Abstract

Transportation infrastructure, especially highway bridges, is characterized by their growing deterioration rate. Increased traffic loads, severity of environmental conditions, and deferred maintenance decisions are the main reasons for this rapid deterioration. Transportation agencies have developed Bridge Management Systems (BMS's) to select optimal maintenance actions and to predict future funding needs. BMS's require bridge deterioration models to predict the future condition of bridges and to estimate their remaining service life. Most of the state-of-the-art BMS's employ Markov Decision Process (MDP) in modeling bridge deterioration. The main shortcomings of Markovian models are that they assume state independence, neglect improvement actions taken in the past, and overlook the interaction of bridge components. Moreover, these models are difficult to update when new data are obtained or a different rating system is used. In this research, a Case-Based Reasoning (CBR) approach was introduced to provide BMS's with a realistic, accurate, and generic deterioration model that eliminates the previous shortcomings. This approach predicts the future condition of a query bridge case by retrieving bridge cases stored in the BMS database that are similar to the query case in their physical features, environmental and operating conditions, and previous inspection and maintenance records. To implement this approach, a CBR shell called CBRMID was developed to be used in modeling the deterioration of any infrastructure facility. This shell was designed to meet a set of requirements, such as hierarchical decomposition of cases, representation of time-dependent data, data accumulation, case adaptation, data derivation, system modularity, versatility, and extensibility. CBRMID consists of three main modules: (i) Case Based Reasoner, for case retrieval and adaptation; (ii) System Maintainer, for case and knowledge accumulation and modification; and (iii) Database, for case and knowledge representation. CBRMID was implemented using the object-oriented programming language C++, the object-oriented database management system ObjectStore 6.0, and the expert system programming language CLIPS 6.10. Cases obtained from the Ministry of Transportation of Quebec (MTQ) database were employed in developing a "proof of concept" CBR application for modeling the deterioration of concrete bridge decks. Comparison between the solutions provided by the developed application and the actual solutions of real-world cases indicated that the CBR approach has a great potential in modeling bridge deterioration and, consequently, may have a serious impact on our transportation system.