FAULT DETECTION AND DIAGNOSIS OF VEHICLE GEARBOXES USING VIBRATION ANALYSIS AND NEURAL NETWORKS

摘要

This paper investigates a fault detection and identification technique applied on a passenger car gearbox system using artificial neural network. The inspected gearboxes are the Nissan Junior 5-speed manual gearboxes collected from the Charkheshgar Company's production line. The novelty of this work is the implementation of a fault diagnosis technique on 14 gearboxes collected from an assembly line which contain completely dissimilar and uncontrolled fault types. The time domain approach is used for feature vector extraction. Eight parameters, which are used extensively in the literature for gearbox fault diagnosis, are calculated and four of them are found to be more informative than the others which are consequently used in feature vector construction. Finally, the multilayer perceptron neural network is used for monitoring the incurred changes in the patterns. The obtained results demonstrate the good performance of the suggested scheme for fault diagnosis of the gearboxes.