Skrabek Strength Reliability

STRENGTE RETJIåBT'TTY OF STTORT ÃND SLESTDER
COS,ÍPOSTTE STEETJ-CO}{CRETE
COLT¡M3üS
BY
BRYAN WILLTAI{ SKRÀBEK
A Thesis
Subnitted to the Faculty of Graduate Studies
in Partial Fulfillnent of the Requirements
for the Degree of
MASTER OF SCIENCE
Department of Cívil Engineering

University of Manitoba
I{innipeg, Manitoba
(c) December, L989

STREI{GTE RELIÂBILITY OF SBOR.T AI{[D SLENT}ER
COHPOS ITE STEET-COI{CRETE COTIIB{I{S
BY
BRYAI{ }ilLLIÄM SKRABEK
A thesis subnrined to thc Facurty of craduare Studies

of
the univenity of Maîroba in partial fulfiilment
of the requirenrerìrs
of the degree of
HASTER OF SCIET{CE
@ ¡.990
Permission has been granred to the LIBR.ARY

oF THE UNIVER-
s¡TY oF MANITOBA ro rend or seil copies of this thesis. ro
thc NATIONAL LIBRARY OF CANADA tO ¡N¡CTOIiIM
this
thesis and to lend or æ[ copies oí rhe firm,
and uNIVERsrry
${ICROFILMS to publish an absr¡acr of ¡his thesis.
The author reserves other pubrication righfs, and

neither thc
thesis nor extensive extracts from it may
be pnntec or other-
wise reproduced without the author's writ,te¡r permission.
i
ABSTR:ACT
Structural reliability analysis requires that the rela-

tionship between the probability distribution of member
resistance and the probability distribution of load effects
be known so that realist.ic member understrength and overload
factors can be computed for design purposes.
In this study, the probability distribution of the ratio
of theoretical to nominal (design cod.e) strength for compos-
ite steel-concrete beam-columns r¡¡as established and the
major variables affecting the probability distribution
identified. The beam-corumns studied r¡rere comprised. of a
st,ructurar steel wide flange shape surrounded by a reinfor-
cing bar cage and completely encased in concrete.
A computer program !ùas used to calculate the theoretical-
and nominal strengths. The accuracy of the theoretical
model was established by comparisons to test specimens docu-
mented in the literature.
Probability distributions of the geometric and. mechani-
cal properties of the variables which detennine the
resistance of the bean-col-umn r^rere established by reviewing
existinE literature"
The Monte Carlo technique was used to simulate the
resistance of typical beam-coLumns in order t,o determine
both the probability distribution of the member resistance
and the variables having the most effect, on the lower tail_
of the probability distribution.
l- l-
For short composite beam-columns, the variables that

affected the probability distributions of the strength hrere
the specifÍed concrete strength, ratio of structural steel
area to gross area of cross-section, slenderness ratio and
end eccentricity ratio"
The same variables were found to affect the strength
probability distributions of slender composite beam-col-umns
except that, the effect of specified concrete strength became
negligible for beam-columns with very large slenderness
ratios and the effect of end eccentricity ratio became neg-
ligible for large eccentricity ratios.
l_ l_ l_
åCK}üOWTJEDG}'ÍENTS
The author would like to acknowledge the assistance of the

National Sciences and Engineering Council of Canad.a for providing
funding through research grants to Dr. S"A. Mirza.
sincere appreciation goes to Dr. s.A" Mirza for his guidance
and extreme patience during this investigation"
Special thanks are due to Mr" C"E" Mickelson, p" Eng" , for
providing support and encouragement.
iv
TÃBTJE OF coStTES{TS
CHAPTER PAGE
1 TNTRODUCTION 1
1"1 Overview of study 2
1.2 Outline of Research Program I

THEORETTCAL BEA¡{-COLUMN STRENGTH t_1
Review of Previous Work
2 " 1- t_3
2"2 Assumptions Used In Theoretical l_8

Strength Model
2"3 Cross-Section Discretization 20
2"4 Cross-Section Strength Model 26
2"5 Slender Beam-Column Strength Model 35
2"6 Stress-Strain Curves for Concrete 48
2. "I6 Unconfined Concrete 49
2.6.2 Partially Confined Concrete 53
2"6"3 Heavily Confined Concrete 66
2.6"4 Tensile Stress-Strain ReIa- 68
tionship of Concrete
2.6.5
Summary of Stress-strain 70
Relationships for Concrete
2"7 Stress-Strain Curves for Steel 72
2.7 "1- Structural Stee1 73
2"7.2 Reinforcing Steel 74
2"8 Residual Stesses in Rolled SÈruc- 77
tural Steel
2"9 Comparison of Experimental Results 91
to Theoretical Mode1
2"lO Calculation of Mode1 Error 1_01
NOMTNAL BEAM-COLUMN STRENGTH l-11_
3"1 Assurnptions 111
3.2 Nominal Strength Program LL6
3.3 Comparison of Design Codes 1-24
PROBABILITY MODELS OF BASIC VARIABLES 1,29
4"1 Concrete 130
4.1 .1 CompressÍve Strength L34
4"L.2 Tensile Strength 137
4"L.3 Modulus of Elasticity 140
f7
4"2 Structural Steel ].43

4"2"I Modulus of Élasticity L44
4"2"2 Yield Strength l.45
4"2"2"L Web Yield Strength 150
4"2"2"2 Flange Yield Strength t54
4.2 "2 "3 ProbabiLity Distribu- L54
tion of Yield Strength
4 "2 "3 UlÈinate Strength L58
4.2"4 Strain at Initiation of 159
Strain-Hardening
4"2"5 Strain Hardening Modulus L59
4 "2 " 6 Dimensional Variations l-60
4"2"6"t Section Depth 161_
4"2"6"2 Flange $ridth ]-62
4 "2 "6 "3 Flange Thickness L62
4"2"6.4 Web Thickness L67
"7 Residual Stresses
4 "2 767
4"3 Reinforcing Steel 172
4 " 3.l- ModuLus of Elasticity 173
4.3"2 Yield Strength L74
4 "3 " 3 Ultinate Strength ]-78
4.3 " 4 Strain at fnitiation of 1,7 I
Strain-Hardenj-ng
4" 3.5 Ultinate Strain L79
4"4 Column Geornetry 1,79
4.4.1, Column Length L80
4"4.2 Co1umn Width and Depth L8L
4.4.3 Concrete Cover 181
4"4"4 Placement of Layers of Verti- 1-82
cal Bars
4"4"5 Spacing of Rectangular Hoops 183
SIMUI,ATION AND ANALYSTS OF COMPOSITE 1-84
BEAM-COLUMN STRENGTH
5"L Monte Carlo Technigue t_84
5"2 Descriptions of Beam-Columns Studied L85
5"2"1- Basic Stucty t_88
5.2 "2 Supplemental Study 190
5"3 Short Composite Beam-Columns L94
Vi
5"3"1- Overall Strength Variations L97
5.3"2 Effect of Variables Used for 2O4
Basic Study
5"3"2"I Effect of Slenderness 2O4
Ratio
5.3"2"2 Effect of Specified 2O5
Concrete Strength
5"3"2.3 Effect of StructuraL ztg
Steel Ratio
5"3"2"4 Effect of End Eccen- 227
tricity Ratio
5.3 " 2.5 Sensitivity Analysis 234
5.3"2.6 Sumnary of Effects of 238
Variables Used for
Basic Study
5 " 3 " 3 Ef fect,s of Variables Used for Z3g
SupplementaL Study
5.3.3.L Effect of Specified 239
Yield Strength of
Structural Stee1
5"3"3"2 Effect of Strain- 243
Hardening of Stee1
5.3.3.3 Effect of Quality of Z4g
Concrete
5"3"3.4 Summary of Effects of ZS3
Variables Used for
Supplemental Study
5"4 Slender Composite Beam-Columns 253
5.4"1- Overall Strength Varíations ZS4
5 "4"2 Effects of Variables Used in 257
Basic Study
5.4.2.L Effect of Slenderness 2SB
Ratio
5"4"2.2 Effect of Specified 266
Concrete Strength
5.4.2.3 Effect of Structural 274
Steel Ratio
5"4.2.4 Effect of End Eccen- 282
tricity Ratio
5"4"2.5 Sensitivity Analysis 29O
5.4"2.6 Summary of Effects of 293
Variables Used for
Basic Study
i/ii
5"4.3 Effect of Variables Used for 294
Supplemental Study
5"4"3"1 Effect of Specified 294
Yie1d Strength of
Structural Steel
5"4"3"2 Effect of Strain- 300
Hardening of Steel
5.4"3"3 Effect of Qualíty of 305
Concrete
5"4"3"4 Summary of Effects of 309
Variables Used for
Supplemental Study
SUMMARY AND CONCLUSIONS 3Lo
6"L Summary 3 i_0
6"2 Conclusions 311
6"2"1- Short Columns 311
6"2"2 Slender Columns 3L2
vii i
Ï,ÏST O5' FTGURES
Figure Description Page

1-"1 Definition of failure, probabiJ-ity of 4
failure, and safety index B
I"2 Loading and Support Conditions of CoI- 7
umns Studied
2"I Flow Chart of Computing Procedure 12
2"2 Material Types in Composite Cross- 22
Section
2.3 Discretization of Composite One-Half 23
Cross-Section for Theoretical Strength
Subroutine
2"4 typical M-q-P Relationships for Compos- 27
ite Cross-Sections
2"5 Typical Composite Cross-section P-M 28
Interaction Diagram
2"6 Strain Gradient in Cornposite Cross- 3l-
Section
2"7 Flow Chart of Calculation of M-$-P 34
Relationship for Composite Cross-Section
2"8 Deflected Shape of Slender Composite 37
Beam-Column under Eccentric ÀxiaI Load
2.9 Flow Chart for Calculating Slender Col- 47
umn M-0-P Relationship
2"1-O Hognestad Stress-Strain Relationship for 50
Unconfined Concrete
2"1"1, Kent and Park Stress-Strain Relationship S0
for Unconfined Concrete
2.1,2 Unconfined Concrete Compressive Stress- 34
Strain Relationship Used in Theoretical
Strength Subroutine
2"1"3 Modified Kent and Park Stress-Strain 57
Relationship for Concrete Confined by
Rectangular Ties
2"L4 Sheikh-Uzuzmeri Stress-Strain Relation- 57
ship for Concrete Confined by Rectangu-
lar Ties
2").5 Comparison of Unconfined and partially 65
Confined Concrete Compressive Stress-
Strain RelaÈionships for Co1umn LH-L00
IX
2"L6 Partially Confined Concrete Compressive 67

Stress-Strain Relationship Used in Theo-
retical- Strength Subroutine
2"L7 Heavily Confined Concrete Compressive 69
Stress-Strain Relationship Used in Theo-
retical Strength Subroutine
2"L8 Concrete lensile Stress-strain Relation- 7l
ship Used in Theoretical Strength Sub-
rout,ine
2"L9 Structural Steel Stress-Strain Relation- 75
ship in Tension or Compression Used in
Theoretical Strength Subroutine
2"2O Reinforcing Steel Stress-Strain Rela- 76
tionship in Tension or Compression Used
in Theoretical Strength Subroutine
2"2L Residual- Stresses in Wide F1ange Shapes 8L
2"22 Residual Stress Distribution in !{ide 90
Flange Shapes Used in TheoreÈical
Strength Subroutine
2"23 Cunulative Frequency of Ratio of Tested 1,O2
to CaÌculated Resistance of Composite
Beam-Column Specimens with 0"0
o"2
2.24 Cumulative Freguency of Ratio of Tested l-03
t,o Calculated Resistance of Composite
with 0.2
Beam-Column Specimens
2"25 Cumul-ative Frequency of Ratio of Tested 104
to Calculated Resistance of Composite
Beam-Column Specimens with e/h = æ
2.26 2,n", and V mod.a¿ Used !O7
3.1 Code Stress-Strain Relationship for Con- l_13
crete
3.2 Code Stress-Strain Relat,ionship for 115
SÈee1
3" 3 Axial Load-Bending Moment Interaction l_19
Diagram for Nominal Strength Subroutine
4.I Definit,ions of YíeId Stress of Steel i-48
4.2 Variation of l{eb Lower yield Strength /rz j-56
4.3 Cumulative Freguency of tfeb Lower yield LS7
Strength /"2
4" 4 Variation of Ratio of Measured t,o Nomi- t-63
nal Flange Iîídth of IrI Sections
4"5 Cumulative Freguency of Ratio of Mea- L64
sured to Nominal Flange Width of W
Secti-ons
4"6 Variation of'Ratio of Measured to Nomi- 165
nal Flange Thickness of hf Sections
4 "7 Cumulative Frequency of Ratio of Mea- 1"66
sured to Nominal Flange Thickness of W
Sections
4 "8 Variation of Ratio of Measured to Nomi- t-68
nal Web Thickness of W Sections
4 "9 Curnulative Freguency of Ratio of Mea- i,69
sured to Nominal t{eb Thickness of I{
Sections
4.10 Probability Density for Static Yield i,76
St,rength of Grade 60 (4i_4 MPa) Reinfor-
cing Bars
4 " Ll- Cumulative Frequency for Static yield i-77
Strength of Grade 60 (4L4 MPa) Reinfor-
cing Bars
5.1- The Monte-Carlo Technique i-86
5.2 Nominal Column Cross-Section Details 1-87
5.3 Axial Load-Bending Moment Strength i_99
fnteraction Curves of Randomly Generated
Samples of 500 Short Cornposite Columns
(Co1unn 6-50-4-22 - Table 5"1)
5.4 Axial Load-Bending Moment Strength 199
Interaction Curves of Randomly Generated
Samples of 500 Short Cornposite Columns
(Co1umn 4-50-8-22 - Tab1e 5.J.)
5"5 ACI 3l-8-83 Ultinate Strength fnteraction ZO3
Diagram for Column 4-50-4-0 (Table 5"L)
5"6 Effect of Slenderness RatÍo on the Ratio 206
of Theoretical to Nominal Strength of
. Short Composite Steel-C'oncrete Beam-
Columns (Colurnns 4-50-4-0 and 4-50-4-22)
5"7 Effect of Slenderness Ratio on the Ratio ZOg
of Theoretical to NominaÌ Strength of
Short Composite Steel-Concrete Beam-
Columns (Co1umns 6-50-8-0 and 6-50-B-22)
5"8 Effect of Specified Concrete Strength on Zj_3
the Ratio of Theoretical to Nominal
Strength of Short Cornposite Steel-
Concrete Beam-Columns (Columns 4-50-4-22
and 6-50-4-22)
XI
5"9 Effect of Specified Concrete Strength on 216

Strength of Short Composite Steel--
Concrete Beam-Columns (Columns 4-5O-B-22
and 6-50-8-22)
5.1-0 Effect of Structural Steel Ratio on the Z2L
Ratio of Theoretical to Nominal Strength
of Short Composite Steel-Concrete Beam-
Columns (Co1urnns 4-SO-4-22 and
4-5O-8-22)
5"L1 Effect of Structural Stee1 Ratio on the 224
of Short Composite Steel-Concrete Beam-
Columns (Columns 6-50-4-22 and
6-50-8-22 )
5"1-2 Range of Ratio of Theoretical to Nominal 229
Strength of Short Composite Steel-
Concrete Beam-Columns
5.13 Range of the Coefficient of Varition of 233
the Ratio of Theoretical- to Nominal
5.14 Effect of Variabilities of properties of 235
Constituent Materials on Overall
Strength Variability of Beam-Column
4-50-4-0 (Table s.1)
5"15 Effect of Specified Structural Steel 24O
Yie1d Strength on the Ratio of Theoreti-
cal 'to Nominal Strength of Short Compos-
iÈe Steel-Concrete Beam-Columns
5"L6 Effect of Strain-Hardening of Steel on 244
Èhe Ratio of Theoretical to Nominal
5"I7 Effect of Concrete euality on the Ratio ZSO
Short Composite Steel-Concrete Beam-
Columns
5"l-8 Axia1 Load-Bending Mornent Strength 2Ss
Samples of 500 Slender Composite Columns
(Column 6-50-4-66 - Tabl_e S"Z)
5.19 Axial Load-Bending Moment Strength 256
Samples of 500 Slender Cornposite Columns
(Co1umn 4-50-g-33 - Table 5"2)
xii
5"2O Effect of Slenderness Ratio on the Ratio ZS9
Slender Composite Steel-Concrete Beam-
Columns (Co1umns 4-50-4-22"L, 4-SO-4-33 t
4-50-4-66 and 4-50-4-L00)
5"2L Effect of Slenderness Ratio on the Ratio 262
Columns (Columns 6-50-8-22 "!, 6-5O-8-33,
6-50-8-66 and 6-50-8-L00)
5.22 Effect of Specified Concrete Strength on 267
Strength of Slender Composite Steel-
Concrete Beam-Co1umns (Columns
4-50-4-22 "I and 6-so-4-22 "t)
5"23 Effect of Specified Concrete Strength on 27O
the Rat,io of Theoretical to Nominal
Strength of Slender Composite Steel-
Concrete Beam-Co1umns (Columns 4-50-4-33
and 6-50-4-33)
5"24 Effect of Structural Steel Ratio on the 275
of Slender Composite Steel-Concrete
Beam-Columns (Columns 4-50-4-33 and
4-50-8-33 )
5"25 Effect of Structural Steel Ratio on the Z7B
Ratio of Theoretical to Norninal Strength
of Slender Composite Steel-Concrete
Beam-Columns (Co1urnns 4-50-4-66 and
4-50-8-66 )
5"26 Range of Ratio of Theoretical to Nominal 284
St,rengÈh of Slender Composite Steel-
5"27 Range of the Coefficient of Variation of ZBg
Strength of Slender Compos-
Beam-Column
ite Steel-Concrete Beam-Columns
5"28 Effect of Variabilities of properties of 2gI
Constituent Materials on Overall
Strength Variability of Beam-Column
4-50-4-66 (Table s"2)
5"29 Effect of Specified Structural Steel 296
Yield Strength on the Ratío of Theoreti-
cal to Nominal Strength of Slender Com-
posite Steel-Concrete Beam-Columns
5"30 Effect of Strain-Hardening of Steel on 3OL
St,rength of Slender Composite Steel-
xiii
5"3L Effect of Concrete Quality on the Ratio 306
of Theoret,ical to Nominal Strength of
Columns
xiv
LÏST OF TÃBTJES
Table Descri-ption Page

2"! Comparison of Strength Ratios Caluclated 63
Using Kent and Park and Sheikh and
Uzumeri Stress-Strain Relations for Con-
crete Confined by Lateral Ties
2.2 Comparison of Measured and Calculated 86
Residual Stresses in W Shapes
2"3 Ratios of Test to Theoretically Calcu- 96
lated Strengths for AIl Specimens
2"4 St,atistical Analysis of Ratios of Test 99
to Calculated Strength
2.5 Variability of Theoretical Strength 108
Model
4.)- fn-batch Variations of Basic Variables l-31
4.2 Overall Variations of Basic Variables L32
4"3 Elastic Modulus of Structural Steel 1,46
4"4 irieb Yie1d Stress Measurements 151_
4"5 Measured Residual Stress in !{íde Flange L7L
Column Shapes
5.1- Specified Properties of Short Beam- 191
Columns Used for Basic Study
5"2 Specified Properties of Slender Beam- L92
Columns Used for Basic Study
5"3 Specified Properties of Short Beam- l-95
Columns Used for Supplemental Study
5"4 Specified Properties of Slender Beam- L96
Columns Used for Supplemental Study
l_
1 TNTRODUCTTON
The probability-based limit states designs are based on
liniting the probability of failure to an acceptable level"
The actual strength of a structural member differs from the
nominal strength calculated by the designer due to varia-
tions in the material strength, variations in dirnensions and
geornetry of the member, and. variations in the accuracy of
eguations used to compute the nominal strength. Sirnilarly,
the load effects upon a member differ from assumed val_ues
due to the variation in loadings over the lifetime of the
structure. To compute the probability of fail-ure due to
load effects being higher than anticipated and/or member
strength being lower than anticipated, the statj-stical-
descriptions of variations of both the load effects and the
member resistance must be known" The statistical combina-
tion of these two variations allows the probability of fail--
ure to be calculated. This procedure is referred to as
reliability analysis.
This study reports the strength statistics reguired for
use in the reliability analysis of composite beam-columns in
which steel shapes are encased in concrete. The factors
cont,ributing t.o the variation of the rat,io of actual (theo-
retical) strength to nominal (design code) strength of com-
posite beam-col-umns are identified. The importance of each
factor to the overall- strength variation and the conditions
under which it applies are analyzed" This work rn¡il-I facil-i-
tate the reliability analysis of representative composite
steel-concrete beam-columns currently underway at Lakehead
University.
The composite beam-columns investigated consist of a
rolled structural steel- wide flange section surrounded by a
cagie of reinforcing bars and entirely encased in concrete.
The column cross-section is rectangular and meets the rein-
forcement reguirements of ACI (American Concrete fnstitute)
3L8-83 (L983) and CSA (Canadian Standards Association)
CAN3-423. 3-MB4 (L984) design codes" Àssumptions regarding
the theoretical behavior of the cross-section and the member
(beam-colunn) are di-scussed in Chapter 2. Assumptions
regarding the behavior of the cross-section and the col_umn
with respect to the design codes are discussed in Chapter 3.
1.I" OVERVIEW OF STUDY
The design procedure for composite beam-columns speci-

fied in ACI Standard 318-83 accounts for the probability of
understrength by the use of capacity reduction factors of
less than 1"0. Sirnilarly, CSA Standard CAN3-A23.3-M84 con-
siders the probability of understrength by assigning mate-
rial- understrength factors of less than 1"0" To account for
the probability of l-oad effects being greater than the
nominal design values, specified loads are multiplied by a
factor g'reater than 1. 0 in both cases. To satisfy the
strength requirements of these standards Eguation 1"1 must
be satisfied"
Factored Resistance (t.t)
Consider a large number of columns, each designed to

have the same resistance to road effects and each assumed to
be subjected to the same specifJ-ed loading. Due to varia-
tions in geornetry and constituents material strengths of the
column, the actual resistance of each col_umn wil1 vary. The
distribution of corumn resj-stance (R) is represented by the
horizontal axis of Figure 1.1(a) . The live and dead l-oads
are also variable and, therefore, each column wiII be sub-
jected to different maxj-mum load effects during its 1ife-
tine. The distribution of maximurn load effects (U) is
represent,ed by the verticaL axis in Figure t-"1(a) . The
4S-degree line represents the condition where the 1oad
effect u equals the resistance R. combinations of u and R
that fa1l above the 4S-degree line represent the fail-ure
condition where R
A functj-on, Y=R/U, is used to sinplify the reliability
analysis (Mirza L985a) . The distribution of the val-ue of y
is represented in Figure 1.i-(b)" The function has a mean
(b)
Probab¡lity[(B/U)<11
or probability of failure
Figure 1"1 - Definition of faj-lure, probability of failure,

and safety index p (Mirza l9g5a)
value Y, and a standard deviation of oi"" A particular col-
umn will fail if its value of Y is less than 1.0" Hence,
the shaded area in Figure l-.1-(b) represents the failure
condition" The ratio of the shaded area to the entire area
under the curve of Y is the probability of failure" The
safety index, B, is a multiple of standard deviation of Y by
which the mean val-ue of Y exceeds the failure IeveI. If
the type of probability distribution of Y is known, the
probability of fail-ure can be calcuLated frorn B. An
increase in ß due to an increase in Y, or a decrease in ûy,
or both, increases the margin of safety and vice versa" The
value B is therefore a measure of structural reliability
(Mirza 1-985a) .
To compute the value of the strength reduction (or mate-
rial understrength) and load factors, the statistical prop-
erties of both the column strength and the load effects must
be considered concurrently" This can be accompli-shed by use
of a step-wise reliability analysís technique as presented
by Mirza (1985a) and reproduced as foLLows:
(a) derivation of statistical models for material strengths
and geometric properties;
(b) selection of equations to predict theoretical strengths;
(c) derivation of the probability distribution of the
strength of representative structural members;
(d) description of load and load effect statistics;
(e) sel-ection of a target safety index B based. on cod.e car-i-
brat,ion studies;
(f) selection of design code format and load factors; and
(g) caLculation of resistance (or material- understrength)
factors based on representative structural- members and
rerative occurrences of different types of roading in
buildings where these members occur.
In this study, items (a), (b) and (c) have been com-
preted for composite steel-concrete beam-columns subject to
the limitations discussed be1ow. Items (d) through (g)
describe the work that is currentry in progress as part of
another study at Lakehead University.
The composite beam-corumns studied meet the forrowing
assumptions regarding loading and support conditions as
shor,r¡n in Figure l"2z
(a) the columns are pin-ended and the effective length is

equal to the actual_ length;
(b) bending is in singre curvature about the najor axis;
(c) end moments are equar and opposite producing a uniform
primary moment distribution along the length of the
beam-column;
(d) no laterar load is applied between the ends of the co1-
umn; and
I
I
I
I
I
I
I
Èq=-¡--éé
P2
P2
(a) Modelof Column in SymmetricalSingle Curvature (b) Forces on Column
,Figure I"2 Loading and Support Conditions-.of

Columns .Studied (Mirza 1990)
(e) loading i-s progressive to failure and is of short dura-
tion such that creep and shrinkage effects of concrete
are not considered"
The loading configuration described by (a) to (d) above
provides for maximum secondary moments due to deflection at
the nid-height of the beam-coÌumn and is an extreme case for
beam-columns in non-sway frames designed in accordance with
ACI 3l-8-83 and CSÀ .A23"3-M84" Shrinkage of concrete, corro-
sion of steel components and increase of concrete strength
due to maturation are ignored. It was assumed. that the
beneficj-al effects due to increase in concreÈe strength with
tine would compensate for any decrease in strength due to
creep and shrinkage of concrete and corrosion of structural_
and reinforcing steel.
L.2 OUTLTNE OF RESEARCIÍ PROGRåM
A computer program vlas used to calculate the theoretical
resistance of composi-te steel-concrete beam-columns. The
program is based on equations and assurnptions consj-dered to
be of greater accuracy than design code equations. The
accuracy (node1 error) of the theoretical strength program
!üas established by comparisons with existing test data of
the ultimate strength of composite beam-columns. The theo-
retical computer program is discussed in detail in Chapter
2.
A computer program designed to calculate the nominal
capacity of composj-te steel-concrete beam-columns was used
to compare the theoretical member strength to that allowed
by the design codes" The nominal capacity is based on the
specified mechanicat and geometríc properties of the column
components and on the eguations given by design codes. The
nominal program subroutine is discussed in Chapter 3 "
Probability distrj-butions of the mechanical and geomet-

ric properties of column components r¡rere taken or derived
from data available in the literature" The probability
distributions are discussed in Chapter 4.
A Monte Carlo technigue (Chapter 5) was used to estab-
lish the statistical properties of the member strength.
This nethod consists of repeated simulations of a chosen
sample col-umn using random serections of the magnitudes of
variables based on the probability distributions mentioned
previously. The theoretical strength of each simulated cor-
umn is cal-culated by the program subroutine described in
Chapter 2.
The ratio of theoret,ical to nominal strength was calcu-
lated for each configuration of composite beam-corumn stu-
died" The resulting ratios r¡rere then analyzed statisticalÌy
to determine the shape of the probabirity distribution and
its properties for each beam-column studied" The probabil-
ity dist,ributions so generated $rere used to investigate the
effect,s of different variabres sÈudied" The contribution of
l_0
variability of the properties of structural steel, concrete,

and model error to the overall strength variations of a few
individual columns r¿as also examined" The computed data is
discussed separately for short and for sl-ender beam-columns
in Chapter 5" A sunmary of the study and conclusions drawn
from it are presented in Chapter 6.
The methodology of the research program is similar to,
although more refined than two earlier studies for rein-
forced concrete beam-coIumns. Grant et aI. (1978) studied
the strength variation of short reinforced concrete col-umns.
l{.írza and MacGregor (1989) studied the strength variation of
slender reinforced concrete columns" The rnajor difference
between this report and the above-noted studies are the
effects of the structural steel core and of concrete con-
finement due to minimum ties specified in ACI Standard
3l-8-83 (and CSA Standard CAN3-423.3-M84) on the strength
variation of beam-columns.
I1
2 TTTEORETTCÃTJ BEAß{-COLU}.Í3{ STRESIGTIT
To analyze the theoretical- strength of a composite beam-

column, a computer program previously developed at Lakehead
University (Mirza L989) was tested and revised as reguired
for use in this study. The changes implemented into the
program for use in this study j-ncluded more efficient tech-
niques to allow for full strain hardening of steel, general-
ized interpolation techniques (Lagrangian), definition of
maximum allowable interpolation errors, accounting for
numerical discrepancies due to column behavior at extreme
values, reduction of computing time reguired and addition of
Monte Carl-o simulations. A brief flow chart of the comput-
ing procedure used is shown in Figure Z.I.
The entire program consists of a main driver program,
CO¡ÍPOSIT, and two major subroutines. The main driver read.s
input, initiates Monte Carlo variations of input data if
reguired, calls the najor subroutines and statisticall-y ana-
lyzes the output data. One of the two major subroutine
progirams, RTIIEO, analyzes for the theoreticaL strength of
the composite column and the other, RNOM, cal-culates the
nominar strength of the corumn following the design require-
ments of ACr 318-83 or csA A23"3-M84" The theoretical model-
and related subroutines are discussed in this chapter" The
nominal- strength model and subroutine are discussed in chap-
ter 3"
T2
Reod lnput Voriob'les
Colculote Ì'lorrrínol Strerrgth

Rn
Monte Corlo
Simulotion rrf Vrrricrlrles
Colculote Theoreticol Strength
Required
Nr-¡mber of
Simuleticns
Attoined
Stotisticcl Anolvsis of
RtlR;
Rotios
Output Doto
figure ?.'i - Flow Chort of ûcmputinç Prccedure

13
2" 1 REVTEW OF PREVIOUS W'ORK
In order to describe the theoreticaL model used in this

study, it was considered useful to see what technigues and
assumptJ-ons have been used by others in previous studies of
composite beam-columns. A revier¿ of the published work on
the theoreticaL analysis of cornposite col-umns is briefJ-y
summarized below.
Bondal-e (L966 a,b,c) tested composite beam-columns and
attempted to duplicate his test results by analyzing the
columns with theoreticaL model-s" Strain compatibility
assumptions r¡rere used to establish the cross-section rel-a-
tionships between axial load, bending moment and curvature.
The tangent modulus theory vras used to analyze the
concentric load case" slender, eccentrically loaded col-umns
were assumed to defl-ect in the form of a cosine curve.
Graphically solving for the eguilibrium relationship between
load end eccentricity and nid-height deflection establ-ished
maximum eccentricities at the fail-ure section for given
axial loads" the assumptions mad.e by Bondale are the same
as described in Section 2.2 with the following except¡-orr=,
(a) tensile strength of concrete was neglected; and (b)
residual stresses in the structural steer were negrected.
Anarysis of the test specimens showed that the tested road
capacities were from l-.1-5 to i-"23 times the theoretical
capacities "
L4
Basu (1-967) presented a computer method to approximate

the capacity of pin-ended composite beam-col-umns subjected
to equar and opposite end eccentricities and uniaxial bend-
ing" The method made efficient use of computer time since
equilibrium needed to be satisfied onry at the ends and at
the nid-height of the beam-column. The method was similar
to that of Bondare except that interporations were made
mathematically by the computer rather than graphically.
comparisons made by Basu with the test results of Bondale
(1-966 a,b,c) predicted the ratio of actual to predicted
strength to be L"2l- to l-.33 when the maximum concrete stress
was assumed to be two thirds of the cube strength. The
results improved to 1.09 to 1"20 when a val_ue of eighty
percent of the cube strength was used. This shows that the
stress-strain curve assumed for concrete significantry
affects the results" The assumptions Basu used in his anal-
ysis are similar to those made by Bondale. Two exceptions
are notable. An initial defLection of the corumn in the
form of a cosine curve is assumed to account for some ini-
tial camber of the col-umn. subsequent deflections due to
secondary moments are assumed to be in the form of a part
cosine v/ave.
Basu and Hill (1968) confirmed the accuracy of Basu
(1967) by developing a more precise numericar integration
nethod of analysis" nguilibriun is satisfied at a number of
points (nodes) between the nid-height and the ends of the
15
beam-column giving a more precise evaluation of the

deflected shape of the column" Comparing runs made by the
approximate method (Basu 3-967) and the numerical- integration
rnethod showed that the naximum error for the approximate
method was five percent on the conservative side" The
approximate method anarysis was completed in only seven per-
cent of the computer time required for the nurnerical- inte-
gration analysis"
virdi and Dowling (1973) extended the work of Basu and
Hill (1968) by applying the numerical integration technigue
to biaxially loaded composite columns" The corumns h¡ere
pin-ended with equal end eccentricitÍes" The column v/as
assumed to be initiaÌl-y deflected arong both axis in the
form of a cosine wave, wi-th additional second-order defrec-
tions forming a part cosj-ne wave" rnitiar deflections v/ere
assumed to be less than those assumed by Basu and Hill
(1968) " rn contrast to previous studies by others, virdi
and Dowring incruded the effect, of residuar stresses in the
structural steel section" They reported that, in some
cases, residual stresses enhanced the strength of rong beam-
columns" the column l-oad capacities caLculated by their
analysis technique r¡ras compared to eight physicar test
specimens" A mean ratio of tested to calculated strength of
1.04 and a coefficient of variation of 10.4 percent resul-ted
when residuar stresses and an initiar out-of-straightness of
0. 001- t,imes the column length were assumed
L6
Virdi and Dowling (1982) present a revisj-on of their

earlier work" The revised method used Gauss guadrature to
integrate axial force in an element of the cross-section
rather than assurning the strain at the center of gravity of
an element applicable to the entire elernent" Residual
stresses are not reported to be used in the analysis" An
initial out-of-straightness of O.OO1 times the length was
assumed for each axis. The authors compared the revj_sed
technique against the physical tests of columns and the
analysis reported earlier (Virdi and Dowling L973) " The
ratio of tested to predicted strength decreased. to O"962
with a coefficient of variatj-on of 9"7 percent. These val-
ues indicate a small overestimatj-on of the ultimate strength
although the coefficient of variation is slightly improved.
Wakabayashi (t976) proposed a superposition method of
independent concrete, reinforcing bar and structural steel
columns as a solution to the cornposite column" The summa-
tion of the tangent moduLus capacity of the independent col-
umns was proposed for use in the concentric 1oading case.
I{akabayashi sugigested that arthough the steel- sectj-on may be
initially cambered, the reinforced concrete encasement can
be constructed nearly straight thereby reducing the effect
of the initial camber of the steel section to the overal_l
capacity of the composite member" This suggests that assum-
ing the entire composite cross-section to be initiaLly out
of plumb is not required. I.Iakabayashi also noted. that the
L1
stress-strain relationship of concrete inside and outsid.e

the flanges of the steel- section nay differ due to the con-
f ining inf l-uence of the f langes "
May and Johnson (1,978) reported a numerical_ technique
designed to analyze biaxialJ-y loaded composite columns for
both pin-ended and restrained end cases" The method is sim-
il-ar to that of virdi and Dowling (L973) except that May and
Johnson used a finite difference technigue while virdi and
Dowling used a Newton-Raphson iteration scheme to converge
to the equilibriun deflected shape" A comparison of their
axial load - midheight deflection computations with those of
Basu and Hill- (l-968) and virdi and Dowring (T973) indicate
very minimal- differences. Àssumptions used by May and John-
son are consistent with those previously mentioned for Basu
and Hill (1-968) and for virdi and Dowring (L973). Residual-
stresses were assumed to have negligibre effect. No mention
is made of what was assumed for initial out-of-straightness
of the beam-column.
LaChance and Hays (1980) studied the errors to be
expected by making or neglecting various assumptions in the
calculation of E}:e M-0-P (moment - curvature - axial road)
reLationship for composite beam-corumn cross-sections. A
strain compatibirity technique hras used with assumptions
similar to those used by previous authors noted above.
since onry the cross-section Ì¡ras considered, the resul-ts are
applicabre only to short beam-corumns. rgnoring the bending
I8
stiffness of reinforcing bars r¡¡as found. to underestimate the

moment capacity by only 0.05 percent" Ignoring the tensil-e
strength of concrete caused an underestirnation of moment
capacity of only O"01 percent" Neglecting to disregard the
concrete area displaced by reinforcing bars and the steel
I-section caused overestimations of the bending moment
capacity by up Eo 25 percent. using different stress-strain
curves for concrete caused differences ín the calcurated
moment capacities of up to g percent when bending about the
rnajor axis was considered. A stress-strain curve with a
descendj-ng branch beyond the point of maximum stress yielded
higher moment capacities than a curve which terrninated at
the peak stress " Residual- stresses in the steel- sectj-on did
not infruence the ultimate strength of the cross section.
However, the authors pointed out that residual stresses may
influence the strength and behavior of rong beam-corumns.
2.2 ÃSSUMPTTONS USED IN THEORETICÃTJ STRENGTH MODEL
The theoretical strength calculations presented in this
study are similar to the v¡ork of Basu (L967). A straj_n
conpatibility solution was used to compute the M-Q-p rel-a-
tionship of the cross-section and is discussed in section
2.4. The capacity of the member (beam-column) \^ras
calculated by sorving for the maximum eccentricity for which
equilibrium could be maintained beÈween the ends and mid-
height of the beam-column. The procedure used to calculate
the beam-column strength is discussed in Section Z"S"
19
The assumptions regardj-ng the loading and the end condi-

ti-ons of the beam-columns r^rere discussed in Section 1" 1"
The assumptions peculiar to the theoretical anal-ysis are
discussed here" These are:
(a) strains are compatible between concrete and steel such
that no slip occurs,"
(b) strain is linearly proportional to the distance from the
neutral axis;
(c) deflections are smaLl such that curvatures can be cal-cu-
lated as the second deri-vative of the deflection,.
(d) shear stresses are small and their effect. on strength
can be neglected;
(e) effects of axial shortening are negligible;
(f) the ro1led steel section is assurned to be made up of
rectangular plates;
(g) residual stresses in the rolLed steel section exist;
(h) Lhe column is perfectly straight before loading,-
(i) the column cross-section is symmetric about the major
and minor axis; and
( j ) the fail-ure takes place due to material- fair-ure and not
by local or torsi_onal buckling;

Assumptions (a) and (b) r,/üere reguired for the strain
compatibility solution of the cross-section M-þ-p rel_a-
tionship. Assumption (c) was needed for the cal-culation of
length effects due to secondary moments. Assumptions (d)
and (e) h/ere used to sirnplify the calculations. Assumption
20
(f) sinplified the discretization of the cross-section into

elernents and is discussed in Section 2"3" Assumpti-on (g)
acknowledges the existence of resj-dual_ stresses in the
rolled steel section and is discussed in Section 2"8"
Assumption (h) was based on !{akabayashi¡s (L976) observation
that the encasement of the steel section in concrete will-
negate any detrimental effects of initial_ camber of the
steel section. Assumption (i) sinprified the cross-section
M-þ -P caLculations since discretization of onJ_y one-
guarter of the cross-section was reguired to model the
entire cross-section. Àssumption (j) was val_id. since a
review of test data in the literature did not indicate any
failure by local or torsional buckling. This assumpti-on was
also rnade by Bondale (1966 a,brc) and would seem to be par-
ticularry varid where rectangular hoops arong with surround-
ing concrete stiffen the coinpression flange of the steel_
section" Further assumptions directly rel_ated to the
stress-strain curve for individual materiars are discussed
in Sections 2"6 and 2"7 "
2.3 CROSS.SECTTON DISCRETTZåTION
The cross-section of a composite column consists of
three materiars (concrete, structural steel and reinforcing
steeL) each possessing a unique stress-strain rerationship.
concrete was subdivided into three distinct types, uncon-
fined, partially confined and highly confined as described
below. Each of these concrete types had d.ifferent
2I
stress-strain characteristics. The rolled. steel section was

separated into the web and the flanges to moder differences
in their stress-strain characterj-stics. Hence, the cross-
sectj-on was comprised of materials with six different
stress-strain curves" The six distinct areas of the
cross-section are shown in Figure 2"2"
the cover concrete (i"e. concrete outside of the rateral_
ties) was unconfined. The concrete inside the periphery of
the ties but outside the flanges of the steel- section was
assumed to be partially confined" The concrete within an
assumed parabola and between the franges of the steel- sec-
tion (Fig. 2.2) was assumed to be highly confined. The
assumed parabola had a vertex intersecting the edge of the
web at the mid-depth of the steel section when the flange
overhang was less than one-quarter of the sectj-on depth
between the franges. The vertex of the parabola was taken
to be at the mid-height of the section and a distance from
the web d,"rtn* dependant on the flange width b, flange thick-
ness f, depth of steer section d, and web thickness ¿ll as
shown in Figure 2.3 and in Equation 2.1-.
b-w d-2t
d unrtn* (2.r)
2
d uertex
22
Steel
Flange
Steel-
Web
qì']c
oo High-ty
¡-ã
'C tr-r Conf i¡ed
Concrete
58 (j¡rside
assuned
parabola)
Partially
Conf ined
Concrete
Unconfined
Concrete
Vertical Rej¡rforcjng Bar
Lateral Tie
Figr:re 2.2 - Material- Types in Conposite

Cross-Section
23
Element 20 Elements
Thickness Along 3Iange
Varies *
Ididrh
a
Ð
L<
0)
Ê
c)
Fl
14
Ð
tl-{
o A.¡
¡{
fr .É.1
o
e
a
á.a
OuÈside
)
ot
Lateral
Ties
Element .thickness varies to ensure that

element boundary coj-ncides with
material boundary.
Figure 2"3 - Discretj-zation of Composite One-Half

Cross-Section Used for Theoretical
Strength Subroutine
24
The dj-stance, parallel to the major axis, from the edge of

the web to the parabola æn" (Figure 2"3) for an elemental
slice is computed by Eguation 2.2
íb -
t-
w
Iz - d unrrnr) d'0"
ü h" d rnrtn* + ld - 2t\2
t/ /l
lr)
in which do" is measured perpendicular to the rnajor axis

from the plastic centroid of the cross-section to the cen-
troid of the element.
The discretization between the three areas of concrete
recognizes the beneficiar effects of increased confinement
on concrete strength and ductiJ_ity. park, priestl_ey and
GiIl (1982) used distj-nct boundaries between the unconfined
and partially confined concrete areas in their analysis of
reinforced concrete beam-columns. potential di-fferences in
stress-strain characteristics between the partially and
highly confined concrete areas r,¡ere noted by Vüakayabashi
(L976) and Mirza (1989). This distinction is logi-car due to
the high confining effect of the steel section as opposed to
that provid.ed by rectangular ties. The rest of the ritera-
ture described in section 2 " l- made no mention of any attempt
to distinguish between these concrete areas" The individual-
stress-strain characterisiics of these three areas of con-
crete are discussed in Section 2"6.
25
The steel section was subdivided into tr¡¡o areas, the web
and the flanges. This accounted for differences in yierd.
strengths of the two components of the steer section as
noted by Galambos and Ravindra (L979) and Kennedy and Gad
Aly (1e80) "
ïn order to calculate the M - 0 - P relatj_onship the
cross-section must be divided into elements smalr enough to
allow the computer to numericarly integrate the forces in
each erement accurately. To accomprish this the program
dj-scretizes the cross-section into finite strips parallel to
the major axis" IÎithin each strip the cross section is
further discretized into the various material categories as
discussed above. The width of the strip perpendicular to
the najor axis is determined by the number of strips
reguested and input to the program" The width of strips are
automatically adjusted so that strip boundaries occur at the
interface between two materials. Fifty el-emental strips for
the entire cross-section thickness hrere specified for the
computer simulations described in Chapter 5"
To account for varying stresses along the width of the
frange due to residual stresses, the flange is discretized
into 20 equal width elements perpendicular to the major
axis" The initiaL strain i-n each element due to residual-
stresses is calculated with subseguent strains beingr added
26
algebraically to each element. Figure 2"3 shows the discre-

tizati-on for a typical L/Z-secEion of a composite cross-
section.
2.4 CROSS-SECTIO¡Û STRENGTS MODETJ
The cross-section strength model deterrnines the rela-

tionship between bending moment, curvature and axial load
(M-þ -P). This information is required for establishing
the cross-section axial load - bending moment (P - M) inter-
action diagram" Figure 2"4 shows typical M-þ -P rel-ation-
ships for several axial loads with key points marked on the
diagram. In this figure, the bending moment is shown on the
vertical axis with curvature shown on the horizontal axi-s.
The peak moment on the curve for the level_ of axial l_oad
considered represents one point on the cross-section P - M
interaction diagram. computation of these points for a suf-
ficient number of axial load levels yields the cross-section
P- M interaction diagram (Figure Z.S) "
The first step required in determining Elr.e M-0-P rela-

tionships is defining the rang:e of axial load to be
examined. The maximum axial load which can be applied on
the cross-section occurs when that load is applied concen,
tric to the plastic centroid of the cross section (pure com-
pression axial load capacity). This loading arrangement
forces all elements to strain equally" Since the
27
P1 (Pz(Ps
P= oxiol lood
= bending moment
M.
ø = curyoture
For Axiol Lood P' l
yielding initioted in tension flonge

spollinq of concretl beqins
"ou"l.
concrete cover spollerl off
stroin-hordening initíoted in tension flonge
rupture of tension flonge
moximurn bending moment

(stroin-hordening neglected)
moximum bendíng moment
(stroin-hordening considered)
Figure 2.4 - Typicol l,'l-ø-P Relotionships for

Composite Cross-Sections
2B
l'il
J" lll
F_[-ì
AxÍa1
Load
P.
ï'
very
sma1l
t_t_l
l_
lr-aÏ
þ_r_jPi
= P,e.
Il_
M.
a
Bending Moment
Figure 2.5 - Typical Composite Cross-section

P-M Interactj_on Diagram
29
cross-section is doubl-y symmetric, the plastic centroid is

.coincident with the intersection of the najor and mi-nor
axis "
The foll-owing procedure $/as used to calculate the con-

centric axial load corresponding to a given strain:
(a) Determj-ne the stress in each eLement from the stress-
strain relations corresponding to the given strain;
(b) determine the force in each element by rnultiplying the
stress by the area of the element; and
(c) sum the forces from each element to obtain the total
axial load.
An iterative technique was used to solve for the pure
compression axial load capacity" The axial load was cal-cu-
lated at a strain that corresponded to the material- with the
lowest strain at peak stress from the stress-strain
relationshíps (sections 2"6 and 2"2). For structural and.
reinforcing steel, the strain corresponding to the yield
point was used" The strain is incremented and the axial-
load level calculated until the strain corresponded to the
materiar with the highest strain at peak stress. sj-nce the
maximum axj-al- force l-ies between these two strain limits,
the maximum axiar Load calculated during these iterations
was taken as the cross-section concentric axial road capac-
ity" This established a point on the P - M interaction
curve that corresponded to the axiar road capacity at zero
bending moment.
30
To determine other points on the P - M interaction curve,

the bending moment capacities corresponding to axia] r-oad
levels between zero and the pure compression axial l_oad
capacity were calculated. The subroutine determines the
M-þ-P relationship for axial loads of O, Zt 4t 6, B, 10,
12, 16 , 20 , 24 , 2g , 32 , 36 , 40 , 44 , 4g , 52 , 56 , 60 , 64 , 7 O ,
76, 82 and 86 percent of the concentric axial r-oad capacity.
The bendinq moment calculated for each axial load level was
plotted on the P - M interaction dj_agram.
To determine an M-þ-P relationship one additional term
must also be known" This is the distance DN A between the
neutrar axis and the rnajor axis as shown in Figure 2.6. For
a given val-ue of axiar load, there are a number of corre-
sponding moments possibre depending on the location of the
neutrar axis and the value of the curvature. This is shown
in Figure 2.4" A unigue relationship exists between the
axial load P,. the bending moment M, the curvature Q and the
location of the neutral axis DN A" Fixing' a varue for
either DN A or O defines the varue of the other term since
only a unigue pair of values for DN.,4 and Q will satisfy
equilibrium of forces for the given axiar road. once the
axial load, the curvature and the location of the neutral-
axis are known, the bending moment can be calcurated. By
31
h/2 h/2
P last ic
Centroid
Point of
Axial Load
Application
T-
e.f
J J Strain
Ê Distribution
.----_,ls-
I
tral Axis
ø = €c
h
(j-oua¡
Figure 2 "6 Strain Gradient in Composite Cross-Section

32
cal-culating a number of these unigue rel-ationships for each

axiar load leveI, the naxinum bending moment for each axiar
load l-evel can be found"
To reduce computing time, the M-þ -P relationships for
all axiar load levers were calculated and anaryzed simulta-
neously. A starting curvature value was assumed. Holding
this value constant, the distance from the plast,ic centroid
to the neutral axis (DNA) r¿as varied and the correspondj-ng
axiar force was calculated at each DN A value selected. The
DNA was varied in such a v¡ay that all calcurated values of
the axial force v/ere within the reguired range. This
creates a matrix of P versus DNA values.
A linear interpolation of values from the P versus DN A
rnatrix provided an approxirnate DNA varue for each given

level of P. using the approximate DN A varue, the Extended
Newton-Raphson Technigue (Kikuchi, Mirza and MacGregor t-979)
was then used to converge to the correct DNr{ value for the
given lever of axial force" since both the starting curva-
ture and the position of the neutrar axis corresponding to
the required axial l-oad vrere now known, the bending moment
could be calcuLated easily. This procedure róras then
repeated for al_I given values of P.
Iîíth bending moments for the starting curvature known,
the curvature was incremented, a new P versus DN A curve hras
plotted and new bending moments r^rere cal_culated.
33
This procedure created the reguired M-þ-P relation-

ships" An outline of the nethod by whích tlne M-þ- P rela-
tionships l{ere calculated for the composite cross-section is
shown in Figure 2"7 " The data, when plotted, is similar to
the data plotted in Figure 2"4"
The maximum bending moment for a given Ievel of axial
force calculated by this rnethod yields one point on the
P-M interaction curve (Figure 2"5)" To ensure that the
maximurn bending moment for a given axial force was calcu-
lated, the curvature hras incremented until the concrete

cover on the compressive side of the cross-section had
spalled off" Review of M-Q curves showed that the maximum
bending moment occurred prior to spalling off of the con-
crete cover except for low axial load levels when strain-
hardening of steel was considered" !{hen the concrete cover
has spalled off at a particular axial 1oad 1evel, the
program ceases to calculate any further points on the monent
versus curvature graph for that axial load level except when
strain-hardening of steel is consj-dered as described below.
when strain hardening of the steel is considered, at low
level-s of axial load (less than 20 percent of the pure com-
pression axial force capacity), the naximum moment may not
be reached until several hundred curvature increments have
been completed. To save computing tine, the curvature
34
Set storting curvoture /
Colculote P-DNA curve
For eoch oxiol lood chosen

Proqfrso
Choose þproximote DNA

corr€sponding to P**¡,*
by lineor interpolotion
Use Extended Newton-Rophson Techniq

to íterote to exoct
Knowing P;Ø, DNA Colculote

Bending Moment. M
Next Pr*a*
M,no* found
for oll
Subroutine
Figure 2"7 - Flow Chort of Colculotion of M-Ø-p Relotionship

for Composite Cross-Seetion
35
increment was increased to 10 and then to 1oo times the

original increment for only the cases in which strain-
hardening of steel at Low axial forces was consid.ered.
At very high curvature values, it is theoreti_cally pos-
sible that the tension flange of the rolLed steer section
may fracture. The strain in the tension flange of the steel_
section is nonitored at each curvature increment. rf rup-
ture of the tension flange is imminent, Do further points
are calculated for that axial 1oad 1eve1"
when the moment versus curvature diagrams have been com-
pi-eted for arl axial road varues to be considered, the ¡raxi-
mum bending moment for each axial load level is stored..
These bending moments paired with the corresponding axial

loads form the P-M interactj-on diagram. rf the column has
an input length greater than zero, the progiram proceeds to
the slender corumn subroutine. rf a cross-sectional_ study
is all that is regui-red, ârr input length equal to zero will
make the computer return to the main program without execut-
ing the slender column subroutine.
2 " 5 STJENDER BEÃH-COIJUI{N STRENGTE MODEL
A beam-coIumn, due to its length, has }ess strength than

its cross-sectional strength. when an eccentric axial load
is applied to the beam-coIumn, the transverse deflection of
the column increases the effective eccentricity of the road
at" ali points along the length" Therefore, the rnaximum end
moment is controlled by the transverse deflection of the
36
column and the effective eccentricity at the point of maxi-

mum deflection" The purpose of the slender beam-column
strength model is to calculate the maximum end moment
corresponding to a particular axial load in order to con-
struct the slender beam-column P - M inEeraction diagram"
In this study, the maximum deflection occurs at the mid-
height of the beam-col-umn due to the assumption of equar end
moments and single curvature as shown in Figure 2.8. For a
beam-column to be stabre, the internar and external forces
acting on it must be in eguilibriun at every section along
its length" This condition yields an eguilibriurn defr-ected
shape for a given combination of axial road and end. eccen-
Èricity. Increasing the end eccentriciÈy causes greater
deflections and thus greater effective eccentricities until
failure of the material- at nid-height causes the collapse of
the column" The maximum bending moment acting at the ends
prior to collapse of the colurnn subjected to a given axial
load is the long column bending moment capacity.
Two methods t,o calculate the maximum end eccentricity
and, hence, maximum end moment have been developed by Basu
(r967) and Basu and Hill (l-969) as previousry discussed in
section 2"1 for the type of beam-column studied in this the-
sis" The numerical integration method (Basu and Hill 1968)
reguires that the column length be divided into a number of
nodes or st,ations. Eguilibrium must be satisfied at each
node" This is achieved by iterating the defrection at each
37
deflected
shape of
col-umn
I
I
ø
,m I
Figure 2.8 - Deflected Shape of Slender Composite

Beam-Col-umn under Eccentric Axia1 Load
3B
stat,ion until all- stations are in equiJ-ibrium, yield.ing an

eguilibriun deflected shape. Increasing the number of sta-
tions improves the accuracy of the solution but also
increases computing tirne required. For the approximate
method (Basu 1,967), the eguilibriurn deflected shape is
assumed to folLow the shape of a part cosine curve" The
deflection at the rnid-height of the column can then be esti-
mated guickly by simply solving a single equation. In both
methodsr ân end eccentricity is assumed, a deflected shape
is found and the cross-section is checked at the point of
maximum deflection to check whether failure has taken place.
rf faj-lure has not occurred, then the end eccentricity is
increased and the process is repeated. The largest end
eccentricíty before failure at the point of maximum defl-ec-
Èion is used to calculate the maximum end bending moment for
the beam-column at that axial- load l-ever. The approximate
method was found to carcuLate column capacities up to 5 per-
cent more conservative than the numerical integration method
(Basu and Hill L968) " Basurs approximate rnethod (1967) with
some modifications was used in this study. These modifj-ca-
tions are discussed in detail_ in this section.
As for the case of the cross-sectional study (section
2"4), the first regul-rement for creating the srender beam-
col-umn P - M interaction díagram is calcul-ation of the con-
centric axial load capacity" The tangent modulus theory was
used to cal-culate this load". The tangent modulus method has
39
been used by lrlakayabashi (1,976) for his composite design

propoåals and is also reconmended by Basu (11967) for ini-
tially straight columns.
Previous analytical studies of composite columns (Basu
1967, Basu and Hill- 1968, Virdi and Dowling L973) assumed an
initiar out-of-straightness of the composite column. Basu
and Hill (1968) showed concentric axial road reductions of 3
to 30 percent, for length to overall depth ratios of t-O to
40, respectivery, when initiar out-of-straightness v/as con-
sj-dered" Basu assumed. an initial out-of-straightness at
mid-height of the column equal to o.ooooó ¿2/d., where I is the
length of the corumn and d is the depth of the steel sec-
tion, which Ì¡ras meant to estimate both the effect of initial
out-of-straightness and the effect of residual- stresses in
the steel- section on the column strength. This method
assumes initial mid-height deflections of the steel sectj-on
greatry in excess of allowable torerances in North Àmerica.
virdi and Dowling (i-973) compared the effect of assuming
no imperfections, residual stresses only, out-of-
straightness of l/ i0oo, residual stresses plus an assumed
out of straightness of l/ lO0O, and the use of only an
assumed out-of-straj-ghtness of o.o0ooó L2/d. as assumed by
Basu" Ä,ssuming no imperfections showed the highest
strengths and the use of Basurs assumptions the lowest
strengths, especially for Ìong coLumns. The differences
between an assumption of residual stresses onry and residual
40
stresses with an initial out-of-straightness of ¿/ IOOO were

very slight" The residual stress only assumption averaged a
2"6 percent greater capacity than that when residual
stresses l,{ere combined r¿ith the initial deflection of l"/IOOO
for the 9 columns tested by Virdi and DowJ-ing (L973) over a
number of biaxial eccentricities" These results show that
assurning an initially straight colurnn wiLl not be in serious
errorr âS long as residual stresses are accounted for.
The assumption of initial out-of-straightness made by
the authors mentioned above considered the entj-re composite
cross-section to have an initial out-of-straightness.
wakayabashi (1976) recognized that whire the steel- shape may
be assumed to have initiar camber, the concrete encasement
is likely to be constructed straight,. since the concrete
encasement gives rateral support to the steel section, he
argued that the effect of the initial camber of the steel-
section would be negligible"
To exactly account for the geometry of the beam-column
as described by wakayabashi (J-97 6) | the prograrn would have
had to be able to be capable of calculating the strength of
cross-sections symmetric about the minor axis on1y, since
the cross-section g:eometry with respect to the maj or axj-s
would change at every point along the 1ength of the beam-
column. This would greatly increase the complexity and com-
puting tiine as M -0-P relations wourd have to be calculated
for a number of different secÈions along the column height.
4I
Due to the small effect reported by Virdi and DowJ-ing (1973)

of initial out-of-straightness of ¿/IOOO for the entire com-
posite cross-section and the fact that the concrete encase-
ment is not likely to be subjected to significant initial-
out-of-strai-ghtness, it was decided not to assume any
initial camber of the steel section for this study. This
assurnptj-on also allows the use of the tangent modul-us
theory, which cannot be correctly applied to columns that
are not perfectly straight"
The concentric axial capacity of a long column is depen-
dant on the buckling strength of the column and not the
material strength as is the case for the cross-section.
This means the column fails by buckling before the materj-al_
strength is exceeded. The ultimate buckling stress for a
column of homogeneous material is given by the tangent
buckling formula:
)-
fl- L,,
r_
r*- ç¡urf (2.3)
Substituting the value of L.O for the effective length

factor k and recalling that the radius of gyration, r, can
be calculated as the square root of the moment of inertia
divided by the area tltlAl, Equation 2.3 can be rewritten
AS:
42
TL2
f",A=78,1 (2.4)
The column buckJ-ing load, P",, is equal to either side of
Equation 2"4 and may be calcul-ated directly"

Since a composite col-umn is made up of materials with 6
independent stress-strain curves, Eguation 2.4 is not

directly applicable. Instead, Equation 2"4 hras applied
independently to each individuar material and the sum of the
tangent buckling strength for all material-s gave the col-umn
tangent buckling l-oad. This procedure is comparable to that
proposed by Vlakayabayshi (1976)" Hence, Equation 2.4 can
be modified to account for the six independent material-s:
i.6 i'ó
I
i- I
(f
",,A,7 = -2
fL l{r,,r,¡ (2.s)
The buckling load ,

",, which simultaneously satisfies
P
both sides of Equation 2"s is the composite column buck]-i_ng

l-oad capacity (maximum concentric axial load) " Eguation 2.5
cannot be solved directly since the tangent elastic modul_us
of an erement is a function of the stress in the element.
Therefore, an iterative solution rras used" The axial strain
of the column was adjusted unti'l the difference in buckling
load calcurated by each side of Equatj-on z"s is less than 1
43
pound (4"45 N) " This established the point on the sl_ender

col-umn P - M interaction curve corresponding to maxi-mum con-
centric load and zero bending moment.
1o establish the points of the beam column P - M interac-
tion curve due to eccentric loading, a nodified version of
the approximate method of Basu (L967) was used. For each
axiar load lever investigated less than the slender column
concentric axial load capacity for which the cross-section
M-O-P relationship !{as previously calculated (Section
2"4), a maximum end eccentricity was sought" The method for
calculating the maximum end eccentricity (and, hence, maxi-
mum end bending moment) can be described as follows:
(a) assume a mid-height deflection of the column;

(b) find the end curvature which corresponds to the desired
deflected shape;
(c) find the bending moment corresponding to the end curva-
ture from the cross-section M-Q-P relationships and
calcul-ate the end eccentricity;
(d) add the end eccentricity to the assumed rnid-height
deflection and calculate a ne!ù bending moment at the
nid-height of the column; and
(e) if the bending moment carculated in (d) is ress than the
maxj-mum bending moment from the cross-section M - O - p
rel-ationship, increase the mid-height deflection and

44
repeat the process starting from item (a) " If the bend-'
ing moment calculated in (d) is greater than the maximum
bending moment from the cross-section M-þ-P
relationship, the previous end eccentricity calcuLated
in item (d) j-s used to compute the rnaximum end bending
moment"
Basu (L967) and Basu and Hill (1968) used a part, cosine
curve for the assumed deflected shape of the composite col-
umn and showed this method was onry srightly more conserva-
tive than the numerical integration method with a maximum
difference of only 5 percent. euast (Ig7O) studied the
deflected shape of pin-ended reinforced concrete corumns
uniaxially loaded in single curvature. After comparing a
number of theoret.ical def l_ected shapes, including a part
cosine curve, ag,ainst more elaborate and time consuming
numerical integration techniques he concruded that the best
approximation of the defrected shape was a 4th order parab-
ola with the rnid-height deflection given by Equation 2.6.
t2( + Õ"\
"l
em ro (2.6)
[*- 4)
45
ridhere Q- and Q" are the curvatures at nid-height and the

column ends, respectively; I is the length of the col-umn;
and e* is the rnid-height deflection of the column as shown
in Figure 2"8"
Quast found that this deflected shape produced a maximum
unconservative error of 2 percent and a maximum conservative
error of 6 percent" The reinforced concrete columns he sim-
ulated had eccentricity ratios of between o.i- and 1"0 with
l/h raLíos of O to 60"
Mirza and MacGregor (l-999) used euastrs method of
approximating the deflected shape to stud.y the strength
variability of reinforced concrete columns. Bol_in (1985)
tested an earrier version of the present analytical prog'ram
with both approximations and found that in general both
assurnptions returned similar resurts, with Basurs assumption
being slightly more conservative. rn consideration of the
above, Quastrs approximation to the deflected. shape of the
beam-column hTas used in this study"
The total mid-height eccentricity e, is the sum of the
assumed rnid-heightdeflection e* from Eguation 2.6 and the

end eccentricity e as shown in Eguation 2"7 "
ete+em (2.7)
46
Substitution of Eguation 2.6 into 2"7 and rearranging to

sol-ve for the end eccentricity yields Equation 2"9"
(2.8)
The nid-height. eccentricity e¿ can be calculated by

dividing the nid-height bending mornent by the axj-al- load as
shown in Eguatíon 2.9.
M^
et p Q'9)
substitution of Equation 2.9 into z.g gives the sirnpre

relationship between the end eccent,ricity (e), nid.-height
moment (M*) , the mid-height curvature (0-) and the end cur-
vature (Q.) shown in Eguation 2"1_0.
e : (+) (#) (*- . ?) (2 .t o)
The program uses Eguation Z"IO and the cross-sectj_on

M-þ-P rel,ations previously calculated to solve for a
cornbination of end eccentricity, mid-height deflection and
nid-height curvature in eguilibriun" Figure 2.g outrines
47
For oxiol lood, P
Choose triol mid-height curvoture
Retrieve Corresponding Mid-height Eending Moment

from Cross-Section M-ø-P Relotíonship
Colculote totol eccentricíty
Choose tríol end curvoture
Retrieve Corresponding End Bending Moment

from Cross-Section M-ø-P Relotionship
o
L
:t
Colculote end eccentricity o
¿
f,
E a)
f
Colculote Mid-height Deflection (o) o
¿ .P
( : Totol - End Eccentricity) J() 0)
.tr'
I
rt !
L
ö E
Colculste Mid-height Deflection (b) P-
L
+É
by Quost (Equoiion 2.6) 0) 0)
E E
a, 0)
L t-
u o
Ê C
Use Bi-Section Technique ond Logrongion lnterpolotion

Reduce end curvoture untÍl (o) = (U)
Store colculoted end moment
Moximum end moment found ?
Choose next oxiol lood level '
- 2,9
Fiqure - Flowchort for ColculotÍng Slender Column
M-ø'P RelotionshiP
4B
the procedure" Values for the nid-heíght curvature are

incremented .from a mj-nimum value until a maximum end bending
moment is calculated" For each mid-height curvature value
assumed, values of the end curvature are test,ed and increm-
ented from a minimum until an eguilibrium combination is
found. The largest curvature that can be attained at
mid-height is the one that corresponds t,o the maximum moment
from the M-$-P diagrarn for the axial Ioad" Once all pos-
sible rnid-height curvatures have been investigated, the
largest end bending moment calculated becomes one point on
the slender beam-column P-M interaction curve. The pro-
cess is then repeated to complete the entire slender beam-
column P-M interaction curve.
2"6 STRESS.STRåTN CURVES FOR CONCRETE
As outlined in Section 2.3, three distinct concrete
areas have been assumed in the discretization of the compos-
ite column cross-section" These distinctions are meant to
account for dissinilarities in the stress-strain
relationship of the concrete due to confining action of the
vertical reinforcing bars, the rectangular LateraL ties and.
the roLled steel section" Confinement of concrete increases
the compressive strength and ductility of concrete in rein-
forced concrete corumns and methods to compute its effect on
the stress-strain rerationship have been deveroped by park,
Priestllr and- Gil-] (L982) , Sheikh and Uzumeri (tge?) | Sheikh
and Yeh (1986), Mander et, al. (L988) " confinement effects
49
on the compressive strength of concrete in composite columns

were not considered by any of the previous studies on com-
posite columns reviewed in Section 2"1" Cornpressive stress-
strain curve characteristics for differenÈ degrees of
concrete confinement are given in Sections 2"6"Lt 2"6"2, and
2"6"3" the effect of confinement on the tensile stress-
strain relatj-onship of concrete is not avaiLable in the lit-
erature searched. It was, therefore, decided to assume
identical tensile stress-strain relations for al1 types of
concrete confinements. A discussion of the tensile strength
of concrete is given in Section 2.6"4" Finally, all stress-
strain curves used for concrete are sunmarized in section
2"6.5" It should be noted that the stress-strain relations
presented in this chapter are based on statj-c loadj-ng condi-
tions.
2"6.1 Unconfined Concrete
The two curves considered of interest for describing the
stress-strain characteristics of the compressive strength of
unconfined concrete are those of Hognestad (1951) and Kent
and Park (l-97L).
The Hognestad curve (Figure 2"L0) consists of a second.
order parabola from the origin to the peak stress" The
sÈrain at peak stress is a function of the initial tangent
modul-us and the concrete strength. Beyond the peak stress
the curve descends linearly to a stress of 85 percent of the
peak st,ress at a strain of o"oo38" The equations describing
50
ff
I o"is r¿
Figure 2"I0 Hognestad (1951) Stress-Strain

Relationship for Unconfined Concrete
frc
0"5f1c
0"2f'c
0.002 C
-5ou
Figure 2 "II - Kent and Park (1971) Stress-strain
Relationship for Unconfj_ned Concrete
51
the portion of the curve between the origin and the peak
stress and between the peak sLress and the stress corre-
sponding to the ulti-mate strain are given in Eguations 2"I7
and 2.1-2, respectively.
f" = f'" (2.Ì r)
€c €o
f" s,r'. (2.r2)
=
t 0.0038 € 6 ]' r
2f''
where €o =
E"
The Kent and Park curve (Figure 2.II) is simj-lar to the

Hognestad curve except that the strain at peak stress is set
at a value of 0"002" The descending branch is tinear and
assumed to faII from the peak stress to a value of ZO per-
cent of the peak stress. The slope of the descending branch
j-s a function of the concrete strength with lower strength
concrete modelled to be less brittle" The equations
describing the portion of the curve between the origin and
the peak stress and between the peak stress and the stress
corresponding to the ulti¡nate strain are given in Equations
2.1-3 and 2 . 14, respectively"
52
r" = r',lsa (o*u)'l (2.rs)
f " = Í'" il Z (e. 0.002)l (2.14)

0.5
where
€sou O.OOz
3 + O.OO2f'"
and E
L 50u
f'" I 000
For SI conversion replace 3 by O"O207 MPa and 1OOO by 6.895

MPa "
Llewellyn (L986) compared results of using the stress-

strain curves by Hognestad and Kent and Park. Essentially,
no difference r¡/as found in the results obtained from the two
stress-strain curves" However, the Hognestad curve occa-
sionally produced higher strength for unconfined concrete
than that obtained for confined concrete based on modified
Kent and Park confined concrete stress-strain curve (Section
2"6"2) " The Kent and Park curve for unconfined concrete
presented no such confLict for obvious reasons.
The Kent and Park curve for unconfined concrete was used
for the stress-strain curve adopted in this study" Two mod-
ifications lrere made" First, the strain at peak stress was
53
aLlowed to vary as a functj-on of the concrete strength,

similar to the method of the Hoginestad curve as shown in
Equation 2"l-5"
2f '"
'o
C:
(2. r s)
E
"
The second rnodification was to assume that the concrete

loses all strength at a strain of o"oo4. This assumption
has previously been used by park, priestry and cirl (1982)
to model the spalling off of the concrete cover. The
stress-strain curve used in the theoretical anarysis of this
study for the unconfined compressive strength of concrete is
shown in Figure 2.L2.
2"6"2 Partial]-v Confined conerete
confinement enhances the strength and ductility of con-
crete. Past efforts to model the behavior of the confined.
concrete core of reinforced concrete columns reinforced with
rectangurar hoops are summarized by sheikh (i,ggz) and Mander
(L983). rn a reinforced concrete column, the concrete core
strains both axially and. transversery under load.ing. As
loading progresses, the transverse strains cause hoop ten-
sion in the horizontal ties. The horizontal ties thus con-
tribute passive confinement to the core. The vertical
54
0 "2f :x +--
I
rl
'I
€'o= 2fL ê* = 0"004 **

E
c
minimum concrete stress

maximum compressive concrete strain
Figure 2.LZ - Unconfined Concrete Compressive

St,ress-Strain Rel-at ionship
used j-n Theoretical Strengbh .'lubroutine
55
reinforcing similarly confines the core since the vertical-

bars are prevented from bowing out by the horizontal ties
(Park and Paulay L975) "
Confinement of the core concrete in a composite steet-

concrete corumn is expected to occur in a similar fashj-on"
In addition, the component plates of the structural steel-
shape inside the core provide additional confinement.
$Iakayabashi (L97 6) recognized that stress-strain rel-ations
of concrete may differ depending on its rocation in the com-
posite section. He did not, however, try to incorporate
these dj-fferences into his proposed design procedure.
To account for enhancement of strength and ductility of
the concrete core, tvro stress-strain relations hrere consj-d-
ered. The Modified Kent and park curve (park, priestry and
Gi]I 1982) and the sheikh - uzumeri curve (LgBz) were both
developed for the confined cores of reinforced concrete col--
umns. since no model directly applicabre to composite col-
umns vras found in the Ij-terature searched, these two curves
v/ere investigated for their compatibility and accuracy to
the composite corumn. A third method proposed by Mander et
al-. (1988) was briefly reviewed but not considered for adop-
tion since a significant amount of work for this thesis had
been completed by the time the Mander paper was pubrished.
The Modified Kent and park curve is a modification of an
earl ier Kent and Park curve for concrete confined by rectan-
gular hoops (Kent Park L97J-) " The original version of the
56
curve allowed an increase in the ductility of the confined

concrete but not an increase in strength. Later tests on
column specimens by Park, Priestley and ciII (L982) quanti-
fied the increase in concrete strength which was included
into the stress-strain relationship" The degree of
confinement is a function of the verticaL spacing of hori-
zontal ties, the ratio of vol-ume of horizontal ties to vol--
ume of concrete core and the yield strength of the
horizontaL ties. fncreasing the confinement increases both
the concrete strength and the ductility. the Modified Kent
and Park stress-strain curve for concrete confined by rect-
angular hoops is shown in Figure 2.13" Eguation 2"16
describes the portion of the curve between the origin and
the peak stress. Eguation 2"L7 describes the descending
branch of the curve.
{- Jc|
K{' f 2r," (t_, e. \rl ,
(2.t 6)
L0.002K \o.oozx )l
where K=l + 9 rf yn
T'
Jc
5l
Kf'c
0.2Kf U'
vê
€o= 0"002K
Figure 2.L3 - Modified Kent and Park Stress-Strain
Rectangular Ties
(Park, Priestley and ci11 L982)
0.3KSCf'
Figure 2.L4 Sheikh-Uzumeri (1982) Stress-Strain

Rectangular Ties
58
1-
Jc Kf', I Z(e, O.OOZK)] o.2Kf" (2.17)
0.5
where
€so, + €so¿ o.oo2K
and E
3 + 0.OOZf'"
'50¿
f'" I 000
J r-,
ln
an d € so¡,
4O"V Ç
In the above Eguations f', is the concrete cylinder

strength, p" is the ratio of vol-ume of lateral ties to vor-
ume of concrete contained within the laterar ties , f vn is
the yield stress of the laterar hoops, h" j-s the out to out
width of the lateral hoops and s¡. is the spacJ_ng of the
l-ateral hoops. For sr conversion replace 3 and l-ooo by
0"0207 MPa and 6"895 Mpa respectively, for computing €sou.
Further tests by scott, park and priestley (Lgg2) showed
that at high strain rates the enhancement of the core
strength by the confinement was further increased. They
conservatively chose a factor of L.zs to account for the
gain in strength in a confined core due to high strain
rates" The tests also demonstrated an increase in the srope
of the descending branch of the sLress-strain curve. An
increase of 1,"25 in the srope of the descend.inq branch was
attributed to the high strain rate" This may resurt in a
sJ-ighÈ reduction in total ductility. The study described in
this thesis involves short term loading, but the rate of
59
loading is assumed to be slow enough that dynamic effects

are negligible (i"e" guasi-st,atic) " This will l-ead to con-
servative descriptíon of strength if the results of this
study are used for dynamic loads. The tests by Scott et aI"
(1982) also indicated that presence of a strain gradient
increased the ductility of the confined core, but the effect
was not quantified.
The Sheikh - Uzumeri curve (L982) was developed from
their earlier experimental tests of reinforced concrete col-
umns (l-980) " The development of their analytical model rec-
ognized the importance of tie spacing, volurnetric ratio of
tie steel to concrete core, and tie yieJ-d strength to the
degree of confinement of the core. They also found that the
configuration of the vertical bars in the cross-section and
the way they vrere tied to the horizontal hoops was al_so a
factor. The Sheikh - Uzumeri (l_982) curve describing the
stress-strain relation of concrete confined by rectangurar
hoops is shown in Figure 2.t4. Eguation 2.LB describes the
portion of the curve between the origin and the peak stress
and Eguation 2"1-9 describes the descending branch.
ft-zr" /." \tl

t-tl (2. r 8)
L e oo [.oo/ -]
where K" = 1.0 +

2.73 82 ,C' _ \(¡t .s r\-r-.l
|
r,
P o"" [(' 5.5 8'l \ 28)
-tt )
60
f K (2.re)
" ".f "
0.5
where Z
ip.GZ
The Sheikh and Uzumeri (I9BZ) curve partly accounts for

the increase in ductility by assurning a horizontal plateau
in the curve from the strain at peak stress to the start of
the descending branch as shown in Figure 2"L4" The mj-nimum
strai-n corresponding to the maximum stress €,, and the maxi-
mum strain correspondíng to the maxj-mum strêss €r2 are given
by Eguations 2"2O and 2"2i,, respectively"
€sr 0.55,("-f '"x i0-u 2.20
a at
¿-,¿- r
fn Equations 2.1-8 to 2"21" €00 = strain in plain concrete at

maximum compressive stress; B : core concrete width measured
to the center line of the lateral tie; C: the clear dis-
tance between raterally supported longitudinal reinforcing
bars î P o.": the area of core concrete nultiplied the
compressive strength of prain concrete i n = number of rater-
aIly supported longitudinar reinforcing bars; s = spacing of
61
Lateral tiesi p" : ratio of volume of lateral ties to volume

of core concrete and, f', = stress in lateral ties. Linear
dimensions are in inches and stresses are in ksi.
Sheikh and Yeh (L986) confirned an earlier finding by
Park, Priestley and GiI1 (L982) that ductility of the con-
fined concrete core increased when a strain gradient (i.e. a
bending moment) v¡as present. The strain gradient did not
increase the maximum strength of the core concrete. For
simplicity, the effect of strain gradient on the stress-
strain relationship of partially confined concrete was
neglected "
In order to decide which of the stress-strain curves
best suited the theoretical analysis of composite columns
reported in this study, three criteria r¡/ere examined. These
are:
l-. Suzuki et. aI" (1983) provided the results from a series
of tests conducted on composite column specimens similar to
those chosen for this thesis. The ratio of structural- steel
section to gross column section was 3 percent for the LH
series and 6 percent for the RH series. The volumetric
ratio of lateral- ties to concrete core varied from 0.6 to 3
percent. All specimens were concentrically loaded. By
apprying the stress-strain curve assumptions for the steel-
portion of the cross-section, âs described in this thesis,
the stress-strain curves for the concrete portions were com-
put,ed" The method of determining the column strength
62
attributable to the core and to that of the cover was based

on the work of Moehle and Cavanagh (i-985) " The core
strength curves were compared to predictions made by both
the Modified Kent and Park and the sheikh - uzumeri curves"
The Modified Kent, and Park curve better predi-cted the peak
strength of the RH series while the Sheikh and Uzumeri curve
v/as found to better predict the strength of the l,II series.
Strain at peak stress was estimat,ed better by the Sheikh
uzumeri curve. Both curves estimated greater ductility than
the data indicated in most instances, although the shape of
the descending branch was basically accurate. Since the
results $¡ere inconclusive, further investigation was done to
estabrish which of the stress-strain curves would be better
suited for composite columns.
2- calibration of the computer model against physical test
results is described in section 2.9. Before final- cal-ibra-
tion, the two confined concrete stress-strain curves \¡¡ere
compared with some of the experimental data to investigate
which one gave the nore accurate predictions. OnIy the
specimens with length to overall depth ratio less than 6.6
v/ere used for this purpose" Both stress-strain curves pro-
duced about the same results. The results are shown in
lable 2"L"
3" fn Section 2"6"1, it was stated ÈhaÈ the Kent and park
moder (L97L) was used to describe the stress-strain charac-
teristics of the unconfined concrete. since the Modified
63
Table 2.I - Comparison of Strengt.h Ratiost

calculat.ed using Modified Kent. and Park and sheikh and uzumeri
Stress-St.rain Relations for Concrete Confined by
Rect.angular Lateral Ties
SLress-sËrain Mean Value CoefficienË of

Curve Used VariaÈion
(1) (2) (3)
Modified Kent t.026 0 " 0763

and Park
Sheikh-Uzumeri t.o29 o "o7 67

(re82)
*Based on all columns with f,/h -< 6.6 in Table 2"3

** (Park et al " L9'82)
64
Kent and Park curve uses the Kent and Park (1,971,) rnodel as a
starting point, there is apparent.ly no possibility of an

overlap between the two models" the Sheikh - Uzumeri
(1-982) model for partially confined concrete was compared to
the Kent and Park (1-97L) rnodel for unconfined concrete for
some of the test specimens discussed in Section 2"9" It was
found that in some cases, the Sheikh - Uzumeri model for
partially confined concrete predicted a lower initiaL tan-
gent elastic moduLus (Figure 2"1.5) than the Kent and park
rnodel for unconfined concrete" It was also found that, in
some cases, the slope of the descending branch of the sheikh
- Uzumeri curve for partially confined concrete was so steep
that its strength was less than that of the unconfined con-
crete, with a strain at peak stress signíficantly less than
Èhat of the unconfined concrete" This behavior has been
neither documented nor expected"
In order to naintain compatibility between the stress-
strain curves of the unconfined and confined portions of the
concrete, it was decided to use the Modified Kent and park
curve (Park, Priestley and cill r9B2) to rnodel the partially
confined portion of the concrete in the composite cross-
section. As discussed in Section 2.6"i_, the strain at peak
stress was al-lowed to vary as a function of the initial
elastic modulus of the concrete (Eguat,ion 2.15). The
65
ê- PartiaJ.ly Confj¡ed Ccncrete

(Shejkh and Uzinrerj- 1982)
I
.rl Ja tl
(fì Unconfi¡red Concrete
g tl (Kent and Park 1971)
a
tl
(fl
0)
I
g
lJ ') I'
(D1
0"0 0.002 0" 004 0" 006 0.008 0.010 0.0I2 0.014
Strain
Fign:re 2"I5 - Conparison of Unconfined and partially Conflned

Concrete Conpress ive Stress-Stra j¡r Rel-at ionships
for Colunn III-100 -B (TabIê 2 . 3 )
66
stress-strain curve used in the theoretical analysis for the

compressive str.ength of partially confined concrete is shown
in Figure 2"t6.
2"6"3 Heavílv Confined Concret,e
In the previous discussion of partially confined con-
crete it was noted that an increase in the amount of con-
finement. provided to the core is accompanied by an increase
in concrete strength and ductility" Both the Sheikh
uzumeri and the Modified Kent and park models increase both
concrete strength and ductility when confinement is
j-ncreased. rn this study, a portion of the concrete between
the flanges of the structural steel section has been assumed.
to be heavily confined as indicated in Figure 2.2. This
area is confined by the rolled steer section on three sid.es
and by the partialJ-y confined concrete and lateral tj-es on
the fourth side" It is reasonable to assume that the con-
crete in this area is under a higher degree of confinement
than the concrete out,side the influence of the flanges. To
account for this higher confinement, the concrete in this
" area has been assumed to folrow the same stress-strain rel-a-
tionship as the partially confined concrete (Modified Kent

and Park model in Figure 2"16), but does not have a
descending branch" The concrete stress is assumed to remain
at the peak stress throughout all strains past the strain at
which the peak stress was first attained. since the peak
stress is predicted from a model- for concrete confined by
67
Kftc
0"2Kf'c
€" = 2Kflc
Eigure 2"16 - Partially Confined Concrete

Compress ive Stress-Strain
Relationship used in
Theoretj-caI Strength Subroutine
6B
rectangurar hoops, the prediction for the heavily confined

concrete peak stress is expected to be on the conservative
side" The assumed stress-strain curve for heavily confined
concrete is shown in Figure 2"1-7.
2"6"4 Tensile St,ress-St,rain ReLaËionsbip Of Concrete
Park and Paulay (L975) state that the tensj_le stress-
strain rerationship of concrete may be assumed to be linear
up to the tensiLe strength (i.e. modulus of rupture) with a
modulus of elasticity egual to the modulus of elasticity in
compression. The tensile stress beyond the peak stress is
assumed to be zero. The assumption of equal initiat modul_us
of elasticity for tension and compression was al-so suggested
by Mirza et a1. (1979c). This stress-strain relationship
was used by Mirza and MacGregior (1989) for reinforced con-
crete columns"
Recent work on this subject has focused on the shape of
the stress-strain curve after the peak tensj-re stress is
reached" Hwang and Rizkalla (1983), carreira and chu (1986)
and zhen-hai and xiu-qin (t997) all report a descending
branch to the tensile stress-strain curve after the peak
stress. rn all cases the stress drops sharply at strains
beyond the strain at peak stress, retaining only a small-
percentage of the peak strength at larger strains.
Lachance and Hays (1980) assumed that the relationship
between concrete tensile stress and. strain was a continu-
ation of a polynomial curve describing the entire stress-
69
Kftc
¿ = zKflc
Figure 2 "L7 lleavily Conf ined Concrete

Compressive Stres s-Strain
Relationship used in
Theoretj-cal- Strength Subroutine
70
strain relationship both in tension and compression" The

peak tensile stress vras assumed to be 1O percent of the
specified concrete cylinder strength and the ultimate ten-
sile strain r¿as set at O"OOOI-25 for 5000 psi (34"5 Mpa)
concrete" The peak stress and the ultimate strain hrere
assumed to occur at the same point" They found that this
assumpt,ion added only 0.01 percent to the strength of a com-
posite column cross-section as opposed to ignoring it aL1
together"
For sirnplicity, it was assumed in this study that the
tensj-re portion of the concrete stress-strain curve is as
suggested by Park and Pauray (L975) and by Mirza and MacGre-
gor (1989). The work of LaChance and Hays (1980) suggests
that the contribution of concrete tensile strength to the
overall column strength is so small that it could be consid-
ered negligible and, therefore, a simpre model was consid-
ered to be sufficient" The assumed stress-strain curve for
the tensi-Le strength of concrete is shown in Fi-gure 2"r8"
2"6'5 Sunuafv Of St,ress-Straín Relationshins F,or eonclrete
rn this study, the stress-strain relationship for uncon-
fined concrete ra¡as based on the Kent and park (].g7r) curve
with slight rnodifications, the major difference being that
the strain at peak stress was assumed to be a function of
the modulus of elasticity and the strength of the concrete
(Eq-uation 2"r5) instead of being a constant value. The por-
tion of the composit,e section considered unconfined is the
7L
r
ti
c
Figure 2 "18 Concrete Tensi-le Stress-strain

Re1atj-onship used in
Theoretical Strength Subroutine
72
concrete outside the perimeter of the lateral ties" The

Modified Kent and Park curve (Park, Priestley and cill L9g2)
was used to model the partially confined concrete. Horarever,
the strain at the peak stress v¡as calculated as not,ed above.
Partially confined concrete is considered to be within the
lateral ties but outside the confining influence of the
steel section flanges" The Modified Kent and Park curve hras
also used to model the heavily confined concrete between the
flanges except that the peak stress hras assurned to be main-
tained at all strains beyond the strain at which the peak
stress was first attained. The concrete areas of the
composite cross-section assumed to be unconfined, partially
confined and heavily confined are shown in Figure 2.2. The
assumed stress-strain curves for unconfined, partiatly con-
fined and heavily confined concrete are shown in Figures
2.L2, 2"L6 and 2"L7, respectively"
The stress-strain curve of the tensile strength of con-
crete is shown in Figure z"fg. The stress-strain relation-
ship was assumed to be linear from the origin to the modulus
of rupture. The modulus of elasticity for tension is
assumed to be egual to the compressive modulus of erastic-
ítv "
2.7 STRESS-STRÃIN CURVES FOR STEETJ
Tvro types of steel are used in the composite cross-
section. These are Èhe roLled structural steel shape and
the reinforcing bars" The assumptions reg:arding the shape
73
of the stress-strain curve are similar for both materials"

The differences occur because different variables are used
to define the stress-strain curve for each type of steel-.
It was assumed that the stress-strain curves for al-l_ steel_
components were identical in tension and compression" This
assumption is consistent, r,¡ith previous works reviewed in
Section 2"L" The variables for which the data were avail--
able to define the stress-strain curves for the rolled steer
section and for the reinforcing steel are given in Section
2"7 "I and 2"7 "2, respectively.
2"7"L Structural Steel
The stress-strain curve used for structuraL steel was
assumed to be bilinear to the onset of strain-hardening.
From the origin to the attainment of the yield stress,
stress was assumed proportional to strain according to Hook-
e¡s Iaw" Between the attainment of the yield stress and the
onset of straj-n hardening the stress lras assumed to be
constant at the yield stress IeveI. The strain-hardening
portion of the curve is assumed to be a second order parab-
o1a. The slope of the strain-hardening portion of the
stress-strai-n curve at the ultinate strain was assumed to be
equal to zero. The variables used to define the entire
stress-strain curve are the elastic modulus Ã", the yield
stress ;f the strain at onset of strain hardening €"",rn, the
"",
initíal tangent slope of the strain hardening curve .E and
"",,^
the ultimate stress /,". The yierd strain €"" and ultimate
74
strain €us are calculated by the program. The schematic

stress-strain curve for structural steel ís shown in Figure
2"L9"
The curve used is similar to what has been assumed in
previous studies. Basu (L967 ) and !{akayabashi (i,976)
assumed a slight curvature at the transition from the elas-
tic to the plastic condition" Only LaChance and Hays (l-9BO)
and Virdi and Dowling (L982) considered the strain-hardening
portion of the curve. LaChance and Hays (1980) used a non-
l-inear curve similar to a parabola" Virdi and Dowling
(l-982) assumed a linear strain-hardening portion to the
stress-strain curve.
"2 Reinforcincr Steet
2 "7
The shape of the stress-strain curve used for reinfor-

cing steel is virtually identical to that used for struc-
turar steel. The difference is in the variables which must
be specified to establish the stress-strain curve. The
modurus of elasÈicity ã., the yield stress /",, the strain at
onset of strain hardeningt €.",.n, the ultinate stress /,, and
the ultimate strain €u,. must atl be specified" The slope of
the initial tangent to the strain hardening curve .F.,,,. is
calculated by the program, as is the yiej_d strain e",. The
schematic stress-strain curve for rej-nforcing bars is shown
in Figure 2 "20 "
75
SSTTN
1""
é sstrn *us
Figure 2"19 Structural Steel Stress-Strain

Relationship in Tension or Compression
used in Theoretical Strength Subroutine
76
rstrn
rstrn
Fig,ure 2 "20 Reinforcing Stee1 Stress-Strain

Relationship in Tension or Compression
used in Theoretical Strength Subroutj-ne
77
The differences in the specified variables required for

structural steel and reinforcing steel is due solely to the
information available in the literature for these steel-s.
Hardly any usefuJ- data was found on the ultimate strain of
structural steel-. However, some data was available for
reinforcing steel ultinate strain (A11an 1-972). Galambos
and Ravindra (1978) published data on the initial strain-
hardening modulus for structural stee1. No sirnilar data was
available for reinforcing steel" Thus, the structural steel-
stress-strain curve requj-red that the initial strain-
hardening rnodulus be specified in order to calcuLate the
ultimate strain" For reinforcing bars, the ul-timate strain
is specified in order to calculate the initial strain-
hardening modulus"
2" 8 IN ROIJLED STRUCTURÂIJ STEEL
RESIDUAL STRESSES
Resi-dual stresses form in rolled structural- steel mem-

bers due to uneven cooling of their component parts during
the manufacturing process" parts cooling first resist
contraction and become stressed in compression. parts cool-
ing last become stressed in tension in order to maintaj-n
eguilibrium" Heat treating can reduce the magnitude of the
stresses but is usually not done" Residuar stresses in com-
posite columns, the various theories for pred.icting their
nagnitude and distribution and how they were account,ed for
in Èhis sÈudy are discussed in this seciion.
78
LaChance and Hays (1980) stated that residual stresses

in the structural- steel had no effect on the ultimate
strength of composite cross-sections. Virdi and Dowling
(L973) reported that, in some cases, residual stresses may
enhance the strength of beam-columns. Mirza (l_999) found
that residual stresses r¡¡ere detrimental to the composite
beam-column strength at end eccentricity ratios less than
L"0 but could have a beneficial effect for larger end eccen-
tricity ratios. Beedle and TaI} (1960) reported that resid-
ual stresses reduced the strength of concentrically roaded
bare steel columns"
It is evident that the effect of residual st,resses on
the strength of a composíte beam-column can vary signifi-
cantly and, therefore, r¡/as accounted for in this study.
Alpsten (1968) used a time-stepped finite difference
technique to simulate the cooling of a rolled structural
steel shape as it is manufactured. By moderl-ing the rate of
cooling of the component parts of the rorled shape he was
able to accurately duplicate measured values. To do thi-s,
the cooling and manufacturing history of the shape had to be
known. Àlpsten demonstrated how factors such as restric-
tions to heat flovr affected the residual stresses and their
distribution" Arpsten stated that, in general, the fJ-ange
tips and rnid-depth of the web will have a compressive resid-
ual- stress and that the juncture of the frange and web wirL
have a t.ensiLe residual stress. rf the rolled shapes are
79
placed on the cooling t.able in such a manner that the webs

are prevented from cooling guickly, the nid-depth of the web
wiLl be in tension instead of compression"
Measured values of residual stresses have a high varia-
tion especially at the inid-depth of the web (Alpsten 1968).
These variations are due to different cooling rates and
manufacturing processes. After a rolted shape has cooled,
it often has to be straightened by rollers or by rrg:agging,'
(rnechanical- bending) " This process usually reduces the
residual stresses at a cross-section but since al-l cross-
sections are not treated in the same manner, this should not
be considered to increase the overall strength of the steel_
section (Alpsten L968). Alpsten also found that the resid-
ua1 stresses varied considerably across the depth of very
thick flanges [ 3 inch (75 nm) thick]. Thj-nner flanges
showed little variation"
The flange thicknesses used in this study T¡Íere 1ess than
L"5 inch (38 rnn)" Therefore, residual stresses across the
depth of the flange rdere assumed constant" The residual
stresses hTere assumed identicaL at every section atong the
length of the beam-column with no al-lowance for reductions
due to mechanicar straightening" ArpsÈenrs method of cal-cu-
lating the residual stresses requires knowledge of the spe-
cific manufacturing process a structural shape is subjected
BO
to and, therefore, is not applicable to a generaÌ study

where the structural shapes are drar¿n from several- manufac-
turers "
several schemes have been proposed in the literature to

model the distribution and magnitude of residual- stresses in
rolled I-shapes. Of these, two models are of interest.
Both models assume compressive residual stresses aL the
flange tips and tensile stresses at the juncture of the
flange and web. one moder assumes compressive stresses at
the nid-depth of the web [Figure 2.21(a) ] while the other
model assumes tensile stresses IFigure 2"21-(b) ]. The dis-
tribution of the residual stresses between these points is
assumed either linear or parabolic, âs indicated on Figures
2"21,(a) and 2"zl(b)"
Lachance and Hays (1980) tried both models described

above but gave no details of the assumed distribution or
magnitude of the residual stresses. Trahair and Kitiporn-
chai (1'972) studied inerastic buckling of steel r-shaped
beams and used a distribution sirnirar to Figure 2.2r(a) " A
magnitude of 50 percent of the steer yield stress for the
residuar stress at the frange tips (compression), 30 percent
at the juncture of the frange and the web (tension) and 30
percent at the mid-depth of the web (compression) was
assumed" Nethercot (1974) arso studied raterar buckring of
steel r-shaped beams. He examined how the differeni assump-
tions of resj-duaL stress rnagnitude and distribution affected
8I
Tension
o+_
Conpression
parabolic
distribution
h-near
distribution
_l_
COT
(a) Residual- Stresses in Flange and Web (Young 197f)
Tension
0 l-_
Conpression
( Ü rr-*
\ !/
rr
ll-
I
o.,^, =frv rfw
L]
I
_t_
COT
(b) Residual Stresses jn Flange and Web (Galambos f963)
Fiq.ure 2.2I - Residual Stresses jrt Wide Flange Shapes

82
the ratio of experimental to cal-culated bending moment

capacity" He compared the distributions of Galambos (l-963),
Massey (1-964), and Young: (1-97L) " Poorest correlation was
found with the distribution by Massey. This was expected
since Masseyts dist,ribution, which was based on test results
of Australian roLled joist. sections, assumed that the
flanges ldere entirely in uniform tension and the web in uni-
form compression with a short linear transition assumed at
each end of the web. Because of this very different
residual stress distribution and the results reported by
Nethercot (1974), Masseyrs model was not investigated any
further. The models from Galambos (1963), Young (I97t) and
Trahair and Kitipornchai (1"972) were examined further and
are discussed below"
Galambos (l-963) proposed a residual stress distribution
based on measurements of American f-shapes, mostly used as
columns" The distributj-on he proposed is shown in Figure
2"21,(b) " The magnitude of the residual stress at the flange
tip was assuned to be 30 percent of the yield strength for
mild steel" The residual stress at the juncture of the
flange and web was assumed egual to the stress at the mid-
depth of the web" The residual- stress at the mid-depth of
the web was calcul-ated as a function of the residual stress
at the flange tip and the geometry of the section. The
residua] s'l--ress in the '.seb (or,) is given by Eguation 2.22"
83
úr, 6rft bt + bt l¿.¿¿ )

t u(d 2t)
In Eguation 2 "22 o,r, is the residual stress at the tip of

the flanges, b is the flange tridth, t is the flange thick-
ness, tu is the web thickness and d is the depth of the
st,ructural steel shape "
Nethercot (Ig74) found that use of the Galambos' (1963)
distribution (a1so known as Lehigh distribution) consis-
tently gave conservative results in his analysis of r-beams.
Young (I97I) collected previously published data on
residual stress measurements and measured residual stresses
in British beam and coLumn shapes manufactured from nir-d
steel" His proposed distribution is shown in Figure
2.21,(a) " The residual stresses at the flange tip, at the
flange-web juncture and at the rnid-depth of the web are all-
cal-cuLated as functions of the geometry of the section. The
distribution of stress between these poi-nts is described by
a polynomial" Youngrs eguations for the residual stresses
at the flange tip, at the juncture of the flange and. web;
and at the mid-depth of the flange are given by Eguations
2"23, 2"24 and 2"ZS respectively"
o,rt = -rós t' I 1... \

uea (2.2s)
l1ñJ
ö4
6 ,rto I oo (0, +
i) MPa (2.24)
6,* -100 t' . _ + A* \ vea (2.2s)

[tt úJ
rn the above Eguations õ rrt is the residual stress at the

flange tips r 6 r¡u is the residuar stress at the juncture of
the flange and web , 6,, is the residuar stress at the mid-
depth of the web, A* ís the area of the web and A, ís the
area of one flange of the steel sectj_on"
Young (1'97r) arso suggested that since his proposars hrere
based on geometric considerations, they would be applicable
to various grades of steel- and not linited to rnild steel "
Nethercot (L974) concluded that the use of youngrs model

provided a reasonabry accurate method of incorporating
residuar stresses into the analysis of the moment capacity
of beams failing by laterar buckring. He also found that
the predictions of bending moment capacity were accurate and
nearry identical- when the Youngts model and the AÌpsten's
(L968) finite difference technigue were used to predict the
residual- stresses" Nethercot further concluded that correct
prediction of the magnitude of the residual stresses at the
B5
flange típs and the juncture of the frange and web vias more
important than the actual- stress distributj-on assumed ( i. e.
linear or non-linear distribution) .
Beedre and TaII (1960) reported measurements of residual
stresses in various American mild steer sections. stresses
v/ere found to vary significantly at each cross-section
test,ed along the J-ength of a member" Distributions simi-l-ar
to both those in Figures 2"2I(a) and 2"21_(b) hrere found.
Attempts to correrate the dimensions of the test section
sizes and the residuar stresses \,/úere unsuccessful. Hovrever,
it was found that the residual stresses in the flange
influenced column strength to a greater degree than the
residuaL stresses in the web.
Average magnitudes of measured residual stress at the
flange tip and at the juncture of the flange and web were
estimated from Figure 1 of Beedle and TaII (1960) for the
nine steel section sizes incLuded in that figure. These
measured averages were compared against estimates made using
the moders of Trahair and Kitipornchai (1972), Gal-ambos
(1963) and Young (1971). The resulting comparison is shown
in Table 2"2" Youngts (rg7a) moder predicts greater resid-
ual- stresses at the tip of the flange as the ratio of flange
to web area increases as indicated by Tab1e 2"2. The
section sizes in Tabl-e 2"2 are, therefore, arranged in
ascending order of flange to web area for sirnpricity of com-
parisons.
Table 2.2 - Comparison of Measured and Calculated
Residual Stress'es in W Shapes
Residual Srresses (ksi)

Measured Values Calculated Values
Beedle and Tall Proposed Model I Young 197 1 Galambos (196 Trahair &
( i 960)
I Ki t iporncha i
I ( 1.e72)
-\A
lÀt
(s) 10) 11 t2
I
1"038 -10.0 12.o I -4.12 7"64 | -+.tz 24.72 -IO "2 3. 53 70.2
-17.0
I I co
2 -10 .0 18.O 7 Oì
"O54 t- 1,4.22 "46 l-t+"zz t7 "2t -ro.2 s.35 -17.O ro.2
I I
2.373 -B .0 10.0 t- r5.52 8.72 l-rs"sz 16"26 -10.2 5. 73 -17.0 ro.2
I I
2.7 24 -12.O 4.0 t- 16.60 9.80 | -ro. oo 15.47 -ro.2 6.o2 to.2
-17 .0
I I
2.7 58 -15.0 -4. o t- 16.70 9. 98 | -ro. zo t5.41 6. 10 70.2
-LO.2 -t7 "o
I I
3"170 -15 .0 i4.0 t- t7.64 11.12 l-u.a+ 1,4.7 2 -1O.2 6.43 IO.2
-17.0
I I
3.199 -15.O 5.o t- 17.67 TI .17 l-tt.oo 14. 68 -10.2 6 .44 10.2
-17.O
I I
3.672 -8.0 -5.0 t- 18.50 t2 " 1,3 l-ra.so 14.10 6.69 to.2
-ro.2 -17 .0
I I
4.283 -20. o 5.o t- t9 .21 1.3.t4 l-tg.zt 73 .54 6.96 to.2
-to.2 -17.O
* Residual stress at flange tips.

** Residual stress at the flange-web juncture.
Note; (*) stands for tension and (-) sLands for compression.
I ks i = 6. 895' MPa
87
The Trahair and Kitipornchai (1-972) nodel is entirely

based on the yield strength of the steel and therefore no
variation of residual stresses for different shapes is cal-
culated" This results in generally conservative predictions
of the residual- stresses at the flange tips. The prediction
of the residual stress at the flange tip by the Galambosl
(l-963) nodel is also based on the yield strength of the
steel" In this case the predicted val-ue is often less than
the measured vaLue" The Galambosr estimates at the flange-
web juncture are not significantly different from the mea-
sured data in most cases.
The predictions made using youngrs (1971") rnoder provided
the best comparison to the frange tip residual stress data
reported by Beedle and TalI. A trend of larger flange tip
residual stresses as the ratj-o of flange to web area
increases is seen in both the measured (Beedle and Tall
l-960) and estirnated (Young i,97r) values" A t,rend towards
lower tensile stresses at the juncture of the flange and web
as the flange to web area increases may be hypothesized by
examining the measured data" This trend is followed by the
values generated from Young¡s model as indicated in Tabl-e
2"2" However, the Younq's estímat,es for the residuar stress
at the flange-web juncture are significantly greater than
the measured values in most cases.
oÕ
The yield strength of the steel is not believed to be a
major variabre in deterrnining the magnitude of resj-dual-

stresses (Alpsten 1,972) " The model by Trahair and Kitiporn-
chai (L972) and the frange tip stress pred.iction of GaLambos
(1963) are based soIely on the yield stress of the material
and, therefor, hrere not considered useful-" On the other
hand, Young¡s (1971,) rnodel predicts residuaL stresses based
on the dimensional- properties of the r-sections as d.oes the
flange-web juncture model of Garambos (1963) " At the same
tine, the combination of the young¡s model for predicting
the residuar stress at the flange tip (Equation 2.23) and
the Galambosr model for predicting the residual- stress at
the flange-web juncture (Eguation 2.22) provided the best
overall prediction to the measured varues reported by Beedle
and TaII (1-960) " This cornbination is d.efined as the pro-
posed model in labre 2.2 where it is compared with other
procedures of computlng residuaL stresses. rt was decided
to use the cornbi-natj-on of the young and Galambos models
(proposed model) to estimate the residual stresses in the
rolled steel section of the composite beam-columns investi-
gated in this study. The dístribution of residual stresses
was assumed to be Linear"
The proposed model described above gave the resid.ual
stress at the flange tip (Equation 2"23) and at the flange-
web juncture (Eguation 2"22) " The program cal-culates the
B9
required residual stress at the rnid-depth of the web to

maintain force equilibrj-um of the steel section by a trial
and error rnethod. The following steps \¡¡ere made:
(a) deterrnine the net force in the flanges due to residual
stresses;
(b) determine whether the mid-depth of the web is in t,ension
or compression in order to achieve eguilibriurn;
(c) calculate the nid-depth residual stress assuming a
triangular stress distribution in the web (Figure
2.22 (a) (i) or 2"22 (b) (i) ) ;
(d) if the residual stress computed in (c) exceeds 50 per-
cent of the web yield stress, try a trapezoidal_ distri-
bution (Figure 2.22 (a) (ii) or 2.22 (b) (ii) ) assuming a
value of 50 percent of the web yield stress as the
mid-depth stress, incrementing the zone of mid-d.epth
stress to a maximum of 9O percent of the web depth (Fiq-
ure 2"22 (a) (iii) or 2"22 (b) (iii) ) or until equilibrium
is achieved;
(e) if eguilibriun is not reached in (d) increase the nid-
depth stress by 5 percent of the r,veb yield stress and
repeat the trapezoidal distribution.
Item (e) is repeated until eguilibriurn is achieved.
This procedure baranced the residual stresses in the steer-
shape cross-section before the residual- stress in the web
reached the yield stress leveI. The theoretical program can
2.201 /f.__
vrw jncreases from 0"5 f--y each cycle
thrrough (i) to (iii) until equilibrium reached
2.2I)
'r*
Í- $'rt (- I tr* ,t
slrr 2út* 0.9 (d-2r)
(i) (ii) (iii)

(a) Tensile Residual Stress at Mid-depth of Web (+) jndicates tension
frru \o
O
- 5 f ttt
--ç[trt -l^
tro,,
o .9 (d-2r)
-T J
. (i) (ii) (iii)
(¡) Conpressive Residual Stress at Mid-depth of Vüeb (-) i¡dicates conpression
Figrrre 2.22 - Residual Stress Distril¡ution j¡ Wide Fl-ange Steel Shapes

Used jn Theoretical Strenqth Sübroutine ,.
91
be used with or without the above-noted residual stresses in

the roIled steel section depending on the specified input
option "
2"9 COMPARTSON OF EXPERIMENTåL RESULÍS TO THEORETTCAL MODEL
To check the accuracy of the theoretical model, the

ultimate strengths of column physicar tests pubrished in the
literature r¡{ere compared to the ultimate strengths pred.icted
by the theoretj-caI subroutine. No ne\,ü physical tests r,{ere
conducted for this study. Load cases examined consisted. of
concentric l-oads, eccentric loads creating bending moment
about the major axis, and pure bending about the najor axis"
Length to overall depth (L/h) ratios varied from 2"2 to
30"0. The sources of the physical tests and a brief
description of the specimen configurations are given in this
section. Finally, the cornparisons of measured and calcu-
lated beam-column strengths are discussed.
Bondare (1966 a, b, c) tested 16 composite col-umn speci-
mens with various configurations. Four of the sixteen tests
(RSl-20. 0, RSI-00. 1, RS80 " 2 and RS6O. 3 ) vrere applicable to
this study. of these, the data for RSI-20.0 was rejected due
to premature failure which was attributed by Bondale to
improper placement in the testing apparatus. The specimens
consisted of a 4-inch (l-Ot-"6 mm) deep British RSJ shape,
four O.2L-inch (5.3 nm) diameter rods and o.r25-inch (3 nn)
diameter rectangular ties spaced at 2 inches (50"8 mrn) cen-
ter to center" concrete encased the section and provided a
92
cover of one inch around the steel section for an overall-

cross-section of 6 inches (1,52"4 mm) deep by 3"75 j-nches (95
run) wide. Length t,o overall depth ratios r,{ere between 1O.O
and 1,6"7 with end eccentricity ratios between O"1-7 and 0"5"
Procter (L967) tested concrete encased British RSJ sec-
tions" The overaLl- dimensions of the composite cross-
sections were LL inches by I inches (280 run by ZOO mrn) and
1,2 inches by 8 inches (305 rnm by 200 rnrn). The RSJ sizes
were 7 inches by q inches (t7B by 1OO mm) (depth by flange
width) and 8 inches by 4 inches (2OO by 100 mrn). No verti-
cal reinforcing bars or lateral_ ties were used. Twelve
specimens (numbered l- to 12) had length to depth ratios of
1l- to L2. Four specimens (S1, 52, S3 and 54) had length to
depth ratios of 2"0 to 2"2. The end. eccentricity ratio for
the L2 longer col-umns ranged from zero (concentric) to 0.9.
Al-1 four shorter col-umns were concentricarry loaded.. To
account for the lack of reinforcing, all concrete was con-
sidered to be unconfined in the computer analysis used in
this report"
May and Johnson (1,978) tested g composite beam-coLumn
specimens with restrained ends" of the specimens tested,
only 3 (RCl-, RC2 and RC4) were applicable to this study.
May and Johnson calcul-ated an effective rength (eguivalent
length of a pin-ended column) which was used as an input to
ErìÊ -uÍì.€oreri-cai- suþroutt_ne. The overall ciimension of the
ùl^^ &L^^-^!:
--1 -!l
cross-section was I inches by g inches (200 rnm by zoo mrn) "

93
The structural steel- sectj-on was a 6 inch by 6 inch (150 mm
by British uc section" The vertical reinforcing

1-50 mrn)
consisted of four 0.25-inch (6 mm) d.iameter rods with

0.L6-inch (4 run) diameter rectangular hoops spaced at 6
inches (150 rnn) " The tirne of testing r¡/as noted to be
approximately 4 hours. The effective rength to depth ratios
ranged from I " l- to L4 " 8 " The end eccentricity ratios hrere
0.LLt O"I4 and O"2"
suzuki et aI" (l-983) tested 16 beam-column specimens in
each of concentric and pure bending loading condj-tions wj_th
an additional 2 columns toaded eccentrically causing'bending
about the major axis. The concentric and pure bending spec-
imens had a length to depth ratio of 2.g. The eccentricaÌly
loaded corumns had a length to depth ratio of 3.8 and end
eccentricity ratios of o"87 and l-.06" The overall- dimen-
sions of the cross-section v/ere 9.3 inches by g.3 inches
(210 nm by 210 mm). The steel sectj-ons hrere 6 inches (150
nm) deep with a frange width of 4 inches (l-oo mn) and vari-
ous flange and web thicknesses. Four grades of steer v/ere
tested" vertical reinforcing consisted of four o"2s inch (6
nm) diameter wires" The wires r¡¡ere greased. Hence, the
vertical reinforcement did not contribute to the strength of
the col-umn but provided support to the rectangular hoops.
The hoops hlere also O"25 inch (6 nm) diamet,er and were
spaced aÈ 0"8, J-.6 anci 4 inches (2e, 40 and J-00 mrn) center
94
to center. Columns without hoops were also tested" The

vertical reinforcing was rnodelled in the computer analysis
of this study by specifying a yield strength of zero.
Morino et al. (1984) tested 8 composite beam-column
specimens applicable to this study" The data from one of
these tests was not incl-uded because it had an unreasonably
high strength which was not consistent with the rest of the
data reported by Morino et aI" All columns were of identi-
ca1 geometry. The overall dirnensions of the cross-sectj_on
were 6.3 j-nches by 6.3 inches (160 mrn by 160 mm) " The steel-
section was 4 inches by 4 inches (1OO mrn by 1OO rnrn) with a
flange thickness of 0.3 inches (8 nm) and a web thickness of
O"25 inches (6rnm). Four 0.25 inch (6 nm) diameter bars \"/ere
used as vertical reinforcing" Rectangular hoops were made
of 0"15 inch (4 rnm) diameter wire and spaced at 6 inches
(l-50 nn) center to center. Length t,o overall depth rati-os
ranged from 6.0 to 30.0" End eccentricity ratios ranged
from O"25 to 0"47 "
The physical tests noted above provided uLtimate
strengths of 63 beam-columns specimens which were used to
measure the accuracy of the theoretical modeL" The dinen-
sions, material properties and other pertinent data supplied
by the authors noted above was used as input wherever
applicable" Strain-hardening of both structural and rein-
forcing sÈeeLs and residuaÌ stresses in Èhe structural steel
section Ìnrere also included in the anaÌysj-s. rn some cases,
95
estimates were made regarding certain geometric or material

properties not provided in the source l-iterature. These
estimates hrere, however, belj-eved to be of sufficient accu-
racy for the purposes of this study" Time-to-failure is a
required input to the progiram to estimate concrete
properties as indicated in section 4.1-. The loading period
of the test specimens rdas assumed to be 2 hours, except for
the May and Johnson (L97e) specimens where the authors noted
a loading time of 4 hours. Table 2"3 shows the ratios of
test to calcuLated urtinate strengths (strength ratios) for
all 63 beam-column specimens.
The mean, standard deviation and coefficient of varia-
tion were carculated for the strength ratios listed. in Tabre
2"3" For the purposes of this study, the test specimens
were sub-divided into two categories with respecL to ¿/h
ratio" short columns are assumed to be those with t/h Less
than 6.6 and long columns have I/h greater than or equal_ to
6"6" The data was further categorized into 3 ranges of end
eccentricity ratio (e/h): (a) e/h of o.o to o"z incl-usive;
(b) e/h greaLer than 0"2 but ress than infinity; and (c)
e,/h eguar to infinity (pure bending case). The mean, stan-
dard deviation and coefficient of variation calcul_ated for
each of these categories and for the overarl sample are
shown in Table 2"4.
Table 2.3 - Description of Specimens Used for
Racio of Tesced Eo Calculated Srrengch
Author Co lumn elh Lln bx h Vo Lume t r ic f' f web f,, flange Tesred Calculated Tesced/Ca Lculated
Designation (in. x in.) RaÈio'r (Fsi) ($st) (Ës i ) StrengÈh)kìr Strengch'trk Screngt.h
Bonda e RS 60,3 .0 0.0700 0.0074 44800 47.4
(1966) RS 80. 2 0.33 13.3 3.75 x 6.0 O.O7O0 0.0074 4480 44800 44800 70. 1 56.8 1.2343
RS 100.1 0.17 16.1 3.75 x 6.0 0.0700 o. oo74 4480 44800 44800 92.3 14.3 7.24r7
May ec RC1 o. 11 .1 7.87 ?o 0.001 43 0B 416 30 310. 1. 1013
( 197 8) RC3 0.14 8.i 7 -87 7.87 0.o746 0.oo1B 3390 42050 41 630 306. o 238.3 1.2841
RC4 0. 20 14.8 7.87 7.81 0.0746 0. 0018 5191 42050 41 630 191.3 218. B 0.8741 LO
Oì
orino 84- o.25 15. o 0.0789 507 50 416 1 91 .3
et al. c4-90 0. 2 5 22.5 6.3 x 6.3 0.0789 0.001 2 3378 4567 5 44660 93.9 80. 4 .r675
( 1984) D4-90 o.25 30. o 6.3 X 6.3 0.0789 0.0012 307 4 52055 42485 64 .7 59.2 .0924
A8-90 o.47 ó.0 6.3 6.3 0.0789 0. oo1 2 487 2 53360 43935 116. 1 94.3 .2320
B8-90 o.41 15 .O 6.3 ó.3 o.0789 0. oo12 4828 53360 4509 5 94. o 79.9 .t765
c8-90 o.47 22.5 6.3 x 6.3 0. o789 0. oo1 2 3 567 53 505 44225 68. o 58 .3 . 1664
D8-90 o.47 30.o 6.3 X 6.3 0.o789 o. oo1 2 3320 53360 43790 50. 1 45.9 .0920
roc cer S1 o. o0 .0 x 11.0 n / a':, 41 22 470.
(1967 ) S2 0. o0 7') 8.O x 11 .0 o.0482 n/a 4t 22 427r2 42712 481 .6 524.4 0.9184
S3 0. oo 2.o 8.0 x 12 .0 o.0522 n/a 5407 42560 42560 698. 9 645.4 1.0829
5¿+ o.00 2.o 8.0 x 12 .0 o.0522 n/a 5407 42560 42560 703.4 645.4 1 .0899
I 0. 53 1L.1 8.0 x 7L.25 0.0402 n/a 41 22 42112 42772 132.2 128. 5 1 .0287
2 0.80 7r.1 8.0 x 1r.25 0.0402 n/ a 47 22 42rt2 421t2
a
87 .4 87.3 1 .001 1
0.00 i1.7 8.0 x tr.25 0.0402 o/a 47 22 42772 421t2 470.4 503. 5 o.9343
4 o. 53 77 .7 8.0 x 11.25 0.0402 n/a 47 22 42t12 42172 743.4 128.5 i . 1159
) 0.80 7t.7 8.0 x 77 .25 0.0402 n/a 5401 42112 42t72 91.8 90. 7 7.0122
6 0.75 11.U 8.0 x 12.o o.0522 nla 5401 42560 42560 129.9 711 .3 r.lo12
l o. 50 11 .0 8.0 x L2.O o. 0522 n/a 5407 42560 42560 r99 .4 169.8 7 .11 47
I 0.00 11.0 8.0 x 12.0 o.0522 n/a 5407 42560 42560 560. o 614.9 0.9107
9 o.27 12.o 8,0 x 17.25 o.0402 n/a 600 7 42112 42772 268.8 234 .3 1.0706
10 o .27 t2.o 8.0 x 11 .25 o.0402 n/a 600 7 42t12 427t2 250.9 234 .3 7.r477
11 0.00 11 .0 8.0 x 12 .0 o.0522 n/a 600 7 42560 42560 533.1 660 .1 0.8069
t2 o .25 11 .0 8.0 x t2.o o.0522 n/a 6007 42560 42560 315.8 288 .1 1 .0939
Table 2 .3 (cont. )
Aut Co I umn e./h bxh Volumetric f web f flanøe Tested cu aled Tes¡ed Calculated
Designation (in. x in.) Ratio* (ðsi) ([si ) S Lrengthtttk Strengthì'rì'.- Srrength
Su zuk i LH-O 0.00 2. o. o1 94 n | ¿-:t 4513 380. 6
ec al. LH-020-C 0. o0 2.9 8.27 x L21 o. 0194 o.o27 4 4747 45135 4s 586 37 4.o 447.1 0.8468
(1983) LH-040-c 0. o0 2.9 8.21 x 8.27 o.0194 0.0137 4747 45135 45586 374.O 404.0 o. 92 58
LH-100-C 0.00 2.9 8.27 x 8.27 0.o194 0.0055 47 47 45135 4s586 385. o 381 .0 1.0105
RH-000-c o. oo 2.9 8.21 x 8.27 0.0439 nla 4840 55355 48159 550. o 458.4 7 . 1997
RH-020-c 0.00 2.9 8.27 x 8.27 0.0439 o.o274 4840 55355 48159 561.0 536.0 r.0467
RH-040-c 0.00 2.9 8.27 x 8.21 0.0439 0.0137 4840 s5355 48159 517.0 499 .4 1.0353 \o
RH- 100-C o.00 2.9 8.21 x 8.21 0.0439 0.0055 4840 55355 48159 \¡
517.0 475.4 1.0875
HT60-000-c 0.00 2.9 8.21 x 8.21 O.0600 nla 4840 83600 83 600 594. o 562 .7 1.057
HT60-020-c 0.00 2.9 8.27 x 8.21 0.0600 O.O214 4840 83600 83 600 6s3 .4 685 .1 0.953
HT60-040-C 0.00 2.9 8.21 x 8.21 0.0600 0.0137 4840 83600 B3 600 660.0 644 .6 1 .024
HT60-1 00-c 0.00 2.9 8.21 x 8.21 O.0600 0.0055 4840 83600 83 600 620.4 614.4 1 .010
HT80-OO0-C 0.00 2.9 8.21 x 8.27 O.0633 nla 4840 1 13406 13406 708 .4 530. 1 1 .3364
HT80-020-c o.00 2.9 8.21 x 8.27 0.0633 o.0214 4840 r 13406 13406 126.o 808. 1 o. 898 5
HT80-040-c o. oo 2.9 8.27 x 8.27 0.0633 0.0137 4840 r 1 3406 I 3406 723.8 164.8 o.9464
HT80-1 00-c o. oo 2.9 8.21 x 8.27 0.0633 0.0055 4840 1 13406 13406 704.0 125.r 0.9709
HT80-000-cB 0.87 3.8 8.21 x 8.27 O.O429 nl a 4409 1 10568 10568 110. o 91 .t 1.1326
HTBO-O2O-CB 1.06 3.8 8.27 x 8.27 o.O429 o.OZi4 4409 110568 10568 110.0 109.6 1.0036
LH-000-B - 2.9 8.27 x 8.27 0.0194 nla 47 47 45 135 45 586 27 .1 26.9
ó ,O
l.oo71
LH-020-B 8.21 x 8.27 o.O194 O.o274 4141 45135 45586 29.6 32.4 0.9133
LH-040-B - 2.9 8.27 x 8.21 O.O194 0.0137 4741 45135 45586 28.9 30. 6 o. 943 5
LH- 100-B - 2.9 8.27 x 8.21 0.0194 o.oo55 414t 45135 45586 28.9 28 .4 1.0165
RH-000-B - 2.9 8.27 x 8.27 0.0439 nla 4840 55355 48159
6?O 49.t 50. o 0.9816
RH-020-B 8.21 x 8.21 0.O439 O.o274 4840 55355 48159
ó 54.9 5ó.8 0.9658
RH-040-B tO 8.21 x 8.27 O.0439 0.0137 4840 553s5 48159 53 .4 52.4 1 .0i 93
RH-100-B - 2.9 8.21 x 8.21 0.0439 o.OO55 4840 55355 48i59 50.2 50. 9 o. 985 5
Table 2.3 (cont.)
Author Co I umn e/h Llh bxh VolumeEric f',c .. f web f.. flange TesLed Calculated Tesred/Calculared
Designat.ion (in. x in.) Ra c io''.- (ps1) (ðsi) (Ësi ) StrengÈh*':r Strengch;krk Sc rength
Suzuki HT6O-000-B ,o a a1 0. oóoo nt a^ 4840
8.27 83 600 83600 69 .7 72.9 0. 9549
eE aI. HT60-020-B 2.9 Q tÌ 0. 0600 o.o214 4840
8.27 83600 83600 78.O 80.7 0. 9658
( 1983) Hr6o-040-B 2.9 I .27 8.27 0. 0600 0.o137 4840 83 600 83600 76.5 76.4 1.0020
(cont. ) HT6o-100-B to I .27 8.27 0. 0600 o.0055 4840 83600 83ó00 71 .8 75.5 0.9513
HT80-000-B 2.9 I .27 8.27 0.0633 n/a 4840 13406 1 13406 93 .8 95.9 0.9787
HT80-020- B 2.9 8.21 8.27 0. 0633 o.o274 4840 13406 113406 102.5 105.9 0. 9682
HT80-040-B )o I .27 L27 0. 0633 0. 013 7 4840 13406 1 1 3406 101 .1 103.3 0.9778
HT80-1 00-B 2.9 8.21 8.27 0. 0633 o.oo55 4840 13406 1 1 3406 91 -4 100.3 0.9714
\o
@
>!Ratio of volume of laceral hoops to gross volume of confined concret.e core. Concrete core is measured to oucside of lateral hoops. Columns without
Iateral hoops are indicated by. n/a.
:"-:"- SErength (measured or calculac.ed) is axial load at faÍlure in kips, except when e/h = - where st.rength (measured or calculated) is bending mornenE af
faÍlure ín fooc-kips.
1in.=25.4mm;1OO0psi=6.895MPa;1kip=1OOOpounds=4.448KN;1fooc-kip=1.356KN-m.
Notes:
1. Ratio of area of verÈical reinforcing bars Èo gross cross-secEion area was O.0062, O.OO28 and O.OO44 for Bondale (1966), May et aI. (1978) and
Morino et a1. (1984), respectively. All ocher columns had no verrical reinforcing,
2. Yield strength for reinforcing bars (vertical and/or laÈeral ties) was assumed ro be 71000 psi for Bondale (1966) and May e¡ al. (1978). yield
srrengt.h was 56115 psi for Morino et aI. (1984) and 48400 psi for Suzuki er aI. (1983).
Table 2.4 - Statistical Analysis of Rarios of Test to
Calculated Strength
all e/h 0 ..( e/h -< 0. O.2<e h=-

Type
(1) (3) (4) (s) (6)
I 3e I 20 1,6
Short ( c"/n < 6.6 ) I 1 .019 l.

l_lr.ooo
I .123 o.975
11 .04 LO.2 2 .86
T1
Long (!" /h >. 6.6) r.o22 L. L¿

\o
\o
18.37 ).v
21 20 I6
ALr !,/h
tt I i.o4 L.02 r
I - "t2 0.975
ixi 10.6e 12.93
l-V-l 6.3s 2 .86
100
The ratj-o of t,est, to calculated ul-timate strength was

L.04 with a coefficient of variation of 10.69 percent when
all 63 specl-mens Ì¡ùere considered (Tab1e 2"4 Column 3).
This is comparable with the mean value of I"O42 and coeffi-
cient of variation of 10.4 percent obtained by Virdi and
Dowling (1973) for their analysis of I biaxially toaded.
composite corumns. The differences in statistics for short
and long columns drawn from the overall_ sample was consid-
ered negligible as indicated by Co1umn 3 in Table 2"4"
There were significant differences in the statistics for
different ranges of end eccentricity ratio (Table 2.4 CoI-
umns 4,5, and 6). The mean strength ratio hras very close to
1.0 and no effect due to slenderness was noticed on mean
st.rength ratios of coLumns in the row eccentricity rangie
(Table 2"4 - Colurnn 4) " In the second eccentricity range
(Tabre 2.4 column 5), the mean value of the strength ratio
was significantly greater than L"o. Hence, the theoretical
model seems to be conservative in this range of end eccen-
tricity" Again, Do length effect on the mean val-ue was
noticed" For the pure bending condition (Table 2"4 col-umn
6), the mean strength ratj_o was slightly 1ower than 1"0.
This was probably due to the strain-hardening assumptions
used for the theoretical strength rnodel"
In surnmary, the mean strength rat.ios in Table 2"4 are
consistent for short and long column specimens and for the
combined sample" The coefficients of variatíon, hora/ever,
10l_
shot¡ some differences between short and long col-umn speci-

mens although a definite trend to lower coefficients of
variation as the end eccentricity ratio increases is
apparent in Table 2"4 (Colurnns 4, 5, and 6) . Considering
the wide range of column sizes, configurations, load.ings and
sources, and the small_ sample sizes for some categiori_es, the
accuracy of the theoretical strength model seems acceptabre"
To examine the probability dist,ribution of the strength
ratios calculated above, the data for each range of end.
eccentricity ratio was plotted on normal probability paper.
A normal distríbution was calculated from each set of data
using the statistics of overall sample given in Table 2.4
(Columns 4t 5, and 6) " Figures 2"23, 2"24 and 2"25 plot the
probability distributions for the ranges of end eccentricity
ratio noted in Columns 4,5 and 6 of Tab1e 2.4, respec-
tively. rt can be reasonabry assumed that the data forl-ows a
normal probability distribution for all three ranges of end
eccentricity ratio"
2 "LO CALCUÍJåTION OF MODETJ ERROR
The strength ratios calculated in section 2"9 represent

the overall variation between the test specimen strength and
the strength predicted by the theoretj-cal model" The coef-
ficient of variation of the ratio of tested to calcurated
strength, Y rr", is attributed to three sourcesi Vmod"t, Vr"..
q¡¡u
=n¿jl \/.
v in-batch. \/v i ^ +-t^^
¡nodet l_:i L¡Ie L:(J€III(JIenE
^^^êê: ^.: ^-! OI
^ê ----.: ^!.:
Vafl_aE,LOn -- fepfe-
senting the variability of the theoretical model. V."", is

L02
r00 c
99.. 99
oo ô(
99.90
99.80
'Normal Probability /
Distribution ,/
99 .50 Ir1ean value
= L,02 \ /
99.00 of variati-on = O.],2g
Coef f icient V
Nunber of specimens = 27
98. 00 /
/
95.00 /"
,/
/"
1 90. 00
6 o/ ,Á
.H
0)
o/
Ê{ o./
70. 00
o,/
o./
;É o./
OJ '. oo,/
a
B,'
50. 00 o"/
g
f¡{
o
o/
o
"Y
.t 30. 00 ar/o
+J
d
- o./
-l
o'/
/
5
O
,4
,/ ^
10.00
/o
5.00
/"
2.00
1.00
0.50
0. 20
0.10
0.05
0. 0l
0 .80 0.88 0.96 1.04 L.t2 1.20 r.28 1.36
Tested / Calculated Resistance

(0.0
-< e,/h 0.2) -<
Figure 2 "23 ' Cr¡rurlative Frequency of Ratio of Tested / Calculated

nes.r¡talce of conqnsite Bean¡{olunn specinens
(Table 2"3) wirh 0"0 _( e/h_( 0.2
103
99.99 ^xI00B
oo oq
99.90
99.80
Norma1 Probabilit.y Distribution
99.50 - Mean value = L.L2 \
. -
Coeff ici-ent of variation = 0.063 \ \
99.00
*"mber of specimens = 20
98 -00 -
95. 00
90. 00 o,/
E ,/
Þ
.0J "/
o
õ
Ê.' ,/"
v
70. 00
>r /:
(') ,, ,/o
v4
L
(,
j 50.00
I' o///
Ë
0) 30. 00
.-l
Ð r',Å
rd
FI
tr/
d
"r/
U 10 .00 'o,/ -/
,/
,/
5. 00 d
2.00
/
1" 00
0.50
0.20 :.
0.10
0.05
0.0I
r.00 1.04 I.0B r.r2 I.16 1.20 I.24 r.28
Tested / Calculated Resjstance

(0.2 < e/h (o-)
Figure 2"24 - cumulative Frequency of Ratio of Tested / carculated

Resjst¡¡rce of Colrposite Bea¡rr{o1unu: Specjnens
(rable 2.3) with 0"2 <e/h ( so
104
I00 c
99.99
99.95
99.90
99. B0
99.50
99.00 Normal Probabilit,y Distribution .,

98.00
Mean value = 0"975 \.
Coef f icient of variation = 0 .02 9 't
Number of specimens = 16 /
95.00 ,/
a/
90.00
îd
0)
o
t..l
o
g 70. 00
o
o
o.
0,
j 50. 00
g
h
0)
.ì
Ð
rd
Fl
H
(J I0"00
5.00
2.00
1" 00
0.50
0. 20
0.
0.05
0. 0l tl
0.90 0.92 0.94 0.96 0.98 r.00 1.02 r.04 1.06
Tested / Calculated Resistance

(e/h =oo¡
Figure 2" 25- Cu¡nrlative Frequency of Ratio of Tested / CalculateA

Resistance of Corposite BeanrColun¡-r Specj¡rens
(Table 2"3) with e/h =oo
105
the coefficient of variation representing uncertainties in

recording the correct failure load during physical tests due
to inaccuracies in measuring eguipment, recording procedures
and different definitions of faj-lure" Vin-barch represents the
coefficient of variation that accounts for the variability
of laboratory material properties as well as differences in
strength between laboratory control samples and the materi-
als in the test specimens. Mirza and MacGregor (rg}z) have
rel-ated these four variabilities as shown in Eguatíon 2.26.
(2.26)
To calculatê Vmoael, Equation Z"26 was rearranged to Eguation

2.27 "
V mod"l (2.27)
v17" wâs found to be significantly affected by eccentric-

ity ratio in section 2"9" Hencer vmoder arso depends on the
eccentricity ratio of the beam-column as described bel-ow.
v."", was assumed to be o. 04 for end eccentricity ratios
between 0.0 and o"2 incrusive. This value was used by Mirza
and MacGregor (1989) for a similar study of reinforced con-
crete beam-columns " v 1ss1 Ì,üâs assumed to eguar o . 02 at pure
106
bending (e/h=*), and was arbitrarily to vary

assumed
linearly at end eccentricity ratios between o.z and infinity
as shown in Eguation 2"29"
V t.st O,OZ + O.OO4h/ e f or e/ h> 0.2 (2.28)
The higher value for vtest ât low end eccentricity ratios
is justified due to the invorvement of a compression fair-

ure, the resurting loss of measuring accuracy, and the dif-
f iculty in def ining a point of fairure. v.u=. used. in this
study is graphically represented in Figure 2"26(a).
Vi,-¡¿1s¡ 'r,{ês calculated using the Monte Car1o techni_que
described in chapter 5. Four beam-columns v¡ere chosen to
calcurate the in-batch variabirity" These beam-corumns hrere
selected from the t,est specimens listed in Tabre 2"3. vari-
ous e/h and l/h ratíos r¡¡ere represented. The beam-columns
selected for computing v¡,-5"¡"¡ ârê shor,,¡n in Tabre 2"5" The
theoretical strength of each corumn shown in Table 2.5 was
simurated 2oo times" Each time the values of the basic
variables \¡rere randomly generated by the computer according
to predefined in-batch probability distributions of each
variabre affecting the strength (Table 4"1) " For sirnpric-
ity, the material strengths and. dimensions reported for each
test specimen hrere taken as the mean values of the variables
for computing the vin-batch for that specimen. The theoretical
107
vtest
vtest = 0 "02+ (0.004 h/e)
(a)
= 0 "025+ (0.018 h/e)
V
model = 0.025
(b)
test and V_^._.

Figure 2"26 - V-__- mooer used
Table 2.5 - Variability of Theoretical Strengt,h Model
Co lumn Vmode v*ode

Designation elh cln v
I I
r/, in-ba Èch test
(Table 2.3) I
(calc) (used )
(1) (2) (3) (4) (s) (6) I
(7) (8)
LH-100-C 2.9 o.129 0.039 0. 040 0. 116 0. 115
RC1 0.11 8"1 o"129 o. o28 0.040 0.119 0"115

I F
o
co
D8-90 o.47 30.0 0 "063 | o. o18 0.029 0. 053 o. 063
I
LH-100-B 2"9 0.019 0. o20

-lo.o2e I 0.008 o. o25
-l
t09
strength samples so simulated were used to compute Vin_batch

for the beam-columns listed j-n Tab1e 2.5" The computed val_-
ues of Vin-batch and V¡ss1 ârê shown in the same table" The
values of V 17" given in Tab1e 2.5 were taken from Table 2.4 .
Vmoder was then calculated using Equation 2"27 for each of
the beam-columns listed in Table 2"5" The resulting values

are shorr¡n in column 7 of Table 2"5" These values provided a
basis for estimating the coefficient of variation associated
with the theoretical strength modeI.
V -o¿"¡ wâs chosen to be constant at O " Ll_5 for end eccen-
tricity ratios between 0"0 and O"2 inclusive. A value of
Vmoder equal to 0.025 was assumed for the pure bending
condition" Vmoder for end eccentricity ratios greater than

O"2 and less than infinity was assumed to vary inverseÌy
with respect to the end eccentricity ratio as shown in Egua-
tion 2 "29 "
Vmodur = 0.025 + O.OI8h/e for 0.2<e/h <co (2.29)
The coefficient of variation of the theoretical strength

model used in this study is shown graphically in Figure
2"26(b) " The values of V*o4"¡ based on this Figure for four
typicar beam-column specimens are given in cotumn g of Tabl-e
110
2.5" A comparison of these values with the calculated val--

ues of V-o¿", in Co1umn 7 of the same Tab1e indicates a rea-
sonable agreement between calculated and used varues of
V model .
A random normal variable with a mean of l-"0 and a coef-

ficient of variation as described above was used to vary the
strength ratios calculated in the Monte carro simul-ations
described in chapter 5. The mean varue of 1.0 was chosen
since it. is a conservative estimate of the strength ratios
of test to calculated strengths described in section z.g and
summarized in Table 2.4" At the pure bending condition,
this assumption for the mean val-ue appears slightly uncon-
servative when strain-hardening of steel is used (Tab1e
2"4) " However, it is expected to be conservative when
strain-hardening of steel is not allowed" A normal- proba-
bility distribution of the strength ratios was assumed based
on Figures 2.23,2"24 and Z"ZS"
r11
3 3TO}4ÏNATJ BEÃ]4:CO&U}43T STRENGTTT

A computer subroutine was used to calcul_ate the design
code strength (referred to as nominal strength) of a compos-
it'e bean-column" This strength was then compared to the
correspondj-ng theoretical strength of the beam-column. The
subroutine cal-culates the design code strength according to
the assumptions and methods of ACr standard 3l-8-93 or csA
standard cAN3-423.3-M84. rn this study, the comparisons of
the theoretical strengths vrere done prirnarily with the
strengths computed according to ACI Standard. 318-83.
This chapter describes how the nominal strength of the
composite beam-col-umn is carcul-ated. The assumptj-ons mad.e
by the design codes regarding naterial strength and strength
analysis of the beam-column are d.iscussed first. The nomi-
nar strength program, R¡[oM, is then described. The d.iffer-
ences between ACI 318-93 and CSA CAN3-A23.3-MB4 are
discussed in the final section of this chapter.
3 " L Ã,SSUMPTIONS
The North American design standards make certain assump-

tions regarding the characteristics of the materials used
and the behavior of the composite beam-corumns in order to
sinprify the design" The assumptions discussed here are
conmon to both ACr 318-93 and csA CAN3-A23.3-MB4 and have
been incorporated into the nominal strength subroutine.
Assumptions regarding the behavi.or of the composite
bearn-column are:
LI2
(a) perfect bond exists between steel and concrete, i.e. no

slip;
(b) strain in the composite cross-section is proportional-
to the distance from the neutral axis;
(c) composite columns with a slenderness ratio, kl/r, less
than ([34 3-2 þ1t/þ72) where M1 is the lesser and M2 is
the greater column end bending moment, were considered
short columns and length effects were neglected.
Assumptions regarding the behavior of the material-s are:
(a) the maximum useable concrete strain at the extreme com-
pression fibre is 0.003;
(b) the tensile strength of concrete was neglected;
(c) the shape of the concrete stress-strain curve v/as
assumed to be an equival-ent rectangular stress bl_ock
with a maximurn stress equal to 85 percent of the speci-
fied 28-day cylinder strength;
(d) the uniformly distributed compressive stress in the
concrete is bounded by the linits of the section and at
a line parallel to the neutral- axis and a distance B,c
from the extreme compression fibre (Figure 3.1) where c
is the distance from Èhe neutral axis to the extreme
compression fibre and B¡ has a value of O.B5 for speci-
fied concrete strengths up to 4OO0 psi (30 Mpa) and
decreases linearly by 0"05 for each 1-OOO psi (0.08 for
each i-0 MPa) to a minirnum value of 0 " 65;
fl-
Neutral axis
DNA
-]*'
F
H
Plastic (¡.)
Centroid =0 " 003
Strain Compressive Stress

Distribution D istribution
Figure 3"1 - Code Stress and Strain Distributíon i

for Concrete ;
i
IT4
(e) Lhe stress-strain relation of steel- (structural shapes

and reinforcing bars) was assumed to be elastic-plastic
with the maximum stress equal to the specified yield
stress (see Figure 3 "Z) ¡
(f) the stress-strain relation of steel is identical in
compression and tension; and
(g) resj-dual stresses in the steel section are neglected.
The above assumptions hrere applied to all composite
beam-columns " In addition several assumptions \^/ere made
which h¡ere applicable onry to the beam-columns invest.igated
in this study" These assumptions are:
(a) the beam-columns were pin-ended so that the effective
length was egual to the actual length;
(b) the column ends were prevented from translation (sides-
rday prevented) ;
(c) no transverse loads vrere applied to the columns;

(d) end moments were egual and opposite such that the beam-
column bent in single curvature and that the ratio of
the end moments was egual to 1; and
(e) the loading tine to faiLure was short so that the
effect of creep Ìrere neglected (ßa : 0) .
The ACI and CSA design codes specify limitations on the
maximum material- strengths as well as on ratios of struc-
tural- steel and reinforcing bars to the total cross sec-

1l- 5
f (tension¡
€. (comp) e ( tens ion )
f
v
I ( compre ss ion )
Figure 3 "2 Code Stress-Strain Relationship

for Steel
116
tional area" These limitation v¡ere observed. when designing

the beam-columns used in this study and described in chapter
Another reguirement of the desigin codes is that ar-l load

assigned to the concrete portion of the composite section
must be transferred to the concrete by members or brackets
attached to the steel core and in direct bearing with the
concrete" Although not specifically addressed in the
strength carculations, this reguirement ensures the val_idity
of the assumption of no srip between sÈeel and concrete.
To calibrate the nominal strength model- and to provide
data in a form that is easiry useabre for future probabiJ-is-
tic assessment of understrength factors, all understrength
factors in this study vrere assigned a value of l_. O "
3 "2 NOMINÃIJ STRENGTH PROGRã¡'T
calcul-ation of the nominal beam-corumn strength requires

the foLlowing procedures:
(a) input of nominal dimensions, nominal material
strengths, selection of design code (ACf Standard
3l-8-83 or CSA Standard CAN3-A23 "3-M84);
(b) discretization of the structural steel section;
(c) calcul-ation of the cross-section axiar road - bending
moment (P- M) interaction diagram;
(d) carculation of the composite slenderness ratio to
determine v¡hether the beam-col-umn is considered short.
or long; and, if needed,
117
(e) cal-culate the long col-umn P - M interaction diagram by

evaluating the concentric capacity and the bending
moment nodifier for long column effect.
All specified dimensions, areas and geometric properties
of the column as well as the components are read into the
program" The specified concrete 28-day cyli_nder strength
and the speci-fied yíeld strength of the structural and rein-
forcing steer are required J-nputs " The modulus of el-astic-
ity of the concrete is calcul-ated by the design cod.e
expression shown in Eguation 3.L"
E
":57, OOOû i psi (s,t )
= Sooo MPa
^[1"
The modulus of elasticity for steel was assumed to be

29t0OO,000 psi (2001000 Mpa). The progran is designed to
include the specified values of the understrength (O) fac-
tors and the sustained load factor (ß¿). As stated earlier,
the understrength factors were al-l assigned a val-ue of 1. o
and ßd r¡/as assigned a value of O.O for this study.
All concrete areas rárere assumed to behave in a simil-ar
fashion, with no distinction made between unconfined and
confined concrete" Discretization of the concrete in the
cross-section was not reguired because the stress in arI
stressed parts of the concrete v¡as assumed to be uniforrn
118
(rectangul-ar stress block) " The stress in vertical rej-nfor-

cing bars was determined by nultiplying the strain calcu-
lated at the centroid of the bar by the modul_us of
elasticity of the reinforcing bar. Discretization of the
structural- steel section was done since the stress varied
signÍficantly along the depth of the section" The steel_
section was discretized into zo elements with boundaries
parallel to the neutral axis. The stress in each element
was calculated by multiplying the strain at the centroid of
each element by the modulus of el_asticity of stee1" The
stresses so calcurated !ìrere reduced by 0.85/'" for reinfor-
cing bars and parts of the steel section that fell within
the concrete stress block. Residual stresses in the
structural steel v¡ere neglected"
The development of the nominal cross-sectj-on P - M inter-
action diagran is much simpler than that used for the theo-
retical- analysis. The strain at the compression face of the
concrete was set at the maximum value allowed by the cod.e
(0.003)" The location of the neutral axis was fixed at a
distance from the compression face and the corresponding
axial road and bending'moment capacities calculated. This
locates one point on the cross-section P-M interaction
curve (Figure 3"3).
The location of the neutrar axis was varied between the
pure bending anci the pure compressj-on cond.j-tions and the
1l- 9
P
n
P
c Short Column
Interaetion Curve
P
ky <22
r
Slender Co1umn
Interaction Curve
22< E{_-< loo
r
M
snort
M
S lender
/ M
snorË
ù ¡ tat"n*t )z
- 1.0
Figure 3"3 Axial- Load-Bending Moment

Interaction Diagram for
Nominal Strenoth Subroutine
L20
corresponding axj-aI load and bending moment capacities hrere

caJ-cu1ated, creati-ng sufficient points to accurately define
the entire P - M diagram. These points were then used. to
interpolate the axial load and bending moment capacities for
the specified end eccentricity ratios.
After the completion of the cross-section P - M interac-
tion diagram, the program determines whether slenderness
effects are required to be accounted for. If the
slenderness effects are to be included, the radius of gyra-
tion r of the composit,e section is estimated using the code
expression shown in Eguation 3"2 in which E" is the modul-us
of elasticity of concrete (Eguation 3"1), ,4n and 1n are the
area and moment of j-nertia of the gross cross-section,
respectively, -F" is the modulus of erasticity of the struc-
tural steel- section and, 24" and 1" are the area and moment
of inertia of the structural steel section, respectively"
'-V @
G"AJs)-Er\ (3.2)
The slenderness ratio of the beam-column is then calcul-ated

using Equation 3"3"
slenderness ratio = kl/ r (J.JJ

L2L
In this study, the effective length factor k in Eguation

3" 3 had a value of 0 due to the assurnption of pinned ends.
l-.
ACI 3l-8-83 and CAN3-423"3-M84 specify that the slenderness
effects may be neglected if the slenderness ratio calculated
by Eguation 3"3 has a value less than that calculated by
Equation 3.4.
kL/r<34-12(M,/Mr) (s.4)
In thj-s study, the ratio of the end rnoments (\Ar/l\,42) is

equal to 1.0. Therefore, beam-columns with a slenderness
ratio of 22 or less v¡ere crassified as short columns and the
slenderness effects were neglected, i"e" the member strength
h¡as consideredto be the same as the cross-section strength.
slenderness effects must be considered for corumns for
which kl/r exceeds 22" For such cases, the critical
strength of the srender beam-column P" must be carculated. by
Equation 3.5.
n2EI
P (3.s)
"= (k D2
To determi-ne the critical

strength by Eguation 3"5, the
flexural stiffness (E I) of the beam-col-umn v/as first calcu-
lated by Equation 3.6"
L22
,,:(t,,tI + {t)*4"1" (3.ó)

ßa
In this study the value of f3o in Eguation 3.6 was taken

equal to 0"0 as stated earlier. ft is interesting to note
that the design codes all-ow the stj-ffness of the vertical
reinforcing bars to be considered for reinforced concrete
columns, but not for composite columns even though the
requirements for percentage of vertical reinforcing bars are
identical for both cases.
Slenderness effects reduce the bending moment capacity
of a beam-column associated with a particular axj-al load.
Hence, for axial- loads lower than the long column concentric
capacity, these effects were accounted for through a bending
moment modifier öo (Equation 3 "7) applied to the cross-
section bending moment capacity obtained at the same load
leve1 of axial load"
c^
Þ (3.7)
t--
' ôP"
t.0
The nodifier has a varue of i-"0 for the pure bending case
and increases until the long column concentric capacity is
reached" The $ factor in Eguation 3"7 was taken equar to
123
l-.0 as explaj-ned earlier and the 6:,,, factor is defined as

the equivalent uniform bending moment diagram factor and was
calculated from Eguation 3.8 "
M,
C^=0.ó+0.4+
M2
(3.8)
> 0.4
In this study, C^ had a val-ue of i_.0 due to the uniform

prirnary bending moment diagram resulting from the assumption
of egual and opposing end moments"
Dividing the cross-section bending moment capacity at a
given axial load by ö6 yielded the slender column bending
moment capacity at the same axial load as shown in Equation
3 .9.
M M õ, (3.e)
"¿nndnr: "oorr/
Thís yielded one point on the slend.er beam-column P - M

int,eraction diagrarn (Figure 3.3). carcuLating the slender
beam-column bending moment capacities for several- leve]s of
axial l-oad provided the data points required to construct
the entire slender beam-column P-M interaction diagram.
L24
These points !'¡ere then used for interporating the r-ong cor-
umn moment and axial- load capacities for specified end
eccentricity ratios.
3"3 COMPåRISO3ü OF DESIGST CODES
ACI 3l-8-83 and CSA CAN3-423.3-MB4both impose lirnits on
the geonetry, material behavior, strength assumptions and
the nominal strength of structural members d.esigned in
accordance with these codes" Generally, the two design
codes are similar in their limitations. The rimitations and
the differences between the two codes are discussed berow.
The most, obvious difference in the ACr and csÀ codes is
with respect to the apprication of understrength factors.
The ACr code calculates the nominal design strength of a
composite beam-column using all specified. material strengths
and cross-section dimensions. An overall understrength fac-
tor (0 < 1"0) is applied to the nominal axial and moment
capacities. The magnitude of the understrength factor
depends on the failure mode of the beam-column which is
defj-ned by the straj-n state rerative to the balanced strain
condition (balance point). Acr 318-83 defines the balanced
strain condition as the point on the cross-section P - M
interaction diagram corresponding to the strain cond.ition in
which the strain at the compressive face of concrete reaches
0. 003 as the tensil-e stress in the verti-cal reinforcing
reaches its yieJ-d point" Failures at axial- loads greater
L25
than the balance point load are compression faj-lures " Fail--
ures at. axial loads less than the balance point, load are
tension failures" Tension failures can be predicted
accurately and, therefore, the understrength factor for ten-
sion failures is greater than that for compression failures.
The definition of the balance strain conditj-on as it applies
to composite beam-columns is discussed in detaiL in Section
5.3.1-. The nominal strength program assumes the point cor-
responding to maximum moment on the cross-section P- M
interactj-on curve as the transition point between the
tension and compression failures and the related under-
strength factors. In this study, the definition of the bal-
ance point does not affect the results since al-l-
understrength factors v¡ere set to 1"0"
CSA Standard CAN3-À23"3-M84 applies material under-
strength factors directly to the specified strengths of the
constituent materials" Ðifferent values are applied to each
material" No difference between compression or tension
failure is made. This method has been used by the CSA code
since the 1984 edition" Prior to this, the rnethod used by
ACf was also used by the CSA code.
Both design codes restrict the cross-section axial load
capacity by irnposing a ceiling on the axial- load level. f n
the preceding section it was stated that the nominal
strength subroutine cal-cuLates Èhe cross-section concentric
capacity on the basis of strain cornpatibility. The design
L26
codes calculate the concentric axial load capacity by assum-

ing that each material in the cross-section contributes to
strength in proportion to its area and its maximum
permissible stress (Note this assumes all- materials reach
their maximum permissible stress at the same strain). ACI
318-83 assumes that the concrete maximum stress is 85 per-
cent of the specified 28-day cylinder strength and the steel-
stress is the specified yield stress of the material" The
strength from each component is summed and multiplied by the
compression failure understrength factor. The strength is
reduced. further by rnultiplying it by 0.85 as shown in Equa-
ti-on 3"L0"
0P^ : 0.850[0.85 f ' A,,) * f (3.1 0)

"(A n- ""A ",J
CSA CAN3-423.3-M84 uses a similar teehnigue by reducing the

factored concentric axial load by 20 percent for composj_te
columns with rectangul-ar horizontal ties as shown in Egua-
tion 3"1-L"
P n: O.80[0.850" f '"(An- A"- A,) * $".f y"1" + þ, f A,l (3.I I )

",
The ceilings on axial load described herein trere not consid-

ered in the Monte Carlo study described in Chapter 5. How-
L27
ever, the axial load given by Eguation 3 " 10

maximum
(Eguation" 3.1-1 for CSA) is recorded in the output of the
program for the readerrs information"
Both codes apply an understrength factor to the cal_cu-
Iated long column critical strength as shown in Eguation
3 "7 " The value for the understrength factor is different
for Èhe two codes. Hov¡ever, for this study, Èhe Q factor in
Equation 3.7 was taken egual to 1"0"
Limitations on material strengths are similar in both
codes" The minimum specified concrete strength is 25oo psi
(I7 "2 MPa) for ACI 3L8-83 and 2900 psi (20 Mpa) for CSA
cAN3-À23"3-M84. The maximum structural steer yield strength
permitted is 50,OOO psi (345 Mpa) for ACf 3t-B-83 and 50,750
psi (350 MPa) for csA CAN3-A23.3-M84. These criteria were
taken into consideration when designing the beam-columns
studied in Chapter 5.
Geometric limitations refer to percentage of steer- area
and to placement of vertical reinforcing bars and spacing of
horizontal, rectangular ties. The ACI code 1imits the
amount of verticar reinforcing bars to a minimum of l-.0 and
a maximum of 8"0 percent of the net concrete area. No limit
is indicated for the structural- steel core. The csA code
reguires that L.o to 8.0 percent of the gross area be verti-
cal reínforcing bars" The maximum percentage of all steel
lstflrnl-lrr¡'l .anÁ rain€nr¡ina\ v¿¡¡y, i o
¿e 1 iai.Þa'l
r¿¡r¡¿9çs .þa ./ì xaç¡aqr- ^€ åL^
9v av À/E!ug¡¡t, ul_ t-I¡g
gross area for the csA code" These limitations vrere also
12B
included in the Monte Carlo analysis described in Chapter 5"

Requirements for spacing of verticaL reinforcing bars and
lateral hoops is sirnilar for both codes and will not be
discussed further here.
L29
4 PROBÃBILTTY MODELS OF BåSTC VÃRTå,BLES

The strength variation of a composite beam-column is due
to the individuaL variatj-ons of strength and size of all
elements of the beam-column. The statístics (probability
models) of the basic variables were compiled and used in the
Monte Carlo simulations (Chapter 5) " Probability distrib-
utions of al-l variabLes hrere derived as part of this study
from data available in the literature or h¡ere taken from
previous studies" No new testing of materials was per-
formed.
Two distinct probability distributions were used for
each basic variable. One probability distribution described
the in-batch variation of the vari-able and the other one
described the globaI variation" rn-batch variations repre-
sent expected variations within one production run of a man-
ufacturer. Hence, in-batch probability distributions hrere
used to cal-cul-ate the in-batch variation (V ;"_ao,"n) of the
strength of Laborat,ory specimens as discussed in Section
2"IO. G1obal probability distributions represent expected
variations in material strength and dimensions due to dif-
ferences in manufacturi-ng practices between manufacturers of
an item" They al-so represent variations in construction
practice between different contractors" The global proba-
bility distributions were used in Monte carlo simul-ations to
calcul-ate the strengÈh variation of the composite
130
beam-column on an industry wide scal_e (Chapter 5) " The

global variations derived were based on North American data
where possible.
Twenty-four basic variables affecting the strength of a
composite beam-corumn were accounted for. The variables
describe material stress-strain relations, size and geornetry
of individual components of the column cross-section, and
overall corumn dimensions and geometry" The basic variabres
specific to concrete, structural steel, reinforcing bars and
column dimensions are discussed in this chapter" Tabre 4.r
summarj-zes the in-batch probability distributions used in
section 2"ro. Tabre 4.2 summarizes the global probabirity
distributions used for the study described in chapter 5.
4"X CONCRETE
Three mechanical properties of concrete most affecting

the strength of composite beam-columns are compressive
st,rength, tensiÌe strength (rnodurus of rupture) and modulus
of erasticity" Descriptions of the probabirity distrib-
utions of these properties have been presented by Mirza et
al. (L979c) and were used in this study. summaries of these
probability distributions are presented in this section.
These probabirity rnoders have been used by Mirza and MacGre-
gor (1982 and l-989) for strength variation studies of rein-
forced concrete members.
131
TabIe 4.I - In-batch Variations of Basic Variables'k
S tanda rd Coefficient
Prope r t y Mean Deviat ion of Variation
========== ================================================ ========= === =
Concrete in Structure (loaded t.o failure in 2 hr.)
Compressive Strength (psi) 3320 166 0.05
Modulus of Rupture (psi) 462 23 0.05
Modulus of Elasticity (ksi ) 3084 108 0.035
Struct.ural Steel
Modulus of ELasticity (ksi) 29000 290 0.01
Static YieId Srrength of tr'Ieb (psi) 53360 ro61 0.02
Initial Tangent Modulus
of Strain Hardening Curve (ksi) 600 150 o.25
t=======
Reinforcing Steel
Modulus of Elasticity (ksi) 29200 292 0.01
Static Yield Strength (psi) 561r5 1 403 0.025
Strain at St.art of
S t ra in-ha rden ing 0. 015 0.0015 0. 10
Ultirnate Strain 0. 15 0.015 0.10
;;;;;;;;=;;=il;;ilil;;il;=;;;=il;;il;ìil;;==================
Cross Section depth and width 0.0 0.08
ConcreE.e Cover t.o Lateral Hoops 0.o 0.055
ìkNotes: ( i ) Data for in-batch variations of basic variables shown are

only for Column D8-90 taken from Table 2.3. Orher columns
used in determination of in-batch variations IColumns
LH-100-C, LH-1OO-8, and RC1 (TabLe 2"3)] used rhe same
coefficients of variation as shoçm for concreLe, structural
steel and reinforcing steel and the same standard deviation
for deviation of dimensions from specified values. Mean
values for the basic variables of those columns vrere
determined from the tesi data and were different from those
shown in Ehis tabIe.
(2) ALI probability dist.ributions were assumed Lo be normal
except for statÍc yield strength of sLructural steel and
reinforcing st.eel where modified Iognormal probability
distribut,ions were used with lower boundaries of 0.75 times
the static yield strength.
1000 psi = 1 ksi = 6.895 MPa; 1 in" = 25.4 mm.
L32
TabIe 4.2 - Overall Variations of Basic Variables
======:========================================================
Standard Coefficient
Prop erty Mean Deviation of Variation
==================================================== ====================
Concrete in Structure (Loaded to Failure in t hr.)
Compressive Strength (psi)
-Average Quality ConLroI
f.' = 4000 psi 3388 596 o.tt6
f? = psi
OoOO 4640 8L7 o.176
-EÌce11ent. Quality Control
ft = 60O0 osi 4640 63L 0. 136
uoSulus of hrrptu.e (psi)
-Average Quality ConLroI
f' = 4000 osi 445 9l 0.218
r1 = Oooo osi 523 Lr4 0. 218
-nÇcellent iuality Control
f'-
6000 osi
'elasticiEy (ksi) 523 111 O.zIt
MoSulus of
-Average Quality Control
Í' - 4oOO osi 3260 3BB 0.119
r1 - Oooo osi 38oo 452 0.119
-elcellent þuality Control
c' = 6000 osi
ç.' 3800 399 0.105
================================================================= =======
Structura I Steel
Modulus of Elasticiry (ksi) 29000 580 0.02
Static Yield Scrength
of (psi)
T,leb f_____
f., = 3ðöòo psi 39240 331 5 0. 086
frtt' = 4400O osi 47 960 41.25 0.086
f r. = 5OO00 psi 54500 4687 0.086
Scatið yield strengE.h
of flanee - f
-rï?i*"."Y5 0.95 f ys¡s n/ a'¡ n/ a^
s rár i. Êr"r,gtn
ofweb-f 1.5 f n f ¿-:' n/ a';
sraric ulrimHES srrengrh Ytn¡s
of flanse
- - f 1.5 f yts nf a-;' n/¿:k
s r r" i" iuå I
".-õ..r
tra in-Hardening
S 0.017 0.004 o.24
Initial Tangent Modulus
of Strain Hardening Curve (ksi) 600 150 0. 25
Residual Stresses (psi)
- IÎ10 x 54 (iÎ250 x B0)
at. flange tip -l$Jf$>tx 27 86 0. 15
ac flange-web juncÈure 1 2089 88 25 0. 73
- w1O x tIZ (1,¡250 x 167)
a t f lange ip L, -lÇlll;r:'r 2897 0. 15
at flange-web junct.ure t6240 11855 0. 73
133
Table 4.2 (continued)
Ratio of Actual to Nominal Dimensions
- secLion depth 1.0 0.0 0.0
- flange width 1.005 0.0136 0.0135
- flange chickness o.916 0.0407 0.041 7
- web thickness 1.016l 0.039 0. 03B
========== ========== =========== ======== = ===== == ==== === === == ==== =- - - - - - --
Reinforcing Steel
Modulus of Elast.iciÈy (ksi) 29000 957 0.033
Static YieId SErengrh -L c
f.-_ = 60,000 psi yrs
6 6800 5520 0.083
StaÈf'c Ultimate Srrengrh -r c 1.55 f yrs n/¿:r n/a*
SErain at Start of urs
S t ra in-ha rden ing 0.015 0.004 0 .267
Ultímate Strain o. 15 0. 03 o.2
LengEh ( in. 0.0 o .67

Cross Section depLh (in. ) +0.0625 o .25
Cross Sect ion widrh ( in. ¡ +O .0625 o.25
Concrete Cover to lateral hoops (in.¡ +0.33 0. 166
Spacing of lateral hoops (in. ¡ 0.0 0. 53
value of this variable is assumed dependent on Lhe value of another

variable.
:k:'r (-) indicates compressive sÈress
Notes:(1) A1l columns had nominal cross-section of. 20 x 2o in. with 1.5
in- clear concrete cover to lateral hoops. Lateral hoop
nominal spacinþ was 10 in.
(2) YÍeld scrength of reinforcing bars r.,râs assumed to follow a
beta probability distribution, whereas the yield strength of
structural steeI, the ratio of acc.ual r.o specif ied f Iange
wÍdth, and the ratio of actual Eo nominal web thickness were
represented by modified lognormal probability distributions
with lower limits of 0.75 times the specified yield srress,
0.884, and 0.813, respectÍvely. All other variables were
assumed to foIIow normal probabiliry distributions.
(3) 1OOO psi = l ksi - 6.895 Mpa; 1 in. = 25.4 mm.
L34
4"åol ComÞressive Strenath

The in-situ compressive strength of concrete differs
from the specified strength due to factors inctuding varia-
tions in materials, mixing, placing and curing techniques,
guality cont,rol and rate of loading (Mirza et al " 1-979c) "
the rnean value of compressive strength of in-situ concrete

is lower than that indicated by standard cylinder tests"
This is recognized in both ACI 318-83 and CSA CAN3-A23.3-MB4
as only 85 percent of the specified 28-day cylinder strength
is allowed for design use" To relate the specified 28-day
cylinder strength to the mean 28-day in-situ strength loaded
at a similar rate (35 psi or 0"241 MPa per second), Mirza et
al. (1-979c) proposed the expression given in Equation 4.1.
I
J cst¡35 o.67s f, r, t00 5 I-l5/,'pst (4.1)
o"675f
"
7.58 s t.l5/,'MPa
Rate of loading affects the strength of concrete. Mirza

et al " (1-979c) studied experimental tests by others in order
to relate the compressive strength of concrete in a struc-
ture at a given rate of loading to concrete in a structure
Loaded at the standard cyrinder test rate" This rel-ation is
given in Eguation 4"2"
135
7,,t,n = 7."r,csto.89(1 + O.0BlogroR)l ps¿ (4.2)
7cs¿,ssto.gg(l + o.oglogrol45R)l Mpa
The loading rate R was cal-culated by dividing the mean con-

cret,e strength cal-culated by Eguation 4 " 1_ by the loading
time to failure.
The coefficient of variation of the in-situ compressive
strength, V""r,r, h¡as calculated by Eguation 4"3 (Mirza et al_.
r.979c) "
V?.r,rr = V?r"o¿ + V?^-r¿tu + Vtr (4.3)
V
",n.,¿
represents the variation in the relatj-on between real
cylinder strength and the specified design strength . V i.n_situ
represents the variation in the relation between in-situ
strength and real- cylinder strength. V p represents the
variation in the relation between concrete load.ed. at R
psi/sec and concrete loaded at 35 psi/sec (O.241_ Mpa/sec) .
Jones and Richart (1935) found only a smarl dispersion in
concrete strength due to rate of loading effects" Al-len
(r97o) suggested that dispersion of concrete strength is
unaffected by the speed of testing. Therefore, I/p can be
considered negrigible (Mirza et ar " r979c) . v in-s¿¿u has been
assumed to equal 1-0 percent by Mírza et aI " (r979c) based on
136
the v¡ork by Davis (L976) and was used in this study as wel-l-"
The strength of concrete in test cylinders varies due to
the real variations, V ",not, and in-batch variations, V,r-o.,,rn.
This relation is shown in Equation 4"4.
V'."", = V?r"o¿ + V?,-oo,"n Ø.4)

Vrn-oor"n tñas estimated as 4 percent by Mirza et aI" (1979c)
and was used in this study. This roughly corresponds to the
suggesti-ons of Àmerican concrete rnstitute committee zr4
(ACI l-965) which reconmend V rn_aor"¿ values of 4-5 percent for
good guality control-, 5-6 percent for average quality and
above 6 percent for poor quality control.
By cornbining equations 4.3 and 4.4 the variation of the
in-situ compressive strength is given by Eguation 4.5.
V'"r,rr V2."r, V?n-aot"n + V?n-rit, + V'r (4.5)
substituting a value of 4 percent for v in-batcht 10 percent for

Vin-situt and zero fot V r, as d.iscussed above, into Eguation
4"5 yields Eguation 4"62
V2.^
' csl¡R
v2
' ccy I O.O42 + O.lO2 (4.6)
L37
The variation in the strength of test cyli-nders V,.y¿
v/as found to be dependant on the degree of guality control

for a particuÌar job" Cornbining data from a variety of
sources , l{}-rza et aI . (I979c) , found that for specified con-
crete strengths up to 4000 psi (27 "6 MPa), the average V,,",
was roughly constant with values of 10, L5 and 20 percent
for excellent, average and poor guality control, respec-
tively. For specified compressj-ve strengths greater than
4OOO psi, V""", decreases due to the higher degree of care
used in the manufacture of higher strength concrete (Mirza

et al . 1979c) " In the Monte Carlo simulatj-ons described j_n
Chapter 5, quality control of concrete was assumed to be
average for specified design strengths of 40OO psi (V",vL =
i-5 percent) and excellent (except v¡here noted) for 6000 psi
concrete (V ""y¿ = Lo percent) " Data studied by Mirza et aI.
(1,979c) suggested a normal distribution for the compressive
strength of in-situ concrete which was also assumed for this
study.
4.1"2 Tensile Strength
Mirza et al. (L979c) studied data from the literature to
estabLish the relationship between compressive cylinder
strength and tensil-e strength (rnodulus of rupture) " The
relationship given by Eguation 4.7 was found to most closely
fit the regression line caLculated from the data (Mirza et
al" L979c) "
138
f, 8.3.8 ps¿ (4.7)
o.68s J Í MPa
"
This value is only slightly larger than the value suggested

by ACI 3l-8-83 for modulus of rupture shown in Equation 4.8.
f, = 7.sJf ps¿ (4.8)
o.6nJf, MPa
The in-situ tensile strength of concrete may differ from

control specirnens due to effects of volume, rate of loading
and effect of concrete being cast-in-situ and not into a
specified control test fonn" Bolotin (1"969) found Lhat the
vorume of the test specimen did not significantry effect the
minimum tensil-e strength of the concrete, although mean
strengths h/ere" Since the ninimum values are unaffected,
the vol-ume effect can be neglected for understrength studj-es
(Mirza et aI " 1979c) " vtright (r9s2) showed that the tensit-e
strength of concrete increased with increasing rate of road-
ing" McNeely and Lash (t-963) suggested a logarithmic rel_a-
tion between the tensile strength and the rate of stress
application. Data on the effect of in-situ casting as
opposed to control specimen castj_ng was not found by Mirza
et aI " (I979c) and they chose to assume the effect as negli-
gibre. using the resul-ts of their anarysj-s of data and. the
139
relat,ion suggested by McNeeIy and Lash, Ní.írza et aI. (I979c)

proposed that the relation given in Equation 4"9 be used to
cal-cuLate the mean value of the rnodulus of rupture for a
given loading rate"
I
J rstr R 8.3/:í:.rtO.e6(t + O.tItog,oR)l ps¿ (4.s)
8.371í:.rto.eó(t + o.tltosrot4sR)l Mpa
Calculating tensj-le strength from a relation based on

a calculation of compressive strength results in consider-
ably larg'er dispersions than for compressive strength alone.
The total variation of the tensile strength calculated from
Eguation 4.9 combines the variations due to the cal-culation
of the compressive strength as discussed in the previous
section and the variabil-ity of the ratio of observed to cal-
cuLated (Equation 4"7) tensile strength. Mirza et al.
(1"979c) caLculated the coefficient of variation of the ratio
of modulus of rupture caLculated by Eguation 4.7 to actual
tested modulus of rupture calculated to be 20 percent" The
total variation may, therefore, be expressed as shown in
Eguation 4"10"
t¡2
/ cstr35
+ o.22 2 v ?,,,* (4. t 0)
140
Vcstr3s v¡as taken equal to V ¿5¿¡p âfid calcul-ated from Equa-

tion 4"6 since the effect of loading rate on the coefficient
of variation was assumed negligible" Substituting Equation
4"6 into Eguation 4.1-0 yields Equation 4"11.
vccvt
f f ¿
v2.^
Y rstrR + o.o42r
vtvaLL >v2
- Y cslrR (4.r t)
4
The probability distribution

of the tensile strength was
assumed to be normal arthough some deviation from normaJ-ity
may be expected (Mirza et aI" L979c) "
4 " " 3 Modulus Of Elast,icitv
X.
Mirza et a1. (L979c) studied data from 139 standard cyl-

inder tests of normal weight concrete from the university of
IL1inois" This data provided measurements of cylinder
strength and initial tangent moduLus of elasticity j_n com-
pression. The relat,ionship of compressive cyrinder strength
t,o initial tangent modulus was found to have a high degree
of correlation. The relationship given in Equation 4.12 v/as
proposed for the mean val-ue of elastic modul-us at a roading
rate of 35 psi/sec (O.241 Mpa/sec) "
ãci¡s = 60 ,4OOJ f" ps¿ (4.t2)

= 5,OI6lf , MPa
141
This relatj-onship yields slightly higher values for the tan-

gent, elastj-c modulus than the ACf 318-83 recommendation
shown in Equation 4.13.
E, = 57,OOO\n ps¿ (4.l3)

= 4,734JT" Mpa
The ratio of modulus of elasticity calculated by Equa-

tion 4"I2 and the observed el-astic modul-us from the test
data was found to have a mean value of 1-"0 and a coefficient
of variation of 8 percent. A normal probability distribu-
tion was found to adeguately approximate the above-noted
ratj-os (Mirza et al " I979c) .
The effect of the rate of loading on the elastic modulus
sras studied by Allen (l-970) " Eguation 4.I4 r,üas proposed to
relate the elastic modulus at any loading rate to the stan-
dard cylinder test rate of 35 psi/sec (O "24L Mpa/sec) .
E,n =( I.ló 0.0Blog,of )F.r, (4.14)

where t = loading duration in seconds.
As Eguation 4"L4 indicates, an increase in the loading time

results in a softening of the concrete to the peak stress.
Conbining the results of the Illinois test data and A1len,
t42
Mirza et. al- " (L979c) proposed the foLlowing eguation for the
mean value of initial tangent moduÌus of elasticity of con-
crete in structure"
F.¡"¿,n óo,4oo1'.'",',""(l.ló o.oBlog,of) ps¿ (4.1s)
s, o r 67::,?""( t .l ó o.oB ros ,o r) M pa
The coefficient of variation of the modulus of elastic-

ity calculated by Eguation 4. t-5 must include al-l of the
variations associated with the compressive strength as well
as the variations inherent to Equation 4.I2" Combining
these factors, the coefficient of variation of the initiar-
tangent modulus of in-situ normâI weight concrete can be
calculated by Equation 4.1-6 (Mirza et al " 1,979c) "
v?"'"" +
' cis¿¡R =
v?,-,-,
4
o.og2 (4.I ó)
Substituting the value of Vz",,,ss from Equation 4.6 into

Eguation 4.L6 related the in-situ coefficient of variation
of iniÈial tangent modulus in compression to the coefficj-ent
of variation of test cylinders as shown in Equation 4"I7.
¿
t./ ^
' cisr¡R
v?..,
uLtL + 0.0085 (4 .IT)
4
1Â1
IIJ
In this study, Equation 4.15 and 4"17 r¡¡ere used to com-

pute the mean value and the coefficient of variation of the
modulus of elasticity of in-situ concrete" The probability
distribution of the modulus of elasticity was assumed to
fo]low a normal- distribution after ltiria et al- . (t979c) .
Mirza et al- " (I979c) found little data on the rnodulus of
elasticity of concrete in tension, but what was found showed
little difference between compressive and tensile el-astic
modul-i" They concl-uded that the tensile and compressive
elastic moduli may be assumed to have equal magnj-tude.
4.2 STRUCTURAL STEETJ
Variations in the mechanical and geometric propertj-es of
the rolled steel- section effect the variation of the over-
aL1 strength of the composite beam-column. The mechanical-
properties that define the stress-strain curve of structural-
steel- described in Section 2"7.1- are the modulus of elastic-
ity, the yield stress, the strain at the start of strain
hardening, the initial tangent slope of the strain hardening
curve and the ultinate stress. The yield strain and the
ul-timate strain can be calculated from thè above properties
and the assumptions described in Section 2.7 "L" Variations
in residual stresses in the rolled steel section al-so influ-
ence the overall strength variation of the beam-column"
Variation of the dimensions of the depth, flange width,
flange thickness and web thickness affect the cross sec-
tional area and the stiffness (moment of inertia) of the
L44
steel section" To properly model the basic variables noted

above for use in the theoretical program (Chapter 2) and for
the Monte Carlo analysis (Chapter 5), the mean val_ue, coef-
ficient of variation (or standard deviation) and the type of
probability distribution were defined for each basic
variable. These definitions \,{ere taken from the li-terature
or derived from data existing in the literature. No new
test data was generated in this study" A description of the
statistical distributions used for each variable is given
below"
4.2"1 Modulus of ELasticity
Galambos and Ravindra (L978) studied existing' experimen-
ta1 dat,a of mechanical properties of rotled structural steel
sections to determine the statistical properties. They
reconmended that a val_ue of Zg.OOO,0oo psi (zoo,o0O Mpa) be
used as the mean value for the modulus of elasticity for
structurar steer shapes and the coefficient of varj-ation be
taken as 6 percent. Kennedy and Gad Aly (1980) considered
only the data from North American sources conpiled. earlier
by Galambos and Ravindra and recommended that a value of 1.9
percent be used for the coeffícient of variation" Bjorhovde
(1972) found very small variations of the modurus of elas-
ticity in the data he studied (0"L to 0.5 percent) and he,
therefore, considered the modulus of elast,icity to be
constant "
r45
The rnean values and the coeffi-cients of variation of the

data collected by Galambos and Ravindra (1"978) is shown in
Tab1e 4"3 (Rows 1 to 6) " The data in Rows 1 to 5 in the
same table was collected from North American sources and in
Row 6 from a European source" The weighted means and coef-
ficients of variation lrere calculated and are presented for
all- of the data (Row 7) and for the data from only the North
American sources (row 8) "
Based on above-noted discussions and Table 4.3 (Row B),
it was decided to use a value of Zg,OOOTOOO psi (2OO,OOO
MPa) as the mean vaLue for the modulus of erasticity with a
coefficient of vari-ation of 2 percent. None of the authors
referenced above commented on the shape of the probability
distribution curve for modul-us of elasticity of structural
steer. Mirza et aL " (L979b) assumed a normar distribution
for the modulus of elasticity of reinforcj-ng stee1. since
no other data was found, a normal probabitity distribution
was used for the modul-us of elasticity of structural steer.
4o2n2 Yield Strength
The yield strength of a rolled structurar steel shape is
dependant on the rate of loading, the location of the spe-
cific element on the cross-section and the thickness of the
material" The infruence of each of these three factors is
discussed below.
L46
1'able 4.3 - Elasric Modulus of Strucrural SreeI
No. of Mean Coefficient

Row Reference Test Va lue of
No Sp ec imens (ksi) Variation
(1) (')\ (3) (4) (s)
1l
I
Lyse & Keyser (1934) 7 29360 0. 010

2l Rao et al. (L964) 56 29437 0.014
3l Julian (1957) 67 29540 0.010
4 Julian (i957) 67 29550 0.010
5 Johnsron and OpÍla (1941) 50 2971 4 0.038
6 Tall and Alpsten (1969) 94 31200 0.060
7 341 3 0013 O.O/.+4
8 ;k;k 247 29562 0.020
Row 7 combines all data from Rows I to 6 inclusive.

ìl:"- Ro!ù' 8 combines data f rom Nort.h American sources only (Rows 1-5 ) .
Notesr (t) gtl data from tension tests except Row 4 which is for
compression tests.
(2) 1 ksi = 6.895 MPa.
L47
Generally, the strength reported by steel mi1ls is

either the upper (f or lower (f yierd strength as shown
"") ",)
on Figure 4"L" The strain rates (about I'OOO micro in. per
in" per second) used for mil1 tests are significantly higher
than those expected under normal loading conditions (Beedre
and TalI l-960). under normal loading applications, the road
may be applied at a very low rate" Therefore, the r¡staticrl
yield stress, defined as the yield stress at a zero rate of
strain /"", is the yield stress of int,erest (Figure 4"1).
The static yield stress /r" is measured by halting the plas-
tj-c strain until the stress drops from the d.ynamic yierd
stress (f ,o) to stabil-ize at the static yield stress as
shown in Figure 4"t. Rao et aL. (]-964) proposed that the
static yield stress may be calculated froin the relationship:
fr" fro (3.2 + 0.001e) ksi (4. r 8)
fva (22.O7 + 0.0Ole) MPa
where e is measured in micro in. per in" per sec. or micro

nm per nm per sec.
Kennedy and Gad Aly (i-990) used Eguation 4"i-B
to calcu-
late the difference between the dynamic and static yierd
stress for the csA standard c4o.2o-L974 maximum test strain
rate of I/16 in. per in" per minute (l_O4O micro
in"/ in./sec.) " The dynanic yierd stress was calcul-ated to
148
start of
strain hardening
f yu upper yield stress

f'yl Iower yield stress
c
I.
yo dynamic yield stress
f ys static yield stress
Figure 4.I Def initions of yield stress of steel-

L49
be 4"2 ksi (29 Mpa) greater than the static yield. stress.
As shown in Figure 4.L, the lower yield stress (f l_ies
",)
between the dynamic and static yierd stress l-evels. Kennedy
and Gad Àly (1980) assumed that the static yield stress was
2 ksi (l-3.8 MPa) less than the 1ower yield point"
Beedle and Tall- (l-960) found that the original location
of the t,est coupon on the roll-ed wide flange shape affected
the yield strength of the specimen. coupons cut from the
web were found to have yield strengths greater than coupons
cut from the flange. Generally, mirl tests are performed on
web specimens (Beedle and Tall 1960). Kennedy and Gad Aly
(L980) attributed the higher strength of the web to
increased work hardening during the roJ-1ing process due to
the smaller thickness of webs.
Alpsten (1,972) found a tendency for thicker plates to
have lower yieJ-d strengths " He attrj-buted this to a coarser
grain structure due to a longer cooling period. He also
commented that during the manufacture of thick plates, steer
producers may alter the chemical composition to account for
lower strengths" rt shourd be noted that the plate sizes
investigated by AJ-psten exceeded one inch (25 " 4 nn) in
thickness" Kennedy and Gad Aly (L980) neglected any varia-
tion in yierd strength directly due to component plate
thickness since the data they anaryzed. included this
variat'ion, The same assumption was made in this study.
150
MilL test data is based only on tensile tests. Horarever,

it has been found that there is no difference in static
yierd stress level for tension or compression tests (Galam-
bos and Ravindra 1978). Hence, the properties derived from
tensile t,ests have been assurned to apply to the compressive
loading case as well in this study.
4"2o2.1 web yieLd strength - Tension tests on r¡¡eb coupons
from North American rolled shapes T^rere summarized by Lay
(l-965), American fron and steel rnstitute (rg72) and Kennedy
and Gad A1y (1980). The mean val-ues and coefficients of
variation of nilI tests tabulated by Lay and Àmerican rron
and Steel Institute for steel with a specified yield
strengths of 33 ksi (228 Mpa) are shown in columns 2 and 3
of Table 4"4, respectively. These nilI tests report dynarnic
yield stress and were also used by Galambos and Ravindra
(Lg78) for statistical analysis of web yierd strength. Ken-
nedy and Gad Aly reported mill- test measurements of lower
yield point stress for canadian steel grade csA G40.21--44w.
The mean value and coefficient of variation of this data are
given in Column 4 of Table 4"4"
To determine the mean value and coefficient of variation
of the ratio of web static yield st,ress to specified yield
stress, the data reported by Lay (1965), American rron and
steel rnst.itute (1972) and Kennedy and Gad ÀIy (1980) were
q..FeÈic.#ìna'l'lr¡ nrrl la¡{ }næa#lanr E.i--+- .¡-t^^ *^^-
¡/s¿Àvs uvYçe¡rç!. r'f!Ðut t-¡ls ttl,EC¡.tr ,-^r,-^
VclJ_t¡ts:, -À--
5LCtl.I-
dard deviation and coefficient of variation of the reported

151
TabIe 4.4 - f¡ieb Yield Stress Measurements
Lay American Iron and Kennedy and

Prope r ty (1965) Steel Institute Gad AIy
( 197 2) (1e80)
(1) (2) (3) (4)
No. of Tests 3t 94 3t24 4501
Specified yield srress 33 33 44

f (ksi)
v
Measured VaIues
-Mean yield stress 40"0 39 .4 50. 6

f yt{ (ksi)
V 0.09 0. 08 0 .064
o (ksi) 3.6 ? ,c 3 .24
-Mean static yield 36"0 35 .4 48 .6
stress f -___-
yurs
(ks i )
\¡
0. 10 0.089 0.067
o (ksi) 3.6 3.15 J " ¿-t
-Mean static yle Id 1.091 1 .073 1 .105

stress ratio ryTntslt y
0. 10 0.089 0.067
o 0"109 0.095 o.074
Note: 1 ksi = 6.895 MPa.

r52
mill strengths were calcuLated for the three sets of data as

shown in Columns 2, 3 and 4 of Tabl-e 4"4" The mean static
yÍeId stress (7"") was then cornputed from Eguation 4.18
assuming a strain rate e of 800 micro in./in./sec. (BOO
micro mm/mm/sec" ) for the data given in Columns 2 and 3 of

Table 4"4. The strain rate assumed was previously used by
Galambos and Ravindra (1-978) when they studied the same data
and was, therefore, considered applicable in this study.
For the data shown in Column 4 of the same table, 7"" r.=
taken as 2 ksi (L3"8 MPa) lower than the milI test mean
lower yield stress as previously assumed by Kennedy and Gad
Aly (l-980). The new coefficients of variation hrere calcu-
lated based on 7r". The mean value of the ratio of static
to specified yield stress (mean stress ratio) v/as then
calculated by dividing the mean static yield stress value
for each data set by the specified yield stress for that
data set. Multiplying the mean stress ratios by the coeffi-
cients of variation gave the standard deviatj-on of the mean
stress rati-o for each data set.
The weighted mean stress ratio was calculated by combin-
tng the data from alL three data sets using Eguation 4"L9
and was calculated to be 1-"092"
153
I ¡u,*,
i- I
X k
(4 .t e)
It'
t- I
The weight,ed standard deviation of the stress ratio was cal--

culated by Eguation 4"20 to be 0"094"
o: (4.20)
The ratio of this weighted standard deviation and the mean

stress ratio gave the coefficient of variation of the ratio
of web static yield stress to specified yield stress and was
calculated to be 0.086" In the Monte Carlo analysis
described in Chapter 5, the mean web static yield stress was
assumed as l-.09 times the specified yield stress with a
coefficient of varj-ation of 8.6 percent. Galambos and
Ravindra reconmended that the mean vaLue be taken as l_.1O
times the specified value with a coefficient of varÍation of
L1- percent, while Kennedy and Gad Aly proposed that the
mean strength of the web was L"l-1 times the specified value
with a coefficient of variation of 6"5 percent. These val_-
ues are somewhat different from those used for this study,
L54
4"2"2"2 Flange yíe1d st,rengt,h - Beedle and TaIl (1960)

reported that flange static yield stress was 4 Eo 7 percent
lower than that for the web. Galambos and Ravindra (1-978)
recolnmended that the mean value for the static yield
strength of the flange be taken as 1"05 times the specified
yield stress with a coefficient of variation of 10 percent.
Kennedy and Gad AIy (l-980), based on the report of Beedl_e
and Tall (1960), assumed that the flange static yield stress
r¡ras 95 percent of the web static yield stress" The assump-
tions made by Kennedy and Gad AIy were also used in this
study. The flange strength was assumed to be directly in
proportion with the web strength and no further variation
t/as applied.
4"2.2"3 Probabílitv distríþution of víeld strencrt,h - Fre-
quency histograms of the yield strength of test specimens
v/ere reported as positively skewed (Alpsten 1972). This is
reasonable since any heat (manufacturing run) of steel_ not
meeting the minimum specified criteria will be rejected,
truncating the lower end of the strength probability distri-
bution" Since the freguency distribution is not symmetri-
caJ-, a normal distribution is not valid. Alpsten (L972)
reconrmended a modified lognormal distribution for yield
strength of structural steel"
To define a distributj-on for the yield strength, a mod.i-
f i cd 'l ocnnrrna'l rìi q.'l-ri lrrr#ì r¡rãe €ì ++ô/q .Fa .l-Ìra t.¡ah Æa.t--
^h trvs uquq ÀJÀv
vided by Kennedy and Gad Aly (Lggo) for csA G4o"2L Grade 44w
155
steel" The histogram and statistical parameters for the

data are given in Figure 4"2" A modified lognormal distri-
bution was derived using the mean val-ue and standard d.evi-
ation of the data and Eguations 4.2L, 4.22 and 4"23"
fflo
r, t x-
;IoUrol J _X I ,
o
)
4
(4.2t)
¿l (x o) + o?
d ro 0.4342945 log
'o
(4.22)
o.4342945 rI log 'o(x- X")- i,o t\

PDF
(x-X.10,offi .,.e (- 2l 0ro I
(4.23)
Eguations 4.2r, 4"22 and 4.23 give respectively the mean

value, standard deviation and cumulative frequency of a
modified lognorrnal distribution. Lower boundary values (x,)
of 36, 38 and 40 ksi (248 "3 | 262 , and 27s.9 Mpa) hrere tested
for the data reported by Kennedy and Gad Ary. The best fit
to the data was with a lower boundary of 36 ksi as shown in
Figure 4 "3 "
t56
Class %of
Interval- Measurenents
0. 40
42-44 0"00
U
..1
44-46 9 "56
Ð
a'l
46-48 15. 73
rd
tt
48-50 2I"19
t¡r 50-52 20.59
r-l
d
0. 30 52-54 16.04
!
..1 54-56 9.25
O
o
56-58 4.53
O 58-60 2.06
0.20 60-62 0.80
62-64 0 .16
0) 64-66 0.09
tt
(.)
tl CSA G40.21 Grade 44W
H
0. 10
Nomi¡al F__
v
= 44 ksi (303.4 MPa)
Total No. of Measurenents = 4507
44 46 60 62 64 66
Lower Yield Streng'th (Fr1) of Web (ksi)
F\/ . = 50.61 ksi = 348.9

I
MPa
VE, = 0.064
'yf
Figure 4"2 - Variation of Web Lor,rer Yield Strengrth (Fo1)

(Kennedy and Gad a.Iy 1980)
L57
oo oo I00 t
99 .95
99.90
99. B0
Modified Lognornral Probability Djstrjbution
99.50
¡rean value = 50.61 ksi (348.9 Mpa)
99.00 coefficient of variation = 0.064
98-00 lower bor¡ndarlz of dj-strjbution
= 36 ksi (248.3 MPa)
o
95.00
JJ 90"00
c")
'lr
o
È
70.00
c.)
j
tr
CJ
50"00
tr
h
o)
.,1 30.00
Ð
rd
-l
c)
10 .00
5.00
2"00
1.00
0" 50
0. 20
0 .10
0.05
0. 0I
vnc
36 40 44" " 48 56 60 b4
Web Lower Yiel-d Strencrth (For)
Figure 4"3 - cumul-ative Frequency of web r-ower yiel_d strength (Fyr)

l5B
To adjust the web data in Figure 4 "2 from mil-I yield

stress to static yield stress the mean values and the lower
boundary \,ùere reduced by 2 ksi (13"8 MPa) as noted pre-
viously" Thus, a lower boundary value (X.) of 34 ksi (234.s
MPa) r{ras calculated for the web static yÍeld stress or 77
percent of the specified yield strength. For simplicity,
however, a value of 75 percent of the specified yield
strength was taken as the l-ower boundary for the web. The
cumulative freguency of the web static yield strength r¡/as,
therefore, computed from Equation 4.23 with X" taken as O.7S
times the specified yield stress" The static yierd strength
of the flange was assumed as 95 percent of the web static
yield. strength at any point along the cumulati-ve frequency
curve.
4"2.3 Ul-timate Stenoth
Little data was found on the ul-timate strength of a
rolled structural steel shape. Alpsten (i.972) presented
data indicating the ratio of ultimate strength to yield
strength ranged from 1.36 to 1.89 with an average val-ue of
1"59 for 4l- samples taken from a single rolled section.
specifications for structural steer indicate minimum ul-ti-
mate tensile strength requirements ranging from r"2g tj-mes
the specified yierd strength for csA c4o.2L-M 35ow steel to
L"61 for ASTM 436 steel. For this study, the static ulti-
159
mate stress Ìras arbitrarily chosen to be 1.5 times the

static yield stress" No further variation was defined for
this parameter.
4nZu4 Straín åt Inít,iatíon Of Strain-Ifardeníng
Arpsten (1972) reported values for the strain at the end
of the prastic prateau (initiat,ion of strain-hardening) for
various grades of structural st,eeI. These values range from
1.1- percent for ASTM AsTz steel- to z"o percent for ASTM A36
steel" The averag'e value reconmended by Alpsten for ASTM
steers was L.72 percent" No information on the variability
of this straj-n was found. For this study, a mean var-ue of
o"0i-7 r¡ras chosen with a coefficient of variation of 26 per-
cent for the strain at. initiation of strain-hardening for
structurar steeL" Note the coefficient of vari-ation used
for structural steel is the same as that measured for rei_n-
forcing steel (Section 4"3"4) "
,4"2.5 Strain Hardening Modulus
The strain hardening modulus defines the initial tangent
slope of the strain-hardeni-ng portion of the structural-
steer stress-strain rerationship. Arpsten (L972) reported
values of 450 to 72o ksi (3,103 to 4,96s Mpa) for tension
and 7oo to 82o ksi (4 t828 to 5 ,6ss Mpa) for compression
strain-hardening modulus of ASTM steel-s" Arpsten noted that
there was very littre information on the strain hardening
properties of steers and that various definitions of the
strain hardening rnodulus coul-d result in large discrepancj_es
I60
in reported val-ues" Galambos and Ravindra (L979) studied

work by Doane at the Unj-versity of Texas at Austin (l_969)"
Doane measured the strain hardening modulus for ASTM Ã,7, A36
and 4441- steels" The tensile strain hardening modulus v/as
found to have a mean of 57O ksi (3,93i_ Mpa) and the corre-
sponding compressive vaLue rdas 670 ksi (4,621- MPa). Gal-am-
bos and Ravindra recommended a mean value of 600 ksi (41_39
MPa) and a coefficient of variation of 25 percent"
In this study it was assumed that the mean value of the
initiar tangent strain-hardening modurus was 6oo ksi and the
coefficient of variation was 25 percent. The same val-ue was
used for both compressive and tensile toading conditions"
4,2.6 Dimensional Varíations
Variations in the dimensions of the rol1ed steel shape
are discussed here to distinguish them from overarl column
dimensional variations" Alpsten (1-g7z) reported that mea-
surements of approximately 5000 rolled shapes from European
nil-Is showed very littLe variation in section depth and
flange width" More variation r,ras noticed in the flange and
web thicknesses. A tendency for flanges to be thinner and
webs to be thicker than the nominal di-mensions Ì¡¡as noted.
Kennedy and Gad AIy (1_980) reported measurements of
flange width, flange thickness and web thickness of wide
flange sections manufactured at canadian milIs. They used
161
these measurements along with assumptions based on code tol-

erance limits to evaluate the mean values and coeffícients
of variation of the steel section geometric properties.
fn the Monte Carlo simulations (Chapter 5), the dimen-
síons of the rolled steel- sectj-on were varied according to
statistics of the data reported by Kennedy and Gad AJ-y
(l-980) " The remaining geometric properties of the steel
section v/ere calculated using the simulated dimensions.
4"2" 6.3. Section depth - Kennedy and Gad AIy (1980) estirnated
the statistical parameters of the ratio of actual to nomj-nal-
secÈion depth by using the tolerance linits of CSA Standard
S16.1- for Stee1 Structures for Buildings" This code allows
the section depth to vary O.2 inch (4 mm) maximum from the
nominal vaLue" Kennedy and Gad Aly assumed that the mean
ratio of actual to nominal depth was equal to l-.0. The
extreme values of the ratio of actual to nominal depth v/ere
then calculated considering the upper and Lower permitted
tolerances" Six standard deviations r¡rere assumed to occur
between the upper and lower values" The coefficient of
variation v¡as calcul-ated by dividing the standard deviatj-on
by the mean va1ue" When this method was appJ-ied to a nomi-
nal 1-0 inch (254 mrn) deep section, the coefficient of varia-
tion was computed to be Less than 1 percent" Since this
coefficient of variation was so smal-I, it was assumed that
the actual depth of the roLled section was eq-ual t,o the
nominal depth and that there was no variation.
1_6 2
4"2"6"2 Flange nidt,h - Data regarding the rat.io of actual to

nominal flange width for 1248 samples was presented by Ken-
nedy and Gad Aly (1980). A histogram of the freguency dis-
tribution of the data from their report is reproduced in
Figure 4"4" The mean value of the ratio is 1"005 with a
coefficient of variation of 1.35 percent" The measured data
was plotted on a normal probability paper (Figure 4"5) "
Normal and rnodified lognorrnal probability distributions v/ere
plotted usJ-ng the calculated mean and coefficient of varia-
tion. The best fit
to the data, especially at the lower
tail, T¡Ias found to be a modif ied lognormal- distribution with
lower boundary of 0"88 as shown in Figure 4"5.

Based on Figure 4.5, a rnodified lognormal distribution
v¡as assumed for the ratio of actual to nominal flange
tdidth" The lower boundary of the ratio was set at 0.98.
The mean value was taken to be 1.005 with a coefficient of
variation 1-.35 percent as calculated by Kennedy and Gad Aty
( r_e8o) "
4"2"6"3 Flange thickness - Kennedy and Gad AIy (1980)

reported data on 2768 measurements of the ratio of actual
to nominal flange thickness. A histogram of the freguency
distribution is shown in Figure 4"6. The mean val-ue of the
ratio was o"976 with a coefficient of variati-on of 4"r7 per-
cent" The data was also plotted on a normal probability
paper along with a normal distribution using the reported
statisÈical parameters (Figure 4.7) " Figure 4"7 indicates
163
Class Zof
fnterva]- Measure¡re,nts
0 .40
0"95 - 0.96 0.00
0"96 - 0"97 0.24
.9
+J
0"97 - 0"98 1.65
O
rd
0"98 - 0"99 8.47
0.30 0.99 - 1"00 32.93
Eq
rl I"00 - 1.01 25.32

q 1.01 - 1.02 12.41
.=
rì 1.02 - 1.03 10.00
o)
Õ
1.03 - 1.04 4.59
o
1.04 - 1.05 1. 88
"20 1.05 -
U
1.06 I.29
0)
1.06 - 1.07 I.18
r-1
ftl
0)
! Total- No" of Measr-lrenents = I24B
F¡t 0.10
0.96 0.98 1.00 I-02 1.04 1.06
Ratio of Measr:red / Nomjnal Flange Width
Mean value of ratio = 1.005

Coefficient of Variation of ratio = 0"0135
Figiurre 4"4 - Varjation of Ratio of }4easured / Nomj¡al

Flange Width of W Sections
(Kennedy and Gad Aly 1980)
L64
r00 t
99.99
99 -95
99.90
99.80
Modjfied Lognornral
Probab iÌ.ity Di-stri-bution
99.50
nean value = L.005
99.00 coefficient of variation = 0.0135
98.00 louer boundarlz of distrjbution = 0.88
95.00
90.00
Ð
. (J
¡ì
gC)
70"00
U
CJ
a 50.00
tt
g
h
CJ
30"00
.¿
Ð
rd
-l
¿
E
c.) 10 .00
5.00
2.00
1" 00
g.so
o.2o
0.Iq
0 .05
0.01_
0*
0.94 0.96 0.98 1.oo L.o2 l_.04 1.06 I. 08
Ratio of Measured / Nonri¡al Flange Width
Figure 4"5 - cunu¡rative Frequency of Ratio of ¡4easuïed / Norninal

Flange l,Vidth of W Sections
r65
Cl-ass Zof
Interval Measure¡rents
0"82 - 0.84 0 .00

0.84 - 0"86 0.16
0.86 - 0.BB 2.48
ã 0"88 - 0"90 4.66
-¡ 0"90 - 0.92 5 .31
-tJ
fú
0.92 - 0.94 7.L5
H
0.30 0.94 - 0.96 r0 .73
r-l
0.96 - 0.98 L6.25
.-E
rd 0.98 - 1.00 2L.67
1"00 - 1"02 15.28
õ I.02 - L.04 L0.29
!
0.20 1.04 - 1.06 3 .36
¡ì
I"06 - 1.08 2.06
Nurnber of
Total
CJ
5
tf
(J Measr:renents = 2,768
H
0"86 0.90 0"94 0"98 r"02 1"04
Ratio of Measured / Nomj¡al- Flange Thicicress
Mean value of ratio = 0.976

Coefficient of Variation of ratio : 0.041-7
Figr:re 4.6 - Variation of Ratio of Measr:red / Nominal

Flange Thickness of W Sections
(Kennedy and Gad AIy f9B0)
l-66
oô oô 100 B
99-95
99.90
99. B0 :, /
Nornral probabiìity Dist¡jbutio
99 -50 /
99.00 ü,î:ii."::iiarion= 0.0417 Y
98 .00
/
95.00
/
90.00 /
/
Ð
o /
o
lr /
A
0)
70. 00
/
/"
O
F
CJ
50. 00 /
/o
õì
LI
F¡{
o 30.00
.-l
Ð
rd q/
-J
5
e
lr
/
U 10-00 "/
o/ /
//
5 .00
o/
2.00
1.00
0. Þ0
0. 20
/
q.IQ /
0.05 //
//
0.01_
08
0.80 0.84 0"88 0"92 0.96 1.oo 1.04 I .06
Ratio of Measured / Nomj¡al F1ange Thickness
Figure 4"7 - cr:nn¡rative Ereguencry of Ratio of Measured

Flange Thickness of W Sections
/ Nomi¡al-
L67
that a normal probability distribution provides a reasonabl-e

estimate of the actual probability dist,ribution" A normal
distribution vrith the mean value and coefficient of varia-
tion as reported by Kennedy and Gad AIy was used in this
study for the ratio of actual to nominal flange thickness.
4"2"6.4 Web thíckness - Kennedy and Gad Aly (1980) reported
measurements of the ratio of actuar to nominal web thick-
ness. Based on a total of 352 measurements, a mean value of
1-.0l-67 and a coefficient of variation of 3"84 percent was
calcurated. A histogram of the freguency distribution of
the web thickness data j-s shown in Figure 4"9. The data was
also plotted on a normal probability paper (Figure 4"9).
Normal and modified lognormal probability distrj-butj-ons
using the reported mean and standard deviation were compared
to the data" A nodified lognormal distribution with a rower
boundary set at 0"8 was found to provide the best fit to the
data as indicated by Figure 4.9"
A nodified lognormaL distribution \,vas, therefore, used.
in this study for the ratio of actuar to nominar- flange
width" The lower boundary of the rati.o was set at 0.8. The
mean and coefficient of variation were set at the val_ues
calculated by Kennedy and Gad Aly (1980).
4"2"7 Residual St,resses
Residuar stresses in steeL sections have a ì-arge varia-
tion associated with thern (Beedle and Tarl 19600 Alpsten
1972) " sources of variat,ion are from different cooJ-ing
168
C1ass Bof
Interval I4easurenents
0"40
0.90 - 0"92 0 .00
0"92 - 0"94 I.42'
É 0"94 - 0"96 5.40
o
.-l 0"96 - 0"98 9.38.
Ð
U 0"98 - 1"00 15.62
fd
tr 0"30
1"00 - 1"02 24"L5
F{ 1"02 - 1"04 20 "45
-lq I.04 - I.06 L2"78
.-l 1"06 - 1"08 4"55
O
dJ
1"08 - 1.10 3 .69
ô 1.10 - 1.r2 1.14
: 0"20
L"I2 - L"IA 0.57
O 1"14 - I"16 0 .85
6
a
tr
c) Tota1 No. of Measurenents = 352
l{ 0 .10
tr{
0 "94 0 "98 1.02 r.06 1.10 1.14
Ratio of Measured / Nonui¡¡a1 Vfeb Thiclcness
Mean of ratio = 1.0167

value
Coefficient of Variation of ratio = 0"0384
Figr:re 4.8 - Variation of Ratio of l4easr:red / Nomj¡al

Web Thiclmess of V'I Sections
(Kennedy and C'ad Aly 1980)
L69
4. 100 t
99.99
99 -95
99.90
99 .80 McdiJied Iognornal ProbabiJ-ity D jstr jòution
nean value = 1.0167
99.50
coefficie¡t of vari-ation = 0.0384
99.00 lor,ver boundarlz of djstrj-bution = 0.80 / "
o
9B-00
95.00
{J 90.00
OJ
()
Tr
gCJ
Þr 70.00
O
0)
a
tt'
d) 50.00
fr1
(.)
Þ
-d 30. 00
Ð
d
F{
a
E
J
r0"00
5.00
2"00
1.00
0.50
0.20
0.10
0. 05
0. 01
Vor
0.90 0.94 0.98 t.O2 t.c6 1.10 I.I4 1.I8
Ratio of I'feasr:red / Nominal Web Thiclrress
Figure 4"9 - Cunulative Freguency of Ratio of Measured / Nomj¡ral-

Þieb Thiclcress of W Sections
170
rates, different manufacturing processes (A1psten 1968), and

dimensional variations of the cross section" Further varia-
tions result from straightening of the steel section through
rollers or by gagging (bending about a point) (Alpsten
1968) " The combined effect of these sources of variation
makes it difficult to accurately predict the variability of
the residual stresses in a structural steel shape"
Beedle and TaIl (1960) reported the results of residual
stress measurements on a large number of American steel wide
flange shapes. The maximum, minimum and averagie values for
the residual stress at the flange tip and at the flange-web
juncture for beam and column shapes v¡ere reported. The data
for column shapes is summarized in Tab1e 4.5. This data was
used to estimate the coefficients of variation of residual-
stresses which hrere then used for the Mont,e Carlo simu-
lations (Chapter 5) " Six standard deviations r¡rere assumed
to occur between the maximum and minimum vaLues. The
coefficient of varj-ation was cal-cul-ated by dividing the
standard deviation by the reported average value given in
Tab1e 4"5" The coefficient of variation of the residual_
stress at the flange tip so calculated was 14"3 percent.
For simplicity, a value of L5 percent was used in this
study" Using the same procedure, the coefficient of varj_a-
tion of the residual stress at the flange-web juncture was
calculated to be 73 percent and was used in this studv" The
L7L
Table 4.5 - Measured Residual Srress in lîide FIange CoIumn Shapes

(Beedle and TaIL ( 1960) )
Residual Stress Residual Stress

at Flange Tip at Flange-VJeb Juncture
(ksi) (KSIi
Minimum Mean I Maximum Mi n imum Mean Maximum

I
-7 .1 -t2.8 I -18.7 16.5 4.7 -4. 1
Note: 1 ksi = 6.895 MPa; (-) indicates compressive st.ress

172
higher coefficient of variation of the stress at the flange-

web juncture is reasonable since this location is more
Iike1y to have its cooling rate affected by external
influences" The mean values used for the residual stresses
ri¡ere calculated from Equations 2"2O and 2"2L given in Sec-
tion 2"8" A normal probability distribution was arbitrarily
assumed for all residual stresses"
4"3 RETNFORCTNG STEEL
Variations in the mechanical- and geometric properties of
the vertical and transverse reinforcing bars affect the
variation of the overall strength of the composite beam-
columns" This is because the variations in the properties
of vert.ical reinforcing bars affect the overall strength of
a beam-column d.irectly in terms of stiffness and in the
development of El:e M-0-P relationships" Similarly, the
variations in the properties of the transverse ties affect
the degree of confinement of concrete and, therefore, indi-
rectly affect the overal-L strength of the beam-column. The
mechanical properÈies that define the stress-strain
relations of the reinforcing bars (described in Section
2"7 "2) are the moduLus of elasticity, yield stress, strain
at initiation of strain hardening, ul-tirnate stress, and.
ultimate strain" The only geometric variation of concern to
this stud.y is the ratio of actual to nominar cross-sectional
area. Strength variations due to variat,ion in placement of
reinforcing bars is discussed in Section 4.4. To properJ-y
L73
model the basic variabl-es noted above for use in the theo-
retical subroutine (Chapter 2) and for the Monte Carlo anal-
ysis (Chapt,er 5), the mean val-ue, coefficient of variation
(or standard deviation) and the type of probability
distribution r¡¡ere defined for each basic variabl-e. These
definitions were either taken directly from the literature
or derived from the data available in the literature. No
new test data hras generated for this study" A description

of the statistical distributions used for each variabl-e i-s
gi-ven below.
4"3"L Modulus Of Elasticitv
The elastj-c modulus for steel in rej-nforcing bars has
been found to have a small dispersíon and is relatively
unaffected by rat,e of loading (Mirza et al " i,979b) " Allen
(1972) found that the variability of elastic nodulus rtras
about the same for in-batch and overall_ variat,ions" Varia-
tions resulting from the ratio of actual to nominal area of
bar were also incorporated into the overall variation.
Allen (1,972) suggested a mean value of 28,500 ksi (t-96,550
MPa) and a coefficient of variation of 2 percent" Mirza et
aI. (1979b) studied Allen's data as wel_l_ as data from oth-
ers" They suggested that the probability distribution of
erastic modulus can be considered normal with a mean value
of 29,2OO ksi (2O1-t380 MPa) and a coefficient of variation
of 3"3 percent"
L74
In this study it was assumed that the elastic modulus of

reinforcing steel followed'a normal distributj-on with a mean
of 29,OOO ksi (2OO,000 MPa) and a coefficient of variation
of 3.3 percent.
4"3"2 YíeId Strengt,h
Factors contributing to the variabiLity of yield
strength of reinforcj-ng steel are the variation in the com-
position of the steel, variation in the actual cross-
sectional area of the bar and the effect of rate of loading
(Mirza et al . 1-979b) " As outlined for structural steel-
(Section 4"2.2), the static yield strength rather than the
dynamic yield strength provj-des a better estimate of the
yield strength of the bars under normal loading conditions
in a building" Static loading reduces the magnitude of the
yield strength and is, therefore, of more interest for reli-
ability analysis.
Yie1d strengths reported from rniLl tests are based on
strain rates of approxj-rnately 1O4O micro in. per in. per
sec (Mirza et al-. L979b) which is the same as that reported
for niIl tests of structural steel (Kennedy and Gad A1y
1980). Rao et al" rs equation (1964) presented earLier
(Eguation 4"18) r¡ras found to correlate wel-l with the yield
strength data at various strain rates (Mirza et al. 1979b).
Ä value of 4 ksi (27 "6 MPa) was suggested as a reasonabl-e
assumpt.ion to describe the difference between the miI1 test
(dynanic) and the static yield strengths" By cornbining the
L75
collected test data and accounting for the differences

betr¿een the mill test and the static yield strength, Mirza
et al " (L979b-) calculated the mean static yield strength of
66.8 ksi (460"7 MPa) and a coefficient of variation of 8.3
percent for Grade 60 reinforcing bars.
The probability distribution of the static yield
strength was studied by Mirza et al " (1"979b) and was found
to be positively skewed. This is reasonable since qual_ity
control practices l-imit the probability of the yield
strength being less than specified. A probabit-ity d.ensity
function of a beta distribution was found to provide the
best fit to the static yield strength data. The probabirity
density function suggested by Mirza and MacGregior for Grade
60 reinforcing steel is shown in Equation 4"242
pDF = 7.sar( f ,, - s+\'o'(l02 - /v,)óes

\ 48 , t 48 I {+'2+)
in which 54 ksi
For SI conversion, multiply the terms 48, 54, and IO2 in

Equation 4"24 by 6"895.
The probability density for static yield strength of
Grade 60 reinforcing bar described by Equation 4 "24 is
plotted in Figure 4.10 while the cumulatíve freguency is
plotted in Figure 4"11. These eurves were used for the
0.07 -
0-06 -
0-05 -
Beto probobitity Distribution:
. \ |
Meon Votue : 66-8 ksi (461 Mpo)
\
=cç', 0.04 - b ^,r' I Coefficient of Voriotíon : O.OBJ
o)
o {OtP
:Ï
.'=
_ô 0.03 -
0
-o
o
L
o_
0.02
0.01
o
50 70 90 110
Stotic Yield Strength (ksi)
Figure 4.10 - Probobilíty Density for Stotíc Yield Strengthof Grode 60 (414 Mpo) Reinforcing Bors
1.0
0.9
0.8
0.7
\o
c(¡)
:J 0.6
u
o
t_
LL {{ts
(D 0.5
-l Beto Probobility Distribution:
o
f 0.4 Meon Volue : 66-8 ksi (461 MPo)
E
:t Coefficient of Voriotion : 0.083
(J
0.3
o.2
0.1
0
50 70 90 110
Stotic Yield Strength (ksi)
Figure 4.11 - Cumulotive Frequency for Stotic Yield Strengthof Grode 60 (414 MPo) Reinforcing Bors
L7B
Monte Carlo simulations reported in this study (Chapter 5) "
It shouLd be noted that the data used to derive Eguation

4"24 was based on nominal areas of bar cross-section.
Therefore, the effect of variation in rati-o of actual_ to
nominar bar cross-sectional area is already included in
Eguation 4"24.
4 "3.3 Ult,inat,e Strencrth
Factors affecting the variation of ultirnate strength are
the same as those affecting the yierd strength. Mirza et
al-" (L979b) reported that the ratio of ultimate to yield.
strength had a mean value of 1"55. The coefficj-ent of vari-
ation was essentially unchanged from the values obtaj-ned for
yield strength data (9"3 percent). since it, is reasonable
to assume that reinforcing steel with a higher yield
strength will also have a higher uLtirnate strength, the
static ultimate strength of reinforcing steel- was simpry
taken as 1.55 times the static yield strength for use in the
Monte Carlo simuLations (Chapter 5).
4.3.4 St,rain At Inítiat,ion Of St,rain llardening
Allen (L972) perfonned controlled tensile tests on vari-
ous sizes of reinforcing bars and determined the strain at
the initiation of strain hardening. The strain val-ue was
found to vary significantly dependíng on the bar diameter
and ranged from a rninimum of 0"7 percent for No " L4 (44.s mm
diameter) bars to a maximum af 2"2 percent for No. 5 (15 nrn
diameter) bars. Allen calculated a strain val-ue of a.49
L79
percent with a coefficient of vari-ation of 26"6 percent for

the strain at the conmencement of strain hardening when all
bar sizes v/ere considered.
In this study, the mean val-ue of the strain at the ini-
Èiation of strai-n hardeningi was taken as 1"5 percent and the
coefficient of variation vlas assumed to be 26"6 percent.
Àl-len (1972) nade no analysì-s of the probability distribu-
tion" A normal probability dístribution was arbitrarily
assumed for this study.
4.3"5 Ultinate Strain
Allen (1972) measured the ultimate strain of reinforcing
bars ranging in size from No" 3 to No. 14 (9.5 nm to 44.5 mm
diameter) " The mean ul-timate strain reported by AIIen was
L5"5 percent and the coefficient of variation hras ZO.3 per-
cent"
In this study the ultimate strain of reinforcing bars
v/as assumed to have a mean value of L5"5 percent" For sj_m-
plicity, the coefficient of variation r¡/as taken as ZO per-
cent. A normal probability distribution was arbitrarily
assumed for ultimate strain of reinforcing bars.'
4"4 COTJUMN GEOMETRY
A composite column is a combination of factory made and
site casted components. rn this sectj-on the variations in
field fabrications affecting the strength of beam-columns
are discussed" The fiefd fabricat,ions for the construction
of a composite column are similar t.o those for a reinforced
180
concrete column. Mirza et al-" (L979a) studied published

data to determine the statistics of the geometric variabl_es
of reinforced concrete columns. For this study, these sta-
tistics were assumed applicable to composit.e columns as
well. The variables described are the column length, the
overall- column width and depth, the concrete cover to the
hoop reinforcing, the spacing of rectangular hoops and the
distance from the geometric centroid of the column to inte-
rior reinforcing bar layers.
To keep the theoretical strength subroutine (RTHEO) as
efficient as possible, the cross-sect,ion of the composite
beam-column r,üas assumed to be synmetric about each axis
(Chapter 2) " This assumption reduces the number of vari-
ables allowed from the naximum number possibre that affect
the strength of a composite column" Unsymmetrj_c variations
about the rninor axis do not, affect the strength of the col--
umns analyzed in this study since major axis bending, with-
out twj-sting, v/as assumed" Unsymmetric varj_ations about the
major axis could affect the strength slightly but were
neglected. A discussion of each of the variables considered
is given in the following subsecti-ons "
4.4o I Column Length
No data for variations of column length was found in the
literature available. To accommodate this variable, the
statisticar description of beam span suggested by Mirza et
aI" (I979a) was assumed to be applicable to the column
181
length" the deviation frorn the specified column length used

had a mean value of 0"0 and a standard deviation of 0"67
inch (L7 nm) " A normal probability distribution was assumed.
for this variable" For cross-section studies, the specified
length of the column was input as 0"0.
4.4.2 Column Ttídt,h And Depth
The overall column width and depth may vary d.ue to inac-
curate forming. Mirza et al-. (L979a) studied data record.ed
by others on overall- dimens j-ons of cast-in-p1ace columns.
They reconmended that for rectangular columns with face
dimensions ranging from 1-l- to 30 inches (2gO to 762 mm),
mean deviation of the face dimension vras r/16 inch (t-"6 mrn)
gireater than the specified value with a standard deviation
of I/4 inch (6 m:n) " A normal probability distribution was
reconmended by Mirza et aI. (t-979a).
In this study, the above-noted stat,istical properties
r^/ere used" The width and depth of the corumn hrere varied
independently"
4"4"3 Concrete Cover
The concrete cover is measured from the face of the col-
umn to the exterior edge of the lateral hoops" This dimen-
sion may vary due to inaccurate fabrication of the hoop,
inaccurate forming of the col-umn t oy both. Measurements of
the concrete cover do not distinguish these sources" since
the variation of column face dirnension is carcur-ated sepa-
rately, this variable can be considered to be a measurement
LB2
of the placement of extreme bar layers (Mirza et al- " L979a) "
Grant (1976) studied data of others and suggested that the

deviation of the concrete cover from the specified value
(C"o) coul-d be described by a normal distribution with a
mean value (õ") given by Eguation 4.25 and a stand.ard devi-
ation of 0"1-66 inches (4"2 mm):
co C r, + 0.25 0.004h in. (4.2s)

C ,o + 6.35 O.ljzh mm
The cover concrete is a function of the face dimension of

the column (h) . These val-ues were used to vary indepen-
dently the amount of concrete cover paralleI to the minor
and major axís"
4,4o4 Placement Of Lavers of Vertical Bars
The dinension from the najor axis to the center line of
the extreme bar layers was defined by the concrete cover.
It was assumed that these bars vJere tied to the lateral
hoops and, therefore, were not independently varied.
The dimension from the major axis to interior bar layers
was described by a normal probability distribution with a
mean value deviation from the specified dimension of +0.04
inch (1 nm) and a standard deviation given by Eguation 4"26
(Mirza et aI. L979a) z
tB3
ú"o O .2O + 0.033 h In. (4.26)

5.08 + 0.838f¿ mm
where o.o is the standard deviation of the dimension from
the najor axis to the bar layer and h is the overall depth
of the colurnn. This statistical description was assumed. in
the Monte carlo si-rnulations. rt should be noted that due to
the double symmetry of the cross-section assumed by the
theoretj-cal strength model (Chapter 2), an interior bar
layer located at the major axis was assumed to have no devi-
ation from the specified 1ocation.
4o4"5 Spacing Of Rectangrular lfooE¡s
Spacing of the rectangular hoops affects the degree of
confinement of the core concrete and., therefore, the
strength of the composite beam-column (Section 2"6"2) " No
data on the variation of rectangular hoop spacing was found.
Mirza and MacGregor (1992) assumed that the spacing of ties
in concrete beams followed a normal probability distribution
r¡¡ith a mean value equal to the specified varue and standard
deviation equal- to 0"53 in" (13.5 nn). These values hrere
assumed to be vaLid for the spacing of ties in columns as
well and were used for this study"
LB4
5 STMTTT,ÃTTON Pi¡{D ÃNåLYSTS OF COMPOSTTE BEÃI"í-COÏ,UM}ü STRENGTHS
The theoretical uLtimate strengths (R,) of a composite
beam-column !ùere simulated 500 times by a Monte Carl-o tech-

nique and were based on the theoretical strength model and
probability distributions of the variables affecting the
strength" These strengths r,üere divided by the nominal ulti-
mate strength (R") of the beam-column which was computed
using the Code procedures and the nominal properties of
variables affecting the strength. This provided the simu-
lated sample of the non-dimensionalized strength ratios
(R¿/R,) for the beam-column. A statistical analysis of the
simulated sample provided the coefficient of variation and
other probability distribution properties of the strength
ratios for the bearn-column under consideration" The beam-
col-umns studied lrrere of various cross-section configurations
and lengths" A general descríption of the Monte Carl-o
technique is given first followed by descriptions of the
beam-column configurations studied" The results of the
study are then discussed with the effect of different vari-
ables on the overall strength variatíons examined" The dis-
cussions are given separately for short and for sl-ender
colurnns "
5" l" MONTE CARIJO TECHNIQUE
The premise of the Monte Carlo technique is that the

overall variabil-iÈy of the performance of a system can be
lBs
synthetical-Iy derived if a deterministic relationship

between the system performance and each variabl-e affecting
the performance exists and the probability distributions of
all- variables affecting the performance are known (Mirza
l-985b). Repeated random choosing of the value of each vari-
abLe according to its individual- probability distribution,
calculating the performance of the system based on each set
of randomly generated vaLues of variables, and statisticalry
analyzing the simulated sampre of system performance wirl-
provide the overarl- variation of the performance of the sys-
tem" A flowchart of the technigue is given in Figure 5.1.
As the number of simulations is increased, the syntheti-
caIly created probability distribution of the system per-
formance wilt tend to its true distribut,ion (Mirza t-985b) .
Mirza (L985b) compared sample sj-zes of 200, 5OO and I_OOO
simulations in a variability analysis of reinforced concrete
beam-columns" rt was found that there was no significant
difference in the statistical properties of the strength
samples obtaj-ned for 500 and looo sinulations. Therefore, a
sample size of 500 simulations v¡as used for aLl- beam-columns
analyzed in this study"
5"2 DESCRTPTTONS OF BEA¡.í-COI,UMNS STUDIED
the specified materiar properties and dimensions of the
beam-columns studied were chosen to give a reasonable repre-
sentation of the range of variables expected. in actuar con-
strucÈion" Figure 5"2 shows the nominaL dimensions of the
186
INPUT:
STATISTICAL
PROPERTIES OF
VARIABLES
SELECT A
RANDOM VALUE
OF EACH
VARIABLE
REPEAT MANY
TIMES
REI-ATIONSHIP
BETWEEN
CALCULATE VALUE OF
VARIABLES
SYSTEM PERFORMANCE
AND SYSTEM
PERFORMANCE
OUTPUT: SUMMARIZE
RESULTING
VALUES OF SYSTEM
PERFORMAÌ.ICE WITH
STATISTICAL ANALYSIS
Fiqure 5,1 The Monte Corlo Technique

(Mirza 1985b)
187
O
-l
WtO x 54
or
W10 x LL2
0 .5 inch día.
O
rt
ties at I0 inches
on centres
1" 5 i-nch 0 " 875 inch diameter

concrete cover vertical bars
Note: I inch = 25"4 rnm
Figure 5 "2 - Nominal Column Cross-Section Detail-s

(al1 dimensions are gj-ven in inches)
lBB
cross-section used" These dimensions conform to the

reguirements of ACI 318-83 and CAN3-A23.3-M84" The vari-
ables selected for study were the specified concrete
strength, specified structural steel yield strength, ratio
of structural steel area to gross area of cross-section,
slenderness ratio, and end eccentricity ratio" The speci-
fied yield strength of the reinforcing bars was chosen to be
60 ksi (414 MPa) for all columns studied" The ratj_o of
vertical reinforcing bar steel area to gross cross-sectj-on
(p,r) area was I"2 percent" The study was divided into two
parts: (a) basic study and (b) supplemental study for the
short as well as for the sl-ender beam-col_umns. The speci-
fied values of the variables used for these studi-es are dis-
cussed in the following two sections.
5.2.1 Basic Study
The basic study investigated the effects of the speci-
fied concrete strength, ratio of area of structural steel to
gross area of cross-section, end eccentricity ratj_o, and
slenderness ratio on the ratio of theoretical to nominal
strength of composite beam-columns. Specified concrete
strengths of 4000 and 6000 psi (27"6 and 41"4 Mpa) v/ere
studied" These values r,rere chosen to represent the commonly
specified values for columns. The quarity contror of 4ooo
psi (27.6 MPa) concrete was assumed to be average (coeffi-
cient of variation of test cyrind.er strength : 1,5 percent) "
The guality control of 6000 psi (4i_"4 Mpa) concrete was

189
assumed to be excellent, reflecting the extra care taken to

mix higher strength concretes (coefficient of variation of
test cylinder strength = l-0 percent) .
Ratios of structuraL steel area to gross cross-sectional
area (p"") of 4 and 8 percent were studied. The smaller
ratio was obtained by assuming a !'I10 x 54 (I{250 x 80) rolled
steel section and the larger one by using a !dl-O x IL2 (W250
x 167) roll-ed steel shape. It was felt that the structural
steel ratios less than 4 percent woul-d not be practical for
composite columns" Structural steel ratios larger than I
percent are difficult to obtain without the use of built-up
stee] sections. The specified yield strength of the struc-
tural steel was chosen as 50 ksi (345 MPa) for the basic
study. This represents the maximum allowable specified
yield strength of structural steel by ACI 3l-8-83 and CAN3-
A23"3-M84 Codes"
Slenderness ratios (kl/r) of O and 22 hrere examined for
the short columns" A slenderness ratio of 0 represents the
cross-section. A slenderness ratio of 22 represents the
upper liinit for beam-columns designed without length effects
by both ACI 3L8-83 and CAN3-423.3-M84 Codes for the type of
beam-columns studied" Slenderness ratios of 22"L, 33, 66
and l-00 were chosen for the sl-ender columns. A slenderness
ratio of 22"I nearly represents the lower lirnit for which
the length effects must be included in design for the type
r90
of columns studied" A slenderness ratio of 1OO is the maxi-

mum slenderness ratio allowed for the evaluation of stabil--
ity effects by the moment rnagnifier method of ACI 318-93 and

cAN3 -A23 " 3 -M8 4
Eccentricity ratios (e/h) of Ot 0"05, 0.1, 0.15, O.2,

o"25, 0.3, O"4, 0"5, 0"6, 0"7, 0"9, 1"0, 1"5, 2"Q, 4"O and
infinity (pure bending) were studied for all- columns. Note,
for the basic study, the theoretical- strength included the
effects of residual stresses in structural steel and con-
crete confinement due to latera] ties but did not incrude
the effect of strain-hardening of structural and reinforcing
steels.
Table 5"L lists the short columns, whereas Table 5.2
lists the slender columns used for the basic study. The
column designations in these tables are made up of four el-e-
ments separated by hyphens: The first element represents
the specified concrete strength in kips per square inch, the
second element identifies the specified structural steel-
yield st,rength in kips per square inch, the third element
represents the approximate ratio of structural steel area to
gross area of cross-section and the fourth el-ement identi-
fies the slenderness ratio.
5"2.2 Supplemental Study
Three additional variables rárere investigated to study
their effect on the probability dist,ribution properties of
l9I
Table 5.1 - Specified Properries of Shorr Beam-Columns

Used for Basic Study:"-
Run Co Iumn f' f kt/t p

Number Designation (ps1 ) (pti I SS
(i) (2) (3) (4) (s) (6)
BO 4-50-4-o 4000 50000 0.040

BO2 4-50-4-22 4000 50000 22 0.040
BO7 6-50-4-0 6000 50000 0.040
808 6-50-4-22 6000 50000 22 0. 040
813 4-50-8-o 4000 50000 0. 082
874 4-50-8-22 4000 50000 22 0.082

819 6-50-8-0 6000 50000 ^ ô1
820 6-so-8-22 6000 50000 O. OB2
:kEach beam-column listed above was studied for nominaL elh values of 0.0,
0"05, 0.1, 0.15, o.2, o"25,0.3, o.4,0.5, 0.6, o.7, o.g, 1.0, 1.5, 2.o, 4.o,
and -. AII columns had cross-section sLze of.20 x 20 in., Grade 60 (414 Mpa)
reinforcing bars, and 9_. = 0"012. Lateral ties conformed to the minimum
requiremenÈs of ACI 318:83 and CAN3-423.3-M84. The qualiry conrrol of 4OOO
psi (27 -6 MPa) concrete r./as assumed to be average, whereas that for 6000 psi
(4I.4 MPa) concrete qTas taken to be excellent.
Note: 1000 psi = 6.895 MPa; 1 in " = 25.4 mm.
L92
TabIe 5.2 - specified Properties of slender Beam-columns

Used f or Bas ic S tudy''.-
, f'c..
Run Co I umn r. kø /t
Number Designat.ion (PS1.) qpXi )
(1) (2) (3) (4) (s)
B03 4-50-4-22.r 4000 50000 )) 0.040

804 4-50-4-33 4000 50000 33 0.040
B05 4-50-4-66 4000 5OOOO 66 0. 040
806 4-50-4- 1 00 4000 50000 1oo 0. 040
809 6-50-4-22.1 6000 50000 22J 0"040

810 6-50-4-33 6000 50000 33 0. 040
811 6-50-4-66 6000 50000 66 0. 040
81.2 6- 50-4- 1 00 6000 50000 OO 0.040

815 4-50-8-22.1 4000 50000 22. r 0.082
816 4-50-8-33 4000 50000 33 0.082
817 4- 50-8-6 6 4000 50000 66 o. 082-
B1 B 4- 50-8- 1 00 4000 50000 OO 0.082
B2 T 6-50-8-22 " I 6000 50000 22.7 0. 082
822 6-50-8-3 3 6000 50000 33 0. 082
823 6- s0-8-66 6000 50000 66 0.082
824 6- 50-8- 1 00 6000 50000 100 0. 082
listed above r^ras sLudied for nominal e/h values of 0.0,

'''-Each beam-column
0.05, 0.1, 0.15, o.2, o.25,0.3, 0.4, 0.5, 0.6, Q.7, O.g, 1 .0, 1.5, 2"o,4.0,
and -. AII columns had cross-section size of 20 x 20 Ln., Grade 60 (414 MPa)
reinforcing bars, and p_^ = 0"012. Lateral ties conformed t.o the min imum
requiremenrs of ACr 3rgr8: and CAN3-A23.3-Mg4. The qualiry control of 4000
psi (27.6 MPa) concrete was assumed to be average, whereas Ëhat for 6000 psi
(41.4 MPa) concrete vras taken Èo be excellent.
Not.e: 1000 psL = 6.895 Mpa; f. in . = 25.4 mm.
193
the ratio of theoretical to nominaL strength of beam-

columns" The three variables rrere the specified yield
strength of structural steel, the straín-hardening of
sLructural and reinforcing steels, and the effect of con-
crete quality control"
Specified structural steel yield strengths of 36 ksi
(ÀsTM À36) (f y:244 MPa) and 44 ksi (CSA c4O " 21-M 44Vt)
(/y:3O3 MPa) are conmon in the United States and Canada,
respectively" To study the influence of specified struc-
tural steel- yield strength on the theoretical to nomj-na1
strength ratios of beam-columns, the results of two short
and two sLender beam-columns selected from the basic study
\,rere compared to the results of identical columns having the
specified structural steel yield strength of 36 and 44 ksi
(248 and 303 MPa) "
Strain-hardening provides an enhancement of the steel

strength. However, the use of this enhanced strength of
steeL is not all-owed by the ACr or csÀ code. To examine the
effect of strain-hardening on the ratios of theoretical to
nominal strengths, the resulÈs of tv¡o short and two slender
coLumns selected from the basic study \,rere compared to the
results of identicar columns in which the strain-hardening
of both structuraL and reinforcing steer beyond the plastic
plateau was permitted for computing the theoretical
strengths "
194
To examine the infl-uence of concrete quality control on

the variation of theoretical to nominal strength ratios, two
short and two sl-ender columns from the basic study with spe-
cified concrete strengths of 6o00 psi (41-"4 MPa) and excel--
lent concrete quality control were compared to identical
columns having averagie concrete guality control" Note the
excellent guality control assumes a coefficient of variation
of 10 percent for test cylinders, whereas average guality
control had a test cylinder coefficient of variation of 15
percent "
the short beam-columns of the supplemental study are

shown in lable 5.3, whereas the slender columns of the
supplemental study are given in Table 5.4" The fifth ele-
ment appearj-ng in some column designations in these tables
represents the inclusion of strain-hardening of both steels
(STH) or Èhe use of averag'e çfuality control for 60OO psi
(41.4 MPa) concrete (A) " The first four elements of the
column designation in Tables 5.3 and 5"4 are identical_ to
those described for Tab1es 5"L and 5"2 of the basic study.
Note the effects of concrete confinement due to lateral- ties
and residual stresses in structural- steel r¡rere also included
in computing the theoretical strengths for the supplemental
study"
5.3 SHORT COT4POSTTE BEÃ$-COTJT'M3{S
This section examines the overall strength variations of

short composite columns. The effects of individual vari-
195
TabIe 5.3 - Specified Properties of Short Beam-CoIumns

Used for Supplemcntal Study''.
I Strain I Concrete I
Run Co Lumn f' r. kø l, o

'ss I Hardenins Quality
Number Designation (PSI, lpUi) i rr,.ludeã Cont ro I
(i) ( )\ (3) (4) (5) (6) I tzl (8)
I
SG 4-36-4-22 4000 3 6000 22 No Ave rage
SG3 4-44-4-22 4000 44000 22 0.040 No Average
SG5 6-36-4-22 6000 3 6000 22 0.040 No ExceIlenE
SG7 6-44-4-22 6000 44000 22 0. 040 No E:<." t 1""t

STH 1 4-50-4-0-sTH 4000 50000 0. 040 Yes Average
STH3 6-50-4-0-srH 6000 50000 o. 040 Yes E"." t t""t

CQ1 6-50-4-22-A 6000 50000 22 0. 040 No Average
CQ3 6-50-8-22-A 6000 50000 22 0.082 No 4"" r" ge
:kEach beam-column listed above was studied for nominal e/h values of 0.0,
0.05, 0.1, 0.15, o.2, o.25, 0.3, o.4, 0.5, 0.6, o.7, 0.9, 1.0, 1.5, 2.o,
4.0, and -. AII columns had cross-section síze of.20 x 20 in. , Grade 60
(4L4 llPa) reinforcing bars, and p = 0"012. Lateral ties conformed to
the minimum requirements of ACI :IB-g: and CAN3-423.3-M84.
Note: 1OOO psi = 6.895 MPa; 1 in. = 25.4 mm.
L96
Table 5.4 - Specified Properties of Slender Beam-Columns

Used Ior Supplemental Study;!
I Srrain I Concrete I
p
Run Co lumn
, f'c., , f v.. kø/t ss I Hardeninp QuaLity
Number Designation (psr/ (ps1) i r.cru¿"ã Cont ro L
(1) (2) (3) (4) (s) (6) I rzl (8)
I
SG2 4-36-4-33 4000 3 6000 33 o. o4o I No Average
sc4
I
I
4-44-4-33 4000 44000 33 o. o4o I No Ave rage
I I
SG6 I 6-364-33 6000 3 6000 33 o. o4o I

No E".e I le"t
I
SG8 6-44-4-33 6000 44000 33 o. o4o No E". I t."t

I
"
I
STH2 4-50-4-66-STH 4000 50000 66 O. O/-O I Yes Ave rage
STH4 6- 50-4-66-S TH 6000 50000 66

_l
o. o4o I Yes E"." t t""t
I
cQ2 6-50-4-33-A 6000 50000 33 o"o4ol No Average

I
CQ4 6- 50-8-33-A 6000 50000 33 o.082 |

No Average
'kEach beam-column listedàbove r¡ras studied for nominaL e/h values of 0.0
0.05, 0.1, 0.15, o"2, o.25,0.3,0.4, 0.5, 0.6, o.7,0.9, 1.0, 1.5, 2.o,
4.O, and -. All columns had cross-section sLze of 20 x 20 in", Grade 60
(414 MPa) reinforcing bars, and p_-_ = O.OL2. Lateral ties conformed to
the minimum requirements of ACI :18-A: and CAN3-423.3-M84.
Note: 1000 psi = 6.895 MPa; 1 in. = 25.4 mm.
L91
ables on the strength ratios are discussed here and the

rnajor variables that affect the variability of sho.rt
composite columns are identified as weI1" The specified
properties of short composite bearn-columns studied are qiven
in Tables 5"L and 5"3.
5 " 3. I Overall Strength Variat,ions
Figures 5"3 and 5"4 are the plots of the simulated axial
load-bending moment interaction diagrams for Columns
6-50-4-22 and 4-50-8-22 taken from Table 5"1. These col-umns
represent the extremes of the structuraL steel index
(p""-f f
"/ ":0.33 and I.O3) for the columns studied. The theo-
retical maximum, mean, one-percentile and minimum strength
curves, obtained from 500 simulations are plotted" Also
plotted is the ACI 3l-8-83 ultimate strength curve which was
calculated assuming a value of l-.0 for understrength fac-
tors.
The factored ACf strength curve shown in Figures 5.3 and
5.4 t'tas obtained by dividing the ACI ultimate strengths by a
safety factor that varied from 1.55/O.7 =2.2I to
1.55/0.9= I.72" The safety factor of 1.55/0.7 was used for
axial- loads that exceeded the load corresponding t,o the max-
imum bending moment on the ACI ultimate strength interaction
curve" Below this axial load, the safety factor was assumed
to vary linearly lrith the axi-al load from i.55/0.7 Eo
1,55/0.9 at the pure bending condition. The val-ue 1.55 used
above represents the average of the ACI load factors for
o
il Note: 1 kîp : 4.448 KN
()
1 ft-kíp : 1.356 KN-m
Co" SkewneÈ
aî)
Kurtosis = 2"8
o_
v 3000
o ts
o
J l.o
æ
Maximum
.9
x l-Percentile
2000
Minimum
e,lh=O "8
I O00
0 200 400 600 800 1000

Bendinþ Moment (Ft-Kips)
Fìgure 5.3 - Axiol Lood-Bending Moment Strength lnteroction Curves of Rondomly Generoted
Somple of 500 Short Composite Columns'[Colurnn 6-50-4-22 (toUte 5-1)J
O
il
0)
(i, !1,"u/z
.s ,Ø#'
:,.ê,
:f ¿,|'
'o
o P
o 3000 \.o
J LO
.g
4x
I 000
0 200 400 600 800 1000 1200 1400

Bendírrg Moment (Ft-Kips)
Figure 5"4 - Axiol Lood-Bending Moment Strength lnteroctíon Curves of Rondomly Generoted
Somplq' of 500 Short Cornposite Columns fcotumn 4-50-8-22 (Îoble 5.1)]
200
dead load (1"4) and for live load (1.7) " The values O"7 and
0.9 are the ACI 3L8-83 understrength factors for cornpression
and tension failure conditions, respectively" The axial_
road level below which ACr 3l-8-83 allov¡s an increase in the
understrength factor greater than 0"7 is further discussed
at. the end of this section"
A comparison of the ACI ultimate strength and. the theo-
retical mean strength interaction curves shown in Figures
5.3 and 5"4 indicates that the ACr procedure overestimates
the mean urtimate strength for e/h ratios less than or equar
to around 1-"5" However, these differences between the ACr
urtimate strength and the mean theoretical- strength (for
e/ä s 1.5) appear to decrease as the structurar- steer- index
(p""-f ,/ Í'") increases. This can be seen by comparing Figures
5"3 and 5.4" For e/h>1.5, the ACf prediction is nearly the
same as the mean theoretical strength when p",-f v / f
": O.33
(Figure 5.3) " !,Ihen p""f y/f": t.OS (Figure 5.4), the ACI Code
underestimates the mean theoretical strength for e/h greater
than 1.5"
comparison of the one-percentile and factored ACr inter-
action curves shor'¡ that the factored ACr curve is well bel_ow
the one-percentile strength curve" rt should be noted that
the factored ACr curve shown in Figures 5"3 and s.4 is on]-y
an approximation and may rise or fall depending on loading
combinations and variations"
20L
Histograms of theoretical beam-column strength at eccen-

tricity ratios of 0"0 (pure compression), 0.2, and infinity
(pure bending) are also plotted on Figures 5.3 and 5"4. fn
Figure 5 " 3 (p""-/ Í'":0.33) , the histograrns appear symmetric
"/
at all three eccentricity ratios. In Figure 5"4
f '": l.03) the histograms are slightly positiveJ-y
(p""-f
"/
,
skewed, indicating the influence of the J-arger structural-
steel area" Note the probability distribution of structural-
steel strength was assumed to be positiveÌy skewed. A
smaLler coefficj-ent of variation at higher e/h ratios is
apparent in both figures.
The bal-ance point is defined by ACI 318-83 as the strain
condition which produces yielding in the tensil-e steel as
the compressive face of the concrete reaches its maximum
useable strain. In a reinforced concrete cross-section, the
balance point defines the transition from compression to
tensj-on failure and occurs at the maximum bending moment on
the axial load - bending moment interaction curve. ACI
318-83 specifies the lower of the axial load at the balance
point or the axial load calculated by Equation 5"1 as the
transition point for defining the value of understrength
factors "
P=O.lf"An (5.1)
in which rlq = gross area of cross-section.
202
Four composite beam-column cross-sections were analyzed in

order to see how the definition of balance point given above
applies Lo composite columns. The axiaL load corresponding
to the yielding of the tension reinforcing bars, incipient
yielding of the tension flange of the steel section and full-
yielding of the tension flange was calcuÌated for al-l- four
cross-sections (L/h= 0.0) shown in Table 5"1" Figure 5"5
plots the ACI ultimate strength interaction curve for a
cross-section (Column 4-5O-4-O in Table 5"1) with axial load
leve1s corresponding to the above-noted yielding conditions
identified in the Figure" The maximum bending moment capac-
ity occurs at an axial load Ievel between those correspond-
ing to incipient and full yielding of the tension flange.
Furthermore, the axial load leve1 at yielding of the tension
reinforcing bars is significantly greater than the axial-
load corresponding to the maximum bending moment capacity"
Thus, the definition of balance point used for reinforced
concrete beam-columns does not seem to be applicable to the
composite cross-section shown in Figure 5"5. Similar con-
clusions v/ere drawn for the remaining three cross-sections
given in Table 5"1-" Hence, defining the balance point for
composite column cross-sections as the strain condition at
which full yielding of the tension flange occurs as the com-
pression face of concrete reaches its maximum useable strain
O - OO? \
I\----_,v!,-¿s r,¡i 'l I ho l¡n{-h anrnrnr-rri a'l-a an¡{ i r¡a €nr
^ñneôñ¡â*
Note: 1 kip : 4.448 KN
1 ft-kíP : 1.356 KN-m
PI : 641 kips Tensile reinforcing bors yielded

P2 : 454 kips Yielding of tension flonge storted
P3 : 384 kíps Tension flonge fully yielded
2000 P4 : 160 kíps o.1 f; A6
û,
.g
)¿
E
o N)
o
J O
UJ
.E
x
I 000
0 200 400 600 800 1000

Bending Moment (Ft-Kips)
. Figure 5.5 - ACI 318-83 Ultírnote Strength lnteroction Diogrom for Column 4-50-4-0 (Toble 5.1)
204
assj-gning understrength factors" This definition of the

balance point has been used in this study and for Figures
5"3 and 5.4"
The axial load leve1 corresponding to Eguation 5"1 was
significantly less than the axíaI loads discussed in the
previous paragraph and is aLso marked in Figure 5.5" There-
fore, the second ACI 3l-8-83 condition on axial load level as
given by Equation 5"1- seems to be overly conservative
particularly when applied to composite beam-columns.
5.3 " 2 Ef fect,s Of Variables Used I'or Basic Studv
The eight short columns in the basic study (Table 5.1)
hlere used to examine the effects of four variables on the
probabiJ-ity distribution properties of the ratio of theoret-
ical to nominal strength (strength ratio). These variables
are the slenderness ratio (kI/r), specified concrete
strength (f prime"), ratio of structural steel area to gross
area of cross-section (p"") and end eccentricity ratio (e/h)"
The comparisons for each of these variables are made at the
one-percentile and five-percentil-e leveIs and at the mean
value. The one-percentile level is of most concern since it,
pertains to the l-ower tail of the strength probability dis-
tribution"
5"3"2.L Effect of slenderness ratÍo - Four direct compari-
sons of the effect of sl-enderness ratio (kt/r) on the proba-
biJ-ity dist.ribution properties of the strength ratio hrere
made" Each comparison contained two short columns having
205
kt/r values of 0 and 22 with all other properties being

identical" The results from two of the comparisons are
shown in Figures 5.6 and 5"7"
At one-percentile IeveI, 3 out of 4 comparisons clearly
showed lower strength ratios for columns with kt/r of 22
over all end eccentricity ratios" The most significant dif-
ferences in one-percentile strength ratios were obtained for
the column set with combination of f'.:4000 psi (27"6 MPa)
and p"" = 0.040 as shown in Figure 5.6(a)" Exception to this
behavior was the column set having f'.= 6000 psi (4L.4 MPa)
and p"" = 0"082 for which the one-percentile strength ratios
for the colurnn with kl/r : 0 were lower in regions of e/h
less than 0"25" This is indicated in Figure 5.7(a). How-
ever, the differences in one-percentile strength ratios for
e/ h < 0.25 in Figure 5.7 (a) seem to be small.
At the s-percentiLe 1evel and at the mean value, the
column with kl/r of 22 yielded lower strength rat.ios than
the columns having,kl/r of 0"0 for all e/h ratios as shown
in Figures 5"6(b) and (c) and Figures 5.7(b) and (c) " From
this it is reasonable to conclude that the col-umns with
slenderness ratio of 22 are more critical for reliability
analysis of short composite beam-col-umns.
5"3"2.2 Effect of specified concret,e strengt,h * Four column
sets q¡ere used to investigate the effect of specified con-
crete strength (f'") on the probability distribution proper-
ties of the strength ratio" Each set had one colurnn with
(o) 1-Percentile Volues
C
v. 0.8
P
v.
.9
t\)
o o
E. oì
€ 0.7 4-50-4-0
c¡ (l<l/r :
c 0)
o
L-
t-t1
4-50-4-22
(kl/r : 22)
D.6
r].5
0 a.2 0.4 0.6 0.8 1.O 1.5 2.0 4.0 æ
Errd Eccentricitv Rotío (ei'h)
Figure 5.6 - Effect of Slenderness Roli,: on the Fot-i,r of Theoreticcl tq l.l,¡rninol StrenEth of
Slrort Corr'posite :Steel-Corrcrete Beorrr -Colunrns
(b) 5-Percentile Volues
û.9
frC 0.8
fr
.9
t\)
o o
fr \¡
c 0.7 4-50-4-0
Et :
cq) (kl,/r o)
t-
Îñ 4-50-4-22
(Jf/r : 22.,\
0.6
0.5
0 0.2 0.4 0.6 0.8 1.O 1.5 2.O 4.0 æ
End Eccentricity Rotío (e,,1h)
Figure 5.6 icont-) - Effect of Slenderness Rotio on the Rotio of Theoreticol to lJominol Strenq¡r of
Short Cornposil:e Steel-Concrete Beorrr-Columns
(c) Meon Volues
c
E, 1.1
É.
.9 N)
o O
v co
c 1.0
cÞ
c
0)
L
#
Llt 4-50-4-O
(kl/r : O)
4-50-4-22
(kl/r : 22)
0 0.2 0.4 0.6 0.8 1.0 1.5 2-0 4-o cr

End Eccentricity Rotío (e/h)
Figure 5.6 (cont-) - Effect of Slenderness Rotio on the Rotio of Theoreticol to Nominol Strenqth of
Short Cornposite Steel-Concrete Beom-Columns
0.9
c
M 0.8
.; t\)
o o
ûl l.o
E 0.7 6-50-8-0
a (kllr : O)
=
Ð
(n 6-5r1-8-22
{kl,/r : 22')
0.5
0 0.2 0.4 0.6 0.8 1.0 I .5 2.0 4.O æ
End Eccentrícit¡r Rotio (e,,'h)
Figure 5.7 - Effect of Slenderness Rotio on the Rotio of Tlreoretícol to l.lomínol Strengtlr o{
Short Composite Steel-Concrete Beom-Columns
0.9
vt- 0.8
+
E
N)
ts
E O
0.7 6-5û-8-0
(¡)
(kllr : 0)
1-
(J',)
6-50-8-22
(kl/r :22)
0.6
0.5
0 a.2 0.4 0.6 0.8 1.0 1.5 2.0 4.0 a0
End Eccentricity Rotio (e,/h)
Figure 5.7 icont.) - Effect of Slenderness Rotio on the Rotio of Theoreticol to FJominol Strengtli of
(c) Meon Volues 6-50-8-0
(kllr : 0)
6-50-8-22
(kt/r - 22)
E
fY,
E
.9
N)
Ð ts
E. ts
t.0
ul
g
ah
0 0.2 0.4 0.6 0.8 1.0 1 .5 2.0 4.0 æ

End Eccentricity Rotio (./n)
Figure 5.7 (cont-) - Effect of Slenderness Rotio on the Rotio of Theoreticol to Nominol Strength of
212
f'"of 4000 psi (27"6 MPa) and one column lrith f'":6000 psi
(4I"4 MPa) with all other properties being identical. As
noted ín Chapter 4, the coefficíent of variation of the
strength of test cylinders rì/as taken as 1_5 percent (average
quality) for 4000 psi (27 "6 MPa) concrete and 1O percent
(excellenL guality) for 6000 psi (43-"4 Mpa) concrete" This
difference in the test cylinder coefficients of variation
reflects the extra care taken in the manufacture of higher
strength concrete" The results from two of the comparisons
are shown in Figures 5"8 and 5.9.
At the l-percentile 1eveI, 3 out of 4 comparisons showed
significantly lower strength rat,ios for the column with 6ooo
psi (4L"4 MPa) concrete" These differences \¡rere especially
apparent for low eccentricity ratios when pss hras 0.040 as
shown in Figure 5.8(a)" The trend appeared less significant
as the structurar steel- ratio increased to 0.082 as shown in
Figure 5"9(a) where there was no difference between the
strength ratios of columns with l'" = 4OOO and 6O00 psi
(27 "6 and 41.4 MPa) aE e/ h of O"2 or less" The columns with
the higher structural steel percentage have the overall

strength less influenced by the concrete strength and,
therefore, less difference is expected between the one-
percentile strength ratios for columns with 4ooo and 6000
psi (27 "6 and 4l-"4 MPa) specj-fied concrete strength.
(o) 1 -Percentile Voluee
c
É 0.8
M
.g
(t N)
E F
UJ
.t- o.7 4-50-+-22
Ðl
L [f' : 40O0 psi (27.6 MPo)l
o
L
tf) 6-50-4-22
[t'= 6000 psí (41.4 MPo)l
0 0.2 0.4 0.6 0.8 1.0 1.5 2.0 4.0 æ

End Eccentricity Rotio ("/h)
Figure 5.8 - Effect of Specified Concrete Strength on the Rotio ofTheoreticol to Nominol
Strength of Short Composite Steel-Concrete Beom-Columns
uc 0.8
E
.9 T\)
ö H
tr è
s 0.7 4-50-4-22
çlt
a)
[f; : +ooo psi (27.6 MPo)l
(.t)
6-50-4-22
[f;: oooo psi (41.4 MPo)l
0 0.2 0.4 0.6 0.8 1.0 r.5 2.0 4.0 cÐ

End Eccentricity Rotio (erlh)
Figure 5.8 (cont.) - Effect of Specified Concrete Strength on the Rotio of Theoreticol to Nominol
(c) Meon Volues
4-50-+-22
[r; : +ooo psi (27.6 MPo)l
6-50-4-22
[f; - oooo psi (41.4 MPo)l
c
v. 't.1
É
.9
N)
o H
E (¡
t.0
c'¡
(¡)
0 0.2 0.4 0.6 0.8 1.0 1 .5 2.0 4.0 @

End Eccentricity Rotio (e/h)
Figure 5-8 (cont.) - Effect of Specified Concrete Strength on the Rotio of Theoretícol to Nominol
1.0
(o) t -Percentile Volues
0.9
c
E, 0.8
K
.e
Ð t\)
E ts
o\
L o.7 4-50-8-22
ç'l
t; [f;: +ooo psi (27.6 MPo)l
d)
tt't
6-50-8-22
0.6
0.5
0 0.2 0.4 0.6 0.8 1.0 1.5 2.0 4.0 @
Fígure 5.9 - Effect of Specified Concrete Strength on the Rotio ofTheoreticol to Nominol
Strength of short composîte steel-concrete Beom-corumns
(b) 5-Percentíle Volues
c
É 0.8
E
.9
N)
o
E {ts
L 0.7 4-50-8-22
c'l
L [f; - +ooo psi (27.6 MPo)l
{)
Ø 6-50-8-22
0 0.2 0.4 0.6 0.8 1.0 1.5 2.0 4.0 æ

Figure 5-9 (cont.) - Effect of Specifíed Concrete Strength on the Rotio of Theoreticol to l.lominol
(c) Meon Volues
4-50-8-22
[f'. : +ooo psi (27.6 MPo)l
6-50-B-22
[f'" - oooo psi (+r.+ MPo)l
c
É 1.1
u
"9 N)
Íro H
@
L 1.O
c't
L
(D
{t)
0 0.2 0.4 0.6 0.8 1.0 1.5 2.0 4.0 æ

Figure 5.9 (cont.) - Effect of Specifíed Concrete Strength on the Rotio of Theoreticol to Ì,,lorninol
Strength of Short Composíte Steel-Concrete Beom-Columns
2l-9
At the S-percentile and mean val-ue leve1s fFigures

5"8(b) and (c) and 5.9(b) and (c) I the columns having 6000
psi (4L"4 MPa) concrete yieJ-ded lorn¡er values of strength
rat.ios at all e/h raLios" Again, the trend vras more signif-
icant for the columns with the lower percentage of struc-
tural steel"
The columns with 6000 psi (41.4 MPa) concrete produced
lower sÈrength ratios due to the lower value of the ratio of
the mean in-situ strength to specified strength of concrete
used for these col-umns than for those having 4000 psi (27.6
MPa) concrete" The in-situ strength is given by Eguation
4"2" The ratio of mean in-situ strength to specified
strength is 0.847 for 40OO psi (27"6 MPa) concrete and O.773
for 6000 psi (41"4 MPa) concrete when it j-s loaded to fail-
ure in l- hour
"
From the above-noted discussions, it is concluded that

the specified concrete strength significantly affects the
strength ratios, especially for low structural steeL per-
centages and, therefore, should be incl-uded in studies of
reliability analysis"
5.3"2.3 Effect, of, structural steel rat,io - Four direct com-
parisons lrere made to examine the effect of the ratio of the
area of structural steel to gross area of the cross-section
(p"") on the probability distribution properties of the
strengÈh rat,ios. Each comparison included two columns, one
220
T¡¡ithp"": 0.040 and one r,/ùith p"": 0"082 with all other prop-
erties being identical" The results from two of the compar-
isons are shown in Figures 5.1-0 and 5.1-l-"
Three out of four sets showed lower one-percentil-e
strength ratios for columns with 4 percent st,ructural steel
than those obtai-ned for col-umns with 8 percent structural-
steel" The only exception was the column set with Í'":
4000 psi (27"6 MPa) and kI/r = 0.0 where this trend was
reversed" For 4000 psi (27.6 MPa) concrete, the column with
P"" = 0"040 had strength ratios significantly lower than
those for the column with p"": 0"082 when e/h ratío felI
betr¿een O"25 and l-"0 as shown in Figure 5.1-0(a). For e/h
greater than 1"0 and less than O.25, there seem to be mj-nor
differences in the one-percentile strength ratios calculated
for the two columns shown in the figure" For 6000 psi (41-.4
MPa) concrete fFigure 5.11(a) ], the colunn with 4 percent
structural steel produced significantly lower one-percentil-e
strength ratios than those obtained for the coLumn having I
percent structural steel when e/h<O.6" For e/h>LO, the
trend reversed and lower one-percentile strength ratios \^¡ere
obtained for the col-umns with I percent steel, âs indj-cated
in Figure 5"11-(a) .
The l-percentile strength ratios plotted in the above-
noted figures may be explained by examination of the rel-a-
tive contributions of structural steel and concrete to the
overall strength of the column" The concrete contributes
1.0
(o) 1 -Percentile Volues
0.9
C
E 0.8
4
E
-o t\)
t\J
I ts
É
sÐ 0.7 1-50--+-22
c'l
g (ß": o.o4o)
Ð
U)
4-50-8-22
(Prr: o-os2)
0.6
0.5
o.2 o.4 0.6 0.8 1.0 1.5 2.0 4.0 co
Figure 5-10 - Effect of Structurol Steel Rotío on the Rotio of Theoreticol to Nominol Strength of
Short Cornposite Steel-Concrete Beom-Columns
(U) S-eercentile Volues
0.9
c
E 0.8
V
.9
N)
0 t\)
V t\)
0.7 B 4'-50-4-22
(¡
L (ß,: o.o4o)
q)
P
{f) 4-50-g-22
((,u= o'o82)
0.5
0 0.2 0.4 0.6 û.8 1.0 1.5 2.O 4.0 c0
Figure 5.10 (cont.) - Effect of Structurol Steel Rotío on the Rotío of Theoreticol to
Nomlnol strength of Short composite steel-concrete Beom-columns
(c) Meon Volues
+-50-4-22
(
ßr= o-o4o)
4-50-8-22
( n : 0.082)
'lse
c
E 1.1
v.
.e
t\)
v0 N)
(,
À 1.0
c'l
L
(D
L
aît
0 0.2 0.4 0.6 0.8 1.0 1.5 2.0 4.0 @

Figure 5.10 (cont.) - Effect of Structurol Steel Rotío on the Rotío of Theoreticol to
Nomínol Strength of Short Cornposite Steel -Concrete Beom-Columns
0.9
E
E. 0.8
P
æ
N)
.9
# N)
o È
E.
! 0.v 6-50-+-22
c'r
L (f'u: o.o4o)
o
L
U' 6-50-8-22
(ßr: o.082)
0.6
0.5
0 0.2 o.4 0.6 0.8 1.0 1.5 2.O 4.0 æ
End Eccentrícity Rotio (e/n)
Figure 5.11 - Effect of Structurol Steel Rotío on the Rotío of Theoreticol to Nominol Strength of
c
IE 0.8
4
v
.e
(t NJ
u. ¡u
Ltl
& 0.7 6-50-4-22
ç'l
L (ør- o.o4o)
Ð
t-
af, 6-50-8-22
(fut= 0.082)
0 0.2 0.4 0.6 0.8 1.0 1.5 2.0 4.0 æ

End Eccentrícity Rotio (e/h)
Figure 5-11 (cont.) - Effect of Structurol Steel Rotío on the Rotío of Theoreticol to
Nomínol Strength of Short Composíte Steel-Concrete Beom-Columns
(c) nøeon Volues
6-50-4-22
(ß, - o.o4o)
6-50-8-22
(ßr:0.082)
vc 1.1
æ.
.9
t\)
o N)
v Or
1.0
c'ì
L.
{)
at
0 0.2 0.4 0.6 0.8 1.0 1.5 2_O 4.0 @

End Eccentricity Rotio (e/f')
Figure 5.11 (cont-) - Effect of Structurol Steel Rotío on the Rotio of Theoreticol to
Nomínol Strength of Short Cornposite Steel -Concrete Beom-Colu mns
227
more to the overaLl strength of the column at low s¡¿ values

than at high e/h val-ues" Hence, ât J-ow e/ h larger varia-
tions take place in the overall strength of the bearn-coLumn
since the concrete strength has a higher variabil-ity than
the structural steel strength. Hence, lower one-percentile
strength ratios are expected for smaller p"" in regions of
Low e/h. At high e./h values, the contribution of the struc-
turaÌ steel becomes predominant and the variability of steel
strength is the primary cause of overaLl- beam-column
strength variations" This produces smaller overall strength
variations and, hence, less spread in one-percentil-e values
for different p"" ratios. These results concur with earlier
findings by Grant et aI. (1,978) where reinforced concrete
columns with low steel ratios v/ere found to have larger
variations than the columns with high steel- ratios.
Minor differences between the strength ratios for col-
umns having 4 percent and for columns having I percent
structural steel r^rere notj-ced at the S-percentile and mean
value levelsr âs indicated by Figures 5"i_0(b) and (c) and
5 " l-1(b) and (c) . This seems to be particularly valid for
f '" = 6000 psi (4]-"4 MPa) [Figures 5"11(b) and (c) ] " Since
the l-percentile leveI is more critical for reliability
analysis, it is reconmended that p"" be incl-uded as a vari-
able for such analyses"
228
5.3.2.4 Ef fect, of end eccent,rícíty rat,ío - Each column from

the basic study (Tab'le 5"1) \,{as investigated for 17 end
eccentricity ratios (e/h:0"0, 0"05, 0.1, 0"1-5, O"2, O.25,
0.3, o"4, 0.5, O"6, O"7, 0.9, 1.0, I"5, 2"O, 4"O and co)"
Figures 5"6 through 5.11 indicate that the strength ratios
drop sharply as e/h increases from 0.0 to O"2. The strength
ratios tend to increase then at a declining rate as e/h
increases from 0.2 to infinity (pure bending condition).
The dip in strength ratios at e/h of O"2 is the most signif-
icant at l-percentile level- and becomes less significant at
S-percentile and mean value Ievels. For l-percentile and
S-percentile strength ratios, the maximum values \^rere
obtained aE e/h æ (pure bending condition) " For mean
strength ratios, hovrever, the maximum values occurred at e/h
: 0.0 and co (pure compression and pure bending conditions),
as indicated by Figures 5"6 5"11"
For further analysis, data from the eight columns of the
basic study (Table 5.1) \úere grouped into two sets according
to the specified concrete strength" For each concrete
streñgth, the range of one-percentile, S-percentil-e, and
mean strength ratios at all e/h. val-ues studied is plotted in
Figure 5"I2 (a), (b), and (c), respectJ-ve1y" The trends for
the effect of e/h raLio stated in the preceding paragraph
are also vaLid for the plots shown in Figure 5.L2. Addi-
tionali-y, it is apparenÈ from Figure S"I2 that the upper
boundary of the range of strength ratios is defined by the
C
v. 0.8
E.
N)
.9 N)
+ \o
o
v.
g 0.7
(]r,
c
o
t-
.P
aJ) NNñ f'o = 4ooo psi (??.6 Mpa)
I 0.2 0.4 0.6 0.8 1.O 1.5 4-0 cr

End Eccentricity Rotío (./n)
Figure 5.12 - !rog" of Ratio of rheoretical to Nom.i.na} strength of
v. 0.8
v.
N)
,9 UJ
+ O
o
u.
c o.7
(t¡
E
o)
L-
P
U'
0 0.2 0.4 0.6 0.8 1.O 1.5 2.O 4.0

End Eccentricity Rotío (e/n)
Figure 5.12 (cont.) - Rangé of Ratio of Theoretiea3 to Nornina[ S{rengt^h of
(c) Meon Volues
[Nl f'"= 4ooo psi {2?.6 Mpa}
ç
É. 1.1
M.
N-)
-9 UJ
# ts
o
v_
..c 1.0
+
u,
c
0)
1-
P
tn
0 0.2 0.4 0.6 0.8 1.O 1.5 2_O 4.0 co

End Eccentricity Rotío (e/n)
Figure 5.12 (cont.) - Raage of Ratio of Theoretican to Nominal Strength of
Short Composíte Steel-Concrete Beorn-Columns
232
columns having 4oo0 psi (27 "6 MPa) concrete and the lower
boundary is defined by the columns with 6000 psi (4I"4 MPa)
concrete" The maximum spread in one-percentile, S-percen-
tile and mean strength ratios occurs aE e/hof 0 and 0"2,
whereas the ninimum spread in these values takes place at
e/h = æ (pure bending), âs indicated by Figure 5"1,2"
The range of the strength coefficients of variation for
the two sets of columns discussed above is plotted on Figure
5"1-3. The coefficients of variation are the greatest al e/h
of 0.0 and decrease only slightly between e/h of 0.0 and
o.2" The coefficients of variation then decÌine sharply at
a declining rate as e/h increases from 0"2 to infinity. The
largest spread between the minimum and maximum values of the

coefficient of variation occurs for columns having 4000 psi
(27"6 MPa) concrete when e/h l-ies between 0 and 0.2.
AE e/h S O"7, the range of coefficients of variation for
columns having 6000 psi (4L"4 MPa) concrete falI within the
range of coefficients of variatj-on for columns with 4000 psi
(27"6 MPa) concrete. This is expected because the 6000 psi
(4I.4 MPa) concrete was assumed to have excellent quality
control as opposed to average quality control for 4000 psi
(27.6 MPa) concrete and because the concrete çfuality has an
influence on variability of beam-concrete strength when e/ h
is not high. As expected for values of e/ ä higher than 0.7 ,
0.20
f
.cl
Þ 0-x.8
c¡
å
U)
0.r8
o
L) 0_n4
I
Þ
d
{¡)
ca t\)
0.12 UJ
o UJ
o
d 0.L0
&
ø
Þ
q{
o 0-08
.a.¡
c)
C)
t¡.{ 0.06
q-1
€)
o
c)
0.04
0 0.2 0.4 0.6 0.8 1.O 1.5 2.0 4.0 m
End Eccentricíty Rotío (e/n)
Figure 5.13 - Ronge of the Coeffícient of Voríotíon of the Rotio of Theoreticol
to Nominol Strength of Short Composite Steel-Concrete Becm-Columns
234
the range of the beam-column strength coefficients of varia-

tion remained practically unaffected by the specified con-
crete strength" This was particularly valid for pure
bending condition (e/h:-), as indicated by Figure 5.1-3"
From the above discussions, it is obvious to concl-ude
that the end eccentricity ratj-o is a variabl-e which needs to
be considered in reliability studies" End eccentricity
ratio effects both the beam-column strength ratios and the
coefficients of varj-ation to a significant extent" The end
eccentricity ratios below 0.55 are especially critical- since
these e/ h raLíos produce one-percentile strength ratios that
fall below O"7, as indicated by Figure 5.1-2(a) "
5.3 2.5 Sensitivity analysis - The portions of the overal-l-
o
variability of the beam-column strength attributable to the

variations in the mechanical properties of concrete, the
mechanical and geometric properties of structural steel-, and.
the theoretical strength model error \¡/ere determined for a
typical beam-column cross-section. Column 4-50-4-O (Tab1e
5"1) was chosen for this analysis" To determine the bearn-
column strength variability due to each of the three sets of
variables noted above, three separate computer runs of 5oo
simulat,ions each hrere made. For each computer run, only the
variables from one of the above-noted sets were allowed to
vary while the remaining variables were kept constant at
their mean value" the portions of the overall variabiJ-ity
o.028
ô
V"colsô
o.024
VE
Ð coþônc
0)
t-
o 0.020 v8rßgd
0
.1,
Ê o V-EUITI
.9 fEquotion (5 2)]
o.o16
o ar8
t- v nR
o N)
L,J
(tl
o o.012
Ê
.9
.9
û) o.008
o o
O
0.004
0 0.2 0.4 0.6 0.8 1.O 1.5 2.O 4-O co

End Eccentricity RotÍo (e/h)
Figure 5.14 - Effect of Voriobilities of Properties of Constituent Moteriols on Overoll
Strength Voriobility of Beom-Column Cross-Section 4-50-4-0 (Toble 5.1)
236
of the beam-column strength so determined v/ere respectively

designat,ed as V V co¿""t and V for the three sets of
"o¿con"t ô¿n¿
variables noted above.
The sguares of the coefficients of variation V2"ot"on",
V?o1",, and. I/floon, computed for column 4-SO-4-O at e/h values

ranging from O.05 to infinity are plotted on Figure 5"I4.
These plots indicate that the overall variability of the
strength of the above-noted column cross-section is mostJ_y
influenced by the variations in the theoretical strength
model and the concrete mechanical properties for e/hS I.O,
and by the variations in the structural steel section prop-
erties for e/h: æ. For 1"0 < e/h1æ, aII three variations
(V V
"o*on"r "orrrt
and I/ mo¿et) seem to contribute to the overall-
strength variabiLity of column cross-section 4-50-4-0. Note
the effects of the variability of the theoretj-cal strength
moder and of the concrete properties decrease significantly
as the end eccentricity ratio increases. The effect of the
variability of the structural steel properties increases
somewhat as e/ h increases from 1.0 t,o higher values.
The values of Vf,u^ plotted in Figure 5. i-4 represent the
sum of the squares of the coefficients of variation of col-

umn 4-50-4-0 strength obtained from individual variabil-ities
of the three sets of variabilities, i.e.
237
v2
' sum
l/2
' colconc
+ v2
" colss
+ v2
" model (s.2)
Also plotted in Figure 5. 14 are the values of V 2p

which is
the sguare of the coefficient of variation of the beam-

column cross-section strength obtained rdhen variabilities
from all sources r,üere included simultaneously in
computations" A comparison of V?,^ and. VzR plotted in Figure
5"1-4 indicates a very good correlation of these values up to
e/h of 0"3. V?u^ only slightly underestimates V'r for e/h
greater than 0"3" This underestirnation by V?,^ at e/hva:.-
ues higher than 0"3 is like1y caused by the variations in
the properties of the reinforcing steel and perhaps by the
variations in cross-section geometry which were not included
in V!,^ fEguation (5.2) ]. This confirms an earlier finding
by Mirza (1-989) that the cross-section dimensions have neg-
ligible effect on composite column strength variabil-ity.
Grant et al. (1978) reported sirnilar conclusions for
reinforced concrete columns. the effects of variations in
properties of reinforcing bars were insignificant in this
study because the reinforcing steel ratio (p," = 0.01-2) was
much smaller than the structural steeL ratio (p"": 0"040 and
0"082) used for the cornposit,e cross-sect.j-on"
238
5"3.2.6 Sumaarv of effects of variables used for basic studv

- The followingi sunmarizes the effects of variables used for
the basic study of short composite beam-columns:
(a) Slenderness ratio of 22 is critical- for the type of
short columns studied;
(b) the specified concrete strength is a major variable;
(c) the ratio of structural steel area to gross cross-
sectional area is a significant variable;
(d) the end eccentricity ratio has a very significant effect
on the strength ratios and is especially critical in the
range from 0"0 to 0.55; and
(e) the overall varj-ations of the beam-column theoretical
strength are primarily due to the variations in the
mechanical properties of the concrete, the geometric and
mechanical properties of the structural steel, and the
theoretical modeI"
5"3"3 Effect,s Of Varíables Used For SuÞÞlemental Studv
From the short colurnns used for the supplemental study
(Tab1e 5"3), the effects of the specifíed yield strength of
structural steel, the strain-hardening of structuraL steel
section and reinforcing bars, and the guality control of
concrete on the beam-column strength ratios ri¡ere studied.
Plots of the one-percentiLe, five-percentile and mean
strength ratios at various values of e/h were made for each
variable studied"
239
5"3"3"1 Effect of specífied vield strenq,t,h of st,ructural-

steel - To study the effect of the specified yield strength
of structural steel on the strength ratios, two short col-
umns from the basic study (Columns 4-50-4-22 and 6-50-4-22
in Tab1e 5"1) were compared to four columns from the
supplementaL study (Columns 4-44-4-22, 4-36-4-22, 6-44-4-22,
and 6-36-4-22 in Table 5"3) " This provided two sets of
three columns, each column with structural stee1 f" of 50,
44, or 36 ksi (345, 303, or 248 MPa) " All other properties
of the three columns in a set were identicat. Figures 5.15
(a), (b), and (c) respectively plot the one-percentile,
five-percentile and mean strength ratio data for one of the
sets noted above.
Àt the 1-percentile level, the lowest strength ratios
vrere found for the columns having 50 ksi (345 MPa) and 36
ksi (248 MPa) structural steel which plot al-most identically
aE e/ h values below O"4" The one-percentile strength ratios
for the columns with 44 ksi (303 MPa) structural steel plot
somewhat higher in this range of e/h. Between e/h of o.4
and 0"8, the one-percentile strength ratio data for the col--
umn with 50 ksi (345 MPa) structural steel- rises slightly
higher than the data for the columns with 36 ksi (248 Mpa)
structural steel. At the pure bending condition, all three
beam-columns have similar one-percent.iLe strength ratios, âs
indicated by Figure 5"1-5(a) "
c
M. 0.8
v
.9
N)
o rÞ
v. o
& 0.7 4-36-4-22
c'l
L [Structurol steel f" - 36 ksi (2aB MPoX
Ð
.P
(rt 4-4+-4-22
fstructurol Steel f" - M ksi (303 MPo)l
4-50-4-22
[Structurol Steel f" : 50 ksi (345 MPo)l
0.5
0 o.2 0.4 0.6 0.8 1.0 1.5 2.0 4.0 @
Figure 5.15 - Effect of Specifíed Structurol steel Yield Strength on the Rotio of
Theoreticol to Nominol Strength of Short Composite Steel-Concrete Beorn-Columns
c
É 0.8
M
.9
N)
o
0l 'Þ
ts
! o.7 +-36-4-22
c'l
(¡)
flstructurol steel f, : 36 ksi (248 MPo)I
L
th +-44-4-22
[Structurol Steel f" - 44 ksi (303 MPo)l
4-50-+-22
[Structurol Steel f" : 50 ksi (345 MPoX
0.5
0 0.2 0.4 0.6 0.8 1.0 1.5 2.O 4.0 @
Figure 5.15 (cont.) - Effect of Specified Structurol Steel Yield Strength on the Rotio of
Theoreticol to Nominol Strength of Short Cornposite Steel-Concrete Beom-Columns
(c) Meon Volues 4-36-+-22
[Structurol steel fu = 36 ksi (2aB MPo)l
4-44*4-22
[Structurol Steel f, : 44 ks¡ (303 MPo)l
+-50-4-22
c [Structurol Steel f" : 50 tsí (3+5 MPo)I
E 1.1
M
.9.
N)
Ð rÞ
v. N)
L r.0
c'l
L
g
Lft
0.8
0 0.2 0.4 0.ô 0.8 1.0 1.5 2.0 4.0 @
Figure 5.15 (cont.) - Effect of Specified Structurol Steel Yield Strength on the Rotio of
Theoreticol to Nominol Strength of Short Composite Steel-Concrete Beom-Columns
a^')
At the S-percentile and mean val-ue leve1s, the data for

columns with all three grades of structural steel plot cl-ose
to each other lFigures 5.15 (b) and (c)1" The only noted
differences were at e,/h values between 0"3 and 1.0" How-
ever, these differences v¡ere not considered significant"
Sinilar conclusions \,rere obtained frorn the strength
ratio dat,a of the remainj-ng set of beam-columns used to
investigate the effect of the specified yield strength of
structural- steel" Since differences between strength ratios
for columns with three different grades of steel are minimal
at the one-percentile l-evel, it is reconmended that 50 ksi
(345 MPa) structural steel be used in future reliability
analysj-s of composite cross-sections. The 50 ksí (345 MPa)
structural steel is the highest steel grade presently
alLowed by the design codes (ACI 3L8-83 and CAN3-A23.3-M84)
for composite columns and will ensure relevancy as conmon
steel grades increase above the present values.
5.3.3.2 Ef f ect, of straín harðentntof steel - To examine the
effects of straj-n-hardening of the structural steel and of
the vertical reinforcing bars on the st,rength ratios of com-
posite cross-sections, the data from two col-umns taken from
the basic study (Columns 4-50-4-0 and 6-50-4-0 in Table 5.1)
hlere compared to the data from the corresponding' columns of
the supplemental study (Co1umns 4-50-4-0-STH and
6-50-4-0-STH in Table 5"3). NoÈe the strain-hardening of
c
u. 0.9
e
É
.g
g N)
E, È
'N
å 0.8
crl
L
Q)
U)
+-50-4-0
(Stroin-ho rdening neglected)
+-50-4-0-sTH
(Stroin-hordeníng considered)
0 0.2 0.4 0.6 0.8 1.0 1 .5 2.O 4.0 @

F-igure 5-16 - Effect of Stroin-Hordening of Steel on the Rotío of Theoreticol to
Nominol Strength of Short Composite Steel-Concrete Beom-Columns
c
IE 0.9
É
.o
t\)
gÐ È
(¡
L 0.8
c't
{t)
1¿
IJ'
+-50-4-0
(Stroin-ho rdenin g neglected)
4-50-4-0-STH
(Stroín-hordening considered)
0.6
0 0.2 0.4 0.6 0.8 1.0 1.5 2.0 4.0 æ
Figure to
5.i6 (cont.) - Effect of Stroin-Hordening of Steel on the Rotio of Theoretícol
Nomínol strength of short composite Steel-concrete Beom-columns
c
E 1.1
É
.9
t\)
0 È
É. Or
1.0
C¡
L
€)
V)
4-50-4-0
(Stroin-ho rdening neglected)
4-50-4-0-STH
0.8
0 0.2 0.4 0.6 0.8 1.0 1.5 2.O 4.0 co
End Eccentricity Rotio (e/n)
Figure 5.16 (cont.) - Effect of Stroín-Hordening of Steel on the Rotio of Theoretícol to
Nomínol strength of Short composíte Steel-concrete Beom-columns
247
both steels was included in theoretical strength computa-

tions of Columns 4-5O-4-O-STH and 6-50-4-O-STH, while the
strain-hardening of steel was neglected for Columns 4-50-4-0
and 6-50-4-0" This provided two sets of columns, each set
having one column in which strain-hardening was included and
one coLumn in which strain-hardening was not pernitted. All
other properties lrere identical for both columns in a set.
The strength ratios for columns from one of these sets are
shown in Figure 5.16"
The plots in Figures 5.16 (a), (b), and (c) show no
effect of strain-hardening of steel on one-percentiJ-e, five-
percentile and mean strength ratios aE e/h values less than
or equal to around 0.6. Between e/h of 0.6 and 1.5, there
is some increase in strength ratios of the column in which
strain-hardening was included in theoretical strength compu-
tations. At pure bending, however, the strain-hardening of
steel produced significantly higher strength ratios (roughly
in the order of L5 percent) at one-percentile, five-
percentile and mean value leve1¡ âs indicated by Figure
s"1-6"
A similar behavior was observed from the remaining set

of composite cross-sections used to investigate the effect
of strain-hardening of both steels" The data clearly showed
that an improvement in the strength ratios for cross-
sections occurred only at or near the pure bending condj-tion
v¡hen strain-hardening of steel r¿as used" This is expected
248
since the steel- must plast.ical-ly strain t.o a Level approxi-

mately ten times the yield strain before the beneficial-
effect of strain-hardening can be obtained which can only
occur at high l-oad eccentricities" It is, therefore, recom-
mended that strain-hardening not be account.ed for in future
rel-iability analysis of short composite beam-coIumns.
5"3"3.3 Effect, of quality of concrete - In the basic study
(Table 5"1), the concrete guality for four columns with /'.
: 6000 psi (41-.4 MPa) v/as assumed to be excell-ent" this
corresponds to the control cylinder strength coefficient of
variation of L0 percent. The average guality control of
concrete assumes a coefficient of variation of 15 percent
for the control cylinder strength. To study the effect of
average guality control on strength of short cornposite beam-
columns having 60OO psi (4L.4 MPa) concrete, the strength
ratios for Column 6-50-4-22 (Table 5"1) were compared to
those for Column 6-50-4-22-A (Table 5"3) " Similar compari-
sons vtere also made for the strength ratio data obtained for
Columns 6-50-8-22 (Tab1e 5"1) and 6-50-8-22-A (Table 5.3).
Note Columns 6-50-4-22 and 6-50-8-22 had excellent guality
concrete while Columns 6-50-4-22-A and 6-50-8-22-A employed
average guality concrete. This produced two sets of col-
umns, each set had one column with excellent and one column
with average guality concrete. All other properties were
ident,ical for both col-umns in a sei. The one-percenÈile,
249
five-percentile and mean strength ratios for columns in one

of the above-noted sets are plotted in Fj-gures 5"L7 (a) ,
(b), and (c), respectJ-vely.
Figrures 5"L7 (a) and (b) show that the average concrete
guality control produces significantly lower one-percentil-e
and five-percentile strength ratios over the entire range of
load eccentricities than does the excellent concrete quality
control. The lower one-percentile and five-percentil-e
strength ratios can be attributed directly to the larger
coefficient of variation associated with average guality
concrete" The effect is significant for end eccentricity
ratios up to roughly L"5 and is negligible at the pure bend-
ing condition" This is expected since the concrete contrib-
utes little to the overall strength of the col-umn under pure
bending. At the mean value IeveI, there is virtually no
difference between the strength ratios obtained for col-umns
with excel-lent and average concrete guality controls. This
is expected since the mean values of the concrete strength
are not affected by the guality control" Sinilar results
were obtained from the analysis of strength ratio data for
the columns in the other set used to investigate the guality
control of concrete.
The comparisons of the strength rat,ios in Figure 5"L7
show that the concrete quality control significantly affects
ihe lower tail of the strength probability distribution and,
(o) t-Percentile Volues
M, 0.8
M.
.9
N)
o (¡
E, o
¡- 0.7
c'r
L
()
tb
tf)
6-50-4-22
(Excellent concrete quolity)
6-50-4-22-A
(Averoge concrete quolity)
0 0.2 0.4 0.6 0.8 1.0 1.5 2.0 4.0 æ

Figure 5.17 - Effect of Concrete Quolity Control on the Rotio of Theoretícol to
Nominol Strength of Short Composite Steel-Concrete Beom-Columns
(U) S-eercentíle Volues
c
M 0.8
tr
.9
t\)
o Ltl
æ. F
& o.7
ç'ì
L
{)
(.t)
6-50-4-22
6-50-4-22-A
0.5
0 0.2 0.4 0.6 0.8 1.0 1 .5 2.0 4.0 æ
Figure 5.17 (cont.) - Effect of Concrete Quolity Control on the Rotío of Theoreticol to
Nomínol Strength of Short composite Steel-concrete Beom-columns
6-50-4-22-A
(Averoge concrete quolíty)
vc 1.1
É
.9
0 N)
LN
v. t\)
g t.0
c'¡
g
U'
0.8
0 0.2 0.4 0.6 0.8 1.0 I .5 2_O 4.0 @
Figure 5.17 (cont.) - Effect of Concrete Quolíty Control on the Rotio of Theoreticol to
Nomínol Strength of Short Cornposíte Steel-Concrete Beom-Columns
253
hence, it is important that the future reliability analysis

considers the concrete quality control as one of the vari-
abl-es "
5"3"3"4 Summarv of effects of variables used for sunnlemen-

tal" study - The following sunmarizes the effects of vari-
ables used for the supplemental study of short composite
beam-columns:
(a) The use of 50 ksi (345 MPa) structural steel produces
strength ratios not significantly different from those
obtained for lesser grades of steel;
(b) the strain-hardening of steel enhances the strength
ratios of the beam-columns only at very high e/h values
and is not recommended for inclusion in the reliability
analysi-s; and
(c) the guality control of concrete significantly affects
the l-ower tail of the strength probability distribution
and should be included in the reliability analysis"
5 " 4 SLENDER COMPOSITE BEã¡'Í-COLUSINS
Thís section examines the overall strength variations of

slender composite beam-co1umns. The effects of individual
variables on the strength ratios are discussed here and the
major variables that affect the variability of slender com-
posite columns are identified as well" The specified prop-
erties of slender bearn-columns studied are given in Tabl-es
5"2 and 5"4"
254
5.4 " IOverall Strength Variat,ions

Axial load-bending moment interaction diagrams simul-ated
for two slender beam-columns (Colunn 6-50-4-66 and Column
4-50-8-33) taken from Table 5"2 are plotted in Figures 5.18
and 5"1-9" These columns represent the upper and lower lin-
its of the st,ructural steel index studied (p""-f '" = 0.33
"/f
and l-"03). Plots included are of the maximum, mean,
one-percentile, and minimum theoretical strengths as wel-l as
the ACI 3L8-83 ultinate and factored strengths computed for
the slender columns" The ACI ultimate strengths were based
on all understrength factors being equal to l-.0" The ACI
factored strengths r¡/ere calculated as outl-ined in Sectj-on
5"3"L. AIso included in these figures are the theoretical-
mean cross-sectional strength curves which are plotted only
for comparison to the mean strength of the slender beam-
columns"
Figure 5. l-8 (plotted for Co1umn 6-50-4-66 having
P""-/v/f '":0"33 and kI/r = 66) shows the ACI ultj-mate
strength significanÈly underestimating the mean theoretical
strength for e/h between 0"0 and 0"3. For e/h greater than
0"3, the ACI ult,imate strength predicts strengths slightly
higher than the mean theoretical strength" At the pure
bending condition, the ACI ultirnate strength prediction is
nearly identical to the mean theoretical strength" The fac-
tored .ACI st,rength curve plott,ed is very significantly less
Note: 1 kip : 4.4+8 KN
Maximum
oo ^9 "4) -"/
">kr''.Y
Ít, ^úï"'2il::;;i;riÏ:"
ô_
v
T
o
o.t?1 o't1t
u*trñ f\)
o 2000 Ltl
J Ltl
-g
i-siþ
x Cross-Sectio
Strength
Mi-nimum
't000
: l--Perceri
ACT Factor
0 200 400 600 800

Bending Moment (Ft-Kips)
Figure 5..l8 - Axio! Lood-Bending Moment Strength tnteroction Curveä of Rondomþ Generoted
Somple,. of 500 Slender Composite Cotumns fColumn 6-50-4-66 ( Toble 5.2)l
Note: 1 kip : 4.448 KN
^^t
sì ! ,/ 1 ft-kíp = 1.356 KN-rn
Maximum
íf ,%- orulrimare
Cross-Section
Strength
Ul-t imate
^ o6s -/
ù,
Ê-
v 3000
-o
o N)
o Mean (¡
J Or
Cross-Section
.g Strength
2000
it = 0'077
I:-ãoå;::: - ;0.à"
Kurtosas =
l--Percen
Mini-mum
e/h=1,.0
Factored
200 400 600 800 1 000 1200

BendÍng Moment (Ft-Kips)
Figure 5.19 - Axiol Lood-Bending Moment Strength lnteroclion Curves of Rondomly Generoted
Somplel of 500 Slender Composíte Columns [Column 4-50-8-33 .( Toble 5.2)]
257
than the 1-percent.il-e strength for all values of e/ h"

However, the load combinations and variations may decrease
the apparent conservativeness of the ACf factored strength.
In Figure 5"1-9, (plotted for Co1umn 4-50-8-33 having
p""-fy/f '": l-"03 and kl/r = 33), the ACI ultirnate strength
prediction somerÀ¡hat overestimates the mean strength Lor e/h
of 0"0 to 2"O" For e/h between 2.0 and infinity, the mean
strength is somewhat underestimated by the ACI ultimate
strength" The differences between the ACI ultinate strength
and the mean theoretical strength are much less in Figure
5.L9 than those displayed in Figure 5.L8 and, perhaps,
reflect the effect of lower slenderness ratio associated
with the column data plotted in Figure 5.19" The factored
ACI strength is less than the L-percentile strength at all
values of e/h as indicated in Figure 5.1-9.
Probability distribution histograms of the simulated
theoretical strengths for a/h of 0.1, O.2 and 1"0 are also
plotted in Figures 5.LB and 5"l-9" The histograms are sym-
metric" This indicates that the positively skewed shape of
the structural steel strength probabil.ity distrj-bution does
not influence the overall slender beam-column strength as
much as iÈ does for short columns" This can be seen by
comparing Figures 5.4 and 5 " t-9 "
5.4"2 Effect,s Of Variables Used For Basíc Study
The 1-6 slender eolumns in the basic study (Table 5.2)
were chosen to investigate the effect of four variabl-es on
258
the probability distribution properties of the ratio of

theoretical to nominal strength (strength ratio). These
variables are the slenderness ratio (kI/r), specified con-
crete strength (f'") , ratio of structural steel area to
gross area of cross-section (p""), and end eccentricity ratio
(e/h) " The strength ratios are plotted for e/h values in
the range of 0"05 to co" The strength ratios at e/h of 0.0
(concentric loading) are not shown since the concentric
capacity prediction of the theoretical strength model was
based on the tangent modul-us approach and the resul-ts were
believed overly conservative"
5.4u2.X Effect of slenderness ratio - The sl-ender beam-
columns shown in Tab1e 5.2 r¡¡ere divided into 4 sets of 4
columns each to investigate the effect of slenderness ratio
(kl/r) on the strength ratio. Each set had one column with
slenderness ratio of 22.1-, 33, 66 or l-OO. All other proper-
ties in each set of columns hrere identical. A kI/ r of just
greater than 22 is the minimum slenderness ratio requiring
the inclusion of length effects by ÀCI 3L8-83 and by CAN3-
423.3-M84 for the types of columns studied" A kl/r of l-00
is the maximum slenderness ratio all-owed by the two design
codes for evaluating the stability effects by the moment
magnifier approach. Figures 5"2O and 5"2I plot the one-
percentile, five-percentile and mean strength ratios for two
of the col-umn sets studied"
1.5
1.4 (o) 1 -Percentile Volues 4-50-4-22-1 4-50-4-66

(kt/r - 22.1) (klrlr = 5'6¡
1.3
4-50-4-33 4-50-4- 1 00
(kllr : 33) (kllr : 100)
1.2
L
M 1.1
V,
"9 1.0
t\)
o ur
æ. to
-c
+ 0.9
olt
c
o
r,Íl 0.8
o.7
0.6
0.5
0 0.2 0.4 0.6 0_8 1.0 1 .5 2.0 4.0 æ
Figure 5-20 - Effect of Slenderness Rotio on the Rotio of Theoreticol to Nominol Strength of
Slender Composite Steel-Concrete Beom-Columns
1.6
r.5 (b) 5-Percentile Volues 4-50-4-22.1 4-50-4-66

(ut¡r : 22.1) (kllr : 66)
1.4 4-50-4-33 4-50-4-100
(tllr - 33) (kllr : 100)
1.3
ç
fts 1.2
v
.9 1.1
t\)
o Or
M. o
.E 1.0
¡'l
c(¡)
v 0.9
an
0.8
0.7
û.6
0 0.2 0.4 0.6 0.8 1.0 1 .5 2.0 4.0 @
End Eccentricity Rotio ("/n)
Figure 5.20 (cont.) - Effect of Slenderness Rotío on the Rotio ofTheoreticol to Nominol Strength of
1.9
(c) Meon Volues
r.8 4-50-4-22.1 4-50-+-66
(ut¡r : 22.1) (kllr : 66)
1.7
4-50-4-5J 4-50-+-100
(kllr : 33) (kllr : I O0)
1.6
vC 1.5
P
M.
.9 1.4
o N)
fr Or
ts
I 1.3
#
ur
tr
E
Ø 1.2
1.1
't.0
0.9
o.2 0.4 0.6 0.8 1.0 1.5 2.0 4.0 @
Figure 5.20 (cont.) - Effect of Slenderness Rotio on the Rotio ofTheoreticol to Nominol Strength of
Slender Composite Steel-Concrete Beorn-Columns
(o) 1-Percentile Volues 6-50-8-22.1 6-50-8-66
(Ut¡r : 22.1) (xtlr : 66¡
ô-50-8-35 6-50-8- 1 00
(t<t/r : 33) (ullr : t 00)
L
V,
frë
.9 0-8
N)
o Or
V N)
c
g
ct,
c
st 0,7
{n
0 0.2 0.4 0.6 0.8 1.0 1.5 2.0 4-O @

Fìgure 5.21 - Effect of Slenderness Rotio on the Rotio of Theoreticol to Nominol Strength of
(U) S-Percentile Volues
c
M,
v
.9
g
0-B
t\)
o Or
fE (¡)
.,c
Ut
c 6-50-8-22.1 6-50-B-66
Ð
L. o.7
t/) (ut¡r : 22.1) (kllr = 66)
6-50-8-J3 6-50-8-10C1
(kllr : 33) (kllr : 100)
0 0.2 0.4 0.6 0.8 1.O 1.5 2.0 4.0 m

Figure 5.21 (cont-) - Effect of Slenderness Rotio on the Rotio ofTheoreticol to Norninol Strength of
(c) Meon Voluee
6-50-8-22. r 6-50-8-66
(Ut¡r : 22.1) (kllr - 66)
6-50-8-33 6-50-8-100
(l<t/r : 33) (kllr - 100)
c
V
É.
.9 1-1
N)
o Or
É. ,À
€
El!
c
o) 1.0
¡-
6
0 0.2 0.4 0.6 0.8 1.0 1.5 2.O 4.0 cÐ

Figure 5.21 (cont.) - Effect of Slenderness Rotio on the Rotio ofTheoretícol to Nominol Strength of
265
In nost cases, one-percentil-e st,rength ratios for col-

umns with kl/r of 33 lrere the lowest, followed by those with
kI/r of 22.1- and 66" The highest one-percentiÌe strength
ratios \dere obtained for col-umns with kl/r : 100. fn all
cases, the differences in strength ratios of colurnns with
differenL k¿/r decreased as e/h íncreased from 0"05 to L.2"
For e/ h values greater than I"2, there \,{ere no significant
differences in the strength ratios of beam-columns with dif-
ferent kl/r ratio regardless of e/h vaIue" The effect of
slenderness ratio described here was typical for all four
sets of beam-columns and is evident from Figures 5.20(a) and
s.21(a) "
Figure 5"20(a) shows that one-percentile strength ratios

for the beam-column with kt/r of 33 are significantly lower
than those for the beam-column with kl/r of ZZ"t when e/h is
less than 0"3" However, for higher values of e/h, the dif-
ferences in one-percentile strength ratios for these two
columns (kt/r = 22"L and 33) are small [Figure S"2O (a) ]"
This behavior is different from the one observed from Figure

5"21-(a) in which one-percentile strength ratios for columns
with kl/r = 22"I and 33 show small- differences over the
entire range of e/h, studied. This difference in behavior of
the two sets of beam-colurnns is, perhaps, due to different
structural steel ratios used [p"" = 0"040 for columns in Fig-
ure 5"20(a) and 0"082 for columns in Figure 5"zt(a) I"
266
The trends noted in the preceding paragraphs were found

to be consistent with the S-percentile and mean values of
the strength ratios for all four sets of columns studied"
This is indicated by Figures 5"20(b) and (c) and Figures
5 " 2l- (b) and (c) "
From the foregoing discussions, it is concl-uded that the

strength ratios for columns r¡ith k|/r = 2Z"I to 33 are more
critical than those for columns r¿ith kl/r = 66 to 1oO" In
most cases, hortever, the lowest strength ratios T¡rere
obtained for columns having kl/r = 33" Slenderness ratio
is, therefore, an important variabl-e for reliability analy-
sis "
5"4,2.2 Effect of specified concrete st,rength - The sl_ender

columns risted in Tabre 5"2 provided 8 sets for investigat-
ing the effect of specified concrete strength on the
strength ratio. Each set contained one column having f', =
4000 psi (27.6 MPa) and one column having f'":6000 psi
(41"4 MPa) with all other properties of both corumns in the
set, being ident,ical. Figures S"2Z and S"Z3 plot the
st,rength ratio data for two of the eight sets.
The strength ratios obtained for slender columns with
the lowest slenderness ratios (kt/r = 22"L) were expected to
be similar to those obtained for the short columns. In
fact, the two set,s of slender columns with kl/r:22. l used
to study the effect, of f'" on strength ratios giave results
nearly identicar to those obtained for the short corumns.
c
æ. 0.8
v
.9
N)
Þ o\
É {
! o.7 4-50-+-22.1
oì
L [f' = 4000 psi (27.6 MPo)l
o
U) 6-50-4-22.1
[f'= 60O0 psi (41.4 MPo)l
0 0.2 0.4 0.6 0.8 r.0 1 .5 2.0 4.0 @

End Eccentricity Rotio (ein)
Figure 5.22 - Effect of Specified Concrete Strength on the Rotio ofTheoreticol to Nominol
Strength of Slender Composite Steel-Concrete Beom-Columns
c
u. 0.8
M,
.9
N)
o Ot
M æ
¿- 0.7 4-50-4-22.1
c't
L [f;: +ooo psi (27.6 MPoX
{)
tf, 6-50-4-22.1
0 0.2 0.4 0.6 0.8 1.0 1.5 2.0 4.0 @

Figure 5-22 (cont.) - Effect of Specified Concrete Strength on the Rotío ofTheoreticol
to Nominol Strength of Slender Composite Steel-Concrete Beom-Columns
1.3
(c) Meon Volues

4-50-4-22.1
[f" - +ooo psi (27.6 MPc)l
1.2
6-50-4-22.1
vc 1.1
É
"9
+) N)
o Oì
E \o
.L 1.O
c'l
g
(¡)
u)
0.9
0.8
o.4 0.6 0.8 1.0 r.5 2.0 4.0 @
Figure 5-22 (cont.) - Effect of Specífied Concrete Strength on the Rotio ofTheoreticol
to Nominol Strength of Slender Cornposite Steel-Concrete Beom-Columns
c
É. 0.8
v
"9 t\)
uo \¡
C)
o.7 4-50-4-33
o'¡
L [t' : +OO0 psi (27.6 MPo)l
o
{-ñ
6-50-4-53
[f'= 6000 psi (+r.+ MPo)l
0 0.2 0.4 0.6 0.8 1.0 1.5 2.0 4.0 cÐ

Figure 5.23 - Effect of Specified Concrete Strength on the Rotio ofÏheoreticol to Nominol
Strength of Slender Composite Steel-Concrete Beom-Columns
1.0
0.9
c
É. 0.8
É
.9
o t\)
E \.1
H
0.7 4-50-4-33
crl
L
q) [f; - +ooo psi (27.6 MPo)l
tfi 6-50-4-33
[f; = oooo psi (41.4 MPo)l
0.6
0.5
0.6 0.8 1.0 1.5 2.0 4.0 @
End Eccentricity Rotio (e/f¡)
Figure 5.23 (cont.) - Effect of Specífíed Concrete Strength on the Rotio ofTheoreticol
to Nominol Strength of Slender Cornposite Steel-Concrete Beom-Columns
(c) Ueon Volues 4-50-4-33
[r; : +ooo psi (27.6 MPo)l
6-50-4-33
[f; - oooo psi (4r.+ MPo)l
c
E 1.1
É
.9
N)
o \¡
É N)
1.O
c'r
L
g
th
0 0.2 0.4 0.6 0.8 1.0 L5 2.0 4.0 c0

Figure 5.23 (cont.) - Effect of Specífied Concrete Strength on the Rotio ofTheoreticot
273
This can be seen by comparing the strength ratio data in

Figure 5"22 for one of the slender column sets to the plots
in Figure 5.8 presented earli-er for short columns.
Figure 5"22(a) plotted for slender columns having kI/r:
22"L and p"": 0"040 shows that the 6000 psi (4L"4 MPa) con-
crete giave lor¿er one-percentile strength ratios than did the
4000 psi (27 "6 MPa) concrete over almost the entire range of
e/h. A similar behavior was observed for one-percentile
strength ratios of slender columns having the same slender-
ness ratio but a higher structural- steel ratio (p"": 0.082),
except that the 4000 psi and 6000 psi (27 "6 and 4I"4 Mpa)
concretes produced identical one-percentile strength ratios
for e/hS 0.25" At S-percent,ile and mean value levels, the
columns with higher strength concrete gave lower strength
ratios for all e/h values, although the effect was more sig-
nificant for columns with lower structural steel ratio as
indicated by Figures 5"22(b) and (c) "
As the sl-enderness ratio v¡as increased to 33, the ef fect
of specified concrete strength on the one-percentile and
five-percentil-e strength ratios tended to disappear as indi-
cated by Figures 5.23(a) and (b)" However, the higher
strength concrete produced lower mean strength ratios for
columns with kl/r = 33 fFigure S"Z3 (c) ] "
For beam-columns having slenderness ratj-os of 66 and
L00, the effect of the specified concreÈe strength on
274
strength ratios was noticeable only at e /h of less than

0.1-5" In this range of e/h, the lower concrete strength
provided slightly lower strength ratios" This effect dissi-
pated rapidly with increasj-ng e / h due to the faj-lure mode
changing from compression to tension caused by increasing
secondary bending moments acting on the beam-coIumn.
From the data and discussions presented in this Section,
it is concluded that for columns having slenderness ratios
greater than or egual to 33, the specified concrete strength
does not significantly affect the strength ratj-o. For slen-
der columns with k¿lr less than 33, the effect of specifi-ed
concrete st.rength seems to be significant"
5"4"2.3 Effect of st,ructural st,eel ratio - Eight comparisons
$/ere made to investigate the effect of ratio of structural
steel area to gross area of concrete cross-section on the
probability distribution properties of the strength ratios
of the beam-columns listed in Tab1e 5"2" Each compari_son
involved two beam-columns, one having p"" = O"O4O and the
other one having p"": 0.082 with all other properties being
identical" Strength ratios from two typicaÌ comparisons are
plotted in Figures 5.24 and 5.25" the data plotted invoLves
columns with specified concrete strength of 4OOO psi (27 "6
MPa) and slenderness ratj-os of 33 (Figure S.Z4) and 66 (Fig-
ure 5.25) "
c
E 0.8
É
.9
N)
o {
E LN
! o.7 4-50-4-33
ol
L (ß,: o-o4o)
€)
U, ¡$-50-8-33
(ß,= 0.082)
0 0.2 0.4 0.6 0.8 1.0 1.5 2.0 4.0 co

Figure 5-24 - Effect of Structurol Steel Rotío on the Rotio of Theoretícol to Nominol Strength of
Slender Composíte Steel-Concrete Beorn-Columns
Írc 0.8
u4
.9
o N)
É {Oì
o.7 4-50-4-33
ctl
L ((rr- o-o4o)
q)
L
Lf)
4-50-8-33
(ß,: 0.082)
0 0.2 0.4 0.6 0.8 1.0 r.5 2.A 4.0 @

End Eccentricity Rotio (erlh)
Figure 5-24 (cont-) - Effect of Structurol Steel Rotío on the Rotío of Theoreticol to
Nominol Strength of Slender Composite Steel-Concrete Beom-Columns
(c) H'{eon Volues
4-50-4-53
(ß,: o.o4o)
4-50-B-33
0.0S2)
Vrrr:
É
E, 1.1
M
.9
l\.)
o
a. -.¡
L
{
1.O
cll
L
{u
t-
U'
0 0.2 0.4 0.6 0.8 1.0 1.5 2-0 4.0 @

Figure 5.24 (cont.) - Effect of Structurol Steel Rotio on the Rotio of Theoreticol to
Nomínol Strength of Slender Composite Steel-Concrete Beom-Columns
(o) t -Percentile Volues
C
É. 0.8
É
.9
N)
vo {æ
L o.7 4-50-4-66
c'r
o
(ßr- o-o4o)
U) 4-50-8-66
(f,r: o-o82)
0 0.2 0.4 0.6 0.8 1.0 I .5 2.0 4.0 @

End Eccentricity Rotio ("/n)
Figure 5-25 - Effect of Structurol Steel Rotío on the Rotío of Theoreticol to Norninol Strength of
Slender Composíte Steel-Concrete Beom-Columns
(U) S-Percentile Volues
C
É 0.8
v
.o
o t\)
E \¡
\o
! 0.7 .+-50-4-66
c't
L (ß,: o.o4o)
Ð
th 4-50-8-66
(ßr: 0.082)
0 0.2 0.4 0.6 0.8 1.0 1.5 2.0 4.0 æ

Figure 5.25 (cont.) - Effect of Structurot Steel Rotio on the Rotio of Theoreticol to
(c) tvteon Volues
4-50-4-66
(f,r: o.o4o)
4-50-8-66
(ßr- 0.082)
c
E 1.1
v
v
.9
N)
oo
ao O
! 1-O
c'l
L
(Þ
U)
0 0.2 0.4 0.6 0.8 1.0 1.5 2.0 4.0 @

Figure 5-25 (cont-) - Effect of Structurol Steel Rotío on the Rotio of Theoreticol to
Nominol Strength of Slender Composite Steel-Concrete Beom-Columns
28L
The data for slender beam-columns with kl/r of 22"1- was

simj-Iar to that presented earlier for short columns (Figures
5.1-0 and 5"LL). These beam-columns vrith smaller structural
steel ratio (p"" : 0.040) produced lower one-percent,íIe
strength rat.ios than did the beam-columns with larg'er struc-
tural steel ratio (P,": 0.082)" At 5-percentile and mean
value level-s, the columns with higher structural steel ratio
provided sIight,ly lower strength ratios but the dj-fferences
in strength ratios were insignificant"
The trends noted above for columns having kl/r : Z2"I are
similar to those for corumns with kl/r = 33" Figure s.24(a)
shows that the beam-column with p"": 0"040 gives signifi-
cantly lower one-percentile strength ratios than the beam-
column having P"" = 0"082 at e/h< O"8" For e/h lrrigr,€r than
0.8, the differences in one-percentile strength ratios in
Figure 5"24(a) seem to be negligible" The effect of p"" is
less significant for S-percentil-e and mean strength ratios
plotted in Figure 5"24(b) and (c), regardless of the end
eccentricity ratio" For columns with kl/r: 33 and /'c :
6000 psi (4I.4 MPa), hovrever, the effect of pss v/as negli-
gible on all strength ratios (l--percentíIe, S-percentile and
mean values) at all e/h values studied"
When the slenderness ratio increased to 66, the col-umns
with p"" = 0"082 demonstrated significantly lower one-
percentile, five-percentile and mean strength ratios than
282
those for columns with p"" = 0.040; the differences in

strength ratios being very significant for e/hS o"4 as
indicated by Figure 5.25. The effect of p,, oD strength
ratios became even more significant when the sl-enderness
ratio was increased t,o l-00 " This is probably because a very
slender coLumn must withstand high second-order bending
moments and, therefore, depends on the structural steel to
provide stiffness after the concrete has cracked"
In summary, the columns with low structural steel ratios
(P"" = 0"040) produced lower 1-percentil-e strength ratios
when kl/r was 33 or less" this behavior is similar to that
described earlier for short composite columns. Columns with
slenderness ratios egual to and greater than 66, on the
other hand, produced lower l-percentile, 5-percentile, and
mean strength ratios when the structural steel ratio was
high (p"" = 0"082). Hence, the structural steel rati-o must
be considered as a required parameter for reliability analy-
sis "
5"4.2.4 Ef fect, of end eccent,ricity ratío - To investigate

the effect of e/h on t.I:e strength ratio, aÌI columns from
Table 5"2 hrere studied for e,/h ratios ranging from 0"05 to
infinity. An examination of Figures 5"2O to 5"25 shows that
the strength ratios of slender columns vary with e,/h" For
further study, the data from the sixteen columns in TabLe
5"2 ta¡ere grouped j-nto three sets according to the slender-
ness ratio" Columns v¡ith kl/r of 22"1 and 33 v/ere grouped
283
together because of the similar values of strength ratios

obtained for these columns, as shown in Figure 5"2O" The
second set contained beam-columns with kl/r = 66, whil_e the
third set included column with kl/r : 1OO" For each set,
the range of l--percentile, S-percentile, and mean strength
ratios are pÌotted against e/h ín Figures 5"26(a), (b), and
(c), respectively.
Figure 5.26(a) shows that the widest range of one-
percentile strength ratios occurs for e/ h less than O "2 with
the highest strength ratios corresponding to k I/r of tOO and
the lowest to kI/r of 22"1- and 33" The one-percentile
strength ratios for columns with kl/r of l-OO drop very sig-
nificantly as e/h increases from O.O5 to O"Z" The beam-
columns with slenderness ratios of 22"1, and 33 are least
affected as e/ h increases from 0"05 to O"2 fFigure 5"26 (a) ].
Between e/h of O"2 and !"2, the ranges of one-percentile
strength ratios decrease significantly for all three sets of
columns with the differences among the three rangies also
decreasing' as e/h increases. For e/hvalues greater than
I"2, all three ranges overlap and remain nearly constant as
indicated by Figure 5"26(a). It is worth noting that the
one-percentile strength ratios significantly lower than O"7
were obtained aEe/hs 0.3 for columns in the set with kl/r =
22"1 and 33 and much higher than O"7 for columns with kl/r =
t_00 "
1.7
1.6
1.5
1.4
1.J
v.
1.2
P
v.
1.1 N)
-9 co
Ð È
v.
1_O
-Ê
oi)
E
0) o.9
t-
+
\n
o.8
o.7
o.6
rl.5
0.2 0.4 0.6 o.8 1.O 2.0 4.0 ct
End Eccentricity Rotío (u/n)
Figure 5.26 - Ronge of Rotío of Theoreticol to Norninol Strength of
1.8
1.7 (b) S-Percentile Volues Ñ(:SN klr/r = 22.L and 33

1.6
l.Tlillffn-fm kl/r = 6E
1.5
f--7- ktlr = noo
1.4
C
v.
1.3
E.
1.2 N)
.q
g co
o (tl
v.
.c 1.1
v
olt
c(u
¡- 1.O
#
a
0.9
o.8
o.7
'1.6
0 0.2 0.4 0.6 0.8 1.0 1.5 2.O 4.0 m
Figure 5.26 (cont.) - Ronge of Rotio of Theoretícol to Nominol Strength of
2_O
(c) Meon Volues
1.9
N\Sï\S\ Þ.ï./r - 22.X. and 33
1.8
[Tfl-TT-rm kl/r = 6S
1.7
v7-'1 kfr = 100
1.6
L
K
1.5
æ.
1.+ t\)
_9 @
o o\
v.
.E 1.3
{J
E)
E
r¡) 1.2
L.
#
lt)
1.1
1.O
o.9
û.8
o.2 0.4 0.6 o.8 1.0 1.5 2.0 4-0 cÐ
End EccentrÍcity Rotio ("/n)
Figure 5.2ô (cont.) - Ronge of Rotio of Theoretícol to Nominol Strength of
281
Figures 5.26(b) and (c) show that the above-noted trends

for one-percentile strength ratios are also followed by the
S-percentile and mean val-ues of the strength ratios. The
only difference is that the five-percentile and mean
strength ratios drop more sharply for columns with kl/r
ratios of l-00 and increase more s1ow1y for columns in the
set with kl/r of 22.L and 33 as e/h increases from 0"05 to
o.2 "
the foregoing discussions, it is concluded that the

From
end eccentricity ratio has the greatest effect on the
strength ratio of very slender columns (kl/r = L00) aE e/h
values below 0"2" fn this range of end eccentricity ratJ-o,
the least effect of. e/ h is experienced by col-umns with kl/r
of 22"1- and 33" For end eccentricity ratios of L.2 and
higher, there is negligible effect of e/h on the strength
ratios, regardless of the slenderness ratio used" There-
fore, the future reliability analysis of slender columns
should concentrate on data for e/h of L.2 and 1ess" The end
eccentricity ratios below 0"45 are particularly critical
because these e,/h values produced some one-percentile
strength ratios that fell below Q.7, as indicated by Figure
5"26(a) " The effect of e/h on the strength ratio for end
eccent,ricity ratios greater than 1,"2 can be neglected"
Figure 5"27 shows the coefficient of variation of
strength of composite beam-columns varying significantly
0.20
ÞF
.a.,
6 0-t8
ø
fd
{J
(I) NSNN kl/n = 22.t a¡d 33
0.n6
o
kl/r = 66
u 0.14
t
E kl/n = 10O
d
4)
EA NJ
0.n2 co
!l{
Õ æ
o
.+.¡
d 0-n0
Í4
d
Þ
q{
Õ 0_08
+=¡
fl(D
Ð
l¡{ 0.06
l¡.{
{)
cl
t)
0.04
0 0.2 0.4 0.6 0.8 1.o 1.5 2.O 4.0 æ
End Eccentricity Rotío (e,¿h)
Figure 5.27 - Ronge of the Coeffícient of Voríotíon of the Rotio of Theoreticol
toNominol Strength of Slender Composíte Steel-Concrete Beom-Columns
289
Tdith respect to end eccentricity ratio" Furthermore, col-

umns with lower slenderness ratios correspond to higher
coefficients of variation of strength and vice versa. This
is especially apparent at e/h values greater than O"Z. For
e/h< O"2, the coefficient,s of variation of strength for alt
three sets of column seem to overlap" The strength coeffj_-
cient of variation initially decreases rapidly and then at a
reduced rate as e/h increases from 0"2 to æ, as indicated by
Figure 5"27 "
The behavior of the strength coefficient of variation
explained in the preceding paragraph can be at least par-
tially attributed to the coefficient of variation of the
theoretical strength model itself" As discussed in Chapter
2, the theoretical strength model error had a coefficient of
variation that remained constant at its maximum value for
e/h< O"2" The model error coefficient of variation then
decreased linearly with increasing e/h, untiL it reached its
mj-nimum value at the pure bending condition. This partly
expJ-ains the shape of the coefficient of variation curves in

Figure 5"27 " The decrease in strength coefficient of varia-
tion with increasing slenderness ratio of the column is
like1y caused by the increase in dependance on the
structural steel to provide resistance against the increased
probability of tension failure due to secondary moment
effects" Since the coefficients of variation associated
with the steel properties are less than those for concrete,
290
the columns deriving more of their streength capacity from

the structural steel will be subjected to lower strength
variations "
5"4"2"5 Sensítivíty anatysís - The portions of the overal-l_
variability of the column strength attributable to the vari-
ation in the mechanical properties of concrete, the mechani-
caJ. and geometric properties of structural stee], and the
theoretical strength model were determined for a typical
slender beam-col-umn. Co1umn 4-50-4-66 (Table 5.2) r¡/as cho-
sen for this analysis. To determine the beam-column
strength variability due to each of the three sets of
variables noted above, three separate computer runs of SOO
simuLations each $rere made. For each computer run, only the
variabres from one of the above-noted sets vrere aLlowed to
vary while the remaining variables were kept constant at
their mean value" The porti-ons of the overall variability
of the beam-corumn strength so determined were respectively
designated as V V and V for the three sets of
"o¿"on"t "o¿""t ô¿n¿
variables noted above"
The squares of the coefficients of variation V|o¿,on,,
V?o¿r, and Z2¡¡6¿¿¿ for the strength of Column

computed
4-50-4-66 at e/h values ranging from O"05 to co are plotted
on Figure 5.28. These plots indicate that the overalr vari-
ability of the strength of the above-noted column is mostry
influenced by the variations in the theoretical strength
o.o28
o Vgcols
0.o24
€€) o
E{
d o tr V8aûþûnc
e
UJ 0.o20 v8í¡od
É
€
éJ o V
*,u,o
d fiEquotion (S.2)]
åd
o.o16
Õ
d
Þ V V; N)
S{ \o
o ts
+, 0.o12
Ë
€)
c)
Ad
qd
(Ð
o o.o0B
L)
o.o04
0 0.2 0.4 0.6 0.8 1.o 1.5 2.0 4.0 cÐ

Figure 5.28 - Effect of Voriobilities of Properties of Constítuent Moteriols on Overoll
Strength Voriobility of Slender Beom-Colurnn 4-50-4-66 (Toble 5.2)
292
modeL and the concreÈe mechanical properties for p¡hS 1.0"

For 1.0 <e/h1æt all three variations (V"o¿.on", V"o¿""t and
v
ô¿nt)
seem to contribute to the overall strength variabil--
ity of Column 4-50-4-66" Â.t pure bendi_ng condition, the
strength variability of Column 4-50-4-66 is identical to
that of the cross-section and the variations in the struc-
tural steer properties mostly influence the overarl vari-
ability of the column strength. Hence, the behavior of
corumn 4-50-4-66 at e/h=æ in Figure 5.29 is identical to
that shown in Figure 5"1-4" Note the effects of the vari-
ability of the theoretical strength model and of the con-
crete properties decrease significantly as the end
eccentricity ratio increases from e/h = o.2" The effect of
the variability of the structural steel properties increases
somewhat as e/h increases from L"O to higher values.
The values of V2,,^ plotted in Figure S.2B represent the
sum of the squares of the coefficients of variation of col-

umn 4-50-4-66 strength obtained from individual variabili-
ties of the three sets of variables as indicated in Equation
(5.2). Also plotted in Figure 5.29 are the values of V2r
which is the square of the coefficient of variation of the
beam-column strength obtained when variabilities from arl_
sources r¡¡ere incruded simultaneously in computations. A .
comparison of values of V!,^ and. V2r plotted in Figure S.Zg

indicates a good correlation of these values, especiarly in
293
the range of e/h : O"2 to 0"6, with V?u^ only slightly

underestinating Vzp on either side of this range of e/h.
The underestimation of V'r by V?u^ aE e/h< o.2 is 1ikely due
to variations in the cross-section dimensions" The underes-
tination by V?,âE e/h> 0.6 is likely due to variations in
the properties of the reinforcing st.eel and the bar
placements" Note the variations in the cross-section dirnen-
sions and reinforcing sÈeel properties were not included for
computing V?u^, as índicated by Eguation (S"Z).
The effects of variations in properties of reinforcing
bars were insignif icant in this study because p rs was smal-l-
compared to p"" used for the composite columns studÍed. An
earlier study (Mirza L989) found that the cross-section
dimensions had insignificant effect on composite column
st,rength variability" Sirnilarly, t{.írza and MacGregor (1999)
not,ed that the strength variability of slender reinforced
concrete columns was sensitive to dirnensional variations
only at Low e/h" From the discussions presented here and
Figure 5"28, it is apparent that the relation proposed in
Equation (5"2) is valid for the type of slender composite
columns studied.
5"4"2"6 Sunmarv of effects of variables used for lrasia qfrrdlrr
- The following sulnmarizes the effects of variables used for
the basic study of slender composite beam-columns:
(a) slenderness rat,ios of 22"L and 33 are more criticar than
66 and L00 for the type of slender columns studied;
294
(b) the specified concrete strength is a significant vari-

able for columns with sLenderness ratios less than 33
but is insignificant for kl/r greater than or egual to
aa.
JJ t
(c) the ratio of structural steel area to gross area of

cross-section ís a najor variable;
(d) the end eccentricity ratio has the greatest effect. on
very slender columns (kt/r = L00) and the l_east effect
on less slender columns studied (kl/r : 22"i. and 33);
(e) the overall variability of the theoretical strength is
prinarily due to the variations in the mechanical prop-
erties of the concrete, the mechanical and geometric
propertJ-es of the structural steel, and of the
theoretical strength model"
5.4 " 3 Ef fect Of Varíables llsed tr'or Sul:nlernental St.udw
From the slender columns used for the supplemental study
(Tab1e 5"4), the effects of the specified yield strength of
st,ructural steel, the strain-hardeníng of structural steel
section and reinforcing bars and the guaJ-ity contror of con-
crete on the beam-column strength ratios hrere studied"
Plots of the one-percent,ile, five-percentil_e and mean
strength ratios at various values of e/hwere made for each
variable studied"
5"4"3.1 Eff,ect of specífíed vield strencrth of struetural
sÈeel - To study Èhe effect of the specified yierd strength
295
of structural steel on the st,rength ratios, two slender col_-

umns from the basic study (Columns 4-50-4-33 and 6-50-4-33
in Table 5.2) vrere compared to four columns from the
supplemental study (Columns 4-44-4-33, 4-36-4-33, 6-44-4-33,
and 6-36-4-33 in Tab1e 5.4). This provided two sets of
three column each" Each set had a column with structural-
steel f = 50 ksi (345 MPa) , a column with structural steel_
"
:
f v 44 ksi (303 MPa) , and a column with structural steel /"
: 36 ksi (248 Mpa) " À11 other properties of the three col-
umns in a set were identical-" Figure 5"29(a) , (b), and (c)
respectively plot the one-percentile, five-percentil-e, and
mean strength ratio data for one of the above-noted sets.
The specified strength of concrete egual to 4OO0 psi (27.6
MPa) was used for the beam-columns shown in Figure 5.29.
At the 1-percentile leve1 [Figure 5"29 (a) ], the lowest
strength ratios hrere found for the column having 50 ksi (34s
MPa) structural- steel over almost the ent,ire range of e/ h.
The one-percentile strength ratios for the column with 5O
ksi (345 MPa) structural steel plotted significantly less
than those for the column with 36 ksi and 44 ksi (248 Mpa
and 303 MPa) structural steel at e/h< 0"5" Between e/h of
1,"2 and infinity, the data for the columns with 50 ksi and
36 ksi (345 MPa and z4g MPa) structural steel plot almost
identically. The highest values of one-percentile strength
ratios were obtained over almost the entire range of e/ h
1.0
(o) t-Percentile Volues
0.9
E
E 0.8
v.
.9
0 t\)
E \o
Oì
! 0.7 4-36-4-33
c't
L [Structurol steel f" = 36 ks¡ (248 MPo)I
q)
rn 4-44-4-33
[Structurol Steel f, = 44 ksi (JOJ MPo)l
0.6
4-50-4-33
[Structurol Steel f" : 50 ksi (545 MPo)]
0.5
o.4 0.6 0.8 1.0 t.5 2.4 4.0 æ
End Eccentricity Rotio (e¡'h)
Figure 5.29 - Effect of Specífied Structurol Steel Yield Strength on the Rotio of Theoreticol
1.0
0.9
c
É 0.8
v
.9
0 t\)
E \o
s 0.7 4-36-4-33 {
c'l
[Structurol steel f" : 36 ksi (ZaB Neo)l
o
U) 4-44-4-33
fistructurol Steel f" : 44 ksi (303 MPo)J
0.6
4-50-4-35
[Structurol Steel f" - 50 ksi (345 MPo)l
0.5
o.2 o.4 0.6 0.8 1.O t.5 2.0 4.0 @
Figure 5.29 (cont.) - Effect of Specífied Structurol Steel Yield Strength on the Rotio of Theoreticot
to Nomínol Strength of Slender Composite Steel-Concrete Beom-Columns
[Structurol steel f" : 36 ksi (2aB MPo)]
4-44-4-53
[Structurol Steel f, : 44 ksi (303 MPo)]
4-50-4-33
uc t.t [Structurol Steel f, : 50 ksi (3a5 MPo)]
a
.9
0 t\)
\.o
É, æ
! 1.0
c'l
L
o
L
tî)
0.8
0 0.2 0.4 0.6 0.8 1.0 1.5 2.A 4.0 æ
Figure 5.29 (cont.) - Effect of Specified Structurol Steel Yield Strength on the Rotío of Theoreticot
to Nomínol Strength of Slender Composite Steel-Concrete Beom-Columns
299
studied for the beam-corumn in which 44 ksi (303 Mpa) struc-

tural steel was used, âs indicated by Figure 5"29(a). A
sirnil-ar trend is observed from the S-percentire strength
ratios shown in Figure 5"29(b), but the differences amongr
the S-percentile strength ratios for columns wj-th different
grades of structural steel are significantly reduced"
The effect of structural steel grade on mean strength
ratios seems to be negligible [Figure 5"29 (c) ] " Only
between e/h of o"4 and 0"8 does the mean strength ratio data
show some spread" In this range of e/h, the lowest mean
strength rat,ios vrere given by the column having 36 ksi (248
MPa) structurar steel and the highest mean strength ratios
by the corumn having 50 ksi (34s Mpa) structural steel, âs
indicated by Figure 5"29(c).
Results from the other set of columns enploying f :
"'
6000 psi (41'"4 MPa) and used to investj-gate the effect of
structural steel grade (Columns 6-50-4-33, 6-44-4-33, and.
6-36-4-33) indicated negligible effects of structural steel
f on mean, five-percentile, and one-percentile strength
"
ratios. This was valid over the entire range of e/h stu-
died"
Since the l-percentile strength ratios are more impor-
t,ant for safety considerations and the rowest values for
these strength ratios $¡ere obtained in Figure s"29(a) for
columns with structural st,ee1 f = 50 ksi (345 Mpa), it is
"
300
reconmended that 50 ksi (345 MPa) structural steel be used

in future reriability analysis of slender composite corumns.
The 50 ksi (345 MPa) structural steel is the highest grade
of steel presently allowed by the North Amerícan design
codes for composite columns and will ensure relevancy as
commonly used steel grades increase above the present val-
ues "
5"4"3.2 Effect, of strain hardening of steer - To examine the

effects of strain-hardening of the structural steer section
and of the vertical reinforcing bars on the strength ratios
of slender composite colunns, the data from two columns of
the basic study (Columns 4-50-4-66 and 6-50-4-66 in Table
5"2) were compared to the data from the corresponding col-
umns of the supplemental study (Columns 4-SO-4-66-STH and
6-5O-4-66-STH in Table 5"4) " Note the strain-hardening of
both steers was included in theoretical strength computa-
tions of columns 4-50-4-66-srH and 6-so-4-66-srH and it was
neglected for columns 4-50-4-66 and 6-50-4-66" This
provided two sets of columns, each set, having one column in
which strain-hardening was included and one column in which
strain-hardening was not perrnitted. All other properties
!üere identical for both columns in a set. The strength
ratios for colurnns from one of these sets are shown in Fig-
ure 5"30"
1.1
1.0
c
v. 0.9
É.
"9
(¡J
ao O
ts
! 0.8
P
(¡
L
(l)
tJ)
CI 6-50-4-66
(Stroin-hordeníng neglected)
0.7
6-50-4-66-STH
0.6
o.2 0.4 o.ô 0.8 1.0 1.5 2-0 4.0 c()
Figure 5.30 - Effect of Stroin-Hordening of Steel on the Rotío of Theoreticol to
Nomiirol strength of Slender cornposite Steel-concrete Beom=columns
1.1
r.0
C
E 0.9
4
E
.9
0 (,
æ o
N)
! 0.8
ç't
L
O
r-
af,
6-50-4-66
(Stroin-hc rdening neglected)
0.7
6-50-4-66-STH
(Stroín-ho rdening considered)
0.6
0.2 0.4 0.6 0.8 1.0 1.5 2.0 4.0 @
End Eccentrícity Rotio ("/r,)
Figure 5.30 (cont.) - Effect of Stroin-Hordening of Steel on the Rotio of Theoret[col to
(c) Meon Volues
6-50-+-66
(Stroin-hordening neglected)
6-50-4-66-STH
(Stroin- hordening consídered)
c
E 1.2
v
.9 (¡J
o
g3 O
(,
Cn
L
o
(f)
0.9
0 0.2 0.4 0.6 0.8 1.0 1.5 2.0 4.0 @
Figure 5.30 (cont.) - Effect of Stroin-Hordening of Steel on the Rotio of Theoretícol to
Nomínol Strength of Slender Composite Steel-Conc rete Beom- Columns
304
Fi-gure 5.30 shows that the effects of strain-hardening

of steel on one-percentile, five-percentile, and mean
strength ratios are insignificant over almost the ent.ire
range of e/ h" Exceptions are the strength ratios close to
pure bending region. The most significant effect of strai-n-
hardening of steel was found on strength ratj-os at pure
bending" Here, the strength ratj-os of a sLender column are
identical Èo those of the cross-section since there is no
additional- bending moment due to eccentric axial loads.
Hence, the effect of strain-hardening on the cross-section
strength in pure bending discussed in Section 5"3"3.2 is
also valid for Figure 5.30. The conclusions derived from
the second set of columns used to study the effect of
strain-hardening of steel on slender composite columns \¡rere
very simil-ar to those stated for Figure 5"30.
Examinati-on of moment - curvature data for columns at
subject to axial loads less than 20 percent of their concen-
tric axial load cross-section capacity showed that the sec-
ondary moments caused by deflection of the column were
greater than any gain in strength due to strain-hardening of
the steel components. Therefore, the highest bending moment
capacities calculaÈed for the columns where strain-hardening
was allowed was identical to the bending moment capacity
calculated for an identical column in ir¡hich strain-hardening
was not pennitted"
305
Since significant effects of strain-hardening on

st,rength ratios of slender col-umns occur only near and at
pure bending condition, it is concluded that the strain-
hardening of steel be neglected for reliability analysis of
slender composite beam-co1umns"
5"4"3"3 Effect of qualíty of concrete - To study the effect
of concrete guality control on the strength variations of
slender composite beam-columns having f'" : 6000 psi (4I"4
MPa), the data for tr¿o columns from the basic study (Columns
6-50-4-33 and 6-50-8-33 with excellent concrete quality in
Tab1e 5"2) were compared to the data for the correspond.ing
columns from the supplement,al study (Columns 6-50-4-33-A and
6-50-8-33-A with average concrete guality in Table 5.4).
This produced two sets of beam-co1umns, each set had one
column with excellent quality concrete and one column with
average quality concrete. ÀI1 other properties r4rere identi-
cal for both columns in a set" The strength ratios for
beam-columns from one of these sets are given in Figure 5.31
and represent the typical behaviour"
At the l-percentile level [Figure 5.3]-(a) l, the beam-
column having average guality concrete produced signifi-
cantly lower strength ratios over the entire range of e/h"
This can be atÈributed directly to the larger coefficient of
variation associated with concrete of average guality" At
the S-percentile level, the effect, of concrete quality on
strength ratios reduces significantly as indicated by Figure
0.9
c
æ 0.8
æ
.9
+t UJ
g O
æ Or
! 0.7
Ë't
4)
U)
o 6-50-4-33
0.6
o 6-50-4-33-A
0.5
0 0_2 0.4 0.6 0.8 1.0 1 .5 2.0 4.0 æ
Fígure 5.31 - Effect of Concrete Quolity Control on the Rotio of Theoretícol to
Nominol Strength of Slender Cornposite Steel-Concrete Beom-Columns
C
É 0.8
Ð
M
.9 (,
o o
É {
o.7
ç¡
L
o
t-
u)
6-50-4-33
6-50-4-33-A
0.5
0 0.2 0.4 0.6 0.8 1.0 1.5 2.O 4.0 @
Figure 5-31 (cont.) - Effect of Concrete Ouolity Control on the Rotio of Theoreticol to
Nominal Strength of Slender Composite Steel-Concrete Beom-Columns
(c) Meon Volues 6-50-+-33
6-50-4-33-A
c
E
v
.q
Lr)
o O
E @
L 1.0
c't
(¡)
f-ft
0.8
0 0.2 0.4 0.6 0.8 1.0 I .5 2.0 4.0 @
Figure 5.31 (cont.) - Effect of Concrete Quolity Control on the Rotio of Theoreticol to
309
5"3L(b). The mean strength ratios for columns with differ-

ent concrete guality tend to be of the same magnitude lFi-
gure 5"31(c) I " This behavior is expected and is very
similar to the effect of concrete guality on strength
variations of short composite columns.
Since the concrete guality has a significant effect on
the rower tail of the strength probability distributions, it
is important that the future reliability analysis considers
the concrete guality as one of the variables"
5.4.3.4 Su$narv of effects of variables used for sunnlemen-
ta] study - The followj-ng sunmarizes the effects of vari-
ables used for supplemental study of slender composite
beam-columns:
(a) The beam-columns having structural stee] of specified
yield strength of 50 ksi (345 Mpa) produced lower 1-per-
centile strength rat,ios than did the columns with rower
grades of structural steel;
(b) the strain-hardening of steel only enhances the strength
of the beam-column at and near the pure bending condi-
tion and, therefore, can be neglected for practical cor--
umns; and
(c) the quality control of concrete significantry affects
the strength ratios at the l-percentiÌe l_evel_.
310
6 SUT.fffå,RY åND CONCTJUSTONS
6.:. SUM¡{ÃRY
The purpose of the study reported herein r¡/as to simul-ate

the probability distributions of ultimate st,rength of com-
posite columns in which steel shapes are encased in concrete
and to define the najor varj-ab1es that affect the strength.
The descriptions of the strength probability distributions
developed in this study will be used in a reliability anal-y-
sis currently underway at Lakehead University for developing
linit states design criteria for composite col-umns in
building structures.
The Monte Carlo technique r¡/as employed to simulate the
statistical properties of the strength of composite beam-
columns. Probability distributions of the geometric and
mechanical properties of column components required for
Monte Carlo sinulation were either taken from the literature
or were derived from avaitable statistical_ data. Àn exis-
ting computer program to calcul-ate the theoretical
resistance of composite columns vras extensively tested for
numeric and logical accuracy and revised wherever required.
The accuracy of the theoreÈical strength program was estab-
lished by comparisons with existing test data of the urti-
mate strength of composite beam-columns" Repeated
simulations of the strength of several column configurations
311
using the probability distributions of variables that affect

the strength resulted in the definition of the strength
probability distri-bution for each beam-column configuration"
An existing computer program to calculate the nominal_
capacity of composite bearn-columns was tested extensively
and modified wherever reguired. The nominal capacity was
based on the specified geometric and mechanical properties
of the column components and on the equations given in the
North American building codes" The same column configura-
tions that were used to define the theoretical strengLh
probability distributions hrere also checked for their
nominal capacities" By comparing the ratio of theoretical
to nominal strength between various col-umn configurations,
the component variables having significant effects on the
strength $¡ere defined"
6"2 CONCLUSIONS
The strength variability of composite corumns is prirnar-

ily due to the variability in concrete mechanical properties
for compression failures and the variability of the steer-
section mechanical and geometric properties for tension
failures.
6u2" I Short Golumns
The ratio of theoretical strength to nominal strength

(strengÈh rat,io) of short composite columns (kt/ r <22) v/as
infl-uenced rnost significantry by the specified concrete
3L2
strength, the ratio of area of structural steel to gross

area of the cross-section and the end eccentricity ratio"
Beam-coLumns with slenderness ratios at the code limit of
kL/r:22produced lower strength ratios than the correspond-
ing cross-sections. The magnitude of the specified yield
strength (f of structural steel did not significantly
")
affect the strength rat,ios for beam-columns with conmon
grades of steel having /" < SOksi (345 MPa)" Strain-
hardening of the reinforcing and structural- steel in a com-
posite column enhanced its strength only aL and near the
pure bending condition. Quality control of the concrete
significantly influenced the lower tail- of the strength
probability distributions of the beam-columns studied.
6,2"2 Slender Columns
The ratio of theoretical- to nominal strength of slend.er

composite columns (kL/r>22) v/as influenced most signifi-
cantly by the slenderness ratio, the ratio of area of struc-
tural steel- to gross area of the cross-section and the end
eccentricity rat.io" Both the slenderness ratio and the end
eccentricity ratio have littre effect on the strength ratios
of columns with end eccentricity ratios greater than l.Z.
The specified.strength of concrete influenced the strength
ratio only for columns with a slenderness ratio (kt/r¡ of
less than or equar to 33. The lowest strength ratios were
obtained for colurnns with slenderness ratios less than or
313
equal to 33 and the highest ratios !{ere found for columns

with a slenderness ratio of 100. Beam-columns with struc-
tural steel yield strength of 50 ksi (345 MPa) produced
lower strength ratios than Èhose produced by beam-columns
having lower yield strength of structural steel" Strain-
hardening enhances the strength only at and near the pure
bending condition" The quality control of the concrete
significantly affected Èhe lower tail of the st.rength proba-
bility distribution of the column.
314
LTST OF REFERENCES
ACI Cornrnittee 2L4. L965. Recommended practice for

evaluation of compression test results of con-
crete" American Concrete Institute, Detroit, MI"
ACI 318-83 " L983. Buitding code requirements for
reinforced concrete" American Concrete fnstitute,
Detroit, MI.
Al-Ien, D.E " L970 " Probabilistic study of reinforced
concrete in bending. Technical Paper NRC 11139,
National Research Council of Canada, Ottawa"
A11en, D.E. L972 " Statistical study of the mechani-
caI properties of reinforcing bars" Building
Research Note No. 85, Division of Building
Research, National Research Council, Ottawa.
Alpsten, G.A" 1968. Thermal residual stresses in
hot-rol-led members. Frj-tz Engineering Laboratory
Report, No. 337 "3, Lehigh University, Bethlehem,
PA.
Alpsten, G.A. 1972. Variations in mechanical and
cross-sectional properties of stee1. planning and
Design of Tall Buildings, rb(9): 755-805.
American Iron and Stee1 fnstitute Committee on
Product Standards, L972. Check tension tests
plates and wide flange beams. proceedings, Spe-
cialty Conference on Safety and Reliability of
Metal Structures, Pittsburgh, pA.
Basu, A"K" l-967 " Computation of failure loads of
composite columns. Procedures Institution of Civi1
Engineers (London) , 36(March) z SS7-578"
Basu, 4.K", and HilI, W.F. l_968. A more exact com-
putat.íon of failure loads of composite columns.
Procedures Institution of Civil Engineers
(London) , 40 (May) z 37-60"
Beed1e, L. S . , and TalI , L" l_960 " Basic column
strength" Journal of the Structural_ Division ASCE,
86(ST7): 1-39-1"73"
Bjorhovde, R" L972" Deterministic and probabilistic
approaches to the strength of steel columns. ph"D"
Dissertation, Lehigh University; Bethlehem pA.
Bolin, P" l-985. Parametric study of the strength
variability of composite columns. B"Eng. thesis,
School of Engineering, Lakehead Univerãity, Thun-
der Bay, Ontario"
Bolotin, V.V. L969. Statistical methods in struc-
tural mechanics, translated by Samuel Aroni.
Holden-Day Inc", San Francisco, California"
315
Bondale S" I966a. Co1umn theory v¡ith special refer-

ence to composite columns" The Consulting Engi-
neer, Julyz 72-77 "
Bondale S. 1966b" Co1umn theory with special refer-
ence to cornposit,e columns " The Consulting Engi-
neer, Augustz 43-48"
Bondale S" L966c. Column theory r¿ith special refer-
ence t.o composiÈe columns" The Consulting Engi-
neer, September: 68-70.
Carreira, D.J. and Chu, K" l-986" SLress-strain
relationship for reinforced concrete in tension"
Àmerican Concrete Institute Journal, 83(L): 2L-29"
CSA" 1984" Design of concrete structures for buitd-
ings a nat,ional standard of Canada. CAN3-423.3-
M84, Canada Standards Association, Ottawa, Ont"
Davis, S.G " 1976" Further investigations into the
strength of concrete in structures" Technical
Report, Cernent and Concrete Association, London,
England.
Doane, J"F" L969. IneLastic instability of wide-
flange steel beams" M.Sc" thesis, University of
Texas at Austin, Austin, fX"
Galambos, T"V. L963. Inelastic lateral buckling of
beams" Journal of the Structural Division ASCE,
8s (sTS) : p "21,7 "
Galambos, T.V", and Ravindra, M"K" i.979. Properties
of steel for use in LRFD" Journal of the Struc-
tural Divisíon ASCE, 104(ST9): L4S9-1-468"
Grant, L"H " L976 " Strength variability of concrete
columns" M"Sc. thesis, University of Alberta,
Edmonton, Albert,a"
Grant,, L.H., Mirza, S"4., and Maccregior, J.G" i-978.
Monte Carlo study of strength of concrete columns.
American Concrete fnstitute Journal,
7s (8) :345-358.
Hognestad, E" L95L" A study of combined bending and
axial- load in reinforced concrete members" Bul1e-
tín 399, Engineering Experiment, Station, Univer-
sity of Illinois, Urbana ILL"
Hwang, L., and Rizkalla, S" 1983. Effective tensile
stress-sÈrain characteristics for reinforced con-
crete" Proceedings, Canadian Society for Civil
Engineeríng Ä,nnual Conference, Ottawa, Ontario:
L29-L47 "
3r6
Johnston, B"G" and Opila, F" 1941-. Compression and

tension tests of structural a1Ioys" ASTM Proceed-
ings, American Society for Testing and Materials,
4X:
Jones, P.G., and Richart, F"E. 1936" The effect of
testting speed on sLrength and elastic properties
of concrete" Proceedings, American Society for
Testi-ng and Materials 36, Part II: 380-391"
Julian, O"G " 1957 " Synopsis of first progress
report on commit,tee on safety factors" Journal of
the Structural Division, ASCE, 83 (ST4) , Proc.
Paper L316z L-22"
Kennedy, D"L.J., and Gad Àly, M" l-980" Linit states
design of steel structures - performance factors"
Canadian Journal of Civil Engineering, ? z 45-77
Kent, D.C", and Park, R" 197L. Flexural members
with confi-ned concrete" Journal of the Structural-
Division ASCE , 97 (ST7) : L968-l-990.
Kikuchi, D"K" , l{irza, S.4., and MacGregior, J.G.
1978. Strength Variability of Bonded Prestressed
Concrete Beams" Structural Engineering Report no.
68, University of A,lberta, Edmonton, Alberta"
Lachance, L. and Hays, C"O" l-980" Accuracy of Com-
posite Section NonLinear Solutions. Journal of the
Structural Division ASCE J.O6 (STLJ-) : 22O3-22L9.
Lay, M"G" L965" Flange local buckling in wide-
flange shapes" Journal of the St.ructural Division,
Â,SCE, 9X(5T6), Proc. paper 45802 49-66.
L1ewellyn, S" 1986" Parametric study of the
strength of composite columns. B.Eng. thesis,
School of Engineering, Lakehead Universíty, Thun-
der Bay, Ontario.
Lyse, Í., and Keyser, C"C" 1,934. Effect of size and
shape of test specimen upon the observed physical
properties of structural steel. ASTM proceeditrgs,
American Society for Testing and Materials, 34:
Mander, J.B. l-983" Seisrnic design of bridge piers"
Ph"D" Thesis, Civil Engineering, University of
Canterbury, Christchurch, New Zealand.
Mander, J.8., Priestley, M"J"N", and park, R. 1999"
Theoret,ical stress-strain model for confined con-
crete" Journal of Structural Engineering, ASCE,
f"o8 (8) : L804-1-826 "
Massey, P.C" 1-964" Effect of residual stresses on
the lateral stability of steel l-beams. New Zea-
land Engineer 1.9 (9) :
3l-7
May, I.M., and Johnson, R.P. Lg78. Tests on

restrained composite columns" The Structural Engi-
neer, 568(2): 21--27"
McNeely, D"J", and Lash, S"D" L963" Tensile
strength of concrete" American Concrete Institute
Journal, 60 (6) : 75L-76L"
Mirza, S"4", and MacGregor, J"G" 1979a" Variations
in dirnensions of reinforced concrete members.
Journal of the Structural Division ASCE I05(ST4):
75I-7 66 "
Mirza, S.4., and MacGregor, J.G. Ig7gb. Variability
of mechanical properties of reinforcing bars.
Journal of the Structural Division ASCE 105(ST5):
92L-937 "
Mirza, S.4., Hatzinikolas, M., and MacGregor, J.G.

I979c. Statistical descrj_ptions of strength of
concrete" Journal of the Structural Division ASCE
x05 (5T6) : to2L-Lo37 "
Mirza, S..A., and MacGregor, J.G. i-gBZ" probabiLis-
tic study of strength of reinforced concrete mem-
bers" Canadian Journal of Civil Engineeri.g, 9 (3) :
431,-448 "
Mirza, S.A" l-985a" Resistance facÈors for concrete

design using ACf 3L8-83. International Journal of
Structures, 5 (1) :1--22
l{T.rza, S.A. L985b" Application of Monte Carlo simu-
lation to structural engineering problems. pro-
ceedings, Second International Conference on
Conputing in Civil Engineering, Hangzhou, China,
June 5-9: 622-635"
Mirza, S.4., and MacGregior, J.G. 1999. Slenderness
and strength reliability of reinforced concrete
columns" American Concrete Institute Structural
Journal, 86 (4) : 4ZB-439.
Mirza, S.A" L989. Parametric study of composite
column strength variability" Journal Construc-
tional Stee1 Research, t4 z t21-l_37"
Mirza, S "4. 1-990 " Flexural stif fness of rectangular
reinforced concrete columns. American Concrete
Institute Structural JournaÌ, 87: (accepted for
publication) .
Moehle, J.P., and Cavanagh, T" L9BS. Confinement
effectiveness of crossties in RC. Journal of the
Structural Division ASCE ttt (1_0): 2LOS-ZL2O.
Morino, S., Matsui, C., and Watanabe, H. j-9g4"
Strength of biaxially loaded SRC columns. Compos-
ite and Mixed Construction, American Society óf
Civil Engíneers: 24L-253"
318
Nethercot, D"A" 1974. Residual stresses and their

influence upon the lateral buckling of rolled
steel beams" The Structural Engineer, Sd(3): g9-96
Park, R., Priestley, M.J.N., and Gill, W.D " 1-982"
Ductility of sç[uare-confined concrete columns.
Journal of the Structural Division ASCE tog(ST4):
929-950 "
Park, R., and Paulay, T" 1975. Reinforced concrete

structures" John Wiley and Sons.
Procter, A"N" L967. FuIl size tests facilitate der-
ivaÈion of reliable design methods" The Consulting
Engineer, 31(8) : 54-60.
Quast, U" I97O" Geeignete vereinfachungen fur die
losung des traglastproblems der ausmittig gedruck-
ten prisrnatischen stahlbetonstutze mit rechtekc-
querschnitt" Dr.-Ing" Dissertation, Fakultat fur
Bauwesen at Technischen Universítat
Carlo-Wilhelmina, Braunschrøeig, FRG.
Rao, N.R.N", Lohrman, M., and TalI, L" L964. Effect
of strain rate on the yield stress of structural
st,eeI" Report No" 249.23, Frítz Engineering Labo-
ratory, Lehigh University, Bethlehem, pA.
Scott, 8"D", Park, R., and priestley, M.J"N. 19g?.
Stress-strain behavior of concrete confined by
overlapping hoops at low and high strain rateè.
American Concrete Institute Journal Zg(L): 13-27"
Sheikh, S.A,., and Uzumeri, S.M. l_980. Strength and
ductility of tied concrete columns. Journal of the
Structural Division ASCE LO6(STS) : t-079-j_t-01.
Sheikh, S"A" and Uzumeri, S.M. LgB2. Analytical
model for concrete confinement in tied columns.
Journal of the Structural Division ASCE tOg(ST1-2):
27 03-2721- "
Sheikh, S . A. , and Yeh, C. C. i-98 6 " Flexural behavior

of confined concrete columns. American Concrete
Institute Journal, B3 (3) : 399-404 "
Suzuki, T., Takiguchi, K., Ichinose, T., and Oka-
moto, T" L983 " Effects of hoop reinforcement in
steel and reinforced concrete composite sections.
Third South Pacific Regional Confãrence on Earth-
quake Engineering" !{e}1ington, Ner¿ Zealand.
Tall, L" and Alpsten, c.A. 1969" On the scatter in
yieJ-d strength and residual stresses in steel mem-
bers" Fínal Report, Symposium on Concepts of
Safety of Structures and Methods of Design,
fnternat.ional Association for Bridge and Struc-
tural Engineering, London, England.
3I9
Trahair, N"S., and Kitipornchai, S. 1972" Buckling

of inelastic I-bearns under uniform moment" Journal
of the Structural Division ASCE , 98 (STl-1) :
255L-2566.
Virdi, K.S., and Doq¡Iing, P"J" L973 " The ultimate
strength of composite columns in biaxial bending"
Procedures InstituÈion of Civil Engineers (Lon-
don), 56 (May) ¿ 25L-272 "
Virdi, K.S., and Dow1ing, P"J" L982. Composite coL-
umns in biaxial loadíng" Axially Cornpressed Struc-
tures, Stability and Strength, edj-ted by R"
Narayanan" Applied Science Publishers, London:
1,29-t47 "
Wakabayashi, M" L976. A proposal for design formu-

las of composite columns and beam-columns" Second
International Colloguium on Stability Of
Structures, Tokyoz 65-87 "
?tright, P"J"F. 1,952" The effect of the method of
t,est cubes on mean strength and standard devi-
ation. Magazine of Concrete Research, 4(LJ-):
67-76"
Youngi, B"W" L971,. Residual stresses in hot-rolled
sections. Report CUED/C-Struct/TR. B. Dept" of
Engineering, University of Canbridge.
Zhen-hai, G" and Xiu-ein, Z" 1,987. fnvestigation of
complete stress-deformat,ion curves for concrete in
tension. American Concrete Institut,e Materials
Journal 84 (4) : 278-285 "
320
Í,TST OF SYMBOÏ,S
b flange width of structural steel section.

c depth of neutral axis measured from compression face
(Figure 3.1)
d depth of structural steel section"
e eccent,ricity of axial load at column ends
e/h eccentricity ratio
em deflection of slender column at rnid-height
et total eccentricity of axial load at nid-height of slen-
der column
f specified yield strength of reinforcing bars
",
f. stress in concrete
f'" specified strength of concrete
7""¿.n mean 28-day in-structure compressive strength of con-
crete loaded at a rate of R.
7cs¿.ss mean 28-day in-structure compressive strength of con-
crete loaded at a rate of 35 psi (0.241 Mpa) per second.
f, modulus of rupture of concrete
7, mean value of modulus of rupture of concrete j-n struc-
ture"
f" specified yield strength of structural steel.
f ", critical column buckling stress.
f"" static yield strength of structural steel"
fya dynamic yield strength of structural steeL.
f"u upper yield strength of steel"
fv¿ lower yield strength of steeL.
f* static ultimate strength of structural steel.
h overall depth of composite section.
kl/r slenderness ratio of column.
k effective column length factor (equal to l_.0 in this
study)
32L
L column length"
r radius of gyration"
t flange tTrickness of structural steel section"
u web thickness of structural steel section"
Ar area of one flange of structural shape (bt).
A. area of web of structual shape(w(d-Zt)) "
As gross area of cross-section"
A" area of structural steel section"
C
^ factor to relate actual bending mornent diagram to an
equivalent uniform bending moment diagrarn (taken equal
to 1.0 in this study).
DN A perpendicular distance from ptastic centroid of column
to neut,ral axis (see Figure 2"6)"
E. initial tangent modulus of elasticity of concrete.

E mean value of initial tangent, mod.uLus of elasticity of
"n
concrete test cylinders Loaded at rate R"
Fc¡s¿r,r mean value of initial tangent modul-us of elasticity of
in-situ concrete loaded at rate R.
F.ss mean value of initíal tangent, modulus of eLasticity of
concrete test cylinders loaded at a rate of 35 psi/sec
(O"24L MPa/sec) "
E
" rnodul-us of elasticityof structural steel.
Et tangent modulus of elasticity of element.
E,,t,n initial tangent modulus of strain-hardening curve of
reinforcing bars"
ðssrrn initial tangent modulus of strain-hardening curve of
' structural steel"
E, modulus of eÌasticity of reinforcing steel.
./ noment of inertia"
Iq gross moment of inertia of cross-section.
I
" moment of inertia of structural steel section.
M bending moment.
M^ bending moment at rnid-height of slender column.
P axial load.
322
P
" nominal- slender column buckling load capacity"
Pn nominal column cross-section axial load capacity.
PDF probability density function"
R¿ theoretical resistance of structural member.
R, nominal resistance of structural member.
V,," coefficient of variat,ion of ratio of tested to cal-cu-
lated member strength"
V r,-oor"n coeff icient of variation of laboratory control specimens
due to ín-batch variations of material strength and
dimensions "
V tn"t coef f icient
of variation of test procedures.
Vmodet coefficient of variation of theoretical strength model.
V csrrR coefficient of variation of in-situ compressive strength
of concrete"
V coefficient of variation of the relation between cylin-
",not der strength and specified design strength of concrete.
VR coefficient of variation in the relation between con-
crete l-oaded at 35 psi/sec (O.241 MPa) and concrete
loaded at a rate of R"
V
""y¿ coeff icient of varj-ation of strength of concrete test
cylinders "
V cstr3s coef f icient of variation of in-situ compressive strength

of concrete Loaded at a rate of 35 psi/sec (O"24I
MPa/sec) "
V cistrR coef ficient of variation of the init,ial tangent modul-us

of in-situ normal weight concrete.
ß safety índex as defined in Chapter 1 and Figure i-.1.
ßr ratio of depth of rectangular compression block to depth
of neutral axis (Figure 3.1) "
ßa absolute vaLue of the ratio of maximum factored dead
load moment to the maximum factored total load moment
(taken equal to 0"0 in this study).
öb moment magnification factor t,o reflect the effects of
member curr¡ature betr*een ends of compression members.
€.c strain in concrete"
€o strain in unconfined concrete at peak compressive
stress "
323
€r" ultimate strain of concrete ín compression.

€ss¿rn strain at start of strain-hardening curve of structural-
steel "
€u" ult,imate strain of structural steeI.
€r" yield strain of structural steel"
€rs¡rn strain at start of strain-hardening curve of reinforcing
bars "
€,, ultimate strain of reinforcing bars.

€," yield st,rain of reinforcing bars.
Ô curvature (inclination of strain gradient) or design
code understrength factor"
0- curvature at mid-height of slender column.
0n curvature at column ends.
P,, ratio of area of vertical reinforcing bars to gross
cross-section area.
P"" ratio of area of structural- steel to gross cross-section
area.
6,, residual stress at centroid of structural steel section.
6 rrt residual stress at, flange tip of structural steel- sec-
tion"
6,J, residual stress at juncture of flange and web of struc-
tural steel section.

Skrabek Strength Reliability

Uploaded by

Copyright:

Available Formats

Skrabek Strength Reliability

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Skrabek Strength Reliability

Uploaded by

Copyright:

Available Formats

STRENGTE RETJIåBT'TTY OF STTORT ÃND SLESTDER

BRYAN WILLTAI{ SKRÀBEK

Department of Cívil Engineering

(c) December, L989

COHPOS ITE STEET-COI{CRETE COTIIB{I{S

BRYAI{ }ilLLIÄM SKRABEK

A thesis subnrined to thc Facurty of craduare Studies

Permission has been granred to the LIBR.ARY

The author reserves other pubrication righfs, and

Structural reliability analysis requires that the rela-

For short composite beam-columns, the variables that

The author would like to acknowledge the assistance of the

1.2 Outline of Research Program I

2"2 Assumptions Used In Theoretical l_8

4"2 Structural Steel ].43

Figure Description Page

2"L6 Partially Confined Concrete Compressive 67

5"9 Effect of Specified Concrete Strength on 216

Table Descri-ption Page

The design procedure for composite beam-columns speci-

Factored Resistance (t.t)

Consider a large number of columns, each designed to

Figure 1"1 - Definition of faj-lure, probability of failure,

(a) the columns are pin-ended and the effective length is

(a) Modelof Column in SymmetricalSingle Curvature (b) Forces on Column

,Figure I"2 Loading and Support Conditions-.of

Probability distrj-butions of the mechanical and geomet-

variability of the properties of structural steel, concrete,

2 TTTEORETTCÃTJ BEAß{-COLU}.Í3{ STRESIGTIT

To analyze the theoretical- strength of a composite beam-

Reod lnput Voriob'les

Colculote Ì'lorrrínol Strerrgth

Colculote Theoreticol Strength

figure ?.'i - Flow Chort of ûcmputinç Prccedure

2" 1 REVTEW OF PREVIOUS W'ORK

In order to describe the theoreticaL model used in this

Basu (1-967) presented a computer method to approximate

beam-column giving a more precise evaluation of the

Virdi and Dowling (1982) present a revisj-on of their

stress-strain relationship of concrete inside and outsid.e

stiffness of reinforcing bars r¡¡as found. to underestimate the

The assumptions regardj-ng the loading and the end condi-

by local or torsi_onal buckling;

(f) sinplified the discretization of the cross-section into

stress-strain characteristics. The rolled. steel section was

Figr:re 2.2 - Material- Types in Conposite

Element .thickness varies to ensure that

Figure 2"3 - Discretj-zation of Composite One-Half

The dj-stance, parallel to the major axis, from the edge of

in which do" is measured perpendicular to the rnajor axis

algebraically to each element. Figure 2"3 shows the discre-

The cross-section strength model deterrnines the rela-

The first step required in determining Elr.e M-0-P rela-

For Axiol Lood P' l

yielding initioted in tension flonge

moximurn bending moment

Figure 2.4 - Typicol l,'l-ø-P Relotionships for

Figure 2.5 - Typical Composite Cross-section

cross-section is doubl-y symmetric, the plastic centroid is