An Efficient Damage Recognition Method Using Optimal Neural Wavelet Module

Damage pattern recognition research represents one of the challenging stages in structural health monitoring. This is attributed to the fact that the process of damage detection is significantly affected by uncertainties in the damage features and how they propagate during the damage detection process. The challenge is also attributed to the vagueness in defining damage and the significant overlap between damage states. Several attempts have been performed by the research community to develop tools for damage detection using means of structural dynamics analysis such as modal update and wavelets. Other efforts have also been reported on damage recognition for several damage features using probabilistic and fuzzy logic techniques. This article introduces an integrated method for damage feature extraction and damage recognition using an optimal neural-wavelet module. The neural-wavelet module is used for damage feature extraction by recognizing the change in the energy of dynamic signals in the structure. Optimization methods are used to establish a classifier that can provide efficient damage pattern recognition. Design of the classifier is based on minimizing the error of classification of the neural-wavelet module. The classification error is optimized through changing parameters related to the architecture of the neural network including the number of layers and neurons per layer of the neural-wavelet module. A case study showing the ability of the proposed technique to detect and pattern damage in the ASCE benchmark structure is presented. It is suggested that such integration can provide efficient damage detection and recognition for accurate structural health monitoring.