Abstract:In the present study, a novel method combined nonlinear spectrum with multiblock principal component analysis (MBPCA) is proposed for fault diagnosis of large-scale electromechanical systems. According to the generalized frequency response function (GFRF) description, the multivariable nonlinear system is decomposed into a number of single variable nonlinear subsystems that the amount of computation of solving nonlinear spectrum is reduced. The relative gain obtained by the first order generalized frequency response function of multivariable system is used as the correlation feature data, which can reflect the correlated characteristic of multivariable nonlinear system. The nonlinear spectrum data of subsystems and the correlation feature data are divided into several blocks, and a multiblock principal component analysis method is used to extract feature so that the feature data can sufficiently reflect the system characteristics. A fault diagnosis simulation example about a multivariable nonlinear system is provided to illustrate the effectiveness of the proposed approach. The results indicate that the proposed diagnosis method has high accuracy and good real-time performance, which can meet the needs of fault diagnosis of large-scale electromechanical system.

Fault Diagnosis of Large-Scale Electromechanical System Based on Nonlinear Spectrum