Hindawi

Mathematical Problems in Engineering

Volume 2020, Article ID 9518369, 12 pages
https://doi.org/10.1155/2020/9518369

Research Article
A Risk Assessment Model for Dam Combining the Probabilistic
and the Nonprobabilistic Methods

Yantao Zhu ,1,2 Xinqiang Niu,3,4 Jimin Wang,5 Chongshi Gu ,1,2 Qiang Sun,6 Bo Li,1,2
and Lixian Huang7
1
State Key Laboratory of Hydrology-Water Resources and Hydraulic Engineering, Hohai University, Nanjing, China
2
College of Water Conservancy and Hydropower Engineering, Hohai University, Nanjing, China
3
Changjiang Institute of Survey, Planning, Design and Research, Wuhan, China
4
National Dam Safety Research Center, Wuhan, China
5
Yalong River Hydropower Development Company, Ltd., Chengdu 610051, China
6
CCCC Third Harbor Engineering Co.Ltd., Shanghai, China
7
Materials Science and Engineering, University of California–Los Angeles, Los Angeles, CA 90095, USA

Correspondence should be addressed to Yantao Zhu; [email protected] and Chongshi Gu; [email protected]

Received 27 January 2020; Revised 12 March 2020; Accepted 18 March 2020; Published 27 April 2020

Academic Editor: Ioannis Kostavelis

Copyright © 2020 Yantao Zhu 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 dam reliability study is essential for dam operation safety, regarding the complexity in dam failure causes. The assessment
of the dam reliability is now mainly probabilistic or nonprobabilistic. The probabilistic method is usually applied to the cases
with sufficient knowledge on dam parameters, while the nonprobabilistic method is suitable for the cases with insufficient
knowledge on dam parameters. Since a dam can contain multiple parameters, information abundancy can vary among those
parameters, and neither the probabilistic method nor the nonprobabilistic method alone is enough for dam reliability as-
sessment. In this paper, the probabilistic method and nonprobabilistic method are modified based on the adjusted first-order
second-moment method and the interval analysis method to suit the dam reliability assessment. Based on characterization on
these two methods and the research of the fusion method, the secondary performance function of the dam is constructed, and
the construction method of the risk assessment model for dam is proposed. Combined with a case study, this paper contributes
to the safe operation of the dam.

1. Introduction simulation method were used to solve the reliability index,

and the instability reliability model was established based on
The risk probability of a dam, which is the calculation of the the reliability index. Kruger et al. [2] calculated the risk
possibility of the risk path failure, is an important part of probability of the RCC gravity dam by using the first-order
dam risk management. Internationally, scholars have second-moment method, improved first-order second-
studied the evaluation method of the risk probability from moment method, and Monte Carlo simulation method and
probabilistic and nonprobabilistic aspects. provided suggestions for maintenance and reinforcement of
the dam based on the calculation results. Leszek [3] believed
that there were various loads in the dam structure that
1.1. Probabilistic Risk Assessment Methods. Malkawi et al. [1] maintained or caused the crash, and the uncertainty of the
established performance functions based on the Swedish load ratio could be estimated. Based on this load ratio, the
circle method, simplified Bishop’s method, simplified Jan- probability of dam instability was calculated. Leszek [4]
bu’s method, and Spencer’s method, respectively. The first- assumed that the antisliding and sliding forces were con-
order second-moment method and the Monte Carlo trolled within the accuracy range of 15%. Moreover, the
2 Mathematical Problems in Engineering

probability of hydraulic structures instability risk was cal- dam and bedrock index and discussed the probabilistic
culated by using the additional error principle and Gaussian reliability index of the gravity dam stability against sliding
distribution. During seismic events of the Kajiwa concrete- failure mode. Zhang [20] proposed the multidimensional
face rock-fill dam, Liang et al. [5] directly established the parallelepiped nonprobabilistic model which can effectively
iterative formula of checking points in generalized stochastic deal with complex uncertainty problems of the coexisting
space, so as to solve the problem of calculating the structural related variables and independent variables. Liu [21] con-
risk probability with related random variables conveniently. sidered the uncertainty caused by the interval of the non-
Ma [6] calculated the reliability of the horizontal displace- linear system in the nonprobabilistic reliability, calculated
ment of the dam crest with the checking point method and the reliability with the grey number, and successfully
took the elastic modulus of concrete as the random variable overcame the influence of the interval operation uncertainty.
in three different sections of the gravity dam. Zheng et al. [7] Su and Hao [22] proposed the dynamic response surface
believed that the correlation of random variables conformed method of Gaussian process for particle swarm optimization
to the Nataf distribution, so they applied the Monte Carlo to calculate the nonprobabilistic reliability analysis of
simulation method of Cholesky decomposition to calculate complex engineering structures. Meanwhile, they analyzed
the probability of collapse of tailings dam slope. Liu et al. [8] and proved that the Gaussian process dynamic response
applied the finite element method (FEM), support vector surface method of particle swarm optimization is more
machine (SVM), and Monte Carlo simulation method to accurate. Aiming to solve the problem of insufficient un-
calculate the accident risk probability, which greatly reduced certain parameters, Liang [23] proposed an interval analysis
the calculation time and workload with good accuracy. method to solve the problem of parameter uncertainty in the
Based on the rigid limit equilibrium safety factor method, structure and established the index of nonprobabilistic re-
Wang et al. [9] proposed the particle swarm optimization liability analysis. To overcome the poor engineering appli-
algorithm to analyze the stability reliability of multislip cability of the traditional reliability analysis method based on
surface of gravity dam foundation and verified that the the convex model, Bi [24] developed a new simulation al-
method is highly practical. Wang et al. [10] proposed a gorithm to match the reliability analysis model and estab-
method to calculate the antislip stability reliability of gravity lished two first-order approximate reliability analysis
dam foundation based on genetic algorithm, which over- methods. Yun et al. [25] combined the improved Kriging
came the limitation that the function needs to be expressed model with the finite element method to establish a non-
as an explicit function and enriched the applicability of the probabilistic reliability model of the beam structure reso-
first-order second-moment method in the current design nance with a superellipsoid convex set. Zheng et al. [26] used
standards and specifications. Their method can be used for the SVM method to reselect and construct the response
the research of antislip stability reliability of gravity dam surface function and used random variables with statistical
foundation under the condition of multiple sliding surfaces. parameters to obtain the structure reliability. Sun et al. [27]
Lai et al. [11] proposed the partial coefficient method of constructed an explicit quadratic response surface function
bounded strength drop, analyzed the stability of deep an- and combined it with the structural nonprobabilistic model
tislip, and verified that their method was reasonable and to analyze the stability of an engineering structure. Li [28]
feasible. Hariri-Ardebil [12] proposed a method to calculate proposed a time-varying reliability index model from the
the probability of concrete dam reliability based on non- interval nonprobabilistic model, considering the influence of
linear finite element analysis using the mixed parameter structural resistance with time.
probabilistic statistics method.

1.3. Motivation. In the probabilistic risk assessment method,

1.2. Nonprobabilistic Risk Assessment Methods. In the case of the nonlinearity of the limit state function of the dam is
insufficient engineering data, Elishakoff [13] used the convex generally high, so the accuracy of analysis results are fre-
model for uncertainty analysis and verified the effectiveness quently affected by random factors. In this case, the prob-
of the method. Ben-Haim [14] considered the uncertainty of ability distribution and the digital characteristics of the
the variables and verified that the variation of the variables random factors can severely influence the assessment ac-
within a certain range had no great influences on the curacy. As a result, the traditional random impact factor
structure reliability. Elishakoff [15] raised the idea that the distribution type of the method has certain subjectivity that
nonprobabilistic reliability index should be a specific interval decreases the assessment precision.
rather than a value. Based on the properties of the convex set Meanwhile, the nonprobabilistic risk assessment method
operation and considering the maximum allowable uncer- is less applicable to the situation where the statistical data of
tainty of the system, Ben-Haim [16] proposed a theory of the the impact factors are deficient and the fluctuation interval is
nonprobabilistic reliability. Pantelides and Ganzali [17, 18] the only known factor. Unfortunately, few application cases
established the ellipsoidal convex model and the uniformly of the nonprobabilistic risk assessment method in the safety
bounded convex model and applied these two models to the assessment of dam exist, and usually the correlation between
engineering design optimization. Then, the validity of the factors is not considered.
two models is verified in comparison with the worst-case A dam contains various parameters, so information
design. Yuan et al. [19] put forward the nonprobabilistic abundancy can be unbalanced. Neither a single probabilistic
reliability method to establish the reliability model of gravity model nor a single nonprobabilistic model is suitable for the
Mathematical Problems in Engineering 3

dam failure risk assessment. As a result, a combination of 2.2. An Adjusted First-Order Second-Moment Method. In
probabilistic model and nonprobabilistic model should be the dam reliability analysis, the first-order second-moment
studied to solve the risk analysis of dam failure under dif- center point method needs the mean and the standard
ferent operation conditions. deviation values of the mathematical model to calculate the
reliability of dam structure. The limit state performance
2. A Probabilistic Risk Assessment function Z � g(x1 , x2 , . . . , xn ) of the dam is expanded by
Model for Dams Taylor series at the point of mean value and linearized the
expanded function to solve the dam reliability.
At present, the calculation methods for dam reliability When calculating the probabilistic reliability index using
evaluation [29–32] are as follows: average first-order second- the basic first-order second-moment method, two kinds of
moment method, second-order second-moment method, problem will exist: (1) when the nonlinear function is ex-
response surface method, direct integration method, Monte panded by Taylor series and the corresponding higher order
Carlo simulation method, and Latin hypercube sample terms are omitted, the error increases along with the distance
method. The next part focuses on the adjusted first-order between some points and the limit state surface. This
second-moment method. The corresponding probability phenomenon results from the average point specific loca-
assessment model for dam reliability evaluation is tion, which is generally in the reliable region, not on the limit
established. surface. (2) When selecting different performance functions,
the values vary with meanings, and the huge error values
provide different reliable indexes. Therefore, the basic first-
2.1. A Characterization Method of the Probabilistic Risk As- order second-moment method is difficult to meet the re-
sessment Model for Dams. The limit state of the dam can be quirements of engineering application.
described by the performance function, which can be In view of the deficiency in the basic first-order second-
expressed as moment method, the following is an improvement of the
g(X) � g x1 , x2 , . . . , xn 􏼁, (1) basic first-order second-moment method. The basic idea is
as follows: the probability function is expanded by Taylor
where X � (x1 , x2 , . . . , xn ) is a set of basic variables and series, and the high-order nonlinear parts are replaced by the
xi (i � 1, 2, . . . , xn ) is the dam structure factor variables. linear expressions, while the lower order terms stay the
The function can be summarized in the following three same. The mean and variance values of the parameter
forms: (1) when g(X) � 0, it means that the dam structure variables can plug in the linear expression. This method can
reaches a limit state which is also easily changed; (2) when help to solve the nonlinear problem of the performance
g(X) < 0, the dam structure is in a state of failure, and the equation effectively and analyze the influence of the mean
structure cannot function correspondingly. The failure state and variance on the dam reliability.
is only theoretical and not realistic because once g(X) < 0, Suppose the limit state equation of the dam is
the structure of the dam shifts until g(X) ≥ 0; (3) when
Z � g(X) � 0. (4)
g(X) > 0, it indicates that the structure is in a safe state, and
the structure can function correspondingly. Generally, the Assume x∗ � (x∗1 , x∗2 , . . . , xn∗ )T is a point on the limit
dam structure is in the safe state. state surface of the dam structure, namely,
Suppose the dam structure has potential risks, which
means the dam may not complete its function corre- g x∗ 􏼁 � 0. (5)
spondingly under the structure safety. Consequently, the
multidimensional integral expression of the corresponding The limit state equation of the dam is expanded in Taylor
risk probability is series at point x∗ , and the first-order part is taken:
n
Pf � 􏽚 f x1 , x2 , . . . , xn 􏼁dx1 dx2 · · · dxn , (2) zg(x∗ )
Ω
ZL � g x∗ 􏼁 + 􏽘 Xi − x∗i 􏼁. (6)
i�1
zX i
where Pf is the risk probability; Ω is the failover domain; and
f(x1 , x2 , . . . , xn ) is the joint probability density function of In the variable factor X space, ZL � 0 is the tangent plane
the vector X. of the limit state plane passing x∗ , and the mean and the
For the dam engineering, it is difficult to calculate the standard deviation values of ZL are as follows:
risk probability index by equation (2) in a probabilistic n
zg(x∗ )
method. Usually, the dam reliability is calculated first, and μZ L � g x ∗ 􏼁 + 􏽘 ∗
􏼐μXi − xi 􏼑, (7)
the probabilistic reliability index is converted into the risk i�1
zX i
probability. Finally, the limit state function of the dam is 􏽶��������������
expressed as the probabilistic reliability index: 􏽴
n 2
zg(x∗ ) 2
μg σ ZL � 􏽘􏼢 􏼣 σ Xi . (8)
β� , (3) i�1
zXi
σg

where μg is the mean of the function g(X) and σ g is the Substituting equations (7) and (8) into equation (3), the
standard deviation of the function g(X). probabilistic reliability index of the dam can be calculated as
4 Mathematical Problems in Engineering

μZL g(x∗ ) + 􏽐ni�1 zg(x∗ )/zXi 􏼁􏼐μXi − xi∗ 􏼑 2.4. Effects of the Mean and Variance of Dam Parameters on
β� � 􏽱������������������� . (9) the Values of the Risk Probability. From equation (9), the
σ ZL 2
􏽐n 􏼂zg(x∗ )/zX 􏼃 σ 2 probabilistic reliability index can be derived from the mean
i�1 i Xi
and variance of parameter variables. Since the environment
For the convenience of calculation, Xi is transformed of a dam system is complex and changeable, and the op-
into the standard normal variable factor Y space, that is, eration life length of a dam is long, the mean and variance
Yi � (Xi − μXi )/σ Xi . Therefore, the probabilistic reliability will inevitably fluctuate in some extent, which will affect the
index is expressed as risk probability value of dam accidents. Therefore, it is
necessary to study the influence of the mean and variance of
μZL 􏽐ni�1 zg(x∗ )/zXi 􏼁σ Xi
β� � −􏽱������������������ �Y i . (10) dam parameters on the values of risk probability.
σ ZL 2
􏽐n 􏼂zg(x∗ )/zX 􏼃 σ 2 In view of the problems mentioned above, since the risk
i�1 i Xi
probability is calculated by the probabilistic reliability index
According to equation (10), the sensitivity coefficients of in the probabilistic assessment model for dam reliability
Xi and Yi can be obtained as evaluation, the influence of the volatility of the mean and
􏽐ni�1 zg(x∗ )/zXi 􏼁σ Xi variance of the dam parameters on the probabilistic reli-
αXi � cos θXi � cos θYi � − �������������������.
􏽱 ability index should be analyzed first.
2
􏽐ni�1 􏼂zg(x∗ )/zXi 􏼃 σ 2Xi Based on equation (9), the partial derivatives of the
reliability index with respect to μXi and σ Xi , respectively, can
(11)
be obtained by
Through equation (11), the expression of the transfor- zβ zg(x∗ )/zXi 􏼁μXi
mation of the design check point to the standard normal Y a� � 􏽱������������������ �,
zμXi 2
􏽐n 􏼂zg(x∗ )/zX 􏼃 σ 2
space is i�1 i Xi

yi∗ � β cos θYi , i � 1, 2, . . . , n, (12)

zβ 􏽨g(x∗ ) + 􏽐ni�1 zg(x∗ )/zXi 􏼁􏼐μXi − xi∗ 􏼑􏽩σ Xi
b� �− 3/2
.
Substitute the results back to X space. Then, zσ Xi 2
􏼐􏽐ni�1 􏼂zg(x∗ )/zXi 􏼃 σ 2Xi 􏼑
xi∗ � μxi + βσ Xi cos θXi , i � 1, 2, . . . , n. (13) (16)

From the theoretical calculation methods mentioned From equations (3) and (4), the fluctuation in the mean
above, combining equations (6), (10), (11), and (13) [33, 34], and variance values has a complex influence on the value of
we can get the probabilistic reliability indexes β and x∗ based probabilistic reliability index. The results of a and b should
on HL-RF iterative solution whose formula is shown in the be calculated from the parameter intervals of the dam
following equation: variables, so the cases vary with the actual project details.
1 T
XK+1 � 􏼌􏼌 􏼌 􏼂∇g Xk 􏼁Xk − g Xk 􏼁􏼃∇g Xk 􏼁 . (14)
􏼌􏼌∇g Xk 􏼁􏼌􏼌􏼌2 2.5. Calculation Steps of Probabilistic Risk Assessment Model
for Dams. Based on the theoretical elaboration mentioned
above, the following steps are summarized to compute the
probabilistic risk assessment model for dams:
2.3. A Calculation Model for Dam Reliability. The probabi-
listic reliability index is solved by an adjusted first-order Step 1. To determine the limit state equation of the dam
second-moment method, and the risk probability is ob- and assume that x∗ � (x∗1 , x∗2 , . . . , xn∗ )T is a point on
tained. Therefore, based on the probabilistic reliability index the limit state surface
model, the relationship between the reliability and the risk Step 2. Obtain the sensitivity coefficient of reliability
probability is established to construct the calculation model function expression
for the dam reliability. Step 3. Expand the limit state equation by Taylor series
The probabilistic reliability index expression of equation at the point x∗ � (x∗1 , x∗2 , . . . , xn∗ )T and calculate its
(10) shows that the higher the probabilistic reliability index mean value and standard deviation to construct the
β, the lower the risk probability of dam failure and the other probabilistic reliable index expression
way around. Therefore, when the variable factors in the
Step 4. Calculate the probabilistic reliability index β
performance function g(X) � g(x1 , x2 , . . . , xn ) conform to
a normal distribution, or when the nonnormal variable Step 5. If the design precision is not satisfied, the ob-
factors have been normalized and the variables are inde- tained x∗ value is substituted into the limit state
pendent of each other, the performance function g(X) also equation and a new x∗ value should be recalculated
conforms to a normal distribution. In this case, the rela- Step 6. Repeat Step 2 to Step 5 until the difference
tionship between the probabilistic reliability index and the between the new and the previous x∗ values is less than
risk probability of the dam can be deduced as follows: the allowable error ε
pf � Φ(−β). (15) Step 7. Calculate the dam reliability from the calculated
value β
Mathematical Problems in Engineering 5

Figure 1 illustrates the calculation flow of the probabi- Zl + Zu

listic assessment model for dam reliability evaluation. η� . (22)
Zu − Zl
3. A Nonprobabilistic Risk Assessment The nonprobabilistic index η of risk probability repre-
Model for Dams sents the expansion space of the interval variable that is the
shortest distance from the origin of coordinates to the failure
When the information of dam impact factors is insufficient, curve or surface. The higher value of η means the stronger
it is difficult to use probabilistic risk assessment model to the safety performance of the dam structure and the lower
evaluate the dam reliability. Therefore, it is necessary to build the risk probability. Even though Z value can be negative in
the nonprobabilistic risk assessment model for dam. In this theory, practically the state of the dam and the performance
section, the interval analysis method is used to analyze the function will change when Z <0, which leads to the
relationship between the nonprobabilistic reliable index and reevaluation of the Z value. Therefore, as can be seen from
the risk probability of dam failure, so as to establish a equation (22), when η > 1, both Zl and Zu are greater than 0
nonprobabilistic risk assessment model for dam. and g(X) > 0, so the dam is in a safe state; when η < 1,
g(X) < 0 and the dam is in a state of failure.
3.1. Model Characterization of a Nonprobabilistic Risk As-
sessment Model for Dams. Ben-Haim and Elishakoff [35] 3.2. An Interval Analysis Method. The interval analysis
proposed the concept of nonprobabilistic reliability based on method [37] is usually used to deal with the problems where
the word “reliability,” and believed that the smaller the range the original data are not accurate, but the values of the data
of system performance fluctuation, the more reliable the fall into a certain range, so the problems can have exact
system will be. However, they did not provide a unified form solutions or interval solutions.
of nonprobabilistic reliability index. Reference [36] estab- When solving the nonprobabilistic reliable index, the
lished the nonprobabilistic model and gave the measure- upper and lower bounds of dam risk factor
ment form of the unified nonprobabilistic reliable index. Xi (i � 1, 2, . . . , n) are set as Xi u and Xi l , respectively, where
Based on the abovementioned literature, this section pres- Xi ∈ XIi � [Xli , Xui ]. Assume that the performance function
ents a nonprobabilistic risk assessment model for the dam g(Xi ) is a continuous function of Xi (i � 1, 2, . . . , n), and the
reliability evaluation. Suppose the dam structure perfor- function value Z is also an interval variable, where
mance function is Z ∈ ZI � [Zl , Zu ]. These assumptions give
Z � g X1 , X2 , . . . , Xn 􏼁 � g(R, S) � R − S, (17) Zl � min g X1 , X2 , . . . , Xn 􏼁Xi ∈ XIi 􏽨Xli , Xui 􏽩, (23)
where Z is the structural performance function; R is the
uncertainty factor of resistance; and S is the load uncertainty Zu � min g X1 , X2 , . . . , Xn 􏼁Xi ∈ XIi 􏽨Xli , Xui 􏽩. (24)
factor.
According to the calculation flow in the interval analysis
Assume the nonprobabilistic measurement of the degree
method, Zl and Zu of the performance function are ob-
of stability and reliability of dam structure is
tained, and the mean Zc and dispersion Zr are obtained by
η � min􏼈‖δ‖∞ 􏼉, (18) substituting equations (23) and (24) to equations (20) and
(21). Finally, the nonprobabilistic index η of dam reliability
where η is the nonprobabilistic reliability index of dam can be obtained from equation (22).
structure; δ is the normalized vector corresponding to the
nonprobabilistic factor X; and ‖δ‖∞ is the infinite norm of δ.
If Z is the interval variable, equation (18) can be re- 3.3. The Target Reliability Index of the Dam. Generally, a
written as larger than one reliability index indicates that the dam
Zc structure is in a reliable state. Given this broad index value
η � r, (19) range, engineering projects prefer a specific value which is
Z
defined as the target reliability index of dam. In probabilistic
where Zc and Zr are the mean and deviation of the per- and nonprobabilistic assessment, this value is regarded as the
formance function, respectively, and can be calculated by the probabilistic target reliability index βT and the non-
following formula: probabilistic target reliability index ηT , respectively.
Zl + Zu In this paper, the probabilistic target reliability index βT
Zc � , (20) is based on SDJ21-78 Design Specification for Concrete
2
Gravity Dams and SDJ20-78 Design Code for Hydraulic
Zu − Zl Concrete Structures. A comprehensive evaluation is per-
Zr � , (21) formed based on both domestic and international technical
2
standards and long-term accumulated engineering practices.
where Zu and Zl are the maximum and minimum of the Since there is not much literature on the value of ηT , this
performance function Z, respectively. paper obtains nonprobabilistic target reliability index
Substitute equations (20) and (21) into equation (19) to through the relationship between the nonprobabilistic and
obtain the probabilistic target reliability index, whose expression is
6 Mathematical Problems in Engineering

Set g (Xi), μXi, σXi

Set μXi as the checking point

of the first experiment design

Calculate sensitivity index by formula (11)

Use μXi, σXi, αXi to calculate probabilistic

reliability β by formula (9)

Recalculate with the

new checking point
Use β to calculate new checking
point X∗ by formula (13)

N
|last β – β | ≤ allowable error

β is the probabilistic reliability

Calculate probability risk by formula (14)

Figure 1: Flow chart of calculation steps of the probabilistic risk assessment model for dam.

βT or η < 1, the dam is in a state of failure and the probability of

ηT � , (25) failure is 100%, namely, the risk probability Pf is 100%.
k
When 1 < β < βT or 1 < η < ηT , the probability of dam ac-
where ηT is the nonprobabilistic target reliability index; βT cident is uncertain. Therefore, in these cases, the probability
is the probabilistic target reliability index; and k is the of dam accident can only be determined by establishing the
multiple of dispersion from standard deviation of the un- relationship between the nonprobabilistic reliability index
certainty factor, which gives Xri � kσ Xi . and the risk probability of dam. The specific calculation
In equation (25), βT is obtained by the specification. expression is as follows:
Therefore, the key to determine ηT is to determine a rea-
⎪
⎧
⎪ 0, η > ηT ,
sonable k value. According to the change reasons of k value, ⎪
⎪
such as errors, uncertainties, and other factors caused by ⎪
⎪
⎪
⎪
data processing, the nonprobabilistic target reliability index ⎨ ηT − η
Pf � ⎪ , 1 < η < ηT , (26)
is selected as shown in Table 1 [38]. ⎪
⎪ ηT − 1
⎪
⎪
⎪
⎪
⎪
⎩
100%, η < 1,
3.4. A Calculation Model of the Dam Reliability. The
abovementioned contents provide a calculation method of where Pf is the risk probability of dam, the risk probability
the nonprobabilistic reliability index of dam reliability and here is different from the reliability expressed by probability
the target reliability index of dam, respectively, but not the and η is the nonprobabilistic reliability index.
estimation of the risk probability of dam. Therefore, this
section builds the nonprobabilistic assessment model for 4. The Combined Model for the Dam
dam reliability evaluation based on the target reliability Risk Assessment
index of dam structure. According to the previous studies,
when η > ηT , the dam is in a safe state and the probability of The probabilistic method can be used to calculate the dam
accident is 0, namely, the risk probability Pf is 0; when β < 1 reliability only if sufficient information of dam impact
Mathematical Problems in Engineering 7

Table 1: Target reliability index βT and ηT of dam.

Structure safety grade I grade II grade III grade
The first kind of failure Probability target 3.7 3.2 2.7
The second kind of failure Reliability index 4.2 3.7 3.2
Damage type
The first kind of failure Nonprobability 1.3 1.2 1.1
The second kind of failure Target reliability index 1.4 1.3 1.2

factors is known; otherwise, the nonprobabilistic method where μZ and σ Z are the mean and standard devi-
can be used for evaluation when there is little or no in- ation of the standard normal distribution perfor-
formation about dam structure impact factors. However, in mance function Z, respectively, and β(Y) is a
dam engineering, the abundancy of information among reliability index containing bounded uncertainty
factors is usually unbalanced; in some cases, some factors factor Y.
contain sufficient information, but these factors are not (3) Consider the uncertainty of bounded uncertainty
suitable to be described by a random model. In these cases, a factor Y and the dam reliability conditions, and
single analysis method is not enough. Therefore, based on establish secondary performance function Z(2) as
the probabilistic and nonprobabilistic assessment models for
dam reliability evaluation, this section studies and proposes
a combined model for dam risk assessment. Z(2) � β(Y) − βT , (29)
Since the probabilistic and nonprobabilistic assessment
models for dam reliability evaluation are based on the
where βT is the target reliability index of the dam, and
performance function, the combined model cannot calculate
this index can be selected according to the safety grade
the performance function simultaneously. Therefore, the
and failure type of the dam.
idea of constructing a combination model by two levels of
“probabilistic model before nonprobabilistic model” is Based on equation (29) and the nonprobabilistic as-
proposed: at first, the bounded uncertainty factor is regarded sessment model for dam reliability evaluation, the non-
as a general variable, and only the uncertainty of the probabilistic reliability index is derived as follows:
probability factor is considered. Then, the calculation model
Z(2)c Z(2)u + Z(2)l
of dam reliability calculates the reliability index function η(2) � � . (30)
with uncertainty factor, providing an unspecified reliability Z(2)r Z(2)u − Z(2)l
index. The secondary performance function is latter estab- It can be seen from equation (30) that, when the extreme
lished based on the probabilistic reliability index and values obtained from the secondary performance function
probabilistic target reliability index. The uncertainty of the are all positive, the reliability index of the combined as-
bounded uncertainty factor is considered in the secondary sessment model is larger than one. When the extreme values
performance function to compute the nonprobabilistic re- are all negative, the reliability index is negative. When the
liability index and the risk probability of dam failure. The extreme values have one positive and one negative, the
steps of the combined model are as follows: reliability index is between zero and one. Therefore, the risk
(1) Assume that the dam performance function Z(1) probability portfolio of dam failure is given as follows:
which contains the probability factor and the ⎪
⎧
⎪ 0, η(2) > 1,
⎪
⎪
bounded uncertainty factor is ⎪
⎪
⎪
⎪
⎪
⎨ (2)
Z(1) � g(X, Y) � g X1 , X2 , . . . , Xn , Y1 , Y2 , . . . , Ym 􏼁, Pf � ⎪ 1 − η , −1 < η(2) < 1, (31)
⎪
⎪ 2
(27) ⎪
⎪
⎪
⎪
⎪
⎪
⎩ 100%, η(2) < − 1.
where X � 􏼈X1 , X2 , . . . , Xn 􏼉 is the probability factor
vector;Y � 􏼈Y1 , Y2 , . . . , Ym 􏼉 is the bounded uncer-
tainty factor vector; andn and m are the number of
probability factors and bounded uncertainty factors, 5. Case Study
respectively.
5.1. Project Profile. A gravity dam belongs to the second-
(2) Consider the bounded uncertainty factor vector Y in class project, and its foundation is horizontal. Its crest el-
equation (27) as a general variable temporarily and evation is 223.0 m and its foundation level elevation is
calculate the probabilistic reliability index β(1) by the 153.0 m. The normal water level of the reservoir is 220.0 m,
probabilistic model: and the design flood level (0.1%) is 221.0 m. The water level
μZ of check flood (0.02%) is 222.0 m, and the dead water level is
β(1) � � β(Y), (28) 185 m. The conceptual model of the gravity dam is shown in
σZ
Figure 2.
8 Mathematical Problems in Engineering

220.0 10.0 223.0 where Zw1 and gw1 (H1 , H2 , H3 , α, c, f′ ) represent the sta-
bility performance function of the dam along the foundation
213.0
surface including the upstream water level H1 , downstream
water level H2 , silting height H3 , uplift pressure coefficient α,
shear-break cohesion c, and friction coefficient f′ .
In this case, the upstream water level is chosen as the
check water level, and the downstream water depth is zero.
1 : 0.8 Therefore, the stability performance function of the project
example along the dam foundation surface can be calculated
from equation (32) as
Z1 � 2140cc − 20010α − 1725􏼁f′ + 58c − 23805. (33)

153.0

58.0 5.3. Calculation of the Assessment Model for Dam Reliability

Evaluation
Figure 2: Typical cross section of nonoverflow dam section of a
gravity dam (unit: m). 5.3.1. Calculation Based on the Probabilistic Risk Assessment
Model for Dams. Based on equations (9) and (11), after five
In the engineering design of the dam, according to iterations, the probabilistic reliability index β � 3.070785 of
SDJ21-78 Design Specification for Concrete Gravity Dams the performance function is obtained. The iteration data are
and SDJ20-78 Design Code for Hydraulic Concrete Struc- shown in Table 3, and the results of reliability index and risk
tures, the concrete bulk density cc obeys normal distribution probability data are shown in Table 4.
N(23.5, 0.47), the reduction coefficient of uplift pressure α It can be seen from the results that the probability of
obeys normal distribution N(0.15, 0.0375), and the shear- instability along the dam foundation is 1.07×10–3. The re-
break friction coefficient f′ obeys normal distribution sults also show that the dam structure has good reliability
N(1.1, 0.18). The shear-break cohesion c is set as a log- and low-risk probability under specified conditions.
normal distribution, whose mean is μc � 1.0 MPa and the In the abovementioned calculation, the mean and var-
standard deviation is σ c � 0.32 MPa. The design data show iance of dam parameter variables are the original design
that the concrete bulk density cc , reduction coefficient of values. Considering the volatility of the mean and variance
uplift pressure α, shear-break friction coefficient f′ , and due to the environmental impact, the following section
shear-break cohesion c ranges from [23.5, 24], [0.3, 0.4], mainly analyzes the relationship between the volatility of
[1.0, 1.3], and [1.0, 1.2] in turn. It is assumed that the four mean and variance and the risk probability of dam failure.
parameters are independent, as shown in Table 2. Here, suppose that the fluctuation of mean and variance are
controlled within 10%.
Under the condition that the fluctuation of the mean and
5.2. Determination of the Performance Function. The per- variance is controlled within 10% offset, the range of the
formance function of the dam is mainly determined by the mean and variance is shown in Table 5 during the interval
failure mode. In the reliability assessment of the gravity dam, analysis of dam reliability.
the failure modes mainly include instability and insufficient According to the detailed information of parameters listed
strength. In this section, the reliability assessment in instability in Table 5, the influence of the mean and variance fluctuation of
is taken as an example to study the dam reliability assessment. dam variable parameters on probabilistic reliability index is
Other failure modes can be discussed in similar ways. analyzed from equations (29) and (30). We can conclude that
The instability along the foundation surface of the gravity the partial derivatives of dam probabilistic reliability index to
dam is one of the main failure modes of the concrete gravity the variance of each variable parameter is always less than zero,
dam. Suppose the area of the sliding surface of the dam is A, the so the variance σ Xi of dam probabilistic reliability index to each
total vertical force on the sliding surface is 􏽐 W and the uplift variable parameter is monotonically decreasing. The partial
pressure acting on the sliding surface is U. According to the derivatives of the probabilistic reliability index to the mean μα
literature [39], under normal conditions, the loads to be of the uplift pressure reduction coefficient are always less than
considered for the instability of gravity dam foundation surface zero, and the partial derivatives to the mean value of other
are upstream water pressure, downstream water pressure, variable parameters is always larger than zero. Therefore, the
silting pressure, and wave pressure. According to the de- probabilistic reliability index has a monotonic decreasing re-
scription of the project, the performance function of the dam lation to the mean value of the uplift pressure reduction co-
stability along the dam foundation surface is as follows: efficient and a monotonic increasing relation to the mean value
Zw1 � gw1 H1 , H2 , H3 , α, c, f′ 􏼁 of other parameters.
From the range of dam parameter variables in Table 5
� cA + f′ 􏼐􏽘 W − U􏼑 − Pwr1 − Pwr2 + Psk + Pwk 􏼁, and the research on the uncertainty of dam parameters in
(32) equations (29) and (30), along with the relations in equations
(10) and (11), the maximum and minimum values of risk
Mathematical Problems in Engineering 9

Table 2: Statistics of dam material parameters.

Parameter variable Mean, μ Standard deviation, σ Distribution type Variable interval
cc 23.5 0.47 Normal distribution [23.5, 24]
α 0.15 0.0375 Normal distribution [0.3, 0.4]
f′ 1.1 0.18 Normal distribution [1.0, 1.3]
c 1.0 0.32 Lognormal distribution [1.0, 1.2]

Table 3: Reliability iteration result statistics.

Iteration times g β (cc , α, f′ , c)
1 24171.75 3.035949 (23.301718, 0.261803, 0.561806, 0.997735)
2 355.4752 3.071286 (23.394813, 0.256261, 0.549460, 0.997648)
3 −3.8288 3.070785 (23.397855, 0.256080, 0.549419, 0.997664)
4 −4.08E − 04 3.070785 (23.397887, 0.256078, 0.549418, 0.997665)
5 −1.36E − 07 3.070785 (23.397887, 0.256078, 0.549418, 0.997665)

Table 4: Results of reliability index and risk probability data.

Reliability index Variable combination Reliable index value Risk rate
β (23.397887, 0.256078, 0.549418, 0.997665) 3.070785 1.07 × 10–3

Table 5: Distribution parameters of random variables.

Variable name Mean Variance Distribution type
cc (21.15, 25.85) (0.423, 0.517) Normal distribution
α (0.225, 0.275) (0.03375, 0.04125) Normal distribution
f′ (0.99, 1.21) (0.162, 0.198) Normal distribution
c(kPa) (0.9, 1.1) (0.288, 0.352) Lognormal distribution

Table 6: Results of reliability index and failure probability.

Reliability Combination of probability factors Reliable index Failure

Variable combination
index Mean Variance value probability
μcc � 25.85μα � 0.225 σ cc � 0.376σ α � 0.03 (25.744593, 0.231274, 0.487096, 3.85 × 10−6
βmax 4.473198
μf′ � 1.21μc � 1.1 σ f′ � 0.144σ c � 0.256 1.097281)
μcc � 21.15μα � 0.275 σ cc � 0.564σ α � 0.045 (21.062255, 0.280223, 0.629367, 3.33 × 10−2
βmin 1.833655
μf′ �0.99 μc � 0.9 σ f′ � 0.216σ c � 0.384 0.898248)

probability can be calculated. The results of the specific

reliability and the corresponding failure probability values
0.18
are shown in Table 6.
0.16
According to Table 6, the fluctuation of the mean and
0.14
variance of the dam variable parameters has a great influence
0.12
on the probabilistic reliability index of the structure and the
0.10
dam reliability. Therefore, taking the above calculation flow
0.08
as an example, we analyze the dam reliability controlled by
0.06
the variable parameter volatility of the mean and variance
0.04
under different values, so as to analyze the fluctuation be-
0.02
tween the mean and variance of the dam variable parameters
0.00
and the risk probability, as shown in Figure 3 for details.
3 6 9 12 15 18
From Figure 3, when the uncertainty of the statistical (%)
characteristic quantity of dam variable parameters is small, the
impact on the dam reliability is small. However, the upper limit Upper limit
of the dam reliability changes dramatically with the increase of Lower limit
uncertainty. Therefore, the volatility of the mean has a great Figure 3: Relationship between the volatility of mean and variance
influence on the results of the dam reliability. and risk probability.
10 Mathematical Problems in Engineering

5.3.2. Calculation Based on Nonprobabilistic Risk Assessment Zu1 � Z1 max � 32986.2, (35)
Model for Dams. From equation (29), solving the non-
probabilistic assessment model for dam reliability evaluation when (cc , α, f′ , c) � (23.5, 0.4, 1.0, 1.0), there is a min-
requires to solve the extreme value of dam performance imum value:
function. The computational process is as follows:
Zl1 � Z1 min � 16814. (36)
zZ1
� 2140f′ , When the upstream water level rises to the check water
zcc
level and the downstream water level is zero, the nonrisk
zZ1 probability reliability index of the dam is as follows:
� −20010f′ ,
zα Zu1 + Zl1
(34) η1 � � 3.08. (37)
zZ1 Zu1 − Zl1
� 2140cc − 20010α − 1725,
zf′ As η1 � 3.08 > ηT � 1.3, the dam is reliable and the risk
probability of failure is very small.
zZ1
� 58.
zc
It can be seen from the analysis mentioned above that, in 5.3.3. Calculation Based on the Combined Model for Dams.
the parameter variable interval, the partial derivatives are According to equation (27), a risk assessment model for dam
always negative only when the performance function is combining the probabilistic and the nonprobabilistic
partial derivate to α, and the other partial derivatives are methods is established. Considering cc , α, f′ , and c as
always positive. Therefore, in the parameter interval, the general variables, the expression of probabilistic reliability
extreme values of the performance function are as follows. index is obtained from equation (28):
When (cc , α, f′ , c) � (24, 0.3, 1.3, 1.2), there is a maxi-
mum value:

2140f′ 23.5 − cc 􏼁 − 20010f′ (0.15 − α) + 1.1 2140cc − 20010α − 1725􏼁 − 23747

β � 􏽱������������������������������������������������������������������������������������� �. (38)
2 2 2 2
1011634f′ 23.5 − cc 􏼁 + 563062.64f′ (0.15 − α)2 + 0.0324 2140cc − 20010α − 1725􏼁 + 344.47(1 − c)2

The dam has a II safety grade, and its stability target (1) In view of the shortcomings of the basic first-order
reliability index is 3.7. The secondary performance function second-moment method, the adjusted method is
is established as studied. Based on the adjusted first-order second-
moment method, the calculation model of dam re-
Z(2) ′
3 � β cc , α, f , c􏼁 − βT . (39)
liability index is established. This paper discusses the
In the parameter interval, the maximum and the min- relationship between the risk probability of dam
imum of the secondary performance function are obtained failure and the probabilistic reliability index, con-
as follows: structs the probabilistic assessment model for dam
reliability evaluation, and analyzes the influence of
Z(2)
max � 209.3074, the mean and variance of the dam physical and
(40)
Z(2) mechanical parameters on the calculation result of
min � 11.9701.
the risk probability of dam failure.
Then, the combined model index of the dam is (2) This paper analyzes the relationship between the
Z(2) + Z(2) bounded uncertainty of dam variable parameters and
η(2) � max min
� 1.1213. (41) the nonprobabilistic reliability index and discusses
Z(2)
max − Z(2)
min the interval analysis method. The calculation model
Because η(2) � 1.1213 > 1, the risk probability of dam of nonprobabilistic reliability index is established.
failure is very small. The relationship between the nonprobabilistic reli-
ability index and the risk probability of dam failure
are also studied. Accordingly, the nonprobabilistic
6. Conclusion assessment model for dam reliability evaluation is
constructed.
This paper studies how to build the probabilistic and
nonprobabilistic assessment model for dam reliability (3) Considering the characteristics of probabilistic
evaluation and also discusses and establishes a combined model and nonprobabilistic model for dam reli-
model of the dam considering the uncertainty of the pa- ability evaluation, the fusion method of the two
rameters. The main research contents are as follows: models is discussed, and the secondary performance
Mathematical Problems in Engineering 11

function of dam failure is established. Based on the [8] H.-q. Liu, F. Xiao, X. Yang et al., “Dam reliability analysis
interval analysis method, the construction method of method based on FEM-SVM,” Water Resources and Power,
the combined model of dam reliability evaluation is no. 10, pp. 43–45, 2015.
studied. [9] G. Wang, Z. Ma, J. Qin, and Q. Lan, “Reliability analysis on
stabtility of gravity dam foundation over multiple sliding
planes using PSO,” Journal of Hydraulic Engineering, vol. 47,
Data Availability no. 2, pp. 219–228, 2016.
[10] G. Wang, J. Qin, and L. Guan, “Reliability analysis on stability
The [Design parameters of the dam] data used to support the of gravity dam foundation over multiple sliding planes based
findings of this study are included within the article. on GA,” Chinese Journal of Rock Mechanics and Engineering,
no. S1, pp. 3153–3161, 2016.
Conflicts of Interest [11] Z. Lai, X. Chang, Y. Cheng et al., “Analysis of anti-sliding
stability in deep foundation of guanyingyan gravity dam based
The authors declare that there are no conflicts of interest on partial coefficient FEM,” Water Power, vol. 43, no. 1,
regarding the publication of this article. pp. 30–34, 2017.
[12] M. A. Hariri-Ardebili, “Analytical failure probability model
for generic gravity dam classes,” Proceedings of the Institution
Acknowledgments of Mechanical Engineers, Part O: Journal of Risk and Reli-
This work was supported by the National Key R&D Program ability, vol. 231, no. 5, pp. 546–557, 2017.
of China (2018YFC0407104), the National Natural Science [13] I. Ellishakoff, “Essay on uncertainties in elastic and visco-
elastic structures from A M Freudenthal’s criticisms to
Foundation of China (Grant nos. 51739003, 51579085,
modern convex modeling,” Computers&Structures, vol. 56,
51779086, 51579086, 51379068, 51579083, and 51609074),
no. 6, pp. 871–895, 1995.
the National Dam Safety Research Center (Grant no. [14] B.-H. Yakov, “A non-probabilistic concept of reliability,”
CX2018Z38), the Fundamental Research Funds for the Structural Safety, vol. 14, no. 4, pp. 227–245, 1994.
Central Public Welfare Research Institute (no. Y119002), [15] I. Elishakoff, “Discussion on: a non-probabilistic concept of
Open Foundation of Changjiang Survey, Planning, Design reliability,” Structural Safety, vol. 17, no. 3, pp. 195–199, 1995.
and Research Co. Ltd (CX2019K01), Project funded by the [16] B.-H. Yakov, “A non-probabilistic measure of reliability of
Priority Academic Program Development of Jiangsu Higher linear systems based on expansion of convex models,”
Education Institutions (YS11001), the Special Project Fund Structural Safety, vol. 17, no. 2, pp. 91–109, 1995.
of National Key Laboratory (20165042112), the Natural [17] C. P. Pantelides and S. Ganzerli, “Design of trusses under
Science Foundation of Jiangsu Province (Grant no. uncertain loads using convex models,” Journal of Structural
BK20160872), the Key R&D Program of Guangxi Engineering, vol. 124, no. 3, pp. 318–329, 1998.
(AB17195074), and the Central University Basic Research [18] S. Ganzerli and C. P. Pantelides, “Load and resistance convex
Project (2017B11114 and 2018B25514). models for optimum design,” Structural Optimization, vol. 17,
no. 2, pp. 259–268, 1999.
[19] M. Yuan, T. Lu, B. Xu, and B. Xu, “Reliability calculation of
References RCC gravity dam based on non-probabilistic method,”
Journal of China Three Gorges University (Natural Sciences),
[1] A. I. H. Malkawi, W. F. Hassan, and F. A. Abdulla, “Un-
no. 6, pp. 29–33, 2013.
certainty and reliability analysis applied to slope stability,”
[20] Z. Qingfei, A New Type of Non-probabilistic Convex Model
Structural Safety, vol. 22, no. 2, pp. 161–187, 2000.
and its Corresponding Structural Uncertainty Analysis Tech-
[2] C. M. Kruger, A. C. Neto, and D. A. V. Kruger, “Structural
nique, Hunan University, Hunan, China, 2013.
reliability analysis and monitoring program applied to a
[21] J. Liu, M. Wang, and Q. Qing, “A new structural non-
roller-compacted concrete dam,” in Proceedings of the 1st
probabilistic reliability analysis method,” Chinese Journal of
International Symposium on Life-Cycle Civil Engineering, Life-
Solid Mechanics, no. 4, pp. 378–383, 2014.
Cycle Civil Engineering, Varenna, Italy, June 2008.
[3] O. Leszek, “The risk of dam stability loss,” in Proceedings of the [22] G. Su and J. Hao, “Gaussian process based dynamic response
International Congress on Large Dam, Beijing,China, October surface method of non-probabilistic reliability analysis for
2000. complicated engineering structure,” Journal of Basic Science
[4] O. Leszek, “The risk of dam instability,” Express Water Re- and Engineering, no. 4, pp. 750–762, 2015.
sources & Hydropower Information, vol. 22, no. 5, pp. 32-33, [23] B. Liang, Study on Stability and Reliability Analysis Method of
2001. Surrounding Rock and Support Structure of Tunnel, Hunan
[5] P. Liang, Z. Wu, J. Chen, Y. Li, and H. Zhang, “Reliability University, Hunan, China, 2015.
analysis of slope stability based on generalized feasible di- [24] Bi Rengui, An Uncertain Convex Set Model and a Non-
rection method,” Rock and Soil Mechanics, vol. 34, no. 6, probabilistic Reliability Analysis Method. Hunan, Hunan
pp. 1753–1759+1785, 2013. University, Hunan, China, 2015.
[6] Z. Ma, “Study on reliability of horizontal displacement at top [25] Y. Yun, J. Chen, and H. Cao, “Non-probabilistic reliability
of gravity dam based on checking point method,” Urban analysis on resonance of thermal-structural coupling of a
Roads Bridges & Flood Control, no. 11, pp. 150–153, 2013. beam based on improved Kriging,” Journal of Harbin Institute
[7] X. Zheng, Q. Li, K. Xu, and L. Geng, “Calculation method on of Technology, no. 10, pp. 131–136, 2016.
probability of tailings dam failure caused by dam slope in- [26] P.-x. Zheng and M. Chen, “A response surface method based
stability considering correlation of variables,” Journal of Safety on SVM for dam reliability analysis,” China Rural Water and
Science and Technology, vol. 2, no. 11, pp. 23–27, 2015. Hydropower, no. 12, pp. 135–137, 2016.
12 Mathematical Problems in Engineering

[27] M. Sun, X. Guo, M. Wang, and J. Wang, “Study on non-

probabilistic reliability of tunnel based on quadratic or-
thogonal test,” China Civil Engineering Journal, no. S2,
pp. 105–111, 2017.
[28] X. Li, Time-dependent Reliability Study of Crane Jib Structure
Based on Interval Non-probabilistic Model, Taiyuan University
of Science and Technology, Taiyuan, China, 2017.
[29] Y. Li, Robust Reliability Theory and Optimization Methods,
Dalian University of Technology, Dalian, China, 2006.
[30] X. Qiao, On Reliability Analysis and Optimization of Uncer-
tain Structures, Xidian University, Xi’an, China, 2009.
[31] B. Xu, R. Pang, and Y. Zhou, “Verification of stochastic
seismic analysis method and seismic performance evaluation
based on multi-indices for high CFRDs,” Engineering Geology,
vol. 264, Article ID 105412, 2020.
[32] R. Pang, B. Xu, D. Zou, and X. Kong, “Stochastic seismic
performance assessment of high CFRDs based on generalized
probability density evolution method,” Computers and Geo-
technics, vol. 97, pp. 233–245, 2018.
[33] W.-h. Mo, “HL-RF Method of gear reliability calculation,”
Journal of Hubei Industrial Polytechnic, no. 4, pp. 102-103,
2013.
[34] Y. Zhang, Y. Xu, Y. Zheng et al., “Multiobjective optimization
design and experimental investigation on the axial flow pump
with orthogonal test approach,” Complexity, vol. 2019, Article
ID 1467565, 14 pages, 2019.
[35] Y. Ben-Haim, Robust Reliability in the Mechancial Sciences,
Springer-Verlag, Berlin, China, 1996.
[36] Y. Ben-Haim, “Robust reliability of structures,” Advances in
Applied Mechanics, vol. 33, pp. 1–41, 1997.
[37] Y. Li, X. Lu, L. Gan, Bo Chen, and H. Zhao, “Study on safety of
cutoff wall of earth-rock cofferdam based on interval analy-
sis,” Water Resources and Power, no. 03, pp. 111–114, 2016.
[38] Y. Peng, Study on Risk Evaluation Model and Early-Warning
Index of Concrete Gravity Dam, Hohai University, Hohai,
China, 2008.
[39] C. Shen, S. Wang, Y. Lin et al., Hydraulic Structure, Water
Resources and Hydropower Press, Beijing, China, 2008.