Abstract:In order to improve the accuracy of tool wear identification in machine working process,an on-line identification method based on dynamic expert meeting algorithm was proposed.The mechanism of tool wear was analyzed,and the recognition framework of tool wear was designed.The source signal was decomposed by CEEMD to obtain IMF components,and multi-index characteristic matrices composing of the improved I-kaz TM coefficient,power spectral entropy,standard deviation of the signal were extracted based on the IMF components.Aiming at the problems of random forest algorithm,a new dynamic expert meeting algorithm was designed by taking the decision tree as a decision expert,and the expert decision right was determined dynamically according to the historical accuracy of expert decision.The PHM2010 tool wear data set verification shows that the spatial distribution intra class cohesion and inter class discrimination of the multi-index feature matrix are well;the accuracy of tool wear identification based on dynamic expert meeting algorithm is 98.44%,which is 17.19% and 11.72% higher than RF and LS-SVM algorithms,respectively, it shows that dynamic expert meeting algorithm is effective in tool wear identification.