Diagnosis of embankment dam distresses using Bayesian networks. Part I. Global-level characteristics based on a dam distress database

Abstract

Dam safety has drawn increasing attention from the public. To ensure dam safety, it is essential to diagnose any dam distresses and their causes properly. The main objective of this paper is to develop a robust probability-based tool using Bayesian networks for the diagnosis of embankment dam distresses at the global level based on past dam distress data. A database of 993 distressed in-service embankment dams in China has been compiled, including general information on the dams, distresses, and causes. Based on the database, general characteristics of embankment dam distresses are studied using Bayesian networks, which can tackle not only the multiplicity of dam distresses and causes, but also the complex interrelations among them. Common patterns and causes of distresses are identified. The interrelations among the dam distresses and their causes are quantified using conditional probabilities determined based on the historical frequencies from the dam distress database. A sensitivity analysis is also conducted to identify and rank the most important factors that cause the distresses. With the prior information of common characteristics extracted from the database, Bayesian networks are further used to diagnose a specific distressed dam at the local level by combining global-level performance records and project-specific evidence in a systematic structure, which is presented in a companion paper.