FAULT DIAGNOSIS METHOD FOR MECHANICAL DEVICE

摘要

A fault diagnosis method for a mechanical device. First, utilizing a signal analysis method to extract fault features of a mechanical device comprised in signals collected by sensors, and forming fault feature vectors; then, utilizing the fault feature vectors to train a fault diagnosis classifier offline, wherein the fault diagnosis classifier adopts a combined classifier, and the combined classifier adopts an Adaboost lifting algorithm to act as a combined method; finally, applying in real time the trained fault diagnosis classifier to perform fault diagnosis on the mechanical device. The present method adopts a combined classification method to achieve the diagnosis of faults in parts of mechanical devices, and may be used for online real-time fault diagnosis, and for promptly discovering the failure of the mechanical device, thus preventing the occurrence of major accidents. Meanwhile, compared to other fault diagnosis methods, the described method significantly improves the accuracy of fault diagnosis.