Fault Diagnosis Of Motor Drive System Based On Multi-Classifier Fusion

With the development of the large-scale,complex and intelligent technology and industrial equipment,electrical machine,as the key equipment,is widely used in electric vehicles,aerospace,medical and military applications.With the development of science and technology,electric power equipment tends to be large,complicated and intelligent.Once the drive system,as a key part of the motor system,breaks down,it will not only affect the whole equipment,but also may even cause huge economic losses,environmental pollution,casualties and other social problems.In order to improve the accuracy of diagnosis,in this paper,parameters of two classifier,(Support Vector Machine(SVM)and BP Neural Network(BPNN)),are firstly optimized by Gravitational search algorithm(GSA)).On this basis,GSA is combined with Partical Swarm Optimization(PSO)to solve the problem that GSA is easy to fall into local Optimization,and chaos mapping and adaptive attenuation factor are introduced to balance GSA's global exploration ability and local development ability.Finally,in order to further improve the diagnostic accuracy and integrate the performance advantages of SVM and BPNN,the pre-diagnosis results of two classifiers were fused at the decision-making level by using d-s evidence theory to obtain the final diagnosis results.The main content of the paper is as follows:1?Introduing the development of fault diagnosis technology,artificial intelligence diagnosis algorithm and information fusion algorithm are introduced.Then describes several fault types of ac motor,such as stator fault,rotor fault,eccentric fault and inverter fault,and analyzes its formation mechanism..2?Aiming at the problem that the parameter selection of classifier directly affects the classification accuracy,some parameters of SVM and BPNN are optimized by GSA,and the effectiveness of this method is verified by simulation and experiment.3?Particle swarm optimization algorithm(PSO)is introduced to solve the problem that GSA is easy to fall into local optimization,and it is fused with GSA to complement each other.The hybrid algorithm GSAPSO is used to optimize the SVM and BPNN,which is used in balancing the global exploring ability and local development ability of GSA.Introducing chaos mapping and gravity coefficient of attenuation factor to improve the GSA,finally,using chaos adaptive GSAPSO-SVM and chaos adaptive GSAPSO-BPNN to classify the fault of motor drive system.4?Types of information fusion are introduced.For single classifier restricted by their own performance and not on the sample data to conduct a comprehensive analysis of the problem,the pre-diagnosis results of two improved SVM and two improved BPNN were fused at the decision level using the d-s(dempster-shafer)evidence theory.Finally,in order to improve the diagnostic accuracy further,an improved SVM and an improved BPNN were fused to obtain the final diagnosis result.