Fault Diagnosis of Nonlinear System based on Simplified GFRF Estimation and LSSVM

For the fault diagnosis of nonlinear system, a novel method by combining generalized frequency response function (GFRF) and least squares support vector machine (LSSVM) model is proposed. The harmonic signal is used as the input to estimate the generalized frequency response functions. In order to simplify the estimation process and improve the estimation accuracy, the GFRFs of nonlinear system are solved directly according to the time-domain data, and the Fourier transforms are not required. The amplitudes of generalized frequency response functions are used as the fault features. After obtaining the fault features, a multi-fault classifier is designed by LSSVM. The classifier is trained based on the conjugate gradient algorithm and a simplified LSSVM model, which improves the training speed and reduces the storage. The fault diagnosis simulation experiment of a nonlinear oscillation system is carried out to verified the proposed method.