The Application Of Wavelet Analysis And Neural Networks In The Monitoring Of Structural Damage

It elaborates the contents and requirements of the structural health monitoring, and the present research and application of damage identification for structure with vibration diagnosis technique. One new method for structure health monitoring based on wavelet neural network which is fallen into close-type is proposed.This paper decomposes the dynamics response of a structure using wavelet analysis method, and the damage character is selected by comparing the damage susceptibility of every response signal. The quantity of energy over frequency band of the displacement and acceleration response of the first story is chosen as the characteristic vector to identify damage, and is used as input variable for BP neural network. The damage time and location in a shear-type structure are monitored using wavelet analysis and BP neural network, and the sensitivity is compared using different response signal. Numerical example shows that the position and degree of damage are accurately identified as well as the damage moment, which proves the proposed method is feasible.Frequency and model of a shear-type structure to the first order sensitivity of the rigidity among layer has been analyzed according to the theory. On this basis, the square of the first six stages frequency changes in structure model as characteristic parameters are chosen to put into BP neural network to identify damage. Numerical example shows that the proposed method is of practical value.Finally, the energy spectrum of acceleration response in the first layer and the square of the first stage frequency changes in structure model as characteristic parameters are chosen to put into BP neural network to identify damage. Numerical example shows that with the proposed method, the location and degree of the damage in the model can be identified accurately; it is also robust to measured errors.