Research On Fault Diagnosis Method And Applications For Electromechanical System

Electromechanical system is widely used as a basic unit of theautomation system. With the progress of science and technology and therising level of modernization, most electromechanical systems used inindustry site are having more and more complex structures. Due to theapplication environment and conditions, the possibility of beingout-of-order will increase. Therefore, research on fault diagnosis method ofelectromechanical system has practical significance and theory value.This paper takes the electromechanical system as the projectbackground and takes the wind power generation system and robots as theresearch object. A deep investigation is done in this paper on the nonlinearfault diagnosis of electromechanical systems, so as to increase therobustness against modeling error or disturbance and the sensitivity of faultdiagnosis.The major contents and conclusions can be summarized as follows:For the power factor of no-load cutting-in process in doubly-fed windgeneration system is not always1, a mathematical model of no-loadcutting-in process in doubly-fed wind generation is established and anintegral sliding mode variable structure controller is designed. The resultsof the PSIM simulation test and the real-time experiments verify theeffectiveness.For the sensor faults of doubly-fed wind generator, a design of aSliding Mode Observer (SMO) combined with a Luenberger observer isproposed. It can demonstrate the effectiveness of the methodologies via thesimulation experiments.For incipient faults of one class of electromechanical system,coordinate transformation is used to transform the electromechanicalsystem into two subsystems. Then for the subsystem free from disturbances,a Luenberger observer is constructed to reconstruct incipient faults and asliding mode observer is constructed for another subsystem. Finally, arobust fault diagnosis of weak incipient faults is realized. In the design, inorder to relax the selection conditions of key parameters, LMI algorithm is used. And the design method is more practical for LMI algorithm. Finally,the effectiveness of the proposed approach is illustrated via a manipulatorexample.