Intelligent detection of stator and rotor faults of induction motor based on improved backstepping sliding mode observer

iming at the influence of nonlinear part and unknown load disturbance on fault detection in induction motor system, an improved backstepping sliding mode observer based on extreme learning machine (ELM) is proposed by this paper. Firstly, the state space mathematical model of induction motor in d - q coordinate system is established. Secondly, in order to reduce the influence of nonlinear function on system tracking performance. ELM is used to adaptively approximate and estimate the nonlinear function. Then, based on the conventional backstepping method, the error integral and saturation function are introduced, and combined with sliding mode control, the traditional and exponential observers are designed. At the same time, the Mayfly Optimization Algorithm (MOA) is used to optimize the control parameters of the exponential sliding mode observer, which can effectively improve the convergence speed and stability accuracy of the system. Finally, the fault between stator and rotor windings is simulated. The comparative experimental results show that the method proposed in this paper has good robustness to unknown load disturbance. Through the analysis of current and speed residuals, it can quickly and sensitively detect early micro-faults and realize the on-line fault detection of induction motor.