Application of GM(1,1) model and improved EMD in fault diagnosis of airborne direct-driven electro-mechanical actuators

摘要

Empirical mode decomposition (EMD) is a data driven self-adaptive signal processing algorithm, and is effective in analyzing nonlinear and non-stationary signal. But EMD has the serious end effect during decomposition, which leads to distortion near two ends of the signal, so end effect is the difficult point to influence this method's precision. Combined with advantages of the grey prediction model and EMD, this paper used GM(I,l)model to modify ends' extension values, and then effectively inhibit end effect by cubic interpolation envelope. Airborne direct-driven electro-mechanical actuator (DDEMA) is a newly developed electro-mechanical system with complex fault mechanism; its fault signal has non-stationary and nonlinear characteristics. It is unideal to adopt traditional time-frequency analysis method to analysis fault signals of DDEMA, because these traditional signal analysis methods can only give statistical average of dynamic signals in time or frequency domain. By applying the improved EMD based on GM(1,I) model to decompose vibration signals from fauh PMSM with unbalanced rotor, obtain all Intrinsic Mode Functions (IMFs) from the decomposition process, and then extract accurately fault characteristics and frequencies from IMFs, which contain fault information. The experimental results show the improved EMD is correct and effective, so the proposed diagnosis method is a new way in diagnosing faults of DDEMA.