Application of Fuzzified-factor Based BP-network in Wear Debris Identification

The program of auto identification of wear particles has been made by means of artificial neural network (ANN) technique,with which a set of morphology descriptors of wear debris are cited to describe the micro features of wear particles. During training of the network, the fuzzified factor based training technique is used, and the training process is accelerated rapidly. Taking the exclusive or problem (XOR problem) as an example, the training speed increases by five to ten times. The network has an identifying accuracy higher than 90% and is effective in identifying various wear particles.