Research On Tool Wear Condition Monitoring Method Based On S Transform And Hidden Markov Model

The tool wear condition monitoring in real-time is an important means to achieve parts processing automation and intelligence,ensure the quality of parts and improve production efficiency.At present,the technology of tool wear condition monitoring is still not completely and effectively realized.Therefore,the research of tool wear condition monitoring is of great significance for the overall improvement of manufacturing level.This paper aimed at the research topic of tool wear condition monitoring under different cutting conditions,and designed a reasonable experimental scheme according the orthogonal test method,and built an experimental bench and conducted cutting experiments,and collected tool wear acoustic emission signals with acoustic emission sensors,and proposed a tool wear condition monitoring method based on S transform and hidden markov model.The specific research includes several parts:1?The acoustic emission signal of the tool wear was denoised based on the ensemble empirical mode decomposition and the correlation coefficient and the kurtosis.Firstly,the non-stationary tool wear acoustic emission signal was decomposed into several stable intrinsic modal functions through the ensemble empirical mode decomposition.Then,the reconstruction criterion was proposed by combining the cross-correlation method and the kurtosis criterion,and the effective intrinsic modality function was screened to reconstruct the signal,and the denoised signal was obtained.Finally,the effect of the proposed method on the denoising of the tool wear acoustic emission signal was verified by the Hilbert spectrum and the improvement of SNR.2?The feature extraction method of tool wear acoustic emission signals was proposed based on S transform and hidden markov model.Firstly,the denoised signal was analyzed by time-frequency analysis through S-transform,and the time-frequency characteristics of the signal in different wear states of the tool was discussed.Then,the gray level co-occurrence matrix algorithm was applied to extract the texture feature parameters of the time-frequency image,and the original feature vector was constructed,and the sensitivity analysis of the feature parameters was performed by the scattering matrix algorithm,and the sensitive feature vector was constructed.Finally,the manifold learning algorithm was applied to fuse the sensitive feature vectors,and further eliminated the redundant features in the feature vectors,so as to achieve the fusion extraction of nonlinear weak features.The analysis of the scatter diagram qualitatively shows that the fusion feature could better identify the tool wear condition.3?The tool wear condition recognition based on discrete hidden markov model was presented.Firstly,the characteristics of tool wear in the cutting process and hidden markov models were both observed through observations and the hidden markov model could reasonably describe the characteristics of the tool wear condition during the cutting process.A discrete hidden markov model for tool wear condition monitoring in the cutting process was established.Then,the process and method of tool wear condition recognition based on discrete hidden markov model were described.Finally,the fusion feature data extracted from different feature extraction methods was used to train the model and test identification.Experimental results showed that the recognition accuracy of the fusion features extracted by the local tangent space alignment algorithm was higher.