Estimating Shaft Crack Specifications Using Shaft Vibration Analysis and Neural Networks

In recent years, many attempts have been made to estimate shaft crack specifications with the least possible error. In this paper, an indirect method of diagnosing a shaft is proposed using neural networks. The shaft natural frequencies which are influenced by crack specifications are obtained by means of a finite element method. The numerical data are then used to train three two-layer feed-forward back-propagation neural networks. Some simulations are carried out to test the performance and accuracy of the trained networks. The simulation results show that the proposed neural networks estimate the location, width, and depth of cracks precisely.