Hindawi

Mathematical Problems in Engineering

Volume 2020, Article ID 3925689, 11 pages
https://doi.org/10.1155/2020/3925689

Research Article
An Adaptive First-Order Reliability Analysis Method for
Nonlinear Problems

Zhiming Wang , Yafei Zhang , and Yalong Song

School of Mechanical and Electronic Engineering, Lanzhou University of Technology, Lanzhou 730050, China

Correspondence should be addressed to Zhiming Wang; [email protected]

Received 2 January 2020; Revised 18 February 2020; Accepted 27 March 2020; Published 14 April 2020

Academic Editor: Carlo Bianca

Copyright © 2020 Zhiming Wang et al. This is an open access article distributed under the Creative Commons Attribution License,
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

The HL-RF algorithm of the first-order reliability method (FORM) is a widely useful tool in structural reliability analysis.
However, the iteration results of HL-RF algorithm may not converge due to periodic cycles for some highly nonlinear reliability
problems. In this paper, an adaptive first-order reliability method (AFORM) is proposed to improve solution efficiency for some
highly nonlinear reliability problems by introducing an adaptive factor. In AFORM, based on the two-parameter approximate
first-order reliability method, the new iteration point and the previous iteration point are used to obtain the corresponding angle,
and the result of convergence is judged by angle condition. According to the convergence degree of the results, two iteration
parameters of the approximate reliability method are adjusted continuously by adaptive factor. Moreover, iteration step size is
adjusted by changing the parameters to improve the efficiency and robustness of FORM. Finally, four numerical examples and one
mechanical reliability analysis example are used to verify the proposed method. Compared with the different algorithms, the
results show that AFORM has better efficiency and robustness for some highly nonlinear reliability problems.

1. Introduction analysis [10, 11]. Among them, the simulation method is a

traditional method. The common simulation methods in-
In engineering practice, with the improvement of product clude direct Monte Carlo simulation (MCS) [12, 13] and
performance requirements and the complexity of service important sampling (IS) [14, 15]. The sampling points of IS
environment, its safety has become a growing concern. In are concentrated on the important area, so IS has a smaller
general, the safety factor method is usually used in engi- number of sampling and a higher calculation efficiency when
neering design where design variables are regarded as de- the accuracy is the same. However, in engineering practice,
terministic variables according to experience. However, since the failure probability of engineering structure is very
because of the influence of machining errors, internal dis- small, large amount of simulations are needed to solve the
persion of materials, and accidental factors, the dimensional reliability problems of a small failure probability when using
parameters, material characteristics, and external loads of the simulation method. Therefore, the simulation method
the structure all have certain uncertainties [1–5]. Therefore, has certain limitations in the reliability engineering analysis.
when there are many uncertain factors in structure design, it Unlike the simulation method, the approximation method is
is difficult to ensure the reliability of the structures when widely used in reliability engineering because of its sim-
using the safety factor method without considering these plicity. The common approximation methods include the
uncertain factors. Unlike the safety factor method, the re- response surface method [16–18] and moment method
liability method is an effective tool to ensure the safety of [19–21], such as the FORM and second-order reliability
structures by considering the impact of structural uncer- method (SORM). The response surface method is mainly
tainties [6–9]. aimed at the reliability evaluation problem in the case of
The reliability methods including the simulation method implicit function, while the moment method is aimed to
and approximation method are mainly used in reliability avoid the iteration procedure and difficulty to obtain the
2 Mathematical Problems in Engineering

design point both for explicit and implicit performance analysis, the failure probability Pf of the structure can be
functions. used to measure the reliability of the structure by
Among moment approximate reliability methods,
FORM is often used in structural reliability analysis [22, 23]. Pf � 􏽚 fX (X)dX, (1)
g(X)≤0
The Hasofer-Lind and Rakwitz-Fiessler (HL–RF) method is
widely used in engineering practice because of its simplicity where g(X) is performance function and fX (X) is proba-
and high efficiency [24, 25]. However, in reliability analysis, bility density function of random variable X.
the HL–RF method may produce periodic and chaotic so- However, in the actual engineering reliability analysis,
lutions for highly nonlinear performance functions. To because the structural performance function contains
improve the convergence of HL–RF in solving nonlinear multiple random variables, the multidimensional integral
problems, Elegbede [26] applied the particle swarm opti- calculation is very complicated, especially for the perfor-
mization method to calculate the failure probability of the mance function of complex structures. On the contrary, the
structure, which can obtain relatively higher accurate results. approximate reliability method is simple and efficient. The
Santosh et al. [27] proposed a modified HL-RF method by approximation for the failure probability can be expressed as
using step size selection criteria. The chaos control method �� ��
Pf ≈ Φ(− β) � Φ􏼐− ��U∗ ��􏼑, (2)
was introduced by Yang [28] in FORM, which can make the
solution results of nonlinear reliability problems converge where V(·) represents the cumulative distribution function
well, but the convergence speed is slow. To control iterative of the standard normal distribution, β is reliability index,
convergence speed, an improved HL-RF method with finite and U∗ is the most probable point (MPP).
step size robust iterative algorithm was proposed by Gong
and Yi [24]. The conjugate gradient optimization technique
was applied by Keshtegar and Miri [29] in HL–RF, and they 2.2. HL–RF Method. The HL–RF method is widely used in
proposed a modified HL–RF in order to overcome the engineering practice, which mainly includes two parts of
problem that the results of HL-RF do not converge in the standard normal space transformation and linear approxi-
reliability analysis of complex structures. Recently, there are mation [34, 35]. Standard normal space transformation is
some new FORM algorithms in the literature. Keshtegar [30] mainly to transform nonstandard normal space into stan-
developed a new FORM which controls instability solutions dard normal space. The conversion principle is
using chaotic conjugate map. Meng et al. [31] proposed a xi − μ′Xi
new directional stability transformation method of chaos μi � , (3)
control for first-order reliability analysis. Besides, Keshtegar σ′Xi
and Chakraborty [32] improved FORM by introducing a
where xi is the ith random variable in the nonstandard
conjugate search direction approach. Moreover, Roudak
normal space (X-space), µi is the ith random variable in the
et al. [33] proposed an approximate first-order reliability
standard normal space (U-space) corresponding to Xi, and
method with two parameters for nonlinear reliability
σ′Xi and μ′Xi are the equivalent standard deviation and mean
problems, which has good robustness and efficiency.
of ith normal random variables, which can be respectively
However, for different nonlinear problems, different pa-
expressed as [36, 37]
rameter values have great influence on the solution results.
Therefore, how to get the proper parameters becomes a φ􏽮Φ− 1 􏽨FXi xi∗ 􏼁􏽩􏽯
problem to be solved. σ′Xi� , (4)
fXi xi∗ 􏼁
Based on the two-parameter approximate first-order
reliability method, an AFORM method for nonlinear
μ′Xi� xi∗ − σ′Xi Φ− 1 􏽨FXi xi∗ 􏼁􏽩, (5)
problems is proposed in this paper. To measure the con-
vergence degree of the result, AFORM judges the conver- where φ(·) represents the probability density function of the
gence degree of iteration by the angle condition. To improve standard normal distribution. xi∗ is the design point, which
the efficiency and robustness of convergence, according to represents the nonstandard random variable. Also, Φ− 1(·) is
the convergence degree of iteration results, the iteration the inverse function of the standard normal distribution.
parameters are constantly updated by introducing adaptive FXi (xi∗ ) and fXi (xi∗ ) are the cumulative distribution
factors. Then, the iteration step size is adaptively changed. function and probability density function of Xi, respectively.
Finally, the method is verified by highly nonlinear numerical The linear approximation of HL–RF mainly uses Taylor
examples and mechanical engineering examples. The results expansion at the most probable point (MPP) of the per-
show that the proposed AFORM is efficient and stable formance function, which can be expressed as
compared with other methods.
g(U) ≈ g(U) � g U∗ 􏼁 + ∇T g U∗ 􏼁 U − U∗ 􏼁, (6)
2. Approximation of Failure Probability and
where g(·) stands for linear approximation expressions of
First-Order Reliability Methods the performance function.∇g(U∗ ) is the gradient vector and
2.1. Approximation of Failure Probability. To ensure the U∗ represents MPP.
safety of structures with various uncertainties, it is usually According to (6), let g(·) � 0, and then the new design
necessary to carry out reliability analysis. In the reliability point is obtained by
Mathematical Problems in Engineering 3

∇T g Uk 􏼁Uk − g Uk 􏼁 2.5. Adaptive First-Order Reliability Method (AFORM).

Uk+1 � �� ��2 ∇g Uk 􏼁. (7) Although the approximate first-order reliability method
��∇g Uk 􏼁�� with two parameters has the characteristics of fast conver-
Equation (7) is an algorithm for searching MPP by gence and high accuracy for nonlinear reliability problems,
HL–RF. When the iteration point Uk+1 converges, the MPP the choice of parameters for different nonlinear problems
can be obtained. greatly affects the results and efficiency. Therefore, how to
choose appropriate parameters is still a problem needed to
be solved. To overcome this problem, this paper updates the
2.3. CC Method. To overcome the problem of non- iteration parameters by using the angle θk+1 between the
convergence when HL–RF is used to solve highly nonlinear iteration points Uk and Uk+1 and by introducing an adaptive
structural performance function, Yang [28] applied the factor δ. Obviously, only the initial values of the parameters
theory of chaotic dynamics in FORM and proposed a FORM ξ S and λS are given, and the iteration value of each parameter
based on chaos control (CC). Although it can converge to can be automatically updated according to the following
the results of nonlinear reliability problems, it has a relatively formula. The algorithm is described as
low efficiency. This method gives a new iteration point which ⎧
⎪
⎪ δλS , δθk+1 > θk ,
can be computed by ⎪
⎪
⎪
⎪
⎪
⎪
Uk+1 � Uk + αC􏼂f Uk 􏼁 − Uk 􏼃, (8) ⎪
S ⎨ S θk
λ �⎪λ , δθk+1 ≤ θk < θk+1 , (14)
⎪
⎪ θk+1
T
∇ g Uk 􏼁Uk − g Uk 􏼁 ⎪
⎪
f Uk 􏼁 � ∇g Uk 􏼁, ⎪
⎪
�� � (9) ⎪
��∇g Uk 􏼁���2 ⎪
⎩
max 􏼐λS , 0.2􏼑, θk+1 ≤ θk ,
where α is the control factor ranging from 0 to 1, C rep-
resents the m × m dimensional involution matrix, usually ⎪
⎧
⎪
⎪ δξ S , δθk+1 > θk ,
⎪
⎪
taken as the identity matrix I, and f(·) stands for the iterative ⎪
⎪
function vector. S ⎨ S θk
ξ �⎪ ξ , δθk+1 ≤ θk < θk+1 , (15)
⎪
⎪ θk+1
⎪
⎪
⎪
⎪
⎩ max 􏼐ξ S , 0.2􏼑, θk+1 ≤ θk .
2.4. Approximate FORM with Two Parameters. To overcome
the instability of HL-RF with highly nonlinear reliability
To avoid nonconvergence of the iteration result as the
problems, Roudak et al. [33] proposed an efficient and
parameters ξ S and λS are too small, let ξ S and λS be not less
robust algorithm by introducing two adjustable parameters
than 0.2. Moreover, to measure the degree of convergence of
based on HL–RF. We call this method proposed by Roudak
the iteration result, θk+1 is used as a condition for updating
as the Roudak method in the following paper. Based on (6),
parameters. From the algorithm, it can be found that θk+1
the parameter λ is introduced in this method, and then
equation (6) is reformulated as reflects the convergence of Uk. When the error between Uk
and Uk+1 approaches 0, the corresponding value of θk+1 also
g Uk+1 􏼁 � g Uk 􏼁 + ∇T g Uk 􏼁 Uk+1 − Uk 􏼁 � λg Uk 􏼁. tends to 0. θk+1 can be expressed as
(10) U ·U
θk+1 � cos− 1 􏼢��� k+1��� ��� k ���􏼣. (16)
The Roudak method also introduces a parameter ξ and �Uk+1 � · �Uk �
defines a auxiliary point by
By continuously adjusting the parameters ξ S and λS in
∇g Uk 􏼁 each iteration, (11) and (12) are rewritten as
Ak+1 � Uk − ξ ��� ��, (11)
�∇g Uk 􏼁�� ∇g Uk 􏼁
ASk+1 � Uk − ξ S ��� ��, (17)
where ξ is the step size in the opposite direction of the �∇g Uk 􏼁��
gradient vector.
To solve the new iteration point, let the direction of Uk AS
VSk+1 � ��� k+1 ��. (18)
cosine vector be �ASk+1 ��
A
Vk+1 � ��� k+1 ���. (12) Equation (13) is reformulated as
�Ak+1 �
∇T g Uk 􏼁Uk − 􏼐1 − λS 􏼑g Uk 􏼁
According to (12), the new iteration point is computed Uk+1 � VSk+1 . (19)
∇T g Uk 􏼁Uk+1
by
∇T g Uk 􏼁Uk − (1 − λ)g Uk 􏼁 In this paper, the initial values of the parameters in the
Uk+1 � Vk+1 . (13) proposed AFORM algorithm are ξ S � 0.6 and λS � 0.8. In
∇T g Uk 􏼁Vk+1
order to improve the efficiency of the algorithm, the initial
4 Mathematical Problems in Engineering

Input ξS = 0.6, λS = 0.8, δ = 0.4, g(X) = 0 and ε = 10–6

Convert X-space to U-space to get g (U) = 0

Calculate g (Uk) and ∇g (Uk) by equation (10)

k = k+1 Compute Ask+1 by equation (17)

ObtainV sk+1 by equation (18)

Update ξS and λS by equations (15) and (14)

Get new iteration point Uk+1 by equation (19)

Calculate θk+1 by equation 16

No

Compute reliability index βk+1 = ||Uk+1||2 according to equation (2)

||βk+1 – βk|| ≤ ε

Yes

U∗ = Uk+1,β = βk+1

Figure 1: Flowchart of AFORM algorithm.

adaptive factor δ � 0.4 is taken. The specific flowchart of the Table 1: Calculation results of different methods in example 1.
algorithm is shown in Figure 1.
HL–RF CC Roudak AFORM
β 1.5355 2.2982 2.2983 2.2983
3. Example Analyses Iteration number 40 144 35 17
3.1. Nonlinear Numerical Example 1. To test the perfor-
mance of the reliability algorithm, a common nonlinear Figure 2, respectively. We can see that it can quickly
function is used to verify it. This example uses a cubic converge to 1.5355 using the HL–RF method to solve the
polynomial function, which can be expressed as [30] reliability index. However, it has a larger error compared
g � x31 + x21 x2 + x32 − 18, (20) with the other methods. The reliability indexes calculated
by the CC method and Roudak method are 2.2982 and
where x1 and x2 follow the normal distribution with means 2.2983, respectively. But the efficiency of these two
μ1 � 10.0, μ2 � 9.9 and standard deviations σ 1 � σ 2 � 5.0. methods is relatively low (the number of iterations is 144
Reliability index calculated by Wang and Grandhi [36] and 35, respectively). Unlike HL–RF, CC, and Roudak, the
and Gong et al. [6] is 2.2983. Meng et al. [31] and reliability index calculated by the AFORM not only has a
Keshtegar and Miri [29] obtain that the result of the small error but also has a high efficiency. The number of
reliability index is 2.2983. In the reliability analysis of iterations is about 8 times and 4 times less than the CC
example 1, the reliability index obtained by Monte Carlo method and Roudak method, respectively. Therefore, for
simulation using 106 samples is 2.5265. The computation nonlinear reliability problems, the proposed AFORM
results and iterative process of reliability index using the method has the characteristics of high efficiency and small
different methods in example 1 are shown in Table 1 and error.
Mathematical Problems in Engineering 5

3.2. Nonlinear Numerical Example 2. This example uses a 3.4. Nonlinear Numerical Example 4. A nonlinear perfor-
nonlinear performance function with a fourth-order poly- mance function with nonnormal random variables is used in
nomial, which is described as [27] this example, which is expressed as [30]
g � x41 + 2x42 − 20, (21) g � x41 + x22 − 50, (23)

where x1 and x2 follow a normal distribution with means where x1 follows the lognormal distribution with mean
μ1 � μ2 � 10.0 and standard deviations σ 1 � σ 2 � 5.0. μ1 � 5 and standard deviation σ 1 � 1, respectively; moreover,
The reliability index calculated in [36] is 2.3633. The x2 follows the Gumbel distribution with mean μ2 � 10 and
calculated result in [6] is 2.3628. Moreover, the reliability standard deviation σ 2 � 10, respectively.
index given in reference [38] is 2.3654. The reliability Table 4 and Figure 5 show the computation results and
indexes obtained in [33, 39] are 2.3655. The reliability iterative process of reliability index with different methods
index obtained after 106 calculations is 2.8404 by Monte for example 4, respectively. As shown in Figure 5, the re-
Carlo simulation. We also give the computation results liability index of the traditional HL–RF method has a pe-
and iterative process of reliability index by different riodic cycle, which produces unstable results as chaotic
methods for example 2 in Table 2 and Figure 3, respec- solutions in example 4. Although the results of CC and
tively. As shown in Figure 3, the results show a second- Roudak methods can reach convergence, their efficiency is
order periodic oscillation by HL–RF. This is because that it relatively low. For the nonlinear reliability problems of
is difficult for HL–RF to solve the reliability model with nonnormal distribution, the results of AFORM can both
high nonlinearity. Unlike HL–RF, all reliability indexes of converge and have high efficiency. The number of iterations
CC, Roudak, and AFORM are converged, and their con- of AFORM is only 17. Therefore, for this example, the
vergence results are 2.3654, 2.3655, and 2.3655, respec- AFORM method is more efficient than other methods.
tively. However, their efficiency is different. CC has the
lowest efficiency with 163 iterations. Also, Roudak is more
efficient than CC. AFORM has the highest efficiency. 3.5. Example 5 about Reliability Analysis of the Two-Degree-of-
Therefore, the AFORM has higher efficiency for nonlinear Freedom Primary-Secondary Dynamic System. This example
reliability problems. uses a highly nonlinear performance function to validate the
proposed method. The performance function represents the
performance of the two-degree-of-freedom primary and
3.3. Nonlinear Numerical Example 3. A highly nonlinear secondary power system based on the force capacity of the
performance function of a pipeline is obtained by response second spring. The dynamic system is shown in Figure 6.
surface fitting, which can be formulated by [6] Moreover, the performance function corresponding to
Figure 6 is expressed as [33]
g � 1.1 − 0.00115x1 x2 + 0.00157x22 + 0.00117x21 + 0.0135x2 x3 0.5
g � Fs − ks P􏽨E􏼐x2s 􏼑􏽩 , (24)
− 0.0705x2 − 0.00534x1 − 0.0149x1 x3 − 0.0611x2 x4
+ 0.0717x1 x4 − 0.226x3 + 0.0333x23 − 0.558x3 x4 where the subscripts p and s represent the primary and
secondary springs, respectively. Fs and ks, which represent
+ 0.998x4 − 1.339x24 ,
the force and stiffness of the secondary spring, are random
(22) variables, P is a deterministic peak factor and equals to 3, and
E(x2s ) is the mean square response of the relative dis-
where x1 follows the extreme-II distribution with means
placement of secondary spring, which can be computed by
μ1 � 10 and standard deviations σ 1 � 5, x2 and x3 are normal
random variables with means μ2 � 10, μ3 � 0.8 and standard πS0 ⎡⎢⎣ ζ sζ a 􏼐ζ p ω3p + ζ s ω3s 􏼑ωp
E􏼐x2s 􏼑 � ⎤⎥⎦,
deviations σ 2 � 5, σ 3 � 0.2, and x4 is a lognormal distri- 4ζ s ω3s ζ p ζ s 􏼐4ζ 2a + η2 􏼑 + υζ 2a 4ζ a ω4a
bution random variable with means μ4 � 0.0625 and
standard deviations σ 4 � 0.0625. According to [33], the (25)
reliability index is 1.3961 computed by MCS using 106
where S0 is the intensity of a white noise base excitation of
samples. The calculating results and iterative process for
the system and ζ p and ζ s represent the damping ratios of the
different methods are shown in Table 3 and Figure 4,
primary and secondary springs, respectively. ωp � (kp/mp)0.5
respectively. Although reliability indexes computed by
and ωs � (ks/ms)0.5 represent the natural frequency of the
CC, Roudak, and AFORM are almost convergent to the
primary and secondary oscillators, respectively. kp is the
same value, the convergence rate is not the same. From
stiffness of the secondary spring, and ms and mp are the mass.
Table 3, it can be seen that the efficiency of the CC and
ζ a � (ζ p + ζ s)/2 and ωa � (ωp + ωs)/2 represent the average
Roudak method is low, but the AFORM method has the
frequency and damping coefficient, respectively.
highest efficiency which is about 2 times faster than CC
η � (ωp − ωs)/ωa and v � ms/mp are the mass ratio and the
and Roudak. Therefore, compared with other methods, the
tuning parameter, respectively, where, S0, kp, mp, ms, ζ p, and
proposed AFORM method has better efficiency and ro-
ζ s are random variables. Also, distribution characteristics of
bustness for example 3 with highly nonlinear performance
each random variable are shown in Table 5.
function.
6 Mathematical Problems in Engineering

2.5

Reliability index
1.5

0.5

0
0 50 100 150
Iteration
Roudak HL–RF
Proposed method CC
Figure 2: Iterative process of reliability index of example 1.

Table 2: Calculation results of different methods in example 2.

HL–RF CC Roudak AFORM
β — 2.3654 2.3655 2.3655
Iteration number — 163 40 18

2.5

2
Reliability index

1.5

1
0.5 0.95
0.9
0 70 80 90 100
0 50 100 150
Iteration
CC Proposed method
Roudak HL–RF
Figure 3: Iterative process of reliability index of example 2.

Table 3: Calculation results of different methods in example 3.

HL–RF CC Roudak AFORM
β — 1.3303 1.3303 1.3304
Iteration number — 110 109 44

For this dynamic system, we give the computation re- show a periodic oscillation. It shows that the HL–RF method
sults and iterative process of reliability index using the has certain limitations in reliability evaluation of the
different methods in Table 6 and Figure 7, respectively. As complex highly nonlinear performance function. Although
shown in Figure 7, the results obtained by HL–RF method CC, Roudak, and AFORM all obtained convergence results,
Mathematical Problems in Engineering 7

1.4

1.3

1.2
Reliability index
1.1

0.9

0.8

0.7

0 20 40 60 80 100 120
Iteration
HL–RF Roudak
Proposed method CC
Figure 4: Iterative process of reliability index of example 3.

Table 4: Calculation results of different methods in example 4.

HL–RF CC Roudak AFORM
β — 3.2593 3.2593 3.2593
Iteration number — 138 33 17

3.5

2.5
Reliability index

1.5

0.5

0
0 20 40 60 80 100 120 140
Iteration
HL–RF CC
Proposed method Roudak
Figure 5: Iterative process of reliability index of example 4.

the CC and Roudak methods converge slowly. This shows In order to select the proper initial values of the pa-
that CC and Roudak methods are inefficient for nonlinear rameters in the AFORM method, the effects of different
reliability problems. According to Table 6, the reliability initial values of λS, ξ S, and δ on the calculation results of
indexes obtained by CC, Roudak, and AFORM methods all different examples are analyzed.
converge to 2.1231. However, the number of iterations of the The influence of the initial value λS on the iteration
CC and Roudak methods is 118 and 116, respectively, which results of each example is shown in Figure 8 when the initial
is about twice of the AFORM method. Therefore, the values ξ S � 0.6 and δ � 0.4. Figure 8 shows that when the
AFORM method is significantly better than the other initial values ξ S � 0.6 and δ � 0.4, the changes of the initial
methods. value λS have little effect on the iteration results of example 2,
8 Mathematical Problems in Engineering

xp
xs
cp = 2ξpωpmp cs = 2ξsωsms

mp ms

kp = ω2pmp ks = ω2sms

Figure 6: Diagram of the primary and secondary dynamic system.

Table 5: Distribution characteristics of each random variable.

Random variable Distribution Mean Standard deviation
kp Lognormal 1 0.2
ks Lognormal 0.01 0.002
mp Lognormal 1 0.1
ms Lognormal 0.01 0.001
ζp Lognormal 0.05 0.02
ζs Lognormal 0.02 0.01
Fs Lognormal 15 1.5
S0 Lognormal 100 10

Table 6: Calculation results of different methods for the dynamic system.

HL-RF CC Roudak AFORM
β — 2.1231 2.1231 2.1231
Iteration number — 118 116 43

5
Reliability index

0
0 20 40 60 80 100 120
Iteration
HL–RF CC
Roudak Proposed method
Figure 7: Iterative process of reliability index of example 5.

but have an obvious impact on examples 1, 3, 4, and 5. When λS � 0.8 and δ � 0.4. When λS � 0.8, δ � 0.4, and ξ S � 0.7, the
the initial value λS � 0.8, the iteration result of each example solution does not converge for example 5, so the number of
is relatively low. Therefore, the value of λS is recommended iterations is defined as 0. Figure 9 suggests that when the
to be 0.8. initial values λS � 0.8 and δ � 0.4, the changes of initial values
The effect of the initial value ξ S on the iteration results of ξ S have little effect on the iteration results of examples 1 and
each example is given in Figure 9 when the initial values 2, but have an apparent effect on examples 3, 4, and 5. Also,
Mathematical Problems in Engineering 9

60 50

50
40

Iteration number
Iteration number

40
30
30
20
20

10
10

0 0
Example 1 Example 2 Example 3 Example 4 Example 5 Example 1 Example 2 Example 3 Example 4 Example 5

λS = 0.9 λS = 0.6 δ = 0.6 δ = 0.3

λS = 0.8 λS = 0.5 δ = 0.5 δ = 0.2
λS = 0.7 δ = 0.4

Figure 8: Influence of different initial values of λS on the iteration Figure 10: Influence of different initial values of δ on the iteration
results of each example. results of each example.

80
4. Conclusions
Traditional HL-RF algorithm in FORM is convergent or has
inefficient solution for some highly nonlinear reliability
60
evaluation problems. In this paper, an adaptive first-order
Iteration number

reliability method is developed to improve the efficiency and

40 stability of FORM for highly nonlinear reliability evaluation
problems. In AFORM, iteration parameters are continu-
ously updated by introducing adaptive factors and angular
20 conditions. Besides, to select the proper initial values of the
parameters in the AFORM method, the effects of different
initial values of parameters on the calculation results of
0 different examples are analyzed. The rationality of the
Example 1 Example 2 Example 3 Example 4 Example 5
method is verified by four numerical examples and one
ζS = 0.7 ζS = 0.4 engineering example. Compared with the other methods
ζS = 0.6 ζS = 0.3 (HL–RF, CC, and Roudak methods), the results show that
the proposed AFORM method has better efficiency and
ζS = 0.5
robustness for some highly nonlinear reliability evaluation
Figure 9: Influence of different initial values of ξ S on the iteration problems.
results of each example.

Data Availability
when the initial value ξ S � 0.6, the number of iterations for
solving each example is low, so it can be used as the initial The data used to support the findings of this study are in-
parameter value for the AFORM method. cluded within the article.
In addition, Figure 10 shows the effect of the initial
value δ on the iteration results of each example when the
initial values λS � 0.8 and ξ S � 0.6. Among them, when
Conflicts of Interest
δ � 0.6 and δ � 0.5, the calculation results of the example 5 The authors declare no conflicts of interest regarding the
do not converge, so the number of iterations of example 5 is publication of this paper.
defined as 0. Figure 10 shows that there is a big difference in
iteration number for each example when using the AFORM
method with the different δ. When δ � 0.4, the number of Acknowledgments
iterations of the AFORM method to solve each example is
This work was financially supported in part by the National
low, so δ � 0.4 can be used as the parameter value of the
Natural Science Foundation of China (51565032) and China
AFORM method.
Scholarship Council (201708625073).
10 Mathematical Problems in Engineering

References structural reliability analysis,” Probabilistic Engineering Me-

chanics, vol. 25, no. 4, pp. 365–371, 2010.
[1] J. O. Lee, Y. S. Yang, and W. S. Ruy, “A comparative study on [17] D. L. Allaix and V. I. Carbone, “An improvement of the
reliability-index and target-performance-based probabilistic response surface method,” Structural Safety, vol. 33, no. 2,
structural design optimization,” Computers & Structures, pp. 165–172, 2011.
vol. 80, no. 3-4, pp. 257–269, 2002. [18] S. Goswami, S. Ghosh, and S. Chakraborty, “Reliability
[2] R. Chowdhury and S. Adhikari, “Reliability analysis of un- analysis of structures by iterative improved response surface
certain dynamical systems using correlated function expan- method,” Structural Safety, vol. 60, pp. 56–66, 2016.
sion,” International Journal of Mechanical Sciences, vol. 53, [19] A. D. Kiureghian and M. D. Stefano, “Efficient algorithm for
no. 4, pp. 281–285, 2011. second-order reliability analysis,” Journal of Engineering
[3] M. Kamiński and P. Świta, “Structural stability and reliability Mechanics, vol. 117, no. 12, pp. 2904–2923, 1991.
of the underground steel tanks with the stochastic finite el- [20] Y.-G. Zhao and T. Ono, “Moment methods for structural
ement method,” Archives of Civil & Mechanical Engineering, reliability,” Structural Safety, vol. 23, no. 1, pp. 47–75, 2001.
vol. 15, no. 2, pp. 593–602, 2015. [21] Z. Zhang, C. Jiang, G. G. Wang, and X. Han, “First and second
[4] Z. Meng, G. Li, B. P. Wang, and P. Hao, “A hybrid chaos order approximate reliability analysis methods using evidence
control approach of the performance measure functions for theory,” Reliability Engineering & System Safety, vol. 137,
reliability-based design optimization,” Computers & Struc- no. 5, pp. 40–49, 2015.
tures, vol. 146, pp. 32–43, 2015. [22] X. Du and Z. Hu, “First order reliability method with trun-
[5] Z. Meng, Z. Zhang, G. Li, and D. Zhang, “An active weight cated random variables,” Journal of Mechanical Design,
learning method for efficient reliability assessment with small vol. 134, no. 9, Article ID 091005, 2012.
failure probability,” Structural and Multidisciplinary Opti- [23] U. Alibrandi and C. Koh, “First-order reliability method for
mization, vol. 61, no. 3, pp. 1157–1170, 2020. structural reliability analysis in the presence of random and
[6] J. X. Gong, P. Yi, and N. Zhao, “Non-gradient–based algo- interval variables,” ASCE-ASME Journal of Risk and Uncer-
rithm for structural reliability analysis,” Journal of Engineering tainty in Engineering Systems, Part B: Mechanical Engineering,
Mechanics, vol. 140, no. 6, Article ID 04014029, 2014. vol. 1, no. 4, Article ID 041006, 2015.
[7] M. A. Shayanfar, M. A. Barkhordari, and M. A. Roudak, [24] J.-X. Gong and P. Yi, “A robust iterative algorithm for
“Locating design point in structural reliability analysis by structural reliability analysis,” Structural and Multidisciplin-
introduction of a control parameter and moving limited re- ary Optimization, vol. 43, no. 4, pp. 519–527, 2011.
gions,” International Journal of Mechanical Sciences, vol. 126, [25] B. Keshtegar and M. Miri, “Reliability analysis of corroded
pp. 196–202, 2017. pipes using conjugate HL-RF algorithm based on average
[8] Z. Meng, D. Zhang, G. Li, and B. Yu, “An importance learning shear stress yield criterion,” Engineering Failure Analysis,
method for non-probabilistic reliability analysis and opti- vol. 46, pp. 104–117, 2014.
mization,” Structural and Multidisciplinary Optimization, [26] C. Elegbede, “Structural reliability assessment based on
vol. 59, no. 4, pp. 1255–1271, 2019. particles swarm optimization,” Structural Safety, vol. 27, no. 2,
[9] Z. Meng, Z. Zhang, and H. Zhou, “A novel experimental data- pp. 171–186, 2005.
driven exponential convex model for reliability assessment [27] T. V. Santosh, R. K. Saraf, A. K. Ghosh, and H. S. Kushwaha,
with uncertain-but-bounded parameters,” Applied Mathe- “Optimum step length selection rule in modified HL-RF
matical Modelling, vol. 77, pp. 773–787, 2020. method for structural reliability,” International Journal of
[10] M. Rashki, M. Miri, and M. Azhdary Moghaddam, “A new Pressure Vessels and Piping, vol. 83, no. 10, pp. 742–748, 2006.
efficient simulation method to approximate the probability of [28] D. Yang, “Chaos control for numerical instability of first order
failure and most probable point,” Structural Safety, vol. 39, reliability method,” Communications in Nonlinear Science and
pp. 22–29, 2012. Numerical Simulation, vol. 15, no. 10, pp. 3131–3141, 2010.
[11] L. Faravelli, “Response-surface approach for reliability anal- [29] B. Keshtegar and M. Miri, “Introducing Conjugate gradient
ysis,” Journal of Engineering Mechanics, vol. 115, no. 12, optimization for modified HL-RF method,” Engineering
pp. 2763–2781, 1989. Computations, vol. 31, no. 4, pp. 775–790, 2014.
[12] A. Naess, B. J. Leira, and O. Batsevych, “System reliability [30] B. Keshtegar, “Stability iterative method for structural reli-
analysis by enhanced Monte Carlo simulation,” Structural ability analysis using a chaotic conjugate map,” Nonlinear
Safety, vol. 31, no. 5, pp. 349–355, 2009. Dynamics, vol. 84, no. 4, pp. 2161–2174, 2016.
[13] E. Jahani, R. L. Muhanna, M. A. Shayanfar, and [31] Z. Meng, G. Li, D. Yang, and L. Zhan, “A new directional
M. A. Barkhordari, “Reliability assessment with fuzzy random stability transformation method of chaos control for first
variables using interval Monte Carlo simulation,” Computer- order reliability analysis,” Structural and Multidisciplinary
Aided Civil and Infrastructure Engineering, vol. 29, no. 3, Optimization, vol. 55, no. 2, pp. 601–612, 2017.
pp. 208–220, 2014. [32] B. Keshtegar and S. Chakraborty, “A hybrid self-adaptive
[14] R. E. Melchers, “Radial importance sampling for structural conjugate first order reliability method for robust structural
reliability,” Journal of Engineering Mechanics, vol. 116, no. 1, reliability analysis,” Applied Mathematical Modelling, vol. 53,
pp. 189–203, 1990. pp. 319–332, 2018.
[15] Z. Liu, J. Zhan, and C. Tan, “Improved reliability approximate [33] M. A. Roudak, M. A. Shayanfar, M. A. Barkhordari, and
method combining Kriging and importance sampling,” in M. Karamloo, “A robust approximation method for nonlinear
Proceedings of the IEEE 2012 Prognostics and System Health cases of structural reliability analysis,” International Journal of
Management Conference (PHM-2012 Beijing), pp. 1–4, Bei- Mechanical Sciences, vol. 133, pp. 11–20, 2017.
jing, China, May 2012. [34] A. M. Hasofer and N. C. Lind, “Exact and invariant second-
[16] S.-C. Kang, H.-M. Koh, and J. F. Choo, “An efficient response moment code format,” Journal of the Engineering Mechanics
surface method using moving least squares approximation for Division, vol. 100, no. 1, pp. 111–121, 1974.
Mathematical Problems in Engineering 11

[35] R. Rackwitz and B. Flessler, “Structural reliability under

combined random load sequences,” Computers & Structures,
vol. 9, no. 5, pp. 489–494, 1978.
[36] L. Wang and R. V. Grandhi, “Safety index calculation using
intervening variables for structural reliability analysis,”
Computers & Structures, vol. 59, no. 6, pp. 1139–1148, 1996.
[37] Z. Meng and B. Keshtegar, “Adaptive conjugate single-loop
method for efficient reliability-based design and topology
optimization,” Computer Methods in Applied Mechanics and
Engineering, vol. 344, pp. 95–119, 2019.
[38] B. Keshtegar and M. Miri, “An enhanced HL-RF method for
the computation of structural failure probability based on
relaxed approach,” Civil Engineering Infrastructures Journal,
vol. 46, no. 1, pp. 69–80, 2013.
[39] G. A. Periçaro, S. R. Santos, A. A. Ribeiro, and L. C. Matioli,
“HLRF-BFGS optimization algorithm for structural reliabil-
ity,” Applied Mathematical Modelling, vol. 39, no. 7,
pp. 2025–2035, 2015.