The Development Of The Fault Diagnosis System For Gearbox Based On Support Vector Machine

As an essential part of connection and force transmission in mechanical equipment, gearbox system have extensive application in the modern industrial development. But because of the reasons, such as complicated structure of their own and abominable working environment, gears and gearboxes are easy to be damaged and broken down, thus cause security events. Therefore, it is significant to monitor running condition of gearboxes, find and fix faults in time, which can guarantee precision of equipments and runing stability and have important engineering applications value and significance.This subject regards developing fault diagnosis system of gearbox as the overall goal, applies support vector machine to the pattern recognition of gearbox fault, aim at realizing the intelligent fault diagnosis of gearbox and improving its precision of diagnosis. This thesis has analyzed the vibration feature of the typical gearbox fault and explained the methods of vibration signal analysis and fault diagnosis for gearbox. On the basis of signal analysis methods, it extracted time-domain parameters, frequency-domain parameters and Multi-scale space energy feature parameters based on wavelet decomposition, that are used as input feature vector for support vector machine, to construct fault diagnosis model of gearbox and realize the pattern recognition of gearbox fault. Finally, a fault diagnosis system for gearbox was developed by using object-oriented Visual C++. The software system contains various functions, such as signal files reading, signal preprocessing, time-domain and frequency-domain analysis, wavelet analysis, feature parameters extraction, data saving and reading, fault diagnosis based on support vector machine, and so on. This system can achieve querying, zooming, moving toward analysis graphs, and provide a good man-machine interface, which cause it is very easy to operate.The function of this system is proved through collecting real fault signal of gearbox. The Experimental results indicate that the extracting methods of feature parameters by combining time-domain parameters, frequency-domain parameters and wavelet parameters and the realization of fault diagnosis for gearbox by using support vector machine have good effects, improving the precision of fault diagnosis for gearbox.