Structure Data Processing and Damage Identification Based on Wavelet and Artificial Neural Network

Structural health monitoring is a multi-disciplinary integrated technology, mainly including signal processing and structural damage detection. The aim of the data processing is to obtain the useful information from large volumes of raw data containing noises. In order to obtain the useful information concerned, denoising method and feature extraction technique based on Wavelet analysis is studied. An improved wavelet thresholding algorithm to eliminate the noise for vibration signals is proposed. The results of analysis show that the method based on Wavelet is not only feasible to signal de-noising, but also valuable and effective to detect the health status of bridge structure. In order to detect the damage status of the structure, a multi-layer neural network models based on the BP algorithm is designed. The model is trained with the data from an engineering beam to filter different transfer function, train function and the unit number of hidden layer by contrast to determine the best network model for damage detection. At last, the model is used to detect the damage of cable-stayed bridge with an improved method of data pre-processing using the square rate of change in frequency as input date of network. The structural damage identification results show that the BP neural network model is easy to identify the damage by the changing of vibration modal frequency and effective to reflect the injury status of the existing structure.