Experimental and finite-element free vibration analysis and artificial neural network based on multi-crack diagnosis of non-uniform crosssection beam

Crack identification is a very important issue in mechanical systems, because it is a damage that if develops may cause catastrophic failure. In the first part of this research, modal analysis of a multi-cracked variable cross-section beam is done using finite element method. Then, the obtained results are validated usingthe results of experimental modal analysis tests. In the next part, a novel procedure is considered to identify the locations and depths of cracks in the multi-cracked variable cross-section beam using natural frequency variations of the beam based on artificial neural network and particle swarm optimization algorithm. In the proposed crack identification algorithm, four distinct neural networks are employed for the identification of locations and depths of both cracks. Back error propagation and particle swarm optimization algorithms are used to train the networks. Finally, the results of these two methods are evaluated.