Improved RBF neural network algorithm for reliability data

Abstract

RBF neural network algorithm refers to a kind of new method promoted according to the fact that failure data volume is small while the distributed model cannot be distinguished. For the problem that non-teaching study in current RBF neural network algorithm should be deter-mined by experts' experience, AP is promoted to improve the current algorithm based on which a new p value is de-signed to make the number of clustering centers in new algorithm determinable for that of original samples. By creating variable groups of data arbitrarily, the BWP and NRMSE value are used for comparing the effects of clustering and extending result repeatedly analyzing new or old algorithms. The failure data of one kind of numerical control machines is analyzed and calculated repeatedly to test the validity of new algorithm in which the identifiable rate of distributed model is promoted, compared with original algorithm.