Owing to that the function mapping relationship between the technological parameters and performance index of wire electrical discharge machining (WEDM) has a non-linear characteristic, the artificial neural networks were incorporated into WEDM calculations. To compensate the disadvantage of the conventional back propagation algorithm (CBPA), an improved learning algorithm, which trained a BP neural network by the genetic algorithm, was developed. A predictive model for surface roughness of ceramic by WEDM was developed based on the evolutionary neural networks (ENN). The results show that the ENN can effectively overcome the problems of easily falling into local minimum point and of weak global search capability. The errors between the prediction values and the practical measured ones are less than 4%.
参考文献
相似文献
引证文献
引用本文
徐小青,骆志高,徐大鹏,丁圣银.陶瓷材料电加工表面粗糙度的预测[J].农业机械学报,2007,38(3):164-167.[J]. Transactions of the Chinese Society for Agricultural Machinery,2007,38(3):164-167.