Performance Evaluation System for Main Reinforced Concrete Girders of Existing Bridges

Abstract

A novel performance evaluation system for main reinforced concrete girders of existing bridges evaluates durability and load-carrying capability by using technical specifications and results of simple visual inspections. Three features of the system are described. The first feature is the application of an expert system. Because the purpose of this study is to develop a novel system to evaluate the performance of a bridge element with limited information provided from simple visual inspections, it is necessary to embed into the system the knowledge of bridge experts. The second feature is a neuro-fuzzy hybrid inference. The system performs a fuzzy inference to simulate the inference mechanism used by experts for evaluating bridges. This process is integrated into a computer as a neural network; therefore, the system improves the knowledge base by using the back-propagation method. The third feature is the visualization of the inference process. The system must clearly explain how the outputs are calculated because the road networks, including bridges, represent important investments that must be carefully monitored. The learning capability of the system is demonstrated on an existing bridge in Japan. The effectiveness is verified by comparing the diagnostic results provided by a bridge expert and those proposed by the system.