Abstract: A new method of classification for mechanical system faults by means of support vector machine is proposed, and the multi-class SVM classifier based on binary classification is developed. Experimentation with vibration signals of fault gears is conducted. The results prove that this method has good classification ability and high efficiency for multi-fault classification in mechanical system even without preprocessing the signal for extracting its features. SVM provides us with a new method to develop the intelligent fault diagnosis.

Key words: Fault diagnosis, Intelligent diagnosis, Multi-fault classification, Support vector machine, operability, evaluation, model framework, production job