Multiple Fault Classification Using Support Vector Machine in a Machinery Fault Simulator

The classification of various faults using a fault simulator and support vector machines (SVMs) has been studied. A database is created for number of faults by measuring vibration signals using seven accelerometers mounted on a machinery fault simulator (MFS). Statistical features are extracted in time domain from the vibration signals. Then, the sensitive features are selected using compensation distance evaluation technique. Multi-class SVMs ensemble algorithm is implemented for classification of the various faults by considering SVMs created by the possible combinations of sensitive features for each class of the fault. The effect of distance evaluation criterion for selection of sensitive features amongst the extracted twelve statistical features has been addressed. By using the developed algorithm, the effective location of accelerometer among seven accelerometers for better classification of the faults has been investigated. Measurements are done at five different rotational speeds. The robustness of the developed algorithm has been tested at different speeds.