Bridge health monitoring methods using neural networks techniques

Abstract

Bridge health monitoring has become an important research topic in conjunction with damage assessment and safety evaluation of bridges. The use of system identification approaches for damage detection has been expanded in recent years owing to the advancements in signal analysis and information processing techniques. Soft computing techniques such as neural networks and genetic algorithm have been utilized increasingly for this end due to their excellent pattern recognition capability. In this study, the neural networks-based damage detection methods were proposed for effective health monitoring of bridge structures, since the neural networks technique is adequate to on-line health monitoring due to quick computation time and ability to deal with various types of input data. The aim of this study is to develop practical techniques of damage identification that is suitable for structural health monitoring of bridge structures, and can be summarized as (I) development of a damage detection method using neural networks technique based on the changes of the modal parameters due to the damages which can effectively consider the modeling errors in the baseline FE model, (2) enhancement of the efficiency in the damage estimation method using two-step approach which can alleviate the issues associated with many unknown parameters faced in the real structures, and (3) application of the presented methods to a real bridge which can verify the effectiveness and the applicability of the proposed methods. First, a neural networks-based damage detection technique which employs the input parameters less sensitive to the modeling errors was presented for element level damage assessments of bridge structures. It was proposed to use the mode shape differences or the mode shape ratios between before and after damage as the input to the NN to reduce the effect of the modeling errors in the baseline FE model, from which the training patterns are to be generated. From two numerical ...