Study On The Fault Diagnosis Of Variable Frequency Speed Regulation System Based On Wavelet Packet Neural Network

The variable frequency speed regulation technology as its significant energy-saving effects, good speed performance and the advantages of a wide range of applicability is widely used in industrial and agricultural production and national defense science and technology and other fields. The variable frequency speed regulation system is made up of the convertor, motor, mechanical transmission device .In order to ensure safe and reliable operation of the system, to improve system reliability of the variable frequency speed regulation system is a core studying content. In response to this core content,in this paper, the fault detection and diagnosis of variable frequency motor speed control system are researched. Wavelet de-noising technique in fault signal pre-processing is studied, the wavelet function, the determination rules of the threshold and threshold conducted method are in-depth studied; The characteristics of convertor's output voltage and current in fault state are discussed,and an improved algorithm of wavelet packet to extract the fault current signal feature is proposed. The characteristics of the motor with rotor broken bars are introduced and the fault diagnosis method which based on wavelet packet decomposition coefficients to determine whether there has been motor rotor broken bars is proposed, while the energy method using wavelet packet frequency bands to study it. In order to improve the accuracy of fault diagnosis , motor vibration signals are gathered and analyzed. The characteristics of vibration signals are extracted by the way of wavelet packet frequency band energy, which as BP neural network input vector,then using the train function Trainlm of which learning algorithm is the Levenberg-Marquadt back-propagation algorithm to train the network, through trial and error analysis to determine the BP network structure . By tested multiple sets of data, the test results basically reach the expected goal. Extensive experiment shows that the scheme on fault diagnosis of the variable frequency speed regulation system which based on wavelet packet neural network is effective and of great feasibility.