Research And Application On Fault Diagnosis Of Rotating Machine Based On Support Vector Machine

In this paper, It has introduced some basic knowledge of fault diagnosis of the rotating machine which includes the background^the development and fault mechanism, summarized the commonly used fault diagnosis methods, focused on a method—upport Vector Machine (SVM),which is developed rapidly in the early1960s and dentally introduced six problems existing in support vector machine:feature extraction and selection; training sample; the problem of parameter optimization; unbalanced sample; multi-classification problem; the problem of multiple faults. And detailed decrypted the solutions of to the six problems.According to feature selection problem, its stability can be judged based on size of information entropy. Firstly, we can select some characteristic of several faults, establish the decision table, calculate information entropy, remove the information entropy and think it as the redundant features. We put the remaining features as feature vectors and put the feature vector in support vector machine (SVM) for fault identification. And comparing with the variance method and removing the instable variance feature, finally we can see the simulation results show that the method of the information entropy can realize fault classification and its effect is good.On the imbalance problem, by introducing the genetic algorithm, through the crossover and mutation of genetic algorithm we can increase a few categories of samples to reach the purpose of balancing with the many kinds of samples,select several attributes as feature vectors putting into the support vector machine (SVM) for fault identification and compared to the unbalance. The experiment shows that the introducing genetic algorithm can realize correct classification of the fault and achieve the purpose of classification.