Research On Data-driven Fault Diagnosis Method For Switched Reluctance Motor Speed Regulation System

Switched reluctance motor drives(SRD)are widely used in new energy vehicles,intelligent home appliances,aerospace and other cutting-edge fields,which have simple structure,good fault tolerance and wide speed range.Power converters and sensors are important parts in SRD.Because of the influence of switching frequency and environment,these two parts are more prone to failure than other parts,which would weaken the control of speed control system or even cause the system to crash.Therefore,the fault rapid diagnosis of SRD is very important,The fault diagnosis methods based on data-driven has been widely used in ships,motors,inverters and other fields.It is good at not only in diagnosis speed and but also in accuracy.According to the analysis of the working principle and fault state of SRD,an adaptive sliding window fault diagnosis method for single-tube short-circuit fault,current and speed sensor fault of power converter in SRD is proposed.The main research content includes the following three parts:Firstly,the structure,control mode,mathematical model and working principle of SRD are analyzed in this paper.According to the mathematical model of SRD,the nonlinear simulation model of SRM based on Ansoft is built.By analyzing the working principle of power converter and sensor fault types that may occur,and comparing the fault characteristics of SRD under different types of fault and its impact,the fault model is built and the original data of fault is obtained to provide data support for data-driven fault diagnosis methods.Secondly,the Fast Fourier Transform(FFT)is used to extract the features of the original fault data,and ReleifF algorithm is used to reduce the dimension of the data for learning of fault classification algorithm in this paper.For the single-tube short-circuit fault and sensor fault of power converter in SRD,an adaptive sliding window fault diagnosis method is proposed by combining k-Nearest Neighbor algorithm(kNN),Support Vector Machine(SVM)and multi classification Extreme Learning Machine(ELM).This method consists of three time-windows: Sliding window I combines integrated ELM classifier 1 with SVM,sliding window II uses integrated ELM classifier 2,and sliding window III uses kNN algorithm.Finally,in order to verify the effectiveness of the proposed method,a 12/8-pole SRD with three-phase asymmetric half bridge power converter is researched.Then the simulation model is built based on MATLAB/Simulink according to the adaptive sliding window fault diagnosis mechanism.The results of off-line test and on-line diagnosis show that the proposed method can diagnose the single-tube short-circuit fault and sensor fault of power converter quickly and accurately,where the accuracy are 98.87%.