APPLICATION OF THE UNSUPERVISED NEURAL NETWORK FOR DIAGONOSING FAULTS IN A SPUR GEAR SYSTEM

Abstract

In transmission, gearbox with normal teeth is used popularly. Now, fault detection of gearbox is made by listening to abnormal noise and opening the gearbox to find the fault. This work takes much time and labour and results are not always exact. In this paper, vibration signal caused by fault are recorded and transformed into spectrum. The spectrum is used to train the unsupervised neural network. After the traning is completed, the unsupervised neural network will recognize the kind of fault which is produced by gearbox. Results of recognition from this method are almost exact when compared with others.