Вы находитесь на странице: 1из 17

Downloaded from ascelibrary.org by Universidad Nacional De Ingenieria on 05/26/18. Copyright ASCE.

For personal use only; all rights reserved.

RELIABILITY BASED OPTIMUM DESIGN


OF CONCRETE FRAMES
By Adang Surahman 1 and Kamal B. Rojiani,2 M. ASCE

ABSTRACT: A reliability based optimum design procedure as a rational ap-


proach to optimum structural design is presented. The reliability analysis is
based on a first-order second moment approach. Uncertainties in the structural
strength and loading due to inherent variability as well as modeling and pre-
diction errors are included in the formulation. The failure modes considered
include failure due to bending, combined axial thrust and bending, and shear.
The structural systems investigated are typical office buildings consisting of four-
and ten-story reinforced concrete frames. The procedure to obtain the optimal
structure is based on a minimization of the total expected cost, and the optimal
risk level is obtained as the value that results in the minimum total expected
cost. The total expected cost is taken as the sum of the initial building cost and
the loss due to failure. An iterative procedure is used to obtain member di-
mensions and reinforcing steel areas for prescribed risk levels ranging from
10"2-lCn6, and upper, and lower bounds on the total expected cost are com-
puted. For the buildings considered, the optimal risk level ranges from 10 - 4 1 -
10~5-4. The proposed optimization procedure can provide an accurate estimate
of the optimal level of safety.

INTRODUCTION

The design of a safe a n d economical structure is one of the m a i n con-


cerns of structural engineers. The safety requirements arises from the
fact that engineering designs are formulated u n d e r uncertainty. Most
structural problems are nondeterministic. Both the loading and the
structural resistance are r a n d o m in nature. Also, there is considerable
uncertainty in the load a n d resistance prediction models. To account for
uncertainty in design, most building codes use a factor of safety which
is usually based on engineering judgment and previous experience with
similar structures. Some building codes also apply separate partial safety
coefficients to the load a n d resistance variables in an attempt to relate
the contribution of each variable towards the establishment of the total
safety factor.
Recognizing that loads a n d resistance are r a n d o m variables, the ra-
tional approach to design w o u l d b e t h r o u g h the use of probabilistic
'Grad. Student, Dept. of Civ. Engrg., Lehigh Univ., Bethlehem, Pa. 18015; for-
merly Grad. Student, Dept. of Civ. Engrg., Virginia Polytechnic Inst, and State
Univ., Blacksburg, Va. 24061.
2
Asst. Prof., Dept. of Civ. Engrg., Virginia Polytechnic Inst, and State Univ.,
Blacksburg, Va. 24061.
Note.—-Discussion open until August 1, 1983. To extend the closing date one
month, a written request must be filed with the ASCE Manager of Technical and
Professional Publications. The manuscript for this paper was submitted for re-
view and possible publication on October 27, 1981. This paper is part of the
Journal of Structural Engineering, Vol. 109, No. 3, March, 1983. ©ASCE, ISSN
0733-9445/83/0003-0741/$01.00. Proc. No. 17816.

741

J. Struct. Eng., 1983, 109(3): 741-757


methods. Many proposals for the introduction of probabilistic concepts
in design codes have been made in recent years. In the probablistic ap-
proach, the loads and resistance are treated as random variables rather
than fixed deterministic constants, and failure is defined as the event
that the load effect exceeds the structural resistance. The safety measure
Downloaded from ascelibrary.org by Universidad Nacional De Ingenieria on 05/26/18. Copyright ASCE. For personal use only; all rights reserved.

that corresponds to this is the probability of failure (or the reliability


index) which is obtained from a systematic analysis of the uncertainties
in all variables.
Economy in design is achieved through an optimization procedure.
Structural optimization within a deterministic format comprises essen-
tially the determination of design variables, such as depth, thickness,
reinforcement ratios (usually for elements), which result in either min-
imum cost or minimum weight, while satisfying code restraints. A major
limitation of deterministic optimization procedures is that the safety fac-
tor, which can have a much greater effect on the cost or weight than
any of the subsequent results of mathematical programming techniques
(19) is not included in the optimization procedure, either as a constraint
or as part of the objective function.
Several schemes for structural optimization based on reliability have
been suggested. These include: (1) Minimization of the cost or weight
for a specified allowable failure probability (17,21,22,30); (2) minimiza-
tion of the probability of failure for a fixed cost (20); and (3) minimization
of the total overall cost (30,31). Most reliability based optimization stud-
ies consider the first two procedures. This is due to the difficulties in
determining the nature of failures and in assigning monetary values to
failure consequences. Also, the analysis is usually limited to a consid-
eration of member failure rather than system failure. Since failure of a
member does not necessarily result in total failure of the structure, it is
felt that it is necessary to consider failure of the entire structure in the
optimization procedure.
This paper presents a method for the optimum design of reinforced
concrete frames which incorporates reliability analysis. The objective
function to be minimized is the total cost of failure, which includes the
initial structural cost as well as the cost of failure. The optimization pro-
cedure consists of two steps. The first step is the minimization of the
initial structural cost for a given risk level expressed in terms of the prob-
ability of failure. Designs are obtained for several values of the proba-
bility of failure. The second step is the minimization of the sum of the
initial structural cost and the cost of failure. The probability of failure,
which results from the minimization of this total cost is the optimal
probability of failure.

RELIABILITY BASED DESIGN FORMULATION

The determination of designs for a prescribed risk level is based on


the assumption that the load S and the resistance R are statistically in-
dependent random variables. The reliability analysis is based on the first
order second moment approach. This method requires only the mean
and coefficient of variation of the variables involved for the determina-
tion of the probability of failure, yet it can adequately provide an ac-
curate measure of the risk level (8,18). For lognormally distributed R and
742

J. Struct. Eng., 1983, 109(3): 741-757


S, the load-resistance relationship for a given risk level can be expressed
as
'1 + n2R ex 2
V-R = M-S ^YTsi ( P^Vin[(i + al)(i + n s)]}) (i)
Downloaded from ascelibrary.org by Universidad Nacional De Ingenieria on 05/26/18. Copyright ASCE. For personal use only; all rights reserved.

in which \xR, |xs = the mean values; ilR and fls = the coefficient of vari-
ation of the load and resistance; and (3 = the safety (or reliability) index.
For a given risk level, it is, thus, possible to determine member dimen-
sions and reinforcing steel area for each member if the statistics of the
load effect and member resistance are known. In this study, the loads
considered are the dead, live, and wind loads. The analysis of member
resistance includes pure bending for beams, combined bending and
compression for columns, and shear for both beams and columns. The
statistics of the load effect and the member resistance in the various fail-
ure modes are obtained as follows:
Load Effects.—Live and wind loads are both time- and space-depen-
dent and are generally modeled as stochastic processes. However, only
simplified formulations are used herein. For design purposes, the life-
time maximum total live load (i.e., the sum of the maximum sustained
live load and the maximum transient live load), and the lifetime maxi-
mum sustained live load are of interest. The statistics of these loads are
obtained from National Bureau of Standards (NBS) load surveys on of-
fice buildings (15). The mean and variance of the lifetime maximum sus-
tained load intensity are (15)
127
fjiLS = 19.3 H——, in pounds per square foot (2a)
/A,
18,900
OLS = 14-2 H , in pounds per square foot squared (2b)

whereas for maximum total live load


520
JJLLT = 18.7 H——, in pounds per square foot (2c)
A
=
CTLT 14.2 H , in pounds per square foot squared (2d)
A'
in which A{ = the influence area. The coefficient of variation of the live
load effect is obtained by considering additional uncertainties due to er-
rors in structural analysis in relating loads and load effects bSA, space-
wise variability and load concentration A sc , time-dependent effects A r ,
and load reduction AR (11,15)
«LL = 82LL + A2SA + A2SC + A2T + A| (3)
in which 8LL = the coefficient of variation of the live load intensity ob-
tained from Eq. 2; ASA = 0.05 (11); Asc = 1.682/A;; AT = 0.10; and AR =
0.15 for beams and 0.20 for columns (15).
The magnitude of the wind load depends on the location of the struc-
ture. Wind load is also time- and space-dependent. The wind pressure
743

J. Struct. Eng., 1983, 109(3): 741-757


acting on the structure at a height z can be expressed as (25)
W = Cpcjz = CPKZG (0.00256 V230) (4)
in which Cp = external pressure coefficient; qz = effective velocity pres-
Downloaded from ascelibrary.org by Universidad Nacional De Ingenieria on 05/26/18. Copyright ASCE. For personal use only; all rights reserved.

sure at height z; Kz = velocity pressure coefficient which depends on


type of exposure and height above ground; G = gust factor which de-
pends on the response characteristics of the structure; and V30 = basic
wind speed measured at 30 ft (9.2 m) above ground level (2). The mean
wind pressure \x.w and the coefficient of variation (cov) of the wind pres-
sure tiw are (25)
(iw = CpK2G (0.00256 V|„) (5)
tfw = n2Cp + n | , + fi2G + 4 n ^ .-...; (6)
For design, the lifetime (50-yr) maximum wind pressure WL is of interest.
The statistics of WL have been obtained by Ravindra and Galambos as
[iWl = 1.17 W5a and ftWi = 0.37 (25), where PV50 is the American National
Standards Institute (ANSI) 50-yr wind pressure (2).
The load combinations considered in the study are: (1) Dead load plus
life-time maximum total live load; and (2) dead load plus lifetime max-
imum sustained live load plus lifetime maximum wind load.
Basic Resistance Variables.—The resistance of a reinforced member
in a given mode is a function of several basic variables, such as concrete
strength, f'c, yield strength of the reinforcing steel, fy; area of steel, As;
and member dimensions. Quality control and workmanship also have
an important effect on the resistance of concrete members. Many com-
prehensive studies on the probabilistic analysis of reinforced concrete
structures have been conducted (10,11,13,14,24) in which uncertainties
in the various parameters used in the analysis of the resistance have
been estimated. The statistics of the basic resistance parameters used in
this study were obtained from the results of the previously mentioned
studies and are summarized in Table 1.

ANALYSIS OF RESISTANCE AND FAILURE MODES

In this section, statistics of the resistance of reinforced concrete sec-


tions in several modes of failure are analyzed. These statistics are re-
quired in the reliability based design procedure for the determination of
member dimensions and steel reinforcement areas.
Pure Bending.—The ultimate flexural capacity of a singly reinforced
concrete beam is

"••M1-^) <7
>
in which r\ = k2 lk\ k3 is a factor describing the concrete stress block prop-
erties. Beam failure in bending can occur in two modes: (1) Tensile fail-
ure occurs if the reinforcement yields before the concrete reaches its lim-
iting strain, ec„; and (2) compression failure occurs if the concrete reaches
its limiting strain before the steel yields. The mode of failure depends
on the steel ratio p = AJbd. The value of p which results in simultaneous
744

J. Struct. Eng., 1983, 109(3): 741-757


TABLE 1.—Statistics of Basic Resistance Variables
Parameter Mean, p. cov (£1)
(1) (2) (3)
Downloaded from ascelibrary.org by Universidad Nacional De Ingenieria on 05/26/18. Copyright ASCE. For personal use only; all rights reserved.

' /, 64ksi 0.14


/; 4.7 ksi 0.22
hh 0.72 0.13
•n 0.59 0.05
Ecu 0.004 0.16
A, — 0.04
b — 0.04
h, — 0.45 /h„
d, s, t, d, — 0.68 IK
Note: 1 ksi = 6.895 Pa.

crushing of concrete and yielding of steel is termed the balanced steel


ratio pi,. The event of failure in bending is the union of the events of
tension and compression failure, where both events are mutually exclu-
sive. Compression failure due to its sudden occurrence is not desired.
Therefore, the probability of compressive failure should be specified such
that it is small compared to the probability of tensile failure. In current
American Concrete Institute (ACI) code provisions (4), a tension failure
is assured by limiting p to less than 0.75pb. For fy = 60 ksi and f'c = 4
ksi, this limiting value is equivalent to a probability that failure is com-
pressive ranging from 0.09-0.145 (13). To maintain the same level of
ductility, a value of 0.10 is assigned to the probability that failure is due
to crushing of concrete. The maximum steel ratio is obtained from pmax
= p 6 /7 in which 7 is obtained from Eq. 1 by substituting pb for R and
pmaxfor S with (3 = (IT1 (0.10):

{ex [p V l n (1 + n
y =— = A / T T ^ P ^(1+ "-»)]} (8)
Pmax \ 1 + lip™
and coefficients of variation expressed as
IlL, = *l% +{%•+*& (9)

V Esecn + fy I \ES€CU + y
The mean value of the ultimate moment capacity of a singly reinforced
concrete beam is

^-u'(i-ifa) <">
and the coefficient of variation of Mu is

*•" ( ^ f )"<"'•+ * J + (T^)'<"« + "'+ ««> + ( A ) ' (12>


in which q = Asfy/bdf'c.
745

J. Struct. Eng., 1983, 109(3): 741-757


An iterative design procedure is used with As as the iterative param-
eter. The assumed value of As is substituted in Eq. 12 to obtain fiMu and
the computed value of As is obtained by solving Eq. 11. If the computed
value of As results in p > p max/ new member dimensions are assumed
Downloaded from ascelibrary.org by Universidad Nacional De Ingenieria on 05/26/18. Copyright ASCE. For personal use only; all rights reserved.

and the iteration is repeated. The iteration for As is not sensitive to vari-
ation in the assumed value of As.
Combined Axial Load and Bending.—Column resistance under com-
bined axial load and bending can be expressed in several ways (12). This
is due to the lack of a closed-form expression for the resistance under
combined loading. Assuming that the variation in axial forces is small
compared to the variation in bending forces, the column resistance may
be expressed as the bending resistance corresponding to a constant axial
force. To eliminate the case in which the bending moment is sensitive
to changes in axial forces, the axial force is limited to a certain minimum
value. This approach is similar to the ACI concept of limiting the design
axial strength to a certain maximum value regardless of the computed
design eccentricity (33). The limiting value of the column axial strength
is P„ = P 0 /7 in which 7 is obtained from Eq. 1, and
P0 = k3(bh, - 2As)f' + 2Asfy (13)
The total coefficient of variation of the limiting compressive force is

"p, = 4 {[h(6h, - 2As)f'cf(Hl + D.I + m, + «£) + 4A2S (tt\ + Of)}... (14)


* o

The column resistance is based on


1
R = - [hhf'cby(ht - hy) + Ast{f's + fs - cyk3f'cj\ • • (15)
2
in which cy = a constant which defines the percentage of compressive
steel area replacing the concrete area; y = the distance of the neutral
axis from the extreme compressive fiber of the column; and As = the
area of reinforcing steel at each face of the column. The column resis-
tance depends on the steel stresses fs and f's which in turn depend on
the value of y. Many combinations of / s and f's are possible; however,
only the three most common cases are examined here. The first case is
a tension failure in which both fs and f's reach the yield stress. The sec-
ond case is a compression failure in which f's reaches the yield stress
while the third case is a tension failure in which f$ reaches the yield
stress.
The location of the neutral axis is obtained from
r 1 (P + cyA5k3f'c)
C a s e 1: 16fl
y= „ , t,u< )

Ui + V « i + 4H2M3
Case 2 and 3: y = (16b)
2M3
in which u3 = kik3bfc and for Case 2:
ux = P + As{cyk3f'c - fy - Esecu) (17a)
746

J. Struct. Eng., 1983, 109(3): 741-757


u2 = EsemdAs (17b)
whereas for Case 3:
u, = P + As{cyk3f'c +fy- E s e cu ) (17c)
Downloaded from ascelibrary.org by Universidad Nacional De Ingenieria on 05/26/18. Copyright ASCE. For personal use only; all rights reserved.

"2 = Esacad,As (17d)


The aforementioned equations show that the column resistance is not
independent of the load variable. Unlike in the deterministic design, the
exact failure mode is not known since the load and the resistance vari-
ables are treated as random variables. An approximation is to evaluate
the failure mode in terms of the mean values of the variables. These
variables depend on the steel stress. The functional relationships of the
column resistance to these variables may be expressed as
For Case 1: R = R(b,k3,ht,As,fc,t,fy,P) (18a)
ForCase2: R = R{b,kx ,k3 A ,AS ,f^ecu,d,t,fy ,P) (18b)
For Case 3: R = R(b,fci ,k3,h, ,A, , # ,e cu ,d, ,t,fy ,P) (18c)
Uncertainties in the resistance variables are obtained by using a first-
order approximation and the chain rule of derivatives. The area of rein-
forcing steel for a prescribed risk level is obtained through an iteration
procedure. The assumed value of As is substituted in Eq. 16 to obtain y.
The coefficient of variation of the column resistance is obtained from Eq.
18 and the mean value from Eq. 15. Based on this column resistance,
the value of As is computed. If the resulting area of steel exceeds the
maximum allowable value, which is taken as 6% of the cross-sectional
area (4), new member dimensions are assumed.
Shear.—The shear resistance of a concrete cross section is determined
from the shear resistance of the concrete Vc and the shear resistance
provided by the reinforcing steel Vs, thus (4)
VU = VC + VS (19)
For beams Vc = 2\fj'cbd (20)
/ N \
and for columns 7 = 2 1+ -— f'bd (21)

in which Nu = the compressive force. The shear resistance of the rein-


forcing steel is determined from

Vs = ^ . . (22)
s
in which Av = the total area of shear reinforcement; and s = maximum
spacing as defined in Section 11.5.4 of the ACI code (4). The coefficient
of variation of the shear resistance is

" v . = -kWM + v?(ni + n\ + n* + a2,)] (23)


^ u

The coefficient of variation of the shear resistance is not derived from


747

J. Struct. Eng., 1983, 109(3): 741-757


Eq. 23 since this is an empirical equation. For all practical purposes, it
can be estimated by ilVc ~ ilfc (11).
Design is performed iteratively with Vs as the iterative parameter. If
the computed value of Vs is larger than Syfcbd/y, new member di-
Downloaded from ascelibrary.org by Universidad Nacional De Ingenieria on 05/26/18. Copyright ASCE. For personal use only; all rights reserved.

mensions are assumed. The iteration is not sensitive to the assumed value
of V,.

DESIGN FOR A PRESCRIBED RISK LEVEL

The iterative procedure described previously is used to determine


member dimensions and reinforcing steel area for a given failure prob-
ability. The failure probabilities considered vary from 10~2-10-6. Since
the design of a concrete structure comprises essentially the selection of
member sizes and reinforcing steel for each member, design is based on
member failure. Two failure modes are considered for each member.
These are specifically: for beams, failure due to bending and due to shear
and, for columns, failure due to combined axial load and bending and
due to shear. Assuming that forces in each member are perfectly cor-
related, selection of member dimensions and reinforcement for a pre-
scribed risk level is based on the failure mode which results in a higher
probability of failure. Shear failure, because of its sudden occurrence, is
undesirable. Therefore, a lower probability of failure is assigned to it.
The probability of failure in shear is taken to be 10% lower than for the
other failure mode. This is sufficient to ensure that failure of beams would
be due to bending and that failure of columns would result from com-
bined bending and compression rather than from shear. Also, this re-
duction is approximately equivalent to the results presented in Ref. 11.
Results of reliability analyses indicate that most columns in reinforced
concrete frames have lower failure probabilities than beams. To include
this in the design, two separate designs are considered. In the first de-
sign, the probability of failure of beams is specified to be equal to that
of the columns, i.e., pfbam = pfcol. In the second case, the probability of
failure of columns is taken to be one order of magnitude lower than
those of the beams, i.e., p/beam = 10p/col.
Initially, member dimensions ana the amount of steel reinforcement
are assumed. Based on these member dimensions, an iterative proce-
dure is used to obtain the steel area such that the member failure prob-
ability just exceeds the specified risk level. Maximum and minimum steel
reinforcement is as defined in the ACI code. However, steel area is not
the only design variable. Other design variables include the width and
the depth of the cross section. The determination of the steel area As for
a prescribed risk level is based on a given width and depth of the con-
crete cross section. The width and depth of th& concrete cross section
are initially assumed as small as allowed by the code and then increased
when the required steel area exceeds the maximum allowable limit. The
cross-sectional dimensions are specified such thaf the depth does not
exceed 2.75 times the width for beams and 2.0 times the width for col-
umns (26).
The iteration procedure, based on the aforementioned steps, will re-
sult in a frame with minimum member dimensions and maximum mem-
ber reinforcement. This does not necessarily result in minimum cost since
748

J. Struct. Eng., 1983, 109(3): 741-757


the unit cost of reinforcing steel is far greater than that of concrete. Fur-
ther cost optimization within the member is not considered since it re-
quires the formulation of explicit relationships among design variables,
maximum forces, and unit costs.
Downloaded from ascelibrary.org by Universidad Nacional De Ingenieria on 05/26/18. Copyright ASCE. For personal use only; all rights reserved.

The cost of the reinforced concrete member per unit length is, for
columns
Cs2A5Vv(b + ht)
C = CslAs + Cclbht + + 2Cn(b + ht) (24)
s
CsiAsv{b + ht)
and for beams C = Cs3As + Cc2bht + •• + Cf2b + 2Cf3h, (25)
s
in which Cs, Q , and Cfw are the unit costs of steel, concrete, and form-
work described in detail in Ref. 5. Asv = the area of shear reinforcement;
and s = the stirrup spacing. Eqs. 24 and 25 imply that the structural cost
depends only on the amount of materials and their unit costs. This may
not be the case in actual construction practice. For instance, uniformity
of structural members, which facilitates the reuse of formwork can result
in lower costs. To investigate such effects on overall structural cost, three
separate member configurations are considered. In the first design,
members are proportioned so as to result in minimum member dimen-
sions for all members. In the second design, all columns and beams within
a given story have the same dimensions. In the third design, all columns
and also all beams of the frame have the same dimensions.

OPTIMIZATION

In reliability based optimum design, the optimization procedure in-


volves the determination of the optimal level of safety which best sat-
isfies the design criteria and constraints. This is accomplished through
the minimization of an objective function, which takes into considera-
tion the probability of failure as well as the consequences of failure in
terms of cost. The objective function for cost optimization is (29)
Z = B(l - pf) - Q - Hpfs (26)
in which B = the benefit derived due to the presence of the intact build-
ing; Q = the initial building cost; H = the loss due to failure; and pis =
the probability of failure of the structural system. The optimal level of
safety is obtained when this objective function is maximized. Eq. 26 is
based on the assumption that the building is not reconstructed after fail-
ure. Since the value of Z should be positive in order to make construc-
tion of the building a profitable event, it is advisable to reconstruct the
building after any failure which would increase the value of the objective
function. For this case, the objective function is
Z = B - C, - (C, + H)pb (27)
Since the benefit is independent of the structural design and level of
safety, it can be eliminated in the optimization procedure. Eq. 27 then
becomes
T = Q + (Q + H)pts = C, + Lpfs (28)
749

J. Struct. Eng., 1983, 109(3): 741-757


in which T = the total cost; and Lpfs = the cost of failure. Optimization
now involves the minimization of the total cost.
Bounds on the Cost of Failure.—The objective function presented in
Eq. 28 is based on the probability of failure of the system. However,
Downloaded from ascelibrary.org by Universidad Nacional De Ingenieria on 05/26/18. Copyright ASCE. For personal use only; all rights reserved.

since design is usually based on member failure or critical section failure,


it is necessary to express the failure probability of the system in terms
of member failure probabilities. The estimation of system failure prob-
ability is quite complicated. Thus, the system failure probability is usu-
ally approximated by means of upper and lower bounds (3,7). Bounds
on the cost of failure CF can be obtained from these upper and lower
bounds on the system failure probability. The lower bound on the cost
of failure is
CF & Lpfi,max ' (29)
in which L = the loss due to failure of the entire structure; and Pfl,max is
the largest failure probability of the members (in this case the failure
probability of the beams).
Failure of one member in statically indeterminate structures does not
necessarily result in total collapse. Assuming that the failure of one
member results in partial damage to the structure, the upper bound on
the cost of failure can be estimated as
ii

Q s 2 LiVa (30)
in which n = the total number of members; pfi is the failure probability
of the z'th member; and L, the loss due to failure of the z'th member. For
practical purposes, L, can be approximated as proportional to the influ-
ence area of the member for the load effect being considered.
Cost Estimates.—Initial building cost is the sum of the structural cost
and the nonstructural cost. Structural cost is the cost of members plus
additional costs which are proportional to the structural cost. These in-
clude the cost of admixtures, finishing, and additional overhead costs
which depend on location, construction manner, and overhead calcu-
lation and can be as high as 25% of the cost of members (5). Nonstruc-
tural cost is obtained from the total cost of the structure by subtracting
the structural cost at a standard probability of failure which is taken as
1(T4. Nonstructural cost is, therefore, independent of the level of safety.
Estimates of total building costs for an office building in terms of unit
price per square foot based on the quality of construction are obtained
from Ref. 9.
Loss due to structural failure is more difficult to determine since it
includes many intangible items which are difficult to express in mone-
tary terms. The items considered in the estimation of loss due to failure
include: (1) Additional replacement cost; (2) damage to property; (3) cas-
ualties; (4) business interruption; and (5) legal services. Additional re-
placement cost is taken as two times the building cost (28). Loss due to
damaged property depends on the type and use of the office building.
It includes furniture, supplies, machinery, and computing equipment.
Unlike in the case of loss due to fire, it is assumed that there is no loss
of priceless documents. Thus, the value of damaged property is small
for typical office buildings. A conservative value for loss due to damaged
750

J. Struct. Eng., 1983, 109(3): 741-757


property is 50% of the cost of the building (16). It should be noted that
in certain unusual situations, e.g., when there is expensive computing
equipment in the building, the loss due to damaged property may be
equal to or even more than, the cost of the building.
Downloaded from ascelibrary.org by Universidad Nacional De Ingenieria on 05/26/18. Copyright ASCE. For personal use only; all rights reserved.

Casualties cannot be predicted before the nature of the structural fail-


ure is specified. As required by building codes, the structural design
should satisfy certain ductility requirements such that failure is preceded
by excessive deformation so as to provide warning of possible collapse.
This allows for a lower expectancy of casualties resulting from structural
failure. Liabilities due to death and injury are taken to only include those
incurred by the people present in the building at the time of collapse.
Statistics of the number of deaths and injuries are described in detail in
Ref. 32. According to Ref. 32, the number of fatalities caused by earth-
quake disasters per million 1970 U.S. dollars is 0.4 person with 40 in-
juries for each death. For tornadoes and severe winds, it is 0.1 person
with 100 injuries for each death. Another report (23) indicates that eight
structural failures due to overcrowding in Great Britain resulted in 125
fatalities and more than 1,000 injuries. Information about the actual
number of people involved in the failures was not provided in the afore-
mentioned report. Assuming that there are 70 persons for every 15,000
sq ft (1,395 m2) of floor area, as suggested in Ref. 23, a conservative
estimate of the probability of fatality due to structural failure is 0.003 for
each person involved, with 100 injuries, eight of them crippling, for each
death.
Loss due to fatality is based on an average death age of 30 and is taken
to be the sum of the person's salary until he would have reached the
retirement age of 65 years. Thus, the loss due to fatality is 35 times the
average annual net income of $15,000 (16). Costs for injuries are divided
into two parts, the first being those injuries of a serious or disabling
nature. The cost for disabling injuries is taken as $350,000 (6,28). The
second type of injury includes those which do not have a permanent
effect and require short periods of hospitalization. An average cost for
this type of injury is taken to be $5,000 (1). Assuming that the average
time to reconstruct office buildings is 4 yr, a conservative estimate of the
loss due to business interruption is of the order of the income of the
employees during that period. A value of 15% of the failure loss is as-
signed to the cost of legal services.

RESULTS

The reliability based optimization procedure is applied to the design


of two reinforced concrete frames—a four-story two-bay frame and a
10-story three-bay frame. The first story of each frame is 16 ft (4.88 m)
high, whereas the upper stories are each 13 ft (3.97 m) high. The spacing
between adjacent columns as well as between adjacent frames is 25 ft
(7.63 m).
As mentioned earlier, three separate member configurations are con-
sidered. For each case, designs are obtained for failure probabilities vary-
ing from 10~2-10~6. Since failure probabilities of columns are usually lower
than for beams, a comparison of frames designed for a uniform proba-
bility of failure for all members and frames with columns designed for
751

J. Struct. Eng., 1983, 109(3): 741-757


a probability of failure of one order of magnitude lower than that of
beams is also made. Member dimensions for the four-story and the 10-
story frame for the three-member configurations considered are shown
in Figs. 1 and 2. These member dimensions were obtained for the fol-
lowing risk levels: p/beam = 10' 4 5 and pu = 10~55.
Downloaded from ascelibrary.org by Universidad Nacional De Ingenieria on 05/26/18. Copyright ASCE. For personal use only; all rights reserved.

10x25
o I „.„
DESIGN 3
Booms 110x25
ColumnS'lOxlG 10X25 2 I ,0,,

10x25 2 I ,o >26
•9
o 2

I
x
25-0
~4» H

FIG. 1.—Member Dimensions for Four-Story Frame; pfbem = 10 " 5 ; p/col = 10

9X23 10x22 . 9 x 2 3 9x23

10x22 I 10x22 .'9x24 ! 9x24

I
10x16

10x22 10x23 J_?x24 9x2 4

10x23 I 10x23 19x24 I 9x24

10x24 I 10x24 |l0x25 I 10x23

1 1
10x24 10x25 {10x25 I 10X25

10X25
f
i
10x25 ;IOx25 1 10X25

10x25 1 10x25 110x26 1 10X26

1
11X20

10x26 I 10x26 il0»26 10x26


12x22

(0X27 (0x26 'lOxZS I 10x26

E2
r I
-
l^ifcS! ,| *°'-° ,), " ' - ° "

4,s
FIG. 2.-—Member Dimensions for 10-Story Frames, p/b(,am = 10 , pfml = 10

752

J. Struct. Eng., 1983, 109(3): 741-757


Initial Stractufal Cost.—For the four-story frame, the difference in the
initial structural cost among the three designs is small, as shown in Fig.
3. The values shown in Fig. 3 are for the case when p^beam = 10pcol. Similar
results were obtained for the case in which pfbam = 10pcoi. Similar results
Downloaded from ascelibrary.org by Universidad Nacional De Ingenieria on 05/26/18. Copyright ASCE. For personal use only; all rights reserved.

were obtained for the case in which PfUm = pcoi • For the 10-story frame,

Design 3

n J
10 10 10"1
Member Failure Probability, pf.

FiG. 3.—Cost of Members for Four-Story Frame, p/b, 10R'/col

Member Failure Probability, pfi max

FIG. 4.-—Cost of Members of 10-Story Frame, p/beam = 10p/col

753

J. Struct. Eng., 1983, 109(3): 741-757


structural cost for the third design is significantly higher than for the
other two designs, as can be seen from Fig. 4. When the structural cost
for the third design was recomputed assuming that formwork could be
reused three times (due to uniformity of members), a reduction in cost
Downloaded from ascelibrary.org by Universidad Nacional De Ingenieria on 05/26/18. Copyright ASCE. For personal use only; all rights reserved.

of about 20% was obtained. This reduction is significant since the av-
erage unit cost of formwork when used three times is only 25% less than
that when formwork is used once (5). Using information contained in
Refs. 5 and 9, the cost of members was found to be between 2.5%-3.5%
of the total cost of failure. The cost difference between the first and sec-
ond design is quite small even for the 10-story frame. This is because
the larger member dimensions result in smaller amounts of reinforcement.
Total Cost and Optimal Risk Level.—Upper and lower bounds on the
expected total cost (computed for various values of Pfi,max) for the four-
story and the 10-story frames are shown in Figs. 5 and 6, respectively,
for the case in which Pfhcm = 10p/col. For frames designed for a uniform
failure probability, i.e., pybeam = pycoI, it was found that the initial costs and
the lower bounds on the expected cost were smaller. However, the up-

10 10"' 10 •
Member Failure Probability, p f .

FIG. 5.—Expected Total Cost for Four-Story Frame, p/beam = 10p/col, Design 1

47000

46000
bound \ \ "PP^
\ \ \ bound
45000

44000

43000 42653 \ \ . y'

42000
42178 \L _Jf

41000

40000_ ! * _ J OP*™ 1 _
1 range
cost of members ' i
39000

38000

37000 i i i i ! i

Member Failure Probability, pf.

FIG. 6.--Expected Total Cost for 10-Story Frame, pfhmm = 10p/col, Design 1

754

J. Struct. Eng., 1983, 109(3): 741-757


TABLE 2.—Optima! Range of Risk Level
EXPONENT OF FAILURE PROBABILITY (p„ m » = 1 0 - )

P/beam ~ P/col P/beam 10p /c „,


Downloaded from ascelibrary.org by Universidad Nacional De Ingenieria on 05/26/18. Copyright ASCE. For personal use only; all rights reserved.

Story Design Upper bound Lower bound Upper bound Lower bound
0) (2) (3) (4) (5) (6)
4 1 4.5 5.0 4.3 4.4
4 2 4.6 5.1 4.3 4.4
4 3 4.4 . 5.2 4.5 4.5
4 3a 4.6 5.2 4.6 4.7
10 1 4.3 5.4 4.4 4.8
10 2 4.4 5.4 4.4 4.8
10 3 4.4 5.2 4.1 5.1
10 3a 4.4 5.2 4.1 5.1
"Form work is used three times.

per bounds on the total expected cost of the frame were higher.
The range of values obtained for the optimal risk level are shown in
Table 2. The computed optimal risk level for the various frames and
designs varies from 10~ 41 -10 -5 ' 4 with an average value around 10~4,5.
The maximum difference between the two bounds on the optimal risk
level is about one order of magnitude with the bounds being closer for
the frames with columns designed for a lower failure probability.
It should be noted that the computed expected cost of failure, and,
thus, the optimal risk level, depends on the assumptions made in es-
tablishing monetary values for the various items which contribute to the
cost of failure. There is some uncertainty in estimating these values. Sen-
sitivity studies indicate that no significant change in the optimum risk
level occurs within a range of one-half the failure cost lower and two
times the failure cost higher than the calculated expected cost of failure.
To change the optimum failure probability from 10~4 5-lCT5'5 would re-
quire approximately a one-order-of-magnitude change in the calculated
expected cost of failure. The optimal risk level is, thus, relatively insen-
sitive to the expected cost of failure. Similar results were obtained in a
reliability based optimization study of five- and ten-story steel frames
(27).

SUMMARY AND CONCLUSIONS

A reliability based optimization procedure for reinforced concrete


structures was presented. The reliability analysis was based on a first-
order second moment method. Statistics of the load and resistance vari-
ables were obtained from available information. An iterative procedure
was used to determine member dimensions and reinforcing steel areas
for prescribed risk levels ranging from 10" 2 -10 -6 . Using available infor-
mation, initial structural cost and the cost of failure were estimated. The
optimal failure probability was determined by minimizing the total cost.
From an analysis of the four- and 10-story reinforced concrete build-
ings considered, the following conclusions can be drawn:
755

J. Struct. Eng., 1983, 109(3): 741-757


1. The optimization procedure p r e s e n t e d in this study can provide a n
accurate estimate of the optimal level of safety provided the structural
and architectural design, the e n v i r o n m e n t a n d the type of structure are
specified.
Downloaded from ascelibrary.org by Universidad Nacional De Ingenieria on 05/26/18. Copyright ASCE. For personal use only; all rights reserved.

2. Optimization to obtain m i n i m u m structural cost by minimizing


member dimensions can be performed in several ways, a n d different
results are obtained d e p e n d i n g on the procedure used. This can be ob-
served from the differences in cost for the three-member configurations
considered. To obtain a final optimized structure, the iteration algo-
rithms should be based on a formulation which explicitly considers j
member dimensions, reinforcing steel areas, formwork, a n d distribution
of forces. I
3. For the structures considered, the range of optimal failure proba-
bility, expressed in terms of m a x i m u m m e m b e r failure probability, w a s
found to be lCT^-lCT 5 , 4 . This range d e p e n d s o n the procedure u s e d to
estimate the u p p e r a n d lower b o u n d s o n the cost of failure. j

APPENDIX !.—REFERENCES i
1. "Accident Facts," The National Safety Council, Chicago, 111., 1975.
2. American National Standard Building Code Requirements for Minimum Design Loads
in Buildings and Other Structures, American National Standards Institute, ANSI
A58.1-1972, New York, N.Y., 1972. j
3. Ang, A. H-S., and Amin, M., "Reliability of Structures and Structural Sys- ,
terns," Journal of the Engineering Mechanics Division, ASCE, Vol. 94, No. EM2, \
Apr., 1968. !
4. Building Code Requirements for Reinforced Concrete (ACI318-77), American Con- )
crete, 1979.
5. Building Cost File, American Concrete McKee-Berger-Mansueto Inc., Central \
Edition, Van Nostrand Reinhold, New York, N.Y., 1979.
6. Campey, I. K., "Types of Cover," The Building Economist, Vol. 17, June, 1978, >
pp 11-13. j
7. Cornell, C. A., "Bounds on the Reliability of Structural Systems," Journal of
the Structural Division, ASCE, Vol. 93, No. ST1, Feb., 1967, pp. 171-220. ',
8. Cornell, C. A., "A Probability-Based Structural Code," ACI Journal, Dec, 1969, '
pp. 974-985.
9. Edgerton, W. H., and Hlibok, A. J., Real Estate Valuation Cost Files, Van Nos-
trand Reinhold, New York, N.Y., 1979.
10. Ellingwood, B. R., "Reliability Based Criteria for Reinforced Concrete De- i
sign," Journal of the Structural Division, ASCE, Vol. 105, No. ST4, Apr., 1979,
pp. 713-737. )
11. Ellingwood, B. R., Reliability Basis of Load and Resistance Factors for Reinforced !
Concrete Design, NBS Building Science Series 110, Feb., 1978.
12. Ellingwood, B. R., "Statistical Analysis of RC Beam-Column Interaction," I
Journal of the Structural Division, ASCE, Vol. 103, No. ST7, July, 1977, pp.
1377-1388.
13. Ellingwood, B. R., and Ang, A. H-S., A Probabilistic Study of Safety Criteria
for Design, Structural Research Series No. 387, University of Illinois, Urbana,
111., June, 1972.
14. Ellingwood, B. R., and Ang, A. H-S., "Risk Based Evaluation of Design Cri-
teria," Journal of the Structural Division, ASCE, Vol. 100, No. ST9, Sept., 1974,
pp. 1771-1788.
15. Ellingwood, B. R., and Culver, C , "Analysis of Live Loads in Office Build-
ings," Journal of the Structural Division, ASCE, Vol. 100, No. ST8, Aug., 1977,
pp. 1551-1559.
16. Greene, M. R., Risk and Insurance, Southwestern Publishing, Cincinnati, Ohio
1973.

756

J. Struct. Eng., 1983, 109(3): 741-757


17. Hilton, H. H., and Feigen, M., "Minimum Weight Analysis Based on Struc-
tural Reliability," Journal of the Aerospace Sciences, Vol. 27, No. 9, pp. 641-652,
Sept., 1960.
18. Lind, N. C , "Consistent Partial Safety Factors," Journal of the Structural Di-
vision, ASCE, Vol. 97, No. ST6, June, 1971, pp. 1651-1669.
Downloaded from ascelibrary.org by Universidad Nacional De Ingenieria on 05/26/18. Copyright ASCE. For personal use only; all rights reserved.

19. Mortimer, T. J., "Insurance: A New Look at an Old Subject," The Building
Economist, Dec, 1979.
20. Moses, V., "Design for Reliability—Concepts and Applications," Chapter 13,
Optimum Structural Design, R. H. Gallagher, ed., and O. C. Zienkiewicz, eds.,
John Wiley and Sons, Inc., 1973.
21. Moses, F., and Kinser, D. E., "Optimum Structural Design with Failure Prob-
ability Constraints," AIAA Journal, Vol. 5, No. 6, June, 1967.
22. Moses, F., Stevenson, J. D., "Reliability-Based Structural Design," Journal of
the Structural Division, ASCE, Vol. 96, No. ST2, Feb., 1970, pp. 221-244.
23. Pauls, J. L., "Management and Movement of Building Occupants in Emer-
gencies," Proceedings of the Second Conference Designing to Survive Severe Haz-
ards, IIT Research Institute, Nov., 1977, pp. 103-130.
24. Portillo, M. G., Ang, A. H-S., Evaluation of Safety of Reinforced Concrete Build-
ings to Earthquakes, Structural Research Series No. 433, University of Illinois,
Urbana, 111., Oct., 1976.
25. Ravindra, M. K., Cornell, C. A., and Galambos, T. V., "Wind and Snow
Load Factors for Use in LRFD," Journal of the Structural Division, ASCE, Vol.
104, No. ST9, Sept., 1978, pp. 1443-1458.
26. Reese, R. C , CRSI Design Handbook, Concrete Reinforcing Steel Institute,
Chicago, 111., 1978.
27. Rojiani, K. B., and Bailey, G. L., "A Comparison of Reliability Based Opti-
mum Designs and AISC Code Based Designs," Proceedings, International
Symposium on Optimum Structural Design, Tucson, Ariz., Oct., 1981.
28. Roos, N. R., and Gerber, J. S., Governmental Risk Management Manual, Risk
Management Publishing, Tucson, Arizona, 1976.
29. Rosenblueth, E., Mendoza, E., "Reliability Optimization in Isostatic Struc-
tures," Journal of the Engineering Mechanics Division, ASCE, Vol. 97, No. EM6,
Dec, 1971, pp. 1625-1640.
30. Switzky, H., "Minimum Weight Design with Structural Reliability," Proceed-
ings, Fifth Annual Structures and Materials Conference, American Institute
of Aeronautics and Astronautics, Palm Springs, California, Apr., 1964.
31. Turkstra, C. J., "Choice of Failure Probabilities," Journal of the Structural Di-
vision, ASCE, Vol. 93, No. ST6, Dec, 1967, pp. 189-200.
32. Wiggens, J. H., "National Losses and Mitigation Effects for Air, Earth and
Waterborne Natural Hazards," Proceedings, Conference on Designing to Sur-
vive Severe Hazards, IIT Research Institute, Chicago, 111., Nov., 1977.
33. Winter, G., Nilson, A. H., Design of Concrete Structures, McGraw-Hill Book
Co., Inc., New York, N.Y., 1979.

757

J. Struct. Eng., 1983, 109(3): 741-757

Вам также может понравиться