Research On Pattern Recognition And Life Prediction Of Tool Wear Based On Multi-sensor Information Fusion

Cutting tool is the direct executor in automatic processing,which will have the most direct effect on cutting quality and machining precision.Serious failure even damage the machine tool,endangering life safety,so it is of great significance to tool wear monitoring.In this paper,the information of different sensors is analyzed and processed by multi sensor information fusion.The results are more reasonable,more comprehensive and more accurate than the single sensor,so as to realize the accurate monitoring of the tool state.Therefore,the following research work has been carried out.In the aspect of signal acquisition,this paper studies and analyzes the characteristics of each signal in the process of cutting,and finally chooses the force signal,vibration signal and acoustic emission signal to set up the multi sensor test platform,and collect and process the related signals,and extract the characteristics of the different wear state of the tool by scientific method,so as to realize the tool wear shape.It provides basis for state monitoring,pattern recognition and life prediction.In this paper,an improved total empirical mode decomposition(Modified-EEMD)algorithm,which is suitable for the feature extraction and preprocessing of turning signals,is studied and proposed in this paper.The method is based on the inherent characteristic of the signal,and decomposes the target signal into a number of implication modal functions(IMF).The method is distinguished from the traditional EMD and EEMD method by adding a white noise signal with zero mean value of one negative and one negative two groups by adding the original signal to the original signal,thus reducing the error of the reconstruction of the white noise,making the decomposition of the EMD more complete and passing through.Whether or not the IMF component is abnormal determines whether to continue the EMD decomposition,to a certain extent,it improves the efficiency of feature extraction,simplifies the steps,and reduces the cost of computing.In pattern recognition,in view of the shortcomings of the traditional decision fusion algorithm,a state recognition information fusion based on support vector machine is adopted.The independent BP neural network and the Elman neural network are fused respectively.The results show that the fusion method can also ensure the accuracy and robustness of the pattern recognition.Efficiency.In the aspect of life prediction,the improved grey hidden Markov prediction model is studied and analyzed.Finally,the accurate prediction of cutting tool wear and life is realized.The difference between the classical grey hidden Markov model is that it does not add directly to the original sequence,but directly establishes an improved method for the calculation of the first order differential equations participating in the calculation and the establishment of the GM model.The experimental results show that the method has higher prediction accuracy than the classic grey Markov prediction model,and it is a tool wear state.Accurate prediction provides a new method for reference.