbert and Special care to Kept moist (covered in wet hessian) for
Nejadi (2004a) prevent shrinkage period of 28 days 0.30% to
UDL Six slab 22
0.60%
0.53% to
Point loads @ L/3 Six beam 10 and 11
Gilbert and Special care to Kept moist (covered in wet hessian) for 0.80%
Nejadi (2004b) prevent shrinkage period of 14 days 0.29% to
UDL Six slab 22
0.59%
Cured under wet canvas 5 days, then concrete
Washa and Higher ρ, Ie → Icr 1.47% to
was exposed to air on all surfaces except UDL 5 20 and 30
Fluck (1952) irrespective of fr 1.58%
bottom
Steel control Two point loads
El-Nemr (2013) Not given 1 0.43% 9
specimen near midspan
Equation based on 0.69% to
Branson (1965) Not given UDL 2 22
test results and full fr 2.07%
Cured in molds 3 to 7 days. “Sealed” spec-
imens painted with bakelite lacquer. “Dry” Two dry and two
Washa (1947) Mcr sensitive to fr UDL 0.80% 30
slabs were only sealed on edges, top, and sealed
bottom exposed to air
Corley and Higher ρ, Ie → Icr Wet burlap 24 hours, fog room 7 days, open 1.36% to
Point loads @ L/4 3 12 and 17
Sozen (1966) irrespective of fr lab 21 days 3.05%
Cast in T-shaped forms, covered with
waterproof membrane for 3 days, forms were 0.50% to
Yu (1959) T-beams UDL 4 14 to 30
removed, controlled climate for 18 days. Lab 0.83%
for 7 days and tested at 28 days
Park et al. Early-age loading, 3 Two point loads 0.50% to
Cured at low temperature, <17°C 4 28
(2012) days or 7 days near midspan 1.04%
Bakoss et al. Higher ρ, Ie → Icr Moist cured 14 days, kept in climate-
Point loads @ L/3 1 1.74% 29
(1982) irrespective of fr controlled lab

ACI Structural Journal/September-October 2017 1307

Table 6(b)—Continuous beam tests by others
No. of ρ+ at ρ– at
Study Comment Curing Applied loading specimens midspan, ρ′ support ρ′ L/h
2.83% to
Long-term “plastic flow” Cured under wet canvas 7 days, then 30, 50
Washa and 1.59% to 3.20%
study equation based on concrete was exposed to air on all UDL 3 and
Fluck (1956) 1.65% (0%) (2.48% to
test results and full fr surfaces except bottom. 70
3.54%)
Equation based on test 0.69% or 0.69% or
Branson (1965) Not given UDL 2 22
results and full fr 2.07% (0%) 2.07% (0%)
FRP study–steel control Concrete surface was kept wet for Point loads at
El Mogy (2011) 1 1.48% (0%) 1.11% (0%) 9
specimen 7 days midspan
FRP study–steel control Wet cured, and covered until date of Point loads at
Habeeb (2008) 1 0.84% (0%) 0.84% (0%) 9
specimen testing midspan
Mahroug FRP study–steel control Wet cured, and covered until date of Point loads at 0.50% or 0.50% or
2 13
(2014a,b) specimen testing midspan 0.75% (0%) 0.75% (0%)
Waterproof membrane stretched
over each specimen. Forms removed
Guaralnick and
after 3 days and beams placed in 0.72% to 0.72% to
Winter (1957, T-beams UDL 8 11
moist room for 18 days. Specimens 1.35% (0%) 1.35% (0%)
1958)
dried in lab at 21 days until testing at
approximately 28 days
Point loads at
Bakoss et al. Higher ρ, Ie → Icr irrespec- Moist-cured 14 days, kept in
midspan, no 1 1.74% (0%) 1.74% (0%) 23
(1982) tive of fr climate-controlled lab
self-weight
Redistribution of design Single point load
Mattock (1959) Not given 1 1.97% (0%) 0.98% (0%) 8
bending moments + self-weight

Fig. 7—Simply supported Beams SB-1 (left) and SB-3 (right) (Branson [1965]).
the reduced fr with the reinforcement ratio. Using the reduced tion calculation procedures when the recommended reduced
fr typically gives conservative results for all reinforcement fr values are adopted.
ratios. Typically, the test-to-predicted ratios corresponding Figure 8(b) shows the variation of the test-to-predicted
to a particular test are superimposed on each other, indi- ratios for the various deflection calculation procedures when
cating the consistency of the accuracy of the various deflec- using the full fr to compute the cracking moment. Clearly

1308 ACI Structural Journal/September-October 2017

Fig. 8—Test-to-predicted ratios for various reinforcement ratios.
Table 7—Test-to-predicted ratios for all members
Branson (0.5fr) Bischoff (0.67fr) Branson (1.0fr) Bischoff (1.0fr)
Third power Fourth power Second power Third power Third power Fourth power Second power Third power
Minimum 0.29 0.29 0.31 0.34 0.66 0.67 0.63 0.63
Maximum 1.19 1.19 1.21 1.20 2.55 2.80 1.96 2.25
Mean 0.82 0.82 0.84 0.84 1.16 1.19 1.04 1.07
Standard deviation 0.18 0.19 0.18 0.18 0.38 0.43 0.26 0.30
Coefficient of variation, % 22.0 22.8 21.1 21.2 33.2 35.8 24.9 28.2

using the full fr to compute Mcr yields unconservative results, using the Bischoff Equation with the full fr to compute Mcr
particularly when using the Branson Equation for cross yields a close but slightly unconservative prediction of the
sections with reinforcement ratios less than 1%. The test- observed deflections. In this case, the coefficients of varia-
to-predicted ratios are also markedly more variable when tion are lower than those for the Branson Equation and are
the reinforcement ratio is less than 1%. For cross sections comparable to those obtained when Mcr is computed using
with reinforcement ratios greater than 1%, the test-to- the recommended reduced value of fr.
predicted ratios approach unity because, as noted previously,
the member is more severely cracked at service loads, so Ie SUMMARY AND CONCLUSIONS
approaches Icr irrespective the effective moment of inertia The research presented in this paper investigates current
equation or idealization used. procedures to compute instantaneous deflections of rein-
Table 7 shows the minimum and maximum test-to- forced concrete beams. In particular, the accuracy of equa-
predicted ratios, overall mean, standard deviation, and tions proposed by Branson (1965) and Bischoff (2007) for
coefficient of variation of the test-to-predicted ratios for members idealized as a single element or using discretized
all 65 simply supported beams and continuous beams elements are investigated for the case where the cracking
investigated. Using the reduced fr to compute Mcr yields moment is computed using the full modulus of rupture or
consistent and conservative results using either the Branson using a reduced moduli of rupture recommended in CSA
or Bischoff Equations with the single-element or discret- A23.3-14 (CSA 2014) for use in the Branson Equation or
ized-element idealizations. Using the Branson Equation with by Scanlon and Bischoff (2008) for use in the Bischoff
the full fr to compute Mcr gives unconservative results with Equation. The impact of the material properties of early-
high coefficients of variation whether the single-element aged concrete on the computed instantaneous deflection is
or discretized-element idealization is used. For these data, also investigated.

ACI Structural Journal/September-October 2017 1309

 The conclusions of this study are as follows: the Toronto office of Read Jones Christoffersen Ltd. are also gratefully
acknowledged.
1. The stress-strain relationship for concretes that are less
than a day old does not exhibit significant strain softening.
For concrete with ages of 1 day or greater, although the equa- NOTATION
As = area of tensile flexural reinforcement
tions specified in ACI 318-14 may overestimate the actual b = width of cross section
elastic modulus, Ec, the conventional methods for computing d = distance from extreme compression fiber to centroid of tension
reinforcement
the cracked rigidity, EcIcr, are reasonably accurate. Ec = elastic modulus of concrete
2. When a concrete member is idealized as a number of EcIcr = flexural rigidity of cracked cross section
discretized elements, the accuracy of the computed deflection f0 = compressive strength of concrete
fc′ = specified compressive strength of concrete
increases as the element length is reduced and the number fc = concrete compressive stress at extreme fiber
of elements increases. For the single span and continuous fr = modulus of rupture of concrete
two-span and three-span cases investigated in this study, Icr = moment of inertia of cracked section transformed to concrete
Ie = effective moment of inertia
use of an idealization with 10 constant-length elements per Ig = moment of inertia of gross section
half-span—that is, 20 elements per span—gave a maximum kd = depth from extreme compression fiber to neutral axis, elas-
unconservative error of 2.2%. The accuracy was not mark- tic-cracked analysis
L = span length
edly improved if shorter elements were used in the negative Ma = applied bending moment (in positive moment region, M+, in
moment regions, where moment gradients are relatively high, negative moment region, M–)
and longer elements used in the positive moment regions. Mcr = cracking moment
Mn = nominal moment capacity
3. The Bischoff Equation can accurately compute the Mo = nominal applied bending moment
effective moment of inertia for members idealized using n = modular ratio (Es/Ec)
discretized elements if the exponent in its denominator is wL/wD = live-to-dead load ratio
∆mid = midspan deflection of member
increased from 2 to 3. εc = concrete compressive strain
4. Deflections predicted using the Branson Equation with ϕ = design resistance factor
the cracking moment computed using the full modulus of ϕMn = design strength
ρ = flexural reinforcement ratio (in positive moment region, ρ+, in
rupture, in accordance with ACI 318-14, give a mean test-to- negative moment region, ρ–)
predicted ratio ranging from 1.16 to 1.19 and a coefficient of σc = concrete compressive stress
variation ranging from 38 to 43% when compared to deflec- Ψ = curvature of flexural member
Ψmax = maximum curvature of flexural member
tions of 65 simply supported and continuous beams observed
by others. This approach is increasingly unconservative and
inaccurate as the reinforcement ratio reduces below 1%. REFERENCES
ACI Committee 318, 2014, “Building Code Requirements for Struc-
5. Using the Branson Equation with the cracking moment tural Concrete (ACI 318-14) and Commentary (ACI 318R-14),” American
computed using half the modulus of rupture, or the Bischoff Concrete Institute, Farmington Hills, MI, 519 pp.
ACI Committee 435, 1995, “Control of Deflection in Concrete Struc-
Equation with the cracking moment computed using tures (ACI 435R-95),” American Concrete Institute, Farmington Hills, MI,
two-thirds the modulus of rupture, give consistent and 89 pp.
conservative deflection predictions whether the beam is Bischoff, P. H., 2007, “Rational Model for Calculating Deflection of
Reinforced Concrete Beams and Slabs,” Canadian Journal of Civil Engi-
idealized as a single element or as a number of discretized neering, V. 34, No. 8, pp. 992-1002.
elements. The test-to-predicted ratios obtained using these Branson, D. E., 1965, “Instantaneous and Time-Dependent Deflections
approaches averaged between 0.82 to 0.84 with coefficients of Simple and Continuous Reinforced Concrete Beams,” HPR Report
No.  7, Part 1, Alabama Highway Department, Bureau of Public Roads,
of variation ranging from 21 to 23%. Montgomery, AL.
Canadian Standards Association (CSA), 2014, “Design of Concrete
AUTHOR BIOS Structures (CAN/CSA-A23.3-14),” Canadian Standards Association,
Caitlin Mancuso is an Engineering Intern at Read Jones Christoffersen Mississauga, ON, Canada.
Ltd., in Toronto, ON, Canada. She received her MEngSc degree at the Cement Association of Canada (CAC), 2016, Concrete Design Hand-
University of Western Ontario, London, ON, Canada. She is the 2015 book, fourth edition, Cement Association of Canada, Ottawa, ON, Canada.
winner of the ACI Katharine and Bryant Mather Scholarship. Her research Hognestad, E., 1951, “A Study of Combined Bending and Axial Load in
interests include reinforced concrete mechanics. Reinforced Concrete Members,” Bulletin 399, University of Illinois Engi-
neering Experiment Station, Urbana, IL.
F. Michael Bartlett, FACI, is Associate Dean (Undergraduate Studies) of Jin, X.; Shen, Y.; and Li, Z., 2005, “Behaviour of High- and Normal-
Engineering at the University of Western Ontario. He received his PhD from Strength Concrete at Early Ages,” Magazine of Concrete Research, V. 57,
the University of Alberta, Edmonton, AB, Canada. He is a member of ACI No. 1, pp. 339-345.
Committee 562, Evaluation, Repair, and Rehabilitation of Concrete Build- Khan, A. A., 1995, “Concrete Properties and Thermal Stress Analysis
ings, as well as a member of ACI Committees 214, Evaluation of Results of Members at Early Ages,” PhD thesis, McGill University, Montreal, QC,
of Tests Used to Determine the Strength of Concrete; and ACI Subcom- Canada.
mittee 318-C, Safety, Serviceability, and Analysis (Structural Concrete Mancuso, C., 2016, “Short-Term Deflections of Reinforced Concrete
Building Code). His research interests include the application of structural Beams,” MESc thesis, Western University, London, ON, Canada.
mechanics and reliability principles to structural design, evaluation, and Scanlon, A., and Bischoff, P. H., 2008, “Shrinkage Restraint and Loading
rehabilitation. History Effects on Deflection of Flexural Members,” ACI Structural
Journal, V. 105, No. 4, July-Aug., pp. 498-506.
Todeschini, C. E.; Bianchini, A. C.; and Kesler, C. E., 1964, “Behavior
ACKNOWLEDGMENTS of Concrete Columns Reinforced with High Strength Steels,” ACI Journal
The authors gratefully acknowledge financial support from the Natural
Proceedings, V. 61, No. 6, June, pp. 701-716.
Sciences and Engineering Research Council (NSERC), in the form of a
Whitney, C. S., 1937, “Reinforced Concrete Members under Flexure or
Canada Graduate Scholarship (CGS-M) awarded to the first author and a
Combined Flexure and Direct Compression,” ACI Journal Proceedings,
Discovery Grant held by the second author. Informal contributions, interest,
V. 33, No. 3, Mar., pp. 483-498.
and suggestions from K. MacLean, T. Kokai, and their colleagues from
Wight, J. K., 2015, Reinforced Concrete, Mechanics and Design, seventh
edition, Pearson, Upper Saddle River, NJ.

1310 ACI Structural Journal/September-October 2017