Abstract Manager

Bridges, as integral components of road networks, serve a crucial function in facilitating efficient transportation and upholding public safety standards. However, their durability and performance require careful attention through regular inspection and strategic maintenance strategies. Bridge Management, as a systematic decision-making process, becomes essential for the development of Maintenance, Repair, and Rehabilitation programs, especially when dealing with an extensive network of bridges. The core of Bridge Management lies in acquiring detailed information on the condition and deterioration rate of bridges. This information guides decisions on maintenance activities, which can significantly impact the remaining life of bridges. The management framework begins with a thorough inspection and definition of bridge conditions, followed by predicting their future states and determining appropriate repair actions for a multi-year period. Accurate prediction of bridge conditions is crucial for the effectiveness of Bridge Management Systems (BMS). However, existing bridge performance prediction methods have limitations. This abstract introduces a novel approach employing Generalized Estimating Equations (GEE) to effectively handle correlated bridge inspection data. GEE, a statistical methodology for analyzing correlated data, enhances precision in predicting bridge conditions by accounting for dependencies within the inspection data. The research employs a Genetic Algorithm (GA) to select statistically significant attributes for bridge performance prediction based on GEE results. Unlike conventional methods that often consider only the age of bridges, this approach identifies attributes that are truly predictors of bridge performance. The application of GA in attribute selection for bridge performance prediction, guided by GEE structures, refines the process, and improves the accuracy of prediction models. Genetic Algorithms, an artificial intelligence technique inspired by the concept of natural selection are employed to solve large optimization problems that classical mathematical approaches struggle to address. The study emphasizes the importance of considering all relevant features without creating unnecessary complexity in prediction models. By identifying the subset of attributes significant for bridge performance prediction, the research aims to strike a balance between model accuracy and simplicity. This subset is determined among 2^n possibilities, where 'n' denotes the number of available attributes. By addressing the limitations of existing methods, the research aims to enhance the accuracy and efficiency of predicting bridge conditions. A case study, based on National Bridge Inventory (NBI) data from the Federal Highway Administration (FHWA), will demonstrate the practical application and effectiveness of the proposed method, showcasing its potential impact on improving the accuracy and efficiency of predicting real-world bridge conditions.