Research On Wear Recognition And Life Prediction Of Tool Based On Extend Hidden Markov Models

Machining tools directly affect the quality of finished products.Real-time monitoring of tool wear status is important to ensure process stability and product quality.This paper takes high-speed milling tool as object,and establishes the online identification and RUL prediction system of tool wear state based on hidden Markov Model with its extension.Studying the principles and methods of feature extraction and selection,pattern recognition,RUL prediction.And based on all the methods above,this paper design and develop a prototype system for tool wear state online identification and RUL prediction.The specific research is as follows:For the feature extraction and selection part,the multi-sensor signal data feature extraction method is studied,and multiple signal feature quantities are extracted from three angles: time domain,frequency domain and time-frequency domain.Extracting a plurality of signal features from three angles: time domain,frequency domain,and time-frequency domain,respectively.Then the Pearson correlation analysis and approximate redundant feature analysis are used to complete the feature selection to ensure the validity of the feature quantity.For the state identification modeling part,based on the hidden Markov model's ability to learn and analyze time series data,a tool wear state recognition model based on hybrid Gaussian hidden Markov model(GMHMM)is established.The BSO algorithm obtained by beetle antennae search algorithm(BAS)optimizing the particle swarm optimization algorithm(PSO)is used to optimize the initial value of the model,which effectively avoids the model results fall into local optimal solution.The state recognition modeling based on BSO-GMHMM model is finally completed.For the RUL prediction part,the Duration-Dependent Hidden Semi-Markov model(DD-HSMM)based on a time-varying state transition probability is studied.Which gives the display distribution probability of each state dwell time.Furthermore introducing time components into state transition probabilities,allowing the historical operation information merged into estimation process of RUL prediction model.The obtained tool wear RUL prediction model is more realistic and more accurate.Finally,based on the wear status recognition and life prediction model,the prototype design of the application system was completed with the support of MATLAB App Designer.