Sdudy Of On-line Fault Diagnosis Of DC Motor Based On Neural Network

In the modern major industrial production,DC motor with its speed regulation performance and starting performance,good performance of overload capacity and large torque,is high and widely used in equipment of rolling mill,hoist,crane,machine tools and electric locomotive.In this paper,a practical and simple method for on-line fault diagnosis of DC motor is studied in order to monitor the running state and fault condition of the motor.In order to meet the requirements of comprehensiveness and accuracy in fault diagnosis,a new method for fault classification of DC motor is presented.In order to realize the real time and convenience of on line fault diagnosis of DC motor,an on-line fault diagnosis method based on current signal analysis is presented.Based on the easy collecting and monitoring signal of the DC motor armature current,using comprehensive utilization of Fourier analysis signal processing method to extract the required fault feature,a method of online fault diagnosis based on artificial intelligence is realized.Based on the artificial neural network and grey correlation analysis,a mathematical model of the on line fault diagnosis of DC motor is established.The 5 fault diagnosis characteristic parameters,such as the input voltage of the motor,the peak value of armature current during the starting process of the motor,the maximum drop rate of armature current,the armature current and the corresponding frequency of the motor are analyzed.The MATLAB platform is used to simulate the normal operation state of the DC motor,four kinds of single fault states,the three kinds of multiple faults states and the operating state of the voltage and load disturbance.Based on the characteristic parameter variation characteristics of various fault states,explained the BP neural network recognition ability for various faults,proved the correctness and validity of the neural network algorithm for on-line fault diagnosis of DC motor.The disturbance of voltage and load to fault diagnosis model is solved.Based on the actual data of DC motor,four kinds of faults are taken as examples to extracting characteristic parameter data and analysis failure mechanism.With the analysis of variation characteristics of fault parameters of DC motor under simulated and real state,the correctness and universality of the fault diagnosis model are proved.The neural network training and fault diagnosis are carried out by using the actual fault samples.By comparison,the optimized neural network fault diagnosis accuracy rate has been greatly improved,the average correct rate of 91.4%.The results show that the on-line fault diagnosis model of the DC motor is reliable and can be used to help the user to monitor the status of the DC motor in real time.The proposed DC motor on-line fault diagnosis method only needs the acquisition of the current and voltage signal of the motor,and it needs no other mechanical and temperature signals.It is small,low cost,highly reliable,and easy to realize hardware and software,of small impact on the motor system.It has a wide application prospect in the field of large electric drive production occasions such as rolling mill.