Tool wear prediction using evolutionary Dynamic Fuzzy Neural (EDFNN) Network

In development of self-organizing fuzzy neural network, selection of optimal parameters is one of the key issues. This is especially so for a system with more than 10 parameters whereby it will be challenging for expert users to determine the optimal parameters. This paper presents a hybrid Dynamic Fuzzy Neural Network (DFNN), and Genetic Algorithm (GA) termed Evolutionary Dynamic Fuzzy Neural Network (EDFNN) for the prediction of tool wear of ball nose end milling process. GA, well known for its powerful search method, is implemented to obtain optimal parameters of DFNN, so as to circumvent the complex time varying property without prior knowledge or exhaustive trials. Degradation of machine tools in ball nose end milling process is highly non-linear and time varying. Benchmarked again original DFNN in the experimental study, EDFNN demonstrates the effectiveness and versatility of proposed algorithm which not only produces higher prediction accuracy, and faster training time, but also serves to more compact and parsimonious network structure.