Focusing on the small samples and nonlinear characteristics of tool wear prediction of 42CrMo steel in pre-cision cutting,the QPSO-CNN-LSTM prediction model based on quantum particle swarm optimization(QPSO),convolution-al neural networks(CNN)and long short-term memory network(LSTM)are proposed.QPSO algorithm is used to optimize the number of hidden layer units,learning rate and convolution kernel of CNN-LSTM model.Combined with the strong fea-ture extraction and memory ability of CNN and LSTM network,the tool wear amounts of actual machining experiment are predicted.By the analysis of error evaluation indexes,it is compared with the single model of CNN,LSTM and BP,and PSO-GRNN combined model for the prediction effect.The results show that the proposes has a higher degree of coincidence be-tween the predicted and true values compared with the single prediction model.The error value of the three error evaluation indexes is reduced by 27％at least compared with PSO-GRNN combined model.It has better generalization and stability,and greater prediction accuracy and nonlinear fitting ability.
维普
万方数据