You are on page 1of 28

See

discussions, stats, and author profiles for this publication at: http://www.researchgate.net/publication/279058604

Probabilistic models for curvature ductility and


moment redistribution of RC beams
ARTICLE in COMPUTERS AND CONCRETE APRIL 2015
Impact Factor: 0.87 DOI: 10.12989/cac.2015.16.2.191

READS

27

2 AUTHORS:
Hassan Baji

H. R. Ronagh

RMIT University

Western Sydney University

21 PUBLICATIONS 5 CITATIONS

166 PUBLICATIONS 692 CITATIONS

SEE PROFILE

All in-text references underlined in blue are linked to publications on ResearchGate,


letting you access and read them immediately.

SEE PROFILE

Available from: Hassan Baji


Retrieved on: 23 November 2015

A probabilistic study on the ductility of reinforced concrete

sections

Hassan Bajia,*, Hamid Reza Ronagha

School of Civil Engineering, University of Queensland, St Lucia, Queensland 4072, Australia

Abstract: Although the current design codes apply reliability-based calibration procedures

to evaluate safety factors for the strength based limit state, the safety factors used to ensure

minimum ductility capacities are rather simple and are not resulted from a probability-based

procedure. This study examines level of safety delivered by the current design codes with

10

regards to providing minimum curvature ductility for reinforced concrete (RC) beams made

11

with normal strength concrete. Reliability analysis results show that with regard to the

12

strength limit state, the considered design codes are in good agreement with one another.

13

However, there is considerable disparity in the level of safety provided for minimum

14

curvature ductility amongst the codes. The provided reliability for the design to remain

15

ductile is too low in some and just about acceptable in the others. This signifies the

16

importance and the need to introduce reliability based methods of design for ductility.

17

Keywords: Reinforced concrete beams, design codes, curvature ductility, ultimate concrete

18

strain, Monte Carlo Simulation, Reliability

19
20
21
22

*Corresponding Author: Tel.: +61-7-33661652; Fax: +61-7-33654599

23

Email: h.baji@uq.edu.au
1

Introduction

In the design of RC structures, checking the strength adequacy often takes priority over the

deformation and the ductility, which are indirectly incorporated in the design process. Design

codes prescribe some limits such as the rebar percentage limit to ensure sections would

possess adequate ductility. Park et al. (1988) assessed the available ductility of doubly RC

beam sections using the moment-curvature analysis. They concluded that the implementation

of the general requirements of the American and New Zealand design codes will ensure a

curvature ductility of more than 2.0, while the application of moment redistribution

requirements will ensure curvature ductility larger than 4 for the sections. Ho et al. (2004)

10

investigated the minimum flexural ductility design of high strength concrete beams. Their

11

results showed that the current practice of providing minimum flexural ductility in existing

12

design codes would not really provide a consistent level of minimum flexural ductility. Kwan

13

and Ho (2010) studied the flexural ductility of high-strength concrete beams and columns by

14

extensive parametric studies using nonlinear moment-curvature analysis. Based on their

15

study, a minimum ductility design method for ensuring the achievement of a minimum

16

ductility of 3.32 was proposed. In order to check the criterion of local ductility in the cross

17

sections, Kassoul and Bougara (2010) have taken into account the recommendations of EC2

18

(2004) regarding the stress-strain relationship for concrete and steel. They developed a

19

methodology for evaluating the available curvature ductility factor in RC beams. All of these

20

studies have used a deterministic approach to assess the minimum ductility requirements of

21

RC beams and columns. In this paper, a probabilistic framework for assessing the minimum

22

ductility requirements of RC sections is proposed.

23

Realistic description of strength and deformation requires probabilistic models and

24

implementation of a reliability based analysis. There have been numerous studies performed
2

on the strength of RC members, resulting in code calibration, which has now been

implemented in many design codes (Bartlett, 2007; Szerszen et al., 2003). In contrast, limited

research can be found on the probabilistic aspects of the inelastic deformation and ductility

(De Stefano et al., 2001; Kappos et al., 1999; Lu et al., 2004; Trezos, 1997). In some of these

studies, probabilistic models for ductility related measures such as curvature ductility have

been proposed. However, none of these studies has directly addressed the issue of minimum

ductility requirements suggested by the current codes of practice. Ito and Sumikama (1985)

study is amongst very few, if not the only one, that are directly related to the reliability

analysis of code provisions with regards to the ductile design of RC beams. They examined

10

the suitability of the reduction coefficient for the balanced steel ratio provided in ACI 318-83.

11

Results of their study showed that using the ACI reduction factor of 0.75 results in high

12

probability of producing over-reinforced cross-sections when concrete is placed in situ.

13

This study aims at investigating the reliability of minimum ductility requirements in the

14

current design codes using a comparative-based approach. The amount of tensile rebar has an

15

inverse relation with section ductility. Therefore, design codes prescribe a maximum rebar

16

percentage limit in order to ensure that RC sections would exhibit adequate ductility. On the

17

other hand, the presence of compressive rebar enhances the section ductility. Noting these,

18

the worst-case scenario for investigating the minimum ductility is to have the maximum

19

tensile rebar at the tension side of the section, while the compression side is reinforced by the

20

minimum rebar. This would result in a lower reliability level for ductility of RC sections, and

21

as such, it is devised in this study.

22
23

Moment-Curvature analysis

In order to determine the load-deformation behavior of a cross section, moment curvature

analysis is performed using nonlinear material stress-strain relationship. In Figure 1, the

typical stress-strain curves for concrete and steel bars are shown. Stress-strain curve for

concrete is based on JCSS probabilistic model code (2012). This model is similar to the

parabolic model proposed by the EC2 to be used in section analysis. The first part of the

curve is a parabola, while the second part is constant. As is shown in Figure 1b, for

reinforcing steel a bilinear relationship is employed.

7
8

Figure 1: Material Models for Concrete and Steel

Although the material stress-strain relationship is nonlinear, the strain variation across the

10

height of the section can be assumed linear. This assumption seems to have adequate

11

accuracy and the experimental results have proven its validity (Park et al., 1975). Fiber model

12

is used to derive the moment-curvature of the cross section. In the limit state design, ductility

13

of a member is usually defined as the ratio of the ultimate deformation to the deformation at

14

first yield. The first yield is the state at which the tensile rebar yields. On the other hand, the

15

ultimate curvature is the state at which either the concrete crushes or the tensile bar ruptures.

16

In case of moment curvature analysis, ductility is the ratio of ultimate to yield curvature as

17

shown in Equation 1.

u
y

(1)

In Equation 1, y , u and represent the yield curvature, the ultimate curvature and the

curvature ductility respectively. Curvature ductility depends on factors such as tensile and

compressive reinforcement ratios, concrete and steel material ductility, axial force and other

parameters.

Code provisions on ductility

In this section, provisions of the American (ACI 318, 2011), Canadian (CSA A23.3, 2004),

Australian (AS 3600, 2009), New Zealand (NZS 3101, 2006) and European (EC2, 2004) with

respect to the important issue of ductility are reviewed.

Figure 2 shows a typical stress-strain diagram of an RC beam cross section. In Table 1, the

10

parameters of the equivalent stress block for different design codes are shown. Furthermore,

11

this Table shows the limiting constraints for ductile design based on the aforementioned

12

codes. In this Table, fc is the concrete compressive strength, t (used with the American

13

code) denotes the strain at foremost tensile bar, Parameters 1 and 1 are the concrete stress

14

block parameters, cu is the ultimate strain of concrete and (used with the European code)

15

refers to the moment reduction factor. Here, the effect of moment redistribution is not

16

considered. Hence, the moment reduction factor is taken as 1.0. All other parameters are

17

shown in Figure 2. Unlike other codes, the American code does not directly use the neutral

18

axis parameter (limit the c/d ratio) to ensure adequate ductility; rather it applies a minimum

19

tensile bar strain (tensile strain >0.005). Nevertheless, its limit on minimum tensile strain

20

could be transformed to this neutral axis parameter format if needed. In Table 1, the result of

21

this transformation is shown. It is worth mentioning that Table 1 only covers normal strength
5

concrete.

2
3

Figure 2: Stress-strain diagrams of a typical rectangular section

Using force equilibrium and geometric compatibility, the limiting neutral axis parameter can

be described as per Equation 2. In this Equation, effect of compressive rebar is neglected.

The limiting c/d values for each design code are shown in Table 1.

fy
c
1 cu
c

'
d 11 f c d Limit S .F . cu y

(2)

S.F. shows the safety factor used to ensure adequate safety margin for a ductile design. This

safety factor is different from code to code. As is noted in Table 1, the New Zealand and the

Canadian codes follow exactly the mentioned format. The limiting ratio depends on yield

10

strain of rebar steel as well as the ultimate concrete strain. Thus, for different steel grades, the

11

limiting ratio varies. In Table 2, based on Equation 2, the relationship between the limiting

12

neutral axis parameter and the safety factor for different design codes is shown.

13

Table 1: Concrete stress block parameters of considered design codes


Code

cu

c / d max

ACI

0.0030

for f c' 30
0.85
for f c' 30

0.85

1.09 0.008 f

t 0.005
'
c

0.65 1 0.85
AS 3600

0.0030

NZS 3101

0.85 0.003 fc'

0.85 0.003 fc'

0.67 1 0.85

0.67 1 0.85

0.85

for f c' 30
0.85
for f c' 30

0.0030

1.09 0.008 fc'

d 0.375

c
0.360
d

cu
c
0.75

d
cu f y / Es

0.65 1 0.85
CSA A23.3 0.0035

EC2

0.85 0.0015 f c'

0.85 0.0025 f c'

1 0.67

1 0.67

1.00

0.80

0.0035

c
700

d 700 f y
c k1 ( 0.44)

d
k2 ( 1.25)

1
2

Design codes that apply the strength reduction factor to material properties rather than

strength component (like the Canadian and European codes) already consider certain amount

of safety margin required for the ductile design. To make results of these design codes

consistence with the other codes, the limit shown in Equation 2 needs to be multiplied by

s/c factor, where s and c are the steel and concrete material reduction factors. In Table 3,

these material reduction factors are shown. It should be noted that in deriving the safety

factors for each code, its own ultimate concrete strain is used, and the steel modulus of

elasticity of 200GP is assumed.

10

Table 2: Neutral axis limiting values and corresponding safety factor


Code

f y (MPa)

300

400
7

500

c

d Limit

S.F.

c

d Limit

S.F.

c

d Limit

S.F.

ACI 318 (2011)

0.375

1.78

0.375

1.60

0.375

1.45

AS 3600 (2009)

0.360

1.86

0.360

1.67

0.360

1.50

NZS 3101 (2006)

0.500

1.33

0.450

1.33

0.409

1.33

CSA A23.3 (2004)

0.535

1.31

0.486

1.31

0.446

1.31

EC2 (2004)

0.343

2.04

0.343

1.86

0.343

1.70

1
2

When only the difference in ultimate strain of concrete is considered, the American,

Australian and New Zealand design codes are more conservative in limiting the cross section

ductility. This conservatism is not transparent in the safety factors shown in Table 2. This is

because the available safety factors for the considered design codes depend on other factors

such as the material reduction factors and the neutral axis parameter limit provided by those

codes. It should be noted that the overall safety depends on other parameters of the equivalent

rectangular stress block, e.g. 1 and 1 parameters as well. Only a complete moment-

curvature analysis in which the curvature ductility is directly derived can reveal the level of

10

safety in any of the mentioned design codes.

11

In this study, in addition to investigating reliability of the curvature ductility of cross sections

12

designed based on different design codes, the safety levels of strength limit state is also

13

considered for the sake of comparison. Due to the different statistical load models used in the

14

calibration of load and resistance factors in each design codes, the load combination that only

15

includes the effect of dead load is considered. The considered design codes agree on the dead

16

load combination. Furthermore, the statistical model for dead load is universally the same. In

17

Table 3 the dead load factor and resistance reduction factors are shown for different design

18

codes.

19

Table 3: Safety factors of the considered design codes for dead load combination
8

Resistance Reduction
Factors
Behavior
Material
0.90
-

DL

Code
ACI 318 (2011)

1.40

AS 3600 (2009)

1.35

0.80

NZS 3101 (2006)

1.35

0.85

CSA A23.3 (2004)

1.40

EC2 (2004)

1.35

c 0.65
s 0.85
c 1/1.50
s 1/1.15

Reliability analysis

4.1 Limit states

In the current study, two limit states are defined. The first one is a strength based limit state in

which the bending capacity of the cross section is treated as strength. The second limit state is

a deformation based limit state. The curvature ductility of the section is compared with 1.0,

which is the boundary between ductile and brittle design for RC beams subjected to bending.

In this limit state, failure is deemed to occur when curvature ductility exceeds 1.0. Equations

5 and 6 show the considered limit state functions.


g1 M R M Q

g2 u

u

R y

(3a)

u
Q y

1.0
R

(3b)

10

In Equation 3, MR and MQ represent the bending capacity and bending resulted from demand,

11

respectively. The parameter u

12

a brittle failure. As previously stated, in this study, only the effect of dead load is considered.

can be taken as 1.0 as any value smaller than 1.0 denotes


Q

Parameters y and u denote the yield and ultimate curvature respectively. In the previous

sections, the yield and ultimate states were defined.

4.2 Statistical models

Two major types of uncertainties exist concerning RC member behavior, namely physical

uncertainty, and model uncertainty. In this study, the majority statistical models for basic

random variables are taken from the JCSS Probability Model Code (2012). Table 4 shows the

statistical data for all considered random variables.

JCSS describes the dimensional deviations of any dimension by statistical characteristic of its

deviation from the nominal value. For concrete cover, two different models are suggested in

10

the JCSS for top and bottom reinforcement. For simplicity only the model reported in Table 4

11

is employed in this study for both top and bottom reinforcement covers. In order to remain

12

within the boundaries of practicality, a common 600400 mm cross section with 60 mm

13

cover to rebar center is selected in this study. Although in the JCSS states that the Normal

14

distribution seems to be satisfactory for dimensional parameters, in this study the lognormal

15

distribution is used instead. For random variables having small coefficient of variation

16

approximating normal distribution with lognormal one would not affect the results

17

significantly (Benjamin et al., 1975).

18

In the JCSS model code, all concrete properties are related to reference property of concrete,

19

which is the compressive strength of the standard test specimens tested according to the

20

standard conditions at the age of 28 days. The other concrete properties are related to the

21

reference strength of concrete according to the following Equations.


In situ compressive strength: fc' ( f c'0 ) Y1

(4a)

Tensile strength: 0.3( fc' )2/3 Y2

(4b)
10

Modulus of elasticity: 10500 f c' Y3


1/3

Strain at peak stress: cu 0.006 fc'

1/6

(4c)
Y4

(4d)

The statistical model for parameters, Y1 to Y4 are given in Table 4. For all these parameters,

the lognormal distribution is employed. For the strength of standard concrete specimen f c'0 ,

the student distribution is suggested in JCSS. However, this distribution can be approximated

by a lognormal distribution. In this study, for the reliability analysis, the lognormal

distribution is employed for the concrete compressive strength. The statistical information for

C25, C35 and C45 ready mix concrete grades are presented in JCSS and are shown in Table 4.

The Statistical model used for the concrete strength has a good agreement with findings of a

recent statistical analysis on the concrete strength (Nowak et al., 2003).

Reinforcing steel generally is classified and produced according to grades. In this study, in

10

accordance with the European standard, grades S300, S400 and S500 are used. These grades

11

are nearly equivalent to grades G40, G60 and G75 rebar steel materials according to the

12

American standard. The statistical models for yield and ultimate strength, modulus of

13

elasticity and ultimate strain of steel are shown in Table 4. For all these parameters,

14

lognormal distribution is considered.

15

Table 4: Statistical properties of random variables


Variable
b

Nominal
300 mm

/Bias
1.003

/COV
4 mm+0.006Nominal

600 mm

1.003

4 mm+0.006Nominal

cover

60 mm

Nominal+10 mm

10 mm

As

max bn dn

1.0

0.02

As'

min bn dn

1.0

0.02

Es

200GPa

1.0

0.04

fy

300/400/500MPa

Nominal+2

30MPa

11

fu

1.08 f y

40MPa

su

0.05

Nominal+2

0.09

f c'0

25/35/45MPa

1.55/1.35/1.20

0.17/0.12/0.07

Y1

1.0

0.06

Y2

1.0

0.30

Y3

1.0

0.15

Y4

1.0

0.15

0.96

0.005

It is assumed that the ratio of ultimate to yield stress of steel material is 1.08. This ratio

corresponds to minimum ratio required for Class A reinforcement in EC2. Also for this steel

grade, the minimum ultimate strain should be greater than 0.05. Therefore, in this study this

value is used as nominal ultimate strain of steel material. The correlation among steel rebar

area, yield strength, ultimate strength and ultimate strain of steel material is considered in this

research. Table 5 shows the correlation among these variables.

Table 5: Correlation among rebar steel material properties

As
fy

As

fy

fu

su

1.00

+0.50

+0.35

0.00

1.00

+0.85

-0.50

1.00

-0.55

fu

su

1.00

The statistical model used for rebar properties in this research can be compared with those

used in the available literature (Bournonville et al., 2004; Nowak et al., 2003). In Table 4, a

10

summary of the statistical models used in this study is shown. All of the random variables

11

used in this study are treated as lognormal distributed random variables. The joint probability

12

density function is a multivariate lognormal distribution with correlated variables.

12

4.3 Method of Analysis

Reliability analysis containing stochastic finite element such as the problem being studied

here are almost universally performed using the MCS technique. In the MCS method, the

probability of failure is calculated using random number generation (Melchers, 1999). In this

study, a Latin Hypercube Sampling technique (Ayyub et al., 1984) is used. In order to obtain

a more accurate estimate of the probability of failure, variance reduction methods including

Antithetic Variates are to be used in conjunction with the method (Ayyub et al., 1991).

8
9

The measure of reliability is conventionally defined by the reliability index , which


is related to the probability of failure pf by Equation 5.
p f ( )

(5)

10

In Equation 5, is the cumulative distribution function of standardized normal distribution.

11

The reliability index corresponds to the design working life of the structure and it has one-to-

12

one correspondence with failure probability.

13

For the purpose of reliability differentiation, the European code (EC2, 2004) establishes

14

reliability classes. According to this code, for the reliability class of RC2 and based on a 50

15

years reference period, the recommended minimum reliability indices for ultimate and

16

serviceability (irreversible) limit states are 3.8 and 1.5, respectively. The reliability class of

17

RC2 could be corresponding to the consequences class CC2, which covers residential and

18

office buildings. With respect to the strength limit state in this study, target reliability index

19

of 3.8 could be used. However, when it comes to the ductility limit state, it is difficult to set

20

an appropriate target reliability based on available literature. The limit state, which is here

21

defined for satisfying adequate ductility could not be treated either as ultimate limit state or

22

as serviceability limit state. Failure in curvature limit state means brittle collapse, which
13

comes without warning. On the other hand, this failure does not lead to structural collapse. In

this study, two target reliability indices of 2.3 and 3.1 are selected and will be used in the

calibration of safety factors for ductility limit state.

Results and discussion

Results of the reliability analysis of moment-curvature curve are presented here. Three

different grades of steel (S300, S400 and S500) as well as three different types of concrete

(C25, C35 and C45) are used in the analysis. The MCS method with Variance Reduction

technique is used to derive the probability of failure and the reliability indices. In previous

sections, the strength and ductility limit states were discussed. These limit state are named g1

10

and g2. Before performing the reliability analysis for considered limit states, using available

11

experimental results the statistical models of model uncertainty are derived.

12

5.1 Model uncertainty

13

The model uncertainty is used to quantify the uncertainties associated with assumptions and

14

simplifications used in derivation of the theoretical model. The model uncertainty associated

15

with a particular mathematical model may be expressed in terms of the probabilistic

16

distribution of a variable X defined in Equation 6.

XM

Actual
Pr edicted

(6)

17

Model error covers the uncertainties in the modeling of a structure as a mathematical

18

model where the uncertainties arise from idealization of different parts of the structure. In this

19

study, both strength and deformation model are of interest. In what follows, the statistical

20

model (mean, coefficient of variation and probability density function) for strength and

21

curvature models are evaluated. Moment and curvature data for RC sections (beams with

14

normal strength concrete) have been collected from published literature (Corley, 1966;

Debernardi et al., 2002; Mattock, 1965). Mattock (1965) and Corley (Corley, 1966) based on

similar test programs investigated the rotation capacity of RC beams. In total, they tested 77

beams with different dimensions, material properties and rebar percentage. In the Debernardi

and Taliano (2002) test program, which was on evaluation of the rotation capacity of concrete

beams, 22 beams were tested. They used two different load arrangements in their

experimental program. In this study, the results of all available 99 test specimens (22+77) are

used to derive a statistical model for the strength and curvature of RC sections. Details of all

these specimens can be found in the mentioned studies.

10

In the moment-curvature analysis of the available test results, the theoretical model and

11

assumptions made are similar to those used in the reliability analysis. However, the stress-

12

strain model for the rebar steel is similar to those used in the corresponding studies. Model

13

errors for yield moment, ultimate moment, yield curvature and ultimate curvature are

14

evaluated. The mean and the standard deviation along with the best-fit lognormal distribution

15

parameters for each set of the experimental data are found. Method of ordered statistics is

16

used to find the best-fit lognormal distribution for model error. Table 6 shows the mean and

17

coefficient of variation for different components of the model error.

18

Table 6: Mean and coefficient of variation of model errors

Mattock
(1965)
Mean COV
1.16
0.09

Mean
1.41

1.03

0.21

My

1.02

Mu

0.89

Component

COV
0.17

Debernardi et
al. (2002)
Mean COV
1.15
0.26

Mean
1.26

COV
0.20

0.80

0.16

0.83

0.23

0.90

0.23

0.04

1.02

0.03

1.05

0.09

1.03

0.06

0.13

0.87

0.09

1.03

0.06

0.91

0.12

Corley (1966)

15

All

In Table 6, y , u , M y and M u refer to yield curvature, ultimate curvature, yield moment and

ultimate moment, respectively. Terms Mean and COV show the average and the

coefficient of variation. As is expected, the uncertainty in evaluating yield point components

(considering all the test results) is lower than that of ultimate components. Furthermore, as is

seen, the theoretical procedure that is used in this study underestimates the yield curvature

and bending moment, while overestimates the ones for ultimate state. In Figure 3, based on

the experimental data and the theoretical results, the best-fit line for the model error is shown

on a normal probability paper. The results show that the model error could be reasonably

modeled by the lognormal distribution.

0.80
0.80

0.40
Log(Model Error)

Log(Model Error)

0.40

0.00

-0.40

0.00

-0.40

-0.80
-3.0

-2.0

-1.0

0.0

1.0

2.0

-0.80

3.0

-3.0

Standard Normal Variable

-2.0

-1.0

0.0

1.0

2.0

3.0

Standard Normal Variable


Best Fit Line

Mattock (1964)

Corley (1966)

Debernardi et. Al (2002)

a) Yield curvature

Best Fit Line

Mattock (1964)

Corley (1966)

Debernardi et. al (2002)

b) Ultimate curvature

10

Figure 3: Best fit statistical models for yield and ultimate curvature model error

11

The average and the coefficient of variation resulted from these fitted lognormal distributions

12

are very close to the sample mean and coefficient of variation. Therefore, in this study the

13

statistical model for model error of all components is modeled using lognormal distribution

14

with mean and coefficient of variation shown in Table 5. The statistical data for ultimate
16

bending moment are only shown for comparison, and they will not be used in the reliability

analysis.

5.2 Level of safety in the current design codes

In this part, the MCS technique is used to simulate the required samples for the reliability

analysis. Based on the simulated samples, moment-curvature curves are developed. Then, the

yield and ultimate curvatures as well as the yield moments are derived from the moment-

curvature graph. Figure 4 depicts typical moment-curvature curves obtained from one of the

considered cases. In this Figure, statistical properties of the normalized yield strength and

curvature ductility (with respect to nominal yield strength and curvature ductility) are shown.

10

It should be noted that by substituting the nominal values of random variables, nominal yield

11

strength and curvature ductility are derived. The statistical properties of flexural capacity is

12

comparable with those of Szerszen and Nowak study (2003). For the ordinary cast-in-place

13

concrete, their results showed 1.19 and 0.089 as the bias factor and the coefficient of

14

variation of the flexural strength of RC beams.

17

Figure 4: Sample set of simulate moment-curvature graphs

Figure 4 shows that the disparity of the ultimate curvature is much higher than that of the

yield curvature. Furthermore, for many of the considered cases the coefficient of variation of

the curvature ductility is about triple that of the strength. The reason behind this considerable

difference is that the curvature ductility depends on all of the concrete stress block parameters

including the ultimate strain of concrete while the strength depends on fewer random

parameters. The concrete stress block parameters depend on the concrete compressive

strength. Therefore, dependence of the ductility to all concrete stress block parameters makes

the reliability of the ductility limit state highly sensitive to the concrete compressive strength.

10

Figure 5 shows the sensitivity of the reliability of ductility limit state to the concrete strength.

11

The moment-curvature curves are used to derive the flexural capacity (yield strength) and the

12

curvature ductility of the cross section. Then, using limit states g1 and g2, reliability indices

13

for the strength and the ductility limit states are derived. Figure 5 shows the reliability indices

14

of strength and curvature ductility limit states for different values of steel and concrete

15

strengths and based on different design codes. Results in Figure 5 show that the reliability of

16

strength limit state is higher than that of ductility limit state. Furthermore, using different

17

design codes in the design procedure almost results in about the same level of safety for

18

strength. The European and ACI code have lower strength reliability indices in comparison

19

with the other design codes. In contrast, safety level of ductility limit state has high disparity

20

for different design codes. The American and Australian codes provide safer design for the

21

ductility based limit state while the Canadian code provides the lowest safety level.

18

4.5

Reliability Index

Reliability Index

4.5

3.0

1.5

0.0

3.0

1.5

0.0
15

25

35

45

15

Concrete Compressive Strength (MPa)

35

45

ACI 318-11

AS 3600-09

ACI 318-11

AS 3600-09

NZS 3101-06

CSA A23.3-04

NZS 3101-06

CSA A23.3-04

EC2-04

EC2-04

a) Ductility, f y 400MPa

b) Strength, f y 400MPa
4.5

Reliability Index

4.5

Reliability Index

25

Concrete Compressive Strength (MPa)

3.0

1.5

0.0

3.0

1.5

0.0
15

25

35

45

15

Concrete Compressive Strength (MPa)

35

45

ACI 318-11

AS 3600-09

ACI 318-11

AS 3600-09

NZS 3101-06

CSA A23.3-04

NZS 3101-06

CSA A23.3-04

EC2-04

EC2-04

c)

25

Concrete Compressive Strength (MPa)

Ductility, f y 500MPa

d) Strength, f y 500MPa

Figure 5: Reliability indices for strength and ductility limit states

For a specific concrete cross section, using a higher concrete strength allows for higher rebar

percentage and accordingly less ductility and this in turn results in a lower reliability index

for the ductility based limit state. Figures 5a and 5c show that in some cases the reliability of

ductility based limit state could drop to less than 1.5 for some design codes such as the

Canadian standard, while the corresponding strength based limit state shows high reliability
19

index of near 4.0.

According to Equation 2, design codes allow lower rebar percentage for higher rebar yield

stress. On the other hand, using higher rebar yield stress results in lower nominal ductility

levels. Using higher rebar yield steel leads into a reduction of both the maximum rebar

percentage and the curvature ductility. As can be seen in Table 7, for all of the considered

design codes, using higher rebar yield steel leads into a reduction of both the maximum rebar

percentage and the curvature ductility. Although reduction in rebar percentage is in favor of

increasing the reliability index of the ductility limit state, the lower nominal curvature

ductility leads to a reduction in the reliability index of the ductility limit state. Therefore, the

10

final reliability indices for the ductility limit state depend on these two contradicting effects

11

of implementing the high yield stress rebar. The statistical models shown in Table 4 indicate

12

that, despite the relatively lower uncertainty in the 500MPa steel, the gap is not large enough

13

to make a considerable impact on the end reliability index comparing to the 400MPa steel.

14
15

Table 7: Maximum allowable rebar percentage and corresponding minimum curvature


ductility for different design codes
f c'
(MPa) (MPa)
fy

300

400

500

25
35
45
25
35
45
25
35
45

ACI 318-11

AS 3600-09

max

max

0.0226
0.0296
0.0346
0.0169
0.0222
0.0260
0.0135
0.0178
0.0208

2.90
2.81
2.78
2.35
2.29
2.27
1.99
1.94
1.93

0.0217
0.0303
0.0390
0.0163
0.0228
0.0293
0.0130
0.0182
0.0234

3.19
3.01
2.84
2.57
2.45
2.32
2.17
2.07
1.98

16

20

NZS 310106

max
0.0301
0.0402
0.0465
0.0203
0.0271
0.0314
0.0148
0.0197
0.0228

2.02
1.94
1.93
1.90
1.83
1.82
1.81
1.76
1.75

CSA A23.304

max
0.0329
0.0440
0.0539
0.0224
0.0300
0.0367
0.0164
0.0220
0.0269

1.86
1.78
1.72
1.74
1.67
1.63
1.65
1.60
1.57

EC2-04

max

0.0229
0.0321
0.0412
0.0172
0.0240
0.0309
0.0137
0.0192
0.0247

3.53
2.99
3.12
2.84
2.67
2.55
2.38
2.26
2.16

According to the results shown in Figures 5a and 5c, the difference in the reliability indices

resulted from 400MPa and 500MPa steel materials is not considerable. As mentioned

previously, this is due to the fact that the effects of lower nominal ductility and lower rebar

percentage on the reliability index act opposite to each other.

5.3 Calibrating the ductility safety factors

Level of target reliability has a big influence on the calibration of safety factors. As

previously discussed, target reliability generally depends on the cost of safety measure and

the consequences of failure. Here, two target reliability indices of 2.3 and 3.1 will be used to

calibrate appropriate safety factors for the ductility based limit state. Equation 7 shows the

10

relation between the safety factor and the maximum rebar percentage.

f c' cu
1
11
f y cu y S .F .

(7)

11

Now, instead of using the code requirements for making a ductile design, Equation 7 is used

12

to calculate the maximum rebar percentage. A wide range of safety factors is used to evaluate

13

the maximum rebar percentage. Then, using reliability analysis, the safety factor

14

corresponding to the desired target reliability is evaluated. In Figure 6, based on two different

15

target reliability indices, the normalized safety factors for different design codes are shown.

16

To derive the normalized safety factor, the safety factor resulted from reliability analysis is

17

divided by the available safety factors currently used by each design code. Therefore, a

18

normalized safety factor greater than 1.0 shows that the considered code does not provide

19

adequate safety margin for that particular case. Results in Figure 6 are based on rebar yield

20

stress of 400MPa.

21

1.50

1.25

1.25

Normalized safety factor

Normalized safety factor

1.50

1.00
0.75
0.50
0.25
0.00

1.00
0.75
0.50
0.25
0.00

15

25

35

45

15

Concrete Compressive Strength (MPa)

25

35

45

Concrete Compressive Strength (MPa)

ACI 318-11

AS 3600-09

ACI 318-11

AS 3600-09

NZS 3101-06

CSA A23.3-04

NZS 3101-06

CSA A23.3-04

EC2-04

(a)

EC2-04

t arg et 2.30

(b)

t arg et 3.10

Figure 6: Normalized safety factors for different design codes

Results for the target reliability index of 2.3, except for the Canadian code, almost all

available design codes provide adequate margin of safety for a ductile design. However,

when the target reliability index is increased to 3.1, none but the American code, provide

sufficient safety margin for ductile design in many of the considered cases. Due to the

allowance for higher rebar percentage, using the Canadian design code leads to the highest

normalized safety factor, and even for the low target reliability of 2.3, the provided safety

level in not acceptable. The only safety factor that the Canadian standard is relying on is a

factor that indirectly comes from strength safety factors. Besides, this design code introduces

10

11

the ultimate concrete strain of 0.0035 instead of 0.0030.

Conclusions

12

The probabilistic analysis of RC members with respect to strength and ductility limit states at

13

the sectional level are investigated for different design codes. Base on the results, the salient

14

features of this study are summarized as follows:

22

1. Current literature does not adequately address appropriate probabilistic models for

evaluating ductility related parameters and thus special attention should be paid to

this area.

2. Setting a target safety level is critical when calibrating safety factors. Current

design codes worldwide are calibrated for the strength limit states and the results

of these study confirms adequacy of the currently in-place calibration. On the

other hand, there is a need for definition of appropriate target safety levels when

dealing with the ductility as a limit state.

3. RC beam sections designed based on different standards show almost uniform

10

reliability for the strength based limit state. However, with respect to the curvature

11

ductility reliability, the results exhibit great disparity. This is somewhat expected,

12

as the minimum ductility requirements of these design codes are different. Except

13

in a few cases, the reliability indices for ductility limit state are considerably lower

14

than those of the strength limit state. The results confirm the understanding that

15

the statistical properties of the flexural capacity of reinforced cross sections does

16

not depend on concrete and rebar strengths; on the other hand, the ductility

17

capacity of RC sections depends on the concrete and steel strengths as well as the

18

equivalent rectangular concrete stress block parameters.

19

4. Current design codes provide different minimum requirements for curvature

20

ductility that are generally not close to each other. These minimum requirements

21

aim to provide a minimum level of curvature ductility. Lower bound values along

22

with a rudimentary safety factor are used for evaluating these requirements. Based

23

on the findings of this study, apart from not being rational, this procedure would

24

not guarantee a minimum safety level.

23

References

Aci 318 (2011) "Building Code Requirements for Structural Concrete and Commentary".

Farmington Hills, MI, USA, American Concrete Institute.

As 3600 (2009) "Australian Concrete Structures Standard", Standard Australia. Sydney,

Australia.

Ayyub, B. M. and Haldar, A. (1984) "Practical Structural Reliability Techniques", Journal of

Structural Engineering, vol. 110, no. 8, pp. 1707-1724.

Ayyub, B. M. and Lai, K.-L. (1991) "Selective Sampling in Simulation-Based Reliability

Assessment", International Journal of Pressure Vessels and Piping, vol. 46, no. 2, pp. 229-

10

249.

11

Bartlett, F. (2007) "Canadian Standards Association Standard A23. 3-04 Resistance Factor

12

for Concrete in Compression", Canadian Journal of Civil Engineering, vol. 34, no. 9, pp.

13

1029-1037.

14

Benjamin, J. and Cornell, C. (1975) "Probability, Statistics and Decision for Civil Engineers",

15

New York: McGraw-Hill.

16

Bournonville, M., Dahnke, J. and Darwin, D. (2004) "Statistical Analysis of the Mechanical

17

Properties and Weight of Reinforcing Bars", University of Kansas Report.

18

Corley, W. (1966) "Rotational Capacity of Reinforced Concrete Beams", Journal of the

19

Structural Division, vol. 92, pp. 121-146.

20

Csa A23.3 (2004) "Design of Concrete Structures". Canadian Standards Association.

21

De Stefano, M., Nudo, R., Sar, G. and Viti, S. (2001) "Effects of Randomness in Steel

22

Mechanical Properties on Rotational Capacity of Rc Beams", Materials and Structures, vol.

23

34, no. 2, pp. 92-99.

24

Debernardi, P. G. and Taliano, M. (2002) "On Evaluation of Rotation Capacity for

25

Reinforced Concrete Beams", ACI Structural Journal, vol. 99, no. 3, pp. 360-368.

26

Ec2 (2004) "Design of Concrete Structures: Part 1: General Rules and Rules for Buildings".

27

Brussels, Belgium, European Committee for Standardization.

28

Ho, J., Kwan, A. and Pam, H. (2004) "Minimum Flexural Ductility Design of High-Strength

29

Concrete Beams", Magazine of Concrete Research, vol. 56, no. 1, pp. 13-22.

30

Ito, K. and Sumikama, A. (1985) "Probabilistic Study of Reduction Coefficient for Balanced

31

Steel Ratio in the Aci Code", ACI Structural Journal, vol. 82, pp. 701-709.

32

Jcss (2012) "Probabilistic Model Code", The Joint Committee on Structural Safety. Technical
24

University of Denmark.

Kappos, A., Chryssanthopoulos, M. and Dymiotis, C. (1999) "Uncertainty Analysis of

Strength and Ductility of Confined Reinforced Concrete Members", Engineering Structures,

vol. 21, no. 3, pp. 195-208.

Kassoul, A. and Bougara, A. (2010) "Maximum Ratio of Longitudinal Tensile Reinforcement

in High Strength Doubly Reinforced Concrete Beams Designed According to Eurocode 8",

Engineering Structures, vol. 32, no. 10, pp. 3206-3213.

Kwan, A. K. and Ho, J. C. (2010) "Ductility Design of High-Strength Concrete Beams and

Columns", Advances in Structural Engineering, vol. 13, no. 4, pp. 651-664.

10

Lu, Y. and Gu, X. (2004) "Probability Analysis of Rc Member Deformation Limits for

11

Different Performance Levels and Reliability of Their Deterministic Calculations", Structural

12

safety, vol. 26, no. 4, pp. 367-389.

13

Mattock, A. H. (1965) "Rotational Capacity of Hinging Regions in Reinforced Concrete

14

Beams", ACI Special Publication, vol. 12, pp. 143-181.

15

Melchers, R. E. (1999) "Structural Reliability Analysis and Prediction".

16

Nowak, A. S. and Szerszen, M. M. (2003) "Calibration of Design Code for Buildings (Aci

17

318): Part 1-Statistical Models for Resistance", ACI Structural Journal, vol. 100, no. 3, pp.

18

377-382.

19

Nzs 3101 (2006) "Concrete Structures StandardPart1the Design of Concrete Structures",

20

Wellington: Standards New Zealand.

21

Park, R. and Paulay, T. (1975) "Reinforced Concrete Structures / R. Park and T. Paulay",

22

New York, Wiley.

23

Park, R. and Ruitong, D. (1988) "Ductility of Doubly Reinforced Concrete Beam Sections",

24

ACI Structural Journal, vol. 85, no. 2, pp. 217-225.

25

Szerszen, M. M. and Nowak, A. S. (2003) "Calibration of Design Code for Buildings (Aci

26

318): Part 2-Reliability Analysis and Resistance Factors", ACI Structural Journal, vol. 100,

27

no. 3.

28

Trezos, C. (1997) "Reliability Considerations on the Confinement of Rc Columns for

29

Ductility", Soil Dynamics and Earthquake Engineering, vol. 16, no. 1, pp. 1-8.

30

25

Notations

As

Tensile rebar area, mm2

As'

Compressive rebar area, mm2

Width of the rectangular section, mm

Neutral axis depth, mm

Effective depth of the rectangular section, mm

7
8

d'

Distance from extreme compression fiber to centroid of compression reinforcement,


mm

Ec

Secant modulus of concrete, mm

10

Es

Modulus of steel, mm

11

f c'0

Concrete compressive strength of a standard speciment, MPa

12

f c'

Concrete compressive strength, MPa

13

ft

Concrete tensile strength, MPa

14

fu

Reinforcement ultimate strength, MPa

15

fy

Reinforcement yield strength, MPa

16

g1, g 2 Limit states

17

Height of the rectangular section, mm

18

A parameter used in defing concrete compressive strength

19

MQ

Bending moment resulted from loads, N-mm

20

MR

Nominal bending capacity, N-mm

21

pf

Probability of failure

22

Load random variable

23

Resistance random variable

24

Xi

Random variables

25

XM

Model error random variable

26

Y1 to Y 4

27

1 , 1 Equivalent stress block parameters

28

Reliability index

29

DL

Safety factor for dead load

Random variables related to concrete properties

26

c0

Concrete strain at peak stress, mm/mm

cu

Extreme fiber concrete ultimate strain, mm/mm

su

Rebar steel yield strain, mm/mm

Extreme tensile rebar strain, mm/mm

Rebar steel yield strain, mm/mm

Curvature ductility

Tensile rebar percentage

Concrete material resistance reduction factor

Steel material resistance reduction factor

10

Ultimate curvature

11

Yield curvature

27

You might also like