A Method for Fault Diagnosis Based on Kohonen Neural Network

Aiming at the shortage of tradgeitional fault diagnosis technology,a method for fault diagnosis based on Kohonen neural network is put forward.During learning,this model adopts an improved SOM algorithm,which decides weight-adjustment degree with neighbourhood function,and avoids judging neighbourhood in basic SOM algorithm.This method is advantageous to increase the speed of learning and improve network's adaptability.Taking an example of wheel gear fault diagnosis,a simulative experiment is done in the Matlab.The result shows that the above-mentioned method is not only feasible,but also with a quick speed and a good accuracy.