Bayesian networks have shown their interest in the field of reliability to model the degradation of dynamical systems. The objective of this work is to set up a Bayesian network to model the degradation of the structural reliability of a spillway. For this purpose, a structural fault tree from the literature was transformed into a Bayesian network using equivalences with the various logic gates. Then, the distributions of the structural reliability of each node, determined in the literature by the condition index method, were implemented, and the Bayesian network was calculated. The inferences were made on the Bayesian lattice using the Monte Carlo algorithm. The Bayesian network built was deployed to assess the temporal evolution of the structural reliability of a spillway over a 100-year period.

Structural components are the least studied in the literature because of their properties that vary from one design to another, which explains the scarcity of statistical data on them. The objective of this work is to determine the Weibull parameters of the reliability laws of the structural components of a spillway using the Condition Index method. To this end, the partial Top-Down Structural Analysis of a spillway was presented. A structural fault tree served as a starting point for the determination of the Weibull parameters of the corresponding component reliability laws. Then, the state condition tables associated with each element were drawn from the analyses carried out by the experts during the dam inspections. The determination of the Condition Indices was carried out by following a succession of steps developed in detail in this paper. After a variable change and a linear regression on the empirical data, a first approximation of the Weibull parameters of each component was performed. Then, the parameters obtained were optimized using the generalized gradient reduction nonlinear solver of the Excel software, on points of distinct reliability, and starting from the previous approximations. The resulting reliability models are very close to discrete data, and generally conservative. Finally, the Kolmogorov-Smirnov conformity test performed in each case validates the law and the determined parameters with a confidence interval greater than 95%. They can be used as a basis for determining the structural reliability of spillways.