In order to improve the robustness and generalization of the tool wear status recognition model,a data-driven tool wear status recognition method with deeply stacking sparse denoising auto-encoder(SSDAE)network is proposed to achiev-e automatic mining of data features hidden in the data at a deep level.First,the original vibration signal is decomposed a series of intrinsic mode function(IMF).The Pearson correlation coeffi-cient method is used to select the optim-al intrinsic mode function to combine a new signal.Secondly,the SSDAE adaptive feature extraction is used to ident-ify the state of the tool wear stage,and the accuracy of the tool wear state identification reached 98%.Finally,the netwo-rk model is experimentally validated and com-pared with the most commonly used tool wear state recognition methods.The experimental results show that the proposed method can handle non-smooth vibration signals well.Therefore,this method has good recognit-ion ef-fect on different tool wear stage states good generalization performance and high reliability.
NETL
NSTL
万方数据
维普
