1. Introduction
Due to the multiple effects of external environmental factors such as calcium leaching, sulfate attack, and freeze–thaw cycles, roller-compacted concrete (RCC) dams inevitably experience aging problems in the long-term operation process [
1,
2,
3,
4,
5,
6,
7,
8]. The deterioration of the dam from normal to pathological is a gradual (i.e., performance decreases over time) and abrupt (i.e., sudden change in the serviceability of the dam) process. Under the action of internal and external loads, some potential factors may cause the operational state of the dam to change abruptly, thus threatening the safety of the dam project [
9]. Therefore, it is necessary to adopt corresponding methods to predict the operational state of the dam, such as establishing a finite element model that conforms to the engineering reality to simulate the performance evolution of the dam during its whole-life-cycle service process, and then conduct a safety evaluation of the dam [
10,
11,
12,
13]. In addition, the service state of the dam can also be evaluated by using the actual engineering monitoring data, but this requires the help of predetermined safety monitoring indices (SMIs). Using SMIs to analyze the working state of dams is a direct way, and the key lies in the determination of dam SMIs. During the long-term operation of dams, their bearing capacity will evolve due to changes in material properties, making the formulation of monitoring indices complex. Additionally, the seepage problem of RCC dams is more complex because of the involvement compaction in its construction process in a series of layers [
14]. Therefore, the determination of seepage and deformation SMIs for RCC dams is one of the difficulties in current dam safety monitoring research.
There are many methods for determining the SMIs of dams. For concrete dams, the typical methods for determining the indices of the service state monitoring effects are as follows: the confidence interval method, the limit state method, and the small probability method [
15]. The confidence interval method is a method that uses statistical theory or the finite element calculation method to establish a mathematical model between the monitoring effect and the load based on the existing dam monitoring data, and the model is used to calculate the SMIs of the monitoring quantity under various loads [
16,
17]. The limit state method, according to the different methods of calculating the total effect
S and the resistance
R of the critical load combination, is divided into the safety factor method, the first-order moment extreme value state method, and the second-order moment extreme value state method [
18]. The SMIs calculated by this method are mainly the extreme values of the effect quantity. The typical small probability method combines the monitoring effects produced by the load combinations that are unfavorable for strength and stability and determines the SMIs of dams based on existing observation data. It is a more mature method for determining dam SMIs. Currently, the small probability method has been widely used in determining SMIs such as dam deformation, seepage, tensile strength, and temperature [
17,
19]. This method qualitatively links the effects produced by the load combinations that are unfavorable for strength and stability and estimates the monitoring indices based on previous measured data [
17]. Therefore, it is more reasonable and accurate than empirical values. However, the small probability method only considers dam deformation occurring under the previous unfavorable loads to estimate the extreme value, which lacks the ability to identify abnormal dam behavior under conventional loads [
17,
20].
With the continuous development of methods for determining SMIs for dams, some emerging and efficient methods have been proposed in recent years [
21,
22,
23,
24,
25,
26,
27,
28,
29,
30] which have greatly promoted the development of dam safety monitoring theory. Currently, in the process of determining the indices for seepage and deformation safety monitoring of concrete dams, the methods can be generally divided into two categories: statistical analysis and model analysis. Statistical analysis methods are usually simple and convenient to operate and are therefore widely used in engineering. However, statistical analysis methods only start from the actual monitoring data, without considering the causes and mechanisms of dam state changes, nor do they relate to the grade and type of dams. As a result, the physical probability is not clear enough [
17]. The model analysis method is based on a mathematical model established via the finite element method, which uses numerical calculation to simulate the seepage and deformation of dams and combines the measured data to invert the seepage and deformation parameters of dam body and foundation. The inverted parameters are then substituted into the numerical model to determine the SMIs of dams. The model analysis method has clear physical concepts and can realize the simulation of dam seepage and deformation under various adverse conditions, as well as the prediction of the dam state throughout the whole life cycle. By doing so, it effectively solves the problem of short monitoring time and short data series of dam measurement. Therefore, the model analysis method is an effective method to determine the comprehensive SMIs of dams.
Despite the extensive work conducted by scholars on the determination of SMIs for dam seepage and deformation, there are still some unresolved issues. These are as follows: (1) The existing SMIs for dams are mostly aimed at dam deformation, and they mostly use mathematical statistics methods. However, these methods cannot consider the causes of dam state changes, and they do not relate to different dam types and dam grades. (2) During the operation of dams, the seepage and stress of the dam body actually affect each other, but the existing SMIs for seepage and deformation are mostly considered from a single perspective; they do not consider the coupling effect of the two fields. (3) Due to the speciality of the construction technology of RCC dams, the rolling layer surface produced by them has different seepage characteristics from conventional concrete dams. Therefore, it is necessary to consider the influence of the rolling layer surface when determining SMIs for RCC dams. For this purpose, this paper considers the influence of the anisotropy of the seepage characteristics of the dam body and the roller layer surface on the seepage and stress fields of the dam. Based on this, a three-dimensional finite element model of an RCC gravity dam is established, and a seepage–stress fully coupled analysis method is constructed and applied in COMSOL Multiphysics commercial finite element software (version: COMSOL Multiphysics 5.2a; COMSOL Co., Ltd.: Stockholm, Sweden) to analyze the seepage and stress fields of the dam. Furthermore, the seepage and deformation SMIs under coupled and uncoupled conditions were proposed and compared with traditional methods.