Damage Identification of Mechanical System with Artificial Neural Networks

Lijuan Cao,Shouju Li,Zichang Shangguan
DOI: https://doi.org/10.4028/www.scientific.net/kem.385-387.877
2008-01-01
Key Engineering Materials
Abstract:The inverse problem of structure damage detection is formulated as an optimization problem, which is then solved by using artificial neural networks. Based on the hybrid optimization strategy, the parameter identification algorithm was presented according to the measured data of vibrating frequency and mode shapes in the damaged structure. The proposed algorithm combines the local optimum method having fast convergence ability with the neural networks having global optimum ability. By doing this, the local minimization problem of the local optimum method can be solved, and the convergence speed of the global optimum method can be improved. The investigation shows that to identify the location and magnitude of the damaged structure by using an artificial neural network is feasible and a well trained artificial neural network by Levenberg-Marquardt algorithm reveals an extremely fast convergence and a high degree of accuracy.
What problem does this paper attempt to address?