Available online at www.sciencedirect.

com

Advances in Engineering Software 39 (2008) 505–513

www.elsevier.com/locate/advengsoft

Structural reliability analysis using Monte Carlo simulation

and neural networks
João B. Cardoso *, João R. de Almeida, José M. Dias, Pedro G. Coelho
Faculty of Science and Technology, New University of Lisbon, 2829-516 Caparica, Portugal

Received 10 September 2003; received in revised form 13 March 2007; accepted 30 March 2007
Available online 22 May 2007

Abstract

This paper examines a methodology for computing the probability of structural failure by combining neural networks (NN) and
Monte Carlo simulation (MCS). MCS is a powerful tool, simple to implement and capable of solving a broad range of reliability prob-
lems. However, its use for evaluation of very low probabilities of failure implies a great number of structural analyses, which can become
excessively time consuming. The proposed methodology makes use of the capability of a NN to approximate a function for reproducing
structural behavior, allowing the computation of performance measures at a much lower cost. This approach seems very attractive, and
its main challenge lies in the ability of a NN to approximate accurately complex structural response. In order to assess the validity of this
methodology, a test function and two structural examples are presented and discussed. The second example is also used to show how this
methodology can be used to perform reliability-based structural optimization.
 2007 Elsevier Ltd. All rights reserved.

Keywords: Reliability-based optimization; Structural reliability; Monte Carlo simulation; Neural networks

1. Introduction • Approximate probabilistic or level 2 methods such as

the first order or second order reliability methods
The structural designer must verify, within a prescribed (FORM/SORM) where the probability of failure is
safety level, the serviceability and ultimate conditions, based on the reliability index b [3].
commonly expressed by the inequality: Sd < Rd, where Sd • Exact probabilistic or level 3 methods, where the prob-
represents the action effect and Rd the resistance. The ability of failure is computed from the joint probability
intrinsic random nature of material properties and actions distribution of the random variables associated with the
is actually considered by the Eurocodes [1,2], which classify actions and resistances.
the methods available to deal with this randomness in three
levels: Theory and methods for structural reliability have been
developed substantially in the last few years and they are
• Semi-probabilistic or level 1 methods, the most used in actually a useful tool for evaluating rationally the safety
common practice, where the probability of failure is of complex structures or structures with unusual designs.
indirectly considered through the definition of charac- Recent developments allow anticipating that their applica-
teristic values and the application of partial safety tion will gradually increase, even in the case of common
indexes. structures.
Monte Carlo simulation (MCS) is a simulation method
of level 3 that presents the following characteristics: it can
*
Corresponding author. Tel.: +351 212948567; fax: +351 212948531. be applied to many practical problems, allowing the direct
E-mail address: [email protected] (J.B. Cardoso). consideration of any type of probability distribution for the

0965-9978/$ - see front matter  2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.advengsoft.2007.03.015
506 J.B. Cardoso et al. / Advances in Engineering Software 39 (2008) 505–513

random variables; it is able to compute the probability of where (x1, x2, . . ., xn) are values of the random variables and
failure with the desired precision; it is easy to implement. fX 1 ;X 2 ;...;X n ðx1 ; x2 ; . . . ; xn Þ is the joint probability density
However, despite the advantages it presents, the use of function.
this method is not widespread in structural reliability The Monte Carlo method allows the determination of
because it is not efficient when compared to level 2 meth- an estimate of the probability of failure, given by
ods. In fact, MCS requires a great number of structural
analyses, one for each sample of the set of random vari- 1 XN
pf ¼ IðX 1 ; X 2 ; . . . ; X n Þ ð2Þ
ables. The number of analyses needed to evaluate the prob- N i¼1
ability of failure of a structure with a prescribed precision
depends on the order of magnitude of that probability. where I(X1, X2, . . ., Xn) is a function defined by
As the values of the probability of failure associated to 
1 if gðX 1 ; X 2 ; . . . ; X n Þ 6 0
the ultimate limit states vary normally between 104 and IðX 1 ; X 2 ; . . . ; X n Þ ¼ ð3Þ
0 if gðX 1 ; X 2 ; . . . ; X n Þ > 0
106, the number of analyses to be performed for ensuring
a 95% likelihood that the actual probability be within 5% According to (2), N independent sets of values
of the computed one must be at least 1.6 · 107 to x1, x2, . . ., xn are obtained based on the probability distribu-
1.6 · 109, according to Shooman [4]. tion for each random variable and the failure function is
These analyses are frequently performed with the help of computed for each sample. Using MCS, an estimate of
finite elements codes. Therefore, the computation time can the probability of structural failure is obtained by
be prohibitively high, especially when the structure exhibits
non-linear behavior or the numerical model is rather NH
pf ¼ ð4Þ
complex. N
To eliminate this drawback, it is proposed here the use where NH is the total number of cases where failure has
of neural networks (NN) to approximate structural occurred.
response. Once properly trained, a NN allows the determi-
nation of the structural performances with a very small
number of operations and at a fraction of the cost of the 3. Neural networks
corresponding structural analysis. This methodology
allows the application of MCS to practical cases of great NN are numerical algorithms inspired in the functioning
complexity where the direct use of this method would not of biological neurons. This concept was introduced by
be feasible. McCulloch and Pitts [6], who proposed a mathematical
To illustrate the techniques proposed, three examples model to simulate neuron behavior. Use of NN has become
are presented, considering respectively a mathematical widespread in several fields of engineering, such as struc-
function and two different structures. In the first example, tural mechanics [7,8] and structural reliability [9]. Papad-
a non-linear analytical function is approximated by means rakakis et al. [10] presented an approach in which a NN
of NN. In the second, a linear elastic steel frame is designed was associated to MCS in order to obtain the probability
according to the Eurocodes 1 and 3 [1,2], and subsequently of structural failure. A similar approach is proposed herein.
the probability of failure of the frame is computed by a The model adopted is shown in Fig. 1, representing
Monte Carlo approach in which NN are used for reproduc- the neuron m that receives an input signal vector
ing the structural response. In the third, a genetic algorithm x = [xi, . . ., xL]T from L input channels. It then computes
is used in combination with NN approximation and Monte the weighted sum of the components of x, multiplying each
Carlo method to perform reliability-based optimization of component xk by a coefficient wmk reflecting the impor-
a steel truss. tance of the input channel k. The neuron m activation,
am, is given by the expression

2. Monte Carlo simulation X

L
am ¼ wmk xk þ bm ð5Þ
k¼1
A reliability problem is normally formulated using a
failure function, g(X1, X2, . . ., Xn), where X1, X2, . . ., Xn are
random variables. Violation of the limit state is defined x1
by the condition g(X1, X2, . . ., Xn) 6 0 and the probability wm1
of failure, pf, is expressed by the following expression [5]: x2
wm2
f (am) sm
pf ¼ P ½gðX 1 ; X 2 ; . . . ; X n Þ 6 0 x3 wm3
Z Z Z .
.
wmL
¼ ... fX 1 ;X 2 ;...;X n ðx1 ; x2 ; . . . ; xn Þ dx1 dx2 . . . dxn .

xL bm
gðX 1 ;X 2 ;...;X n Þ60

ð1Þ Fig. 1. Artificial neuron.

 J.B. Cardoso et al. / Advances in Engineering Software 39 (2008) 505–513 507

It should be noticed that Eq. (5) includes a corrective the exact value of the function for a prescribed set of input
term bm, called bias, allowing for the possibility of having values, named as the training set.
a non-negative activation am when all xk are zero. The error for the training set is evaluated using the fol-
The output signal of neuron m, sm, is the numerical lowing equation:
value that results from the computation of an activation
function, f(am). In the present work, the logistic function 1X t
1X r
E¼ ðsij  oij Þ2 ð7Þ
is adopted. This function is obtained making a = 1 in the t i¼1 r j¼1
expression of the sigmoid function, which has the following
form: where t is the number of elements of the training set, r is the
number of neurons in the output layer, sij and oij represent,
1 respectively, the computed and the exact value of the func-
sm ¼ f ðam Þ ¼ ð6Þ
1 þ eaam tion for the j neuron and i element of the set.
Due to the network architecture, the error (7) to be min-
Several neurons can be assembled to form a network.
imized often presents several local and global minima.
Different network architectures for such an assembly have
Unlike gradient-based algorithms, which may be trapped
been studied, together with algorithms to perform the
in local minima, genetic algorithms are generally successful
determination of coefficients wmk and bm, cf. Rosenblatt
for performing multimodal optimization. In this work, a
[11] and Rumelhart et al. [12]. In this work, it was decided
hybrid algorithm is used, combining gradient-based and
to adopt a well-known network arrangement, the multi-
genetic algorithms. Minimization with genetic algorithms
layer perceptron [10], which is composed of several neurons
is performed at an initial phase and the solution obtained
arranged in layers, as shown in Fig. 2.
is used subsequently as a starting point by a conjugate gra-
Fig. 2 represents a NN with three layers: an input layer
dient algorithm.
composed by three neurons, in which no transformations
As the coefficients to be determined, wmk and bm, are
are performed; one hidden layer with four neurons; an out-
non-integer, the genetic algorithm is adapted to work with
put layer with three neurons. It is possible to prove [13]
this type of values. The genetic algorithm uses binary chro-
that this type of network with sigmoid activation functions
mosomes where each coefficient associated with hidden and
in the hidden and output layers can approximate any con-
output layers is represented by eight genes, implying that it
tinuous function with satisfactory precision, provided that
can assume one of 28 = 256 different values. The real value
it has enough neurons in the hidden layer.
r, corresponding to any combination of the eight genes
The process of obtaining the unknown coefficients wmk
expressed by the integer z, depends on the following trans-
and bm required to approximate the prescribed function
formation between z and r:
is called training and it is a somewhat challenging task.
 
The most common training process, called supervised rmax  rmin
training [12], consists of proposing some initial values to r ¼ rmin þ z ð8Þ
28  1
the coefficients and then adjusting those values in order
to minimize the error between the predicted output pro- where rmin and rmax are the lower and upper limits of r,
duced by the NN and the exact value of the function. respectively. In order to perform the optimization of a
Therefore, to perform this training, it is necessary to know function of real variables, the strategy followed consists

f(a11)
w111
x1 w112 s12
= w121w 1 1 f(a12)
13 b1
w122
w211 w132 w 2
b12
1
f(a21) 14

w22
w212
x2 w231 w222 s22
= b2 1
f(a22)
w311 w232 2
1 w24
w32
f(a31) 2 b22
w331 w32w
2 31

x3 w332 s32
w421 w41
1
b31
= f(a32)
w342
w431
f(a41)
b32

b 41

Fig. 2. Multi-layer perceptron.

508 J.B. Cardoso et al. / Advances in Engineering Software 39 (2008) 505–513

of restarting successively the genetic algorithm, modifying Table 1

progressively the domain bounds rmin and rmax in order Errors obtained with NN approximation of function F(x, y)
to circumscribe the optimization to a closer vicinity of s1 Training error, E Max error with
the optimal solution obtained in the previous iteration. test set, e (%)
1 1.55 · 102 114
4. Example 1: test function 6 7.15 · 104 32
12 9.22 · 105 26
18 5.49 · 106 2.5
To evaluate the capability of a NN to approximate accu- 24 1.79 · 106 1.9
rately a non-linear function, an analytical test function was 30 1.19 · 106 1.5
first considered: 36 1.97 · 107 0.93

F ðx; yÞ ¼ 0:3 þ ð2 sinðxÞ cosðyÞ þ sinðxyÞÞ=6 ð9Þ

with x 2 [3.5, 5.5] and y 2 [2.0, 4.0]. The function is repre- for each variable. Such large number of test points was
sented graphically in Fig. 3. It is observed that the test adopted in order to get a good assessment of the precision
function value in the prescribed domain is always within achieved throughout the domain. The training error, com-
the range of the logistic function [0, 1], and so normaliza- puted with Eq. (7), and the testing error, given by Eq. (10),
tion is not needed. It may be assumed that non-linear struc- are presented in Table 1.
tural response is generally described by functions simpler Fig. 4 presents graphically the approximation to the
than this one. Consequently, if a good approximation of function F(x, y), computed by each one of the first four net-
(9) is achieved through NN, this would suggest that this ap- works considered in Table 1. It can be seen that, as the
proach can be applied successfully to structural problems. number of hidden neurons increases, the approximation
Several different NN were considered, keeping constant gradually converges to the exact function. Given the fine
the number of neurons in the input and output layers, s0 match achieved, the 14 · 14-point training set adopted
and s2, respectively equal to 2 and 1, and varying the num- seems adequate. Hence, no further effort was made in order
ber of neurons in the hidden layer, s1. It is worth noticing to refine the training set.
that a greater number of elements of the training set
increases both the accuracy of the response and the training 5. Example 2: single story steel frame
time of the NN, so its size must result from a compromise
between time and quality of the approximation. For this As an example of the methodology proposed for com-
example, a set of 196 different values of the function were puting the probability of failure, the single bay steel frame
obtained, considering a uniformly spaced grid of 14 points represented in Fig. 5 is considered and analysed using the
for each variable, and then presented to the network for methodology currently adopted in Europe for structural
training. After training, the performance of each network design. Applied actions (dead load, live load and wind)
was evaluated with the help of the following error measure: are defined according to the Eurocode 1 [1], and the steel
  structure is designed according to the Eurocode 3 [2]. The
sij  oij 
e ¼ max    for i ¼ 1; . . . ; t and j ¼ 1; . . . ; r ð10Þ design must satisfy safety conditions related to serviceabil-
oij 
ity and ultimate limit states. This is done according to a
where all symbols have the same meaning as in Eq. (7). level 1 methodology and the structural response is deter-
Since a large number of elements in the test set does not mined using a finite element code. After choosing the
represent an inconvenient, a test set of 1024 function values appropriate rolled-steel shapes, safety conditions for ulti-
was considered, using a uniformly spaced grid of 32 points mate states are again checked, but now following a level
3 methodology in which NN are used to approximate the
Z structural response and MCS is performed for computing
Y the probability of failure.
X
5.1. Structural design according to Eurocodes (level 1)

The frame considered is part of an industrial warehouse

0.8 with dimensions 20 · 10 · 4 m. The structure of the ware-
0.6 house consists of five independent and identical frames
0.4
spaced 5 m apart from each other. The geometry of each
3.5 4 frame is shown in Fig. 5. Applied actions are defined
4 3.5 according to the Eurocode 1, and their characteristic values
4.5 3
are:
5 2.5
5.5 2
• Dead load – 0.5 kN/m2.
Fig. 3. Test function F(x, y). • Live load – 2 kN/m2.
 J.B. Cardoso et al. / Advances in Engineering Software 39 (2008) 505–513 509

Z Z

Y Y

X X

0.8 0.8
0.6 0.6
0.4 0.4

3.5 4 3.5 4

4 3.5 4 3.5

4.5 3 4.5 3

5 2.5 5 2.5

5.5 2 5.5 2

s1 =1 s1 =6

Z
Z
Y
Y

X
X

0.8 0.8
0.6 0.6
0.4 0.4

3.5 4 3.5 4
4 3.5 4 3.5
4.5 3 4.5 3
5 2.5 5 2.5
5.5 2 5.5 2

s1 = 12 s1 = 18

Fig. 4. Approximations of test function obtained with NN.

B C functions considered, according to the Eurocode 3. In this

table, NSd and My,Sd are the design axial force and bending
moment; A and Wpl,y are the cross-section area and the
L2 = 4 m
plastic bending modulus; fy is the nominal yield stress of
the material (steel S235); vz, vmin and vLT are reduction
coefficients for the following cases: out-of-plane buckling,
A D minimum between coefficients for in-plane and out-of-
plane buckling, and lateral torsional buckling.
L1= 10 m
For the cross-sections of the elements, standard rolled-
steel shapes are adopted. It can be concluded that a frame
Fig. 5. Steel frame. built with HEA 260 columns and a HEA 300 beam verifies
safety for all limit state functions. The corresponding dis-
placements and internal forces, obtained using a linear
• Wind – For the region where the structure is going to be
elastic finite element code developed by the authors, are:
built (Portuguese seashore), the prescribed dynamic
pressure of the wind is equal to 0.456 kN/m2. From this
pressure, the wind loading is obtained following the pro- Table 2
Limit state functions for steel frame
cedure prescribed in the Eurocode 1 [1].
Limit function Description
Load combination is performed according to the level 1 dH 6 L2/150 Maximum horizontal displacement of the
methodology followed in Eurocode 1, where the character- column
dV 6 L1/300 Maximum vertical displacement of the beam
istic values for the loads are multiplied by combination N Sd 1;5M y;Sd
vmin Afy =1;1 þ W pl;y fy =1;1 61 Resistance to bending and compression of the
coefficients (to reflect the non occurrence of simultaneous column
N Sd 1;5M y;Sd
loads) and by partial safety coefficients (to ensure, indi- vz Afy =1;1 þv
LT W pl;y fy =1;1
61 Resistance to bending and compression of the
rectly, the appropriate safety level for each limit state). beam, considering also lateral torsional
Table 2 shows the serviceability and the ultimate limit state buckling
510 J.B. Cardoso et al. / Advances in Engineering Software 39 (2008) 505–513

• dH = 1.32 mm. random variables that influence structural response. As a

• dV = 16.8 mm. linear stress–strain relation is assumed in this example, only
• NSd(COLUMN) = 98.6 kN (on top of right column – sec- the random loads and the Young’s modulus affect the
tion C). structural response (the yield stress is considered only in
• My,Sd(COLUMN) = 122.5 kN m (on top of right column – the limit state functions) and so s0 = 4. The number of neu-
section C). rons in the output layer is equal to the number of displace-
• NSd(BEAM) = 46.7 kN (on right end of beam – section C). ments and internal forces needed to compute the limit state
• My,Sd(BEAM) = 122.5 kN m (on right end of beam – sec- functions, that is, s2 = 5.
tion C). A first set of NN was trained and subsequent Monte
Carlo simulations were performed. Results revealed that
5.2. Probabilistic model the probability of failure for the serviceability limit states
was insignificant compared to the allowable one (a value
Subsequently, the probability of failure of the frame of 1 · 108 was obtained and the admissible Pf is of the
considered above is evaluated. For that purpose, statistic order of magnitude of 102). Therefore, the two service-
models must be defined for each random variable involved. ability limit state functions considered could be disregarded
Hence, normal distributions are assumed for the Young’s and the problem could be simplified. This was expected,
modulus and dead load, whereas for the live load, wind since the values of the horizontal and vertical displace-
load and yield stress lognormal distributions are consid- ments, dH, dV, obtained in Section 5.1 were much lower
ered. The probabilistic parameters adopted for each than their limiting values shown in Table 2. Knowing also
variable are chosen appropriately, according to the charac- that the internal forces in a linear elastic structure made of
teristic percentiles prescribed in the Eurocode 1 and a single material do not depend on the material Young’s
making the resulting characteristic values coincide with modulus, this variable is removed from the probabilistic
the previously used when level 1 design was carried out. model since it does not have any influence on the ultimate
This way, both methodologies (levels 1 and 3) can be limit state functions considered.
compared. The simplified networks with s0 = s2 = 3 are trained
The statistic parameters adopted for the random vari- with a training set representative of the structural response.
ables are shown in Table 3. In this table, the characteristic More relevance is given to values situated in the upper
values for live load and wind load are the percentile 98 of extreme of the distributions, because it is important to
the probability distributions used. With respect to the yield obtain a precise approximation for those values, as they
stress, its characteristic value corresponds to the percentile influence greatly the probability of failure. A set of 216 ele-
5. For the Young’s modulus and the dead load, their char- ments is considered, combining six possible values for each
acteristic values are considered equal to the mean values of applied load, related to the mean, l, and the standard devi-
the respective probability distributions. ation, r, of the probabilistic distribution for the load. The
In order to determine the ultimate limit state functions, values adopted are: l  2r, l  r, l + r, l + 2r, l + 3r,
it is necessary to know the values of the axial force, NSd, l + 4.2r. In addition to these elements, the design point
and bending moment, My,Sd, both in the critical section previously considered in level 1 methodology is also used.
of the columns (top of the right column) and in the critical A test set containing 12 possible values for each load, total-
section of the beam (right end). Since the maximum bend- ing 123 = 1728 elements, is also created, to measure the
ing moments in the column and in the beam are identical, precision of the approximation obtained with the trained
only three internal forces are required for computing those networks.
functions. Five networks were trained, each with a different number
of neurons in the hidden layer, s1. Table 4 presents the val-
5.3. Neural networks ues of the quadratic mean error for the networks trained,
obtained by Eq. (7) and the time consumed in the process.
The NN that approximates the structural response has The values of the maximum relative error, computed for
input, hidden and output layers. The number of neurons the test set by Eq. (10), are presented in Table 5 for the five
in each layer is designated by s0, s1 and s2, respectively. networks. It can be observed that the NN approximate the
The number of input neurons is equal to the number of structural response with very good precision.

Table 3
Random variables for steel frame example
Variable Distribution Mean Standard deviation Coefficient of variation Characteristic value
Young’s modulus (GPa) Normal 210 10.5 0.05 210
Dead load (kN/m2) Normal 0.50 0.05 0.10 0.50
Live load (kN/m2) LogNormal 1.06 0.366 0.35 2.0
Wind pressure (kN/m2) LogNormal 0.241 0.084 0.35 0.456
Yield stress (MPa) LogNormal 280 28 0.10 235
 J.B. Cardoso et al. / Advances in Engineering Software 39 (2008) 505–513 511

Table 4 Table 7
Quadratic mean error and training time for NN (steel frame example) Probability of failure pf and reliability index b for steel frame
s1 – number of neurons Error obtained Error obtained Time (s) Limit MCS + NN MCS FORM SORM
in hidden layer with training set with test set function conventional
4 6.28 · 106 2.97 · 106 95 Column pf 1.277 · 105 1.255 · 105 1.229 · 105 1.285 · 105
6 3.85 · 107 3.44 · 107 342 failure b 4.21 4.21 4.22 4.21
8 1.01 · 107 8.27 · 108 890
10 3.20 · 109 2.20 · 109 1912 Beam pf 8.630 · 106 8.607 · 106 8.275 · 106 8.650 · 106
12 1.77 · 109 1.53 · 109 3959 failure b 4.30 4.30 4.31 4.30

Table 5
Maximum relative error for the test set obtained with NN (steel frame
adopted for the random variables are in general greater
example) than what is generally suggested by other authors [14,15].
s1 – number e (%)
If smaller coefficients of variation were used, this would
of neurons in certainly result in a smaller probability of failure.
NSd(COLUMN) NSd(BEAM) My,Sd(COLUMN,BEAM)
hidden layer For each limit state function considered, the corre-
4 1.42 2.14 1.11 sponding probability of failure was compared to that
6 0.771 1.11 0.711 obtained using conventional MCS (without NN), with
8 0.545 0.516 0.503 computation of internal forces by finite elements for each
10 0.105 0.107 0.071 sample of the set of random variables, and using 108 sam-
12 0.067 0.069 0.067
ples. These probabilities were also evaluated using FORM
and SORM methods, built in the program COMREL-TI
5.4. Results and discussion [16]. The results for all alternative techniques are presented
in Table 7 and show a very good agreement between the
Considering, for the probabilistic model presented, NN methodology proposed by the authors and the other proce-
with s1 = 12, several Monte Carlo simulations are per- dures considered.
formed. The corresponding values of the probability of For this example, the application of a conventional MC
failure and computation time are shown in Table 6. Failure approach performing 108 computer analyses lasted 6524 s,
is supposed to take place if at least one of the ultimate limit being much slower than the NN-based approach, as seen
state functions is violated. It should be noted that the MCS by the corresponding times shown in Table 6. However, it
enables to conclude that beam failure always occurs simul- is important to stress that the most substantial amount of
taneously coupled with column failure; hence the probabil- time involved in a NN-based procedure is usually spent in
ity of failure of the frame, pf, is equal to the probability of the training phase, as shown by the NN training and simu-
violating the ultimate limit state function associated with lation times for this example, given respectively in Tables 4
column failure. Denoting column failure by event C with and 6. Hence, a comparison based uniquely on the time con-
probability pf(C) and beam failure by event B with proba- sumed during the simulation phase can be misleading.
bility pf(B), it was detected that pf = pf(C [ B) = pf(C)
because pf(C \ B) = pf(B). The possibility of making this 6. Example 3: six element truss
type of system reliability analysis can be seen has a major
advantage of MCS [5]. The next structure considered is a six-bar simply sup-
The probability of failure of the structure, pf, is found to ported truss with random load and random material yield
be 1.277 · 105, and the corresponding reliability index b is stress, also examined in Murotsu et al. [17] and Burton and
equal to 4.21. It is therefore clear that the level 1 method- Hajela [18]. The initial geometry is shown in Fig. 6. The x
ology prescribed in the Eurocodes produces, in this case, a and y coordinates of node 4 always stay equal and are con-
design safer than the limit b = 3.8 recommended for com- sidered as the first design variable, nx. All the six-bars have
mon structures [1]. It is important to outline that the struc- the same cross-sectional area, and this area is the second
tural shapes were chosen considering a design point located design variable, A. Limits on these variables are specified
very close to the infeasible domain, and so this safety mar- in Table 8. The mass of the truss is optimized considering
gin cannot be justified by a conservative initial design. It two random variables, a load P applied at node 3 and the
should also be pointed out that the coefficients of variation material yield stress, Ty. The statistic parameters adopted
are shown in Table 9. A Young’s modulus E = 206 GPa
Table 6 and a specific weight q = 7800 kg/m3 are considered. Six fail-
Probability of failure of steel frame computed with MCS + NN consid- ure functions (one for each bar) are defined, gi = Ty  |Ti|,
ering s1 = 12
where Ti is the stress in bar i shown in Fig. 6. Structural col-
Sample size Probability of failure Time (s) lapse is supposed to take place when at least one bar fails.
6 5
10 1.233 · 10 4 The optimization is performed assuming that the probability
107 1.368 · 105 42 of failure should not exceed pf = 0.001, corresponding to a
108 1.277 · 105 424
reliability index b = 3.090, as was used in [18].
512 J.B. Cardoso et al. / Advances in Engineering Software 39 (2008) 505–513

P 3 5 4

3m 2 4

1 2
1

3m

Fig. 6. Six-bar truss.

The design space for the problem is shown in Fig. 7. To

obtain this figure, a mesh of 41 · 41 equally spaced values
of the design variables is considered, and for each point the
probability of failure of all six-bars is computed by MCS
and the truss mass is also calculated. Using graphic soft- Fig. 7. Design space for the six-bar truss.
ware and through interpolation it is then possible to have
a discrete approximation for the six probability functions
and for the structural mass over the prescribed domain. for each design variable and one for the random load P
The isolines for the mass are represented by dashed lines (as before, the yield stress is considered only for the limit
and the lines corresponding to the probability of failure state functions). The number of neurons in the output layer
pf = 0.001 for each bar are represented by solid lines and is equal to the number of failure functions plus the mass,
named according to the related failure function gi. It can but, as the stresses in elements 1 and 2 always remain equal
be seen that the design space is not convex, having both and the same happens in elements 4 and 5, only five neu-
a local minimum and a global minimum; hence classic opti- rons are needed for the output layer.
mization methods experience difficulties to handle it, which A training set of 83 = 512 elements and a test set con-
is not the case when genetic algorithms are adopted. How- taining 143 = 2744 elements are created. Four networks
ever, such algorithms require the computation of the fitness were trained, and Table 10 presents the values of the qua-
for a great number of points of the design space, and in this dratic mean error obtained by Eq. (7). The values of the
case the probability of failure for each point must be com- maximum relative error, computed for the test set by Eq.
puted using Monte Carlo method. This can be an excessive (10), are presented in Table 11. As before, it can be
computational task unless neural networks are used to observed that the NN approximate the structural response
approximate the response. with very good precision.

6.1. Neural networks 6.2. Mass optimization

The NN has a number of neurons in each layer desig- As the probability of failure is of order of magnitude
nated by s0, s1 and s2, respectively, for the input, hidden 103, a Monte Carlo simulation with 105 independent sets
and output layers. Three input neurons are needed, one is sufficient. A simple genetic algorithm [19] with 24 genes
for each chromosome (12 for each design variable) and a
Table 8 population of 48 chromosomes was used. Probabilities of
Design variables for the six-bar truss
crossover and mutation were taken, respectively, equal to
Design variable Lower limit Upper limit 0.8 and 0.008. After evolving for 48 generations, results
Node 4 coordinates, nx (m) 1.5 4.0 very close to [18] were obtained, as shown in Table 12.
Cross-sectional area, A (m2) 1.0 · 104 4.0 · 104

Table 10
Table 9 Quadratic mean error and training time for NN (steel frame example)
Random variables for the six-bar truss s1 – number of neurons Error obtained with Error obtained
Variable Distribution Mean Standard Coefficient of in hidden layer training set with test set
deviation variation 8 7.4621 · 106 4.4976 · 106
Load, P (kN) Normal 30 3 0.10 16 4.1077 · 107 2.9092 · 107
Yield stress, Ty Normal 172 8.6 0.05 24 5.3745 · 108 4.6667 · 108
(MPa) 32 1.7546 · 108 3.2032 · 108
 J.B. Cardoso et al. / Advances in Engineering Software 39 (2008) 505–513 513

Table 11 efficiently implemented in such scheme. Consequently, the

Maximum relative error for the test set obtained with NN (six-bar truss corresponding computing time was kept within very rea-
example)
sonable limits.
s1 – number of neurons e (%) From the research presented herein, it can be concluded
in hidden layer
eM e1,2 e3 e4,5 e6 that the application of NN in conjunction with MCS seems
8 1.84 4.71 6.22 5.58 4.58 very promising and appears to offer great potential for
16 0.67 2.19 1.58 1.80 1.35 structural reliability and optimization.
24 0.29 0.64 0.51 0.59 0.81
32 0.23 0.63 0.53 0.42 0.57
Acknowledgements

Table 12
The present work was conducted in the framework of
Reliability-based optimization results for the six-bar truss example the Research Project POCTI/ECM/36055/99, financed by
Design variable/ GA + NN Burton and Difference (%)
the Portuguese Foundation for Science and Technology.
objective function Hajela [18] This support is gratefully acknowledged.
nx (m) 2.359 2.366 0.28
A (m2) 1.563 · 104 1.557 · 104 0.36 References
Mass (kg) 22.512 22.450 0.28
[1] Eurocode 1: basis of design and actions on structures. European
Committee for Standardization, Brussels; 1994.
[2] Eurocode 3: design of steel structures. European Committee for
The number of Monte Carlo simulations performed was
Standardization, Brussels; 1992.
482 = 2304, requiring a total computational time of 329 s [3] Hasofer AM, Lind NC. Exact and invariant second-moment code
in a parallel cluster with six nodes composed of Pentium format. ASCE, J Eng Mech Div 1974;100(EM1):111–21.
3.0 GHz PC’s. [4] Shooman ML. Probabilistic reliability: an engineering
approach. McGraw-Hill; 1968.
[5] Mahadevan S. In: Cruse TA, editor. Monte Carlo simulation:
7. Conclusions
reliability-based mechanical design. Marcel Dekker; 1997. p. 123–46.
[6] McCulloch WS, Pitts W. A logical calculus of ideas immanent in
Whenever it is possible to define and model adequately nervous activity. Bull Math Biophys 1943;5:115–33.
the random variables that influence structural behavior, [7] Berke L, Hajela P. Applications of artificial neural nets in structural
the use of level 3 reliability methods such as MCS allows mechanics. Structural optimization, vol. 4. Springer-Verlag; 1992, p.
90–8.
a more rigorous design approach than the semi-probabilis-
[8] Bento JP. Modeling mechanical behavior without mechanics. In:
tic level 1 methods presented in the Eurocodes. However, Tasso C, de Arantes e Oliveira ER, editors. Development of
the amount of time required by conventional MCS makes knowledge-based systems for engineering. Springer-Verlag; 1998. p.
its application not practical. 37–58.
In recent years, NN have become a successful means for [9] Hurtado JE. Neural networks in stochastic mechanics. Arch Comput
Meth Eng 2001;8(3):303–42.
performing function approximation. An example in which
[10] Papadrakakis M, Papadopoulos V, Lagaros ND. Structural reliabil-
a non-linear function was approximated by NN confirmed ity of elastic–plastic structures using neural networks and Monte
the accuracy of this technique. Carlo simulation. Comp Meth Appl Mech Eng 1996;136:145–63.
The application of NN to structural problems, while [11] Rosenblatt F. The perceptron: a probabilistic model for information
leading to satisfactory precision, enables to speed up signif- storage and organization in the brain. Psychol Rev 1958;65:386–408.
[12] Rumelhart DE, Hinton G, Williams R. Learning internal represen-
icantly the computation of structural response. This feature
tations by error propagation. PDP Research Group. Cambridge:
is particularly relevant in reliability analysis, where a very MIT Press; 1986.
substantial computing effort is normally required in order [13] Cybenko G. Approximation by superposition of sigmoidal functions.
to evaluate accurately the probability of failure. Math Control Signal Syst 1989;2(3):303–14.
In the second example considered (reliability analysis of [14] Thoft-Christensen P, Baker MJ. Structural reliability theory and its
applications. Springer-Verlag; 1982.
a single-storey steel frame), MCS was performed using NN
[15] Neves LAC, Cruz PJS, Henriques AAR. SFEM reliability of steel
for describing structural behavior. The results were similar frames. In: Proceedings of the Eurosteel conference; 2002. p. 1531–40.
to those obtained by conventional MCS and the computing [16] COMREL-TI Users Manual, RCP GmbH, Munich; 1992.
time was reduced. www.strurel.de.
In the third example (reliability-based optimization of a [17] Murotsu Y, Shao S, Kogiso N, Tomioka H. Optimal shape of truss
structure based on reliability. In: Proceedings of the US–Japan joint
six-bar truss), NN were used in conjunction with genetic
seminar on structural optimization; 1996. p. 145–56.
algorithms and MCS. In spite of the problem analysed [18] Burton SA, Hajela P. A variable-complexity approach to second-
being non-convex, this methodology made possible to get order reliability-based optimization. Struct Multidiscip Optimization
a solution for the optimum structural design with reliability 2003;25:237–50.
constrains. A parallel computing scheme was adopted, hav- [19] Goldberg DE. Genetic algorithms in search, optimization and
machine learning. Addison-Wesley; 1989.
ing been found that genetic algorithms can be easily and