Fault diagnosis of complex system based on nonlinear frequency spectrum fusion

A novel fault detection and diagnosis approach is proposed for nonlinear complex systems by combining nonlinear frequency spectrum characteristics and evidence theory. In order to overcome the problem of calculated amount expansion of generalized frequency response functions, single-dimensional nonlinear output frequency response functions are used to obtain nonlinear frequency spectrum, from which, features of nonlinear frequency spectrum are extracted. The fault diagnosis model of multiple faults is given based on evidence theory. The mass functions of evidences are obtained according to the similarity among different modes. In order to solve the problem of evidence fusion in the situation of evidence confliction, a dynamic parameter conflicting evidence combination method is proposed based on the average credibility. Fault diagnosis of the transmission system of numerical control machine tool is studied through nonlinear frequency spectrum. Simulations indicate that the proposed approach has simple calculation and high recognition rate for faults.

► NOFRF is used to extract nonlinear frequency spectrum feature. ► Mass functions are obtained according to the similarity among different models. ► Dynamic parameter conflicting evidence combination method. ► Fault diagnosis simulation experiment of the transmission system of CNC machine tool.