Research On Tool Wear Status Intelligent Monitoring Based On The Power And Vibration Signals

Tool wear is inevitable in cutting process.Tool wear will directly affect the surface quality and the dimensional precision.Tool condition monitoring technology can monitor tool wear status in real time,which is of great significance to improve product quality and processing efficiency,reduce production cost,protect mechanical equipment,and promote the transformation and upgrading of manufacturing industry to "intelligent manufacturing".In this paper,the recognition of tool wear state in cutting process was taken as the research object,the machining characteristics of cemented carbide turning tools under different wear states were analyzed,and the mapping relationship between wear states and monitoring features extracted from power and vibration signals was established.The main research is as follows:(1)In this paper,the appropriate signal processing method was applied to analyze the vibration signal,and a method of fusing the extracted monitoring features based on kernel principal component analysis was applied.The vibration signal in the cutting process is nonlinear and non-stationary,modern signal processing methods,such as wavelet analysis,Hilbert Huang transform,empirical mode decomposition,etc.,should be chosen.In this paper,wavelet packet analysis combined with ensemble empirical mode decomposition method was used to process vibration signals.Monitoring features were extracted from time-field,frequency-field and time-frequency field respectively,and the original high-dimensional features were dimensionally reduced by kernel principal component analysis,redundant features were removed to simplify the data,which was convenient for subsequent monitoring model recognition.Based on the power signal,the tangential cutting force coefficient and the tangential edge force coefficient extracted by the least square regression method are only affected by the degree of tool wear.Under the condition of variable cutting parameters,they can be used as the effective features to recognize tool wear states.The research results show that the above feature extraction methods combined with data fusion has higher recognition rate than single feature extraction method for tool monitoring.(2)Support vector machine(SVM)and particle swarm optimization(PSO)were used to classify and recognize the tool wear states.Compared with the shortcomings of artificial neural network training,such as a large number of samples and slow convergence speed of learning algorithm,this paper selected support vector machine and particle swarm optimization algorithm to optimize the parameters of monitoring model.Taking the wear monitoring of carbide turning tools as an example,the validity of recognition model is verified.The experimental results show that in the case of small samples,SVM method shows better classification performance.Compared with other classification algorithms,such as BP,k-NN,the model has higher classification accuracy.(3)Based on the recognition model of tool wear states,the tool wear state online monitoring software was built under Lab VIEW platform.Based on the practical production application and considering the cost,reliability and stability of the monitoring system,a complete and practical tool wear monitoring system was developed.The functions of the system include data acquisition,storage and reading,signal analysis,waveform display,tool wear status recognition and early warning.The experimental results show that the system can run stably and has excellent classification accuracy.In conclusion,aiming at the problem of tool wear monitoring,this paper designs and builds a monitoring system and monitoring method to realize the intelligent recognition and real-time online monitoring of tool wear status.This study provides technical support for tool wear monitoring in the actual industrial environment.