GRINDING MODE-IDENTIFICATION AND SURFACE QUALITY PREDICTION USING NEURAL-NETWORK IN GRINDING OF SILICON-NITRIDE

A simple neural network model for grinding mode identification and sur face quality prediction in grinding of silicon nitride has been establ ished. Brittle to ductile grinding mode transition has been experiment ally analysed. Based on this analysis, model inputs and outputs were c arefully chosen and off-line model training using backpropagation algo rithm was carried out. Various models were verified and the optimum mo del configuration was tested. Simulation results show a good agreement with experimental results for a wide range of working speed and depth of cut. Based on these results, the possibility of using this neural network model for grinding mode identification and ground surface qual ity prediction has been confirmed. On-line prediction concept computat ion was introduced in order to save machining time and costs. This mod el can be improved by modifying internal structures such as the learni ng rule, hidden layers and neuron numbers.