An Improved Method for Dynamic Load Identification Based on Tikhonov Regularization

Dynamic load identification is an inverse problem that the dynamic loads are identified based on the responses and transfer functions (TFs). It involves the inverse of the TF matrix. The TF matrix is a pathological matrix in most cases, so the inversed problem is generally ill-posed. At present, the Tikhonov regularization method is widely used in dynamic load identification. To improve the accuracy of the Tikhonov regularization method, the paper proposes an improved method. In this work, the dynamic load identification results with the TFs and responses under different signal-to-noise (SNR) are given to investigate the effects of filtering and the genetic algorithm (GA). The results indicate that the filtering of the TF before the regularization and the adoption of GA to obtain the regularization parameters can improve the accuracy of the result. Therefore, it can be concluded that there are three main processes in the improved method. Firstly, the TFs with noise are filtered to reduce the noise in them. Secondly, GA is used to obtain good regularization parameters. Finally, the loads are obtained by the Tikhonov regularization.