Training and Application of Radial-Basis Process Neural Network Based on Improved Shuffled Flog Leaping Algorithm

摘要

A radial basis process neural networks can be established, which is based on expanding the traditional radial basis function neural network to the time domain. Combined with the excellent characteristics of cloud model transformation between qualitative and quantitative, a improved shuffled flog leaping algorithm based on cloud model theory is presented. It is applied to training the radial basis process neural network. The neural network after optimization is used in pumping unit fault diagnosis. The diagnostic results between the CCHSFLA and BP algorithm were compared. The conclusion is that the RBPNN based on CCHSFLA has better training performance, faster convergence rate and higher accuracy.