Research On Tool Wear Monitoring Technology Based On Feature Fusion

The tool wear in cutting process will directly affect the surface quality and dimensional accuracy of the workpiece,and the tool breakage may cause the workpiece to be scrapped,even cause damage to the machine tool and bring about huge economic losses.Real-time monitoring of the tool wear status is of great significance for improving product quality,achieving automatic tool change,increasing productivity and reducing production costs.In order to accelerate the pace of intelligent manufacturing,the tool wear in cutting process is studied in this paper.The full text is summarized as follows:Firstly,the research background and significance of tool wear monitoring are outlined.By taking monitoring methods and decision-making systems as the breakthrough point,it summarizes the research status at home and abroad,and determines to take the cutting forces as the monitoring signal and take the tool flank wear width as the evaluation index of tool wear.Two artificial intelligence(AI)models are utilized to build the decision-making systems(identification models).A large number of turning and milling experiments are carried out respectively to obtain the experimental data of turning and milling tools under different wear conditions.This provides data support for the subsequent research.Afterwards,three types of monitoring features for tool wear monitoring are introduced,i.e.,the time-domain features,frequency-domain features and wavelet-domain features.Besides,the integrated radial basis function based kernel principal component analysis(KPCA_IRBF)technique and neighborhood preserving embedding(NPE)technique are utilized to fuse the extracted monitoring features so as to to weaken or remove the effects of noises and improve the performance of the subsequent identification models.Among them,the KPCA_IRBF is a kind of new nonlinear dimension-increment technique and firstly proposed for feature fusion.Common AI models,such as Artificial Neural Networks(ANN)and Support Vector Machines(SVM),can only provide a single predictive value.However,Gaussian Process Regression(GPR)can both provide the tool wear predictive value and the corresponding confidence interval.Besides,GPR performs better than ANN and SVM in prediction accuracy since the Gaussian noises can be modeled quantitatively in the GPR model.Nevertheless,the existence of noises will affect the stability of the confidence interval of GPR seriously.In order to ameliorate the confidence interval of GPR,the KPCA_IRBF technique is utilized to fuse the extracted monitoring features so as to remove the noises and weaken its negative effects,which makes the confidence interval of GPR compressed greatly and more smoothed.This is more conducive for monitoring the tool wear accurately.This paper takes the wear monitoring of turning tools as a case to verify the effectiveness of the constructed tool wear prediction model(KPCA_IRBF+GPR).Next,Support Vector Machines(SVM)coupled with Whale Optimization Algorithm(WOA)is proposed for tool wear classification.The NPE technique is utilized to fuse the extracted monitoring features so as to remove the noises,enhance the computational efficiency and improve the identification accuracy.This paper takes the wear monitoring of milling tools as a case to verify the effectiveness of the constructed tool wear classification model(NPE+WOA-SVM).Experimental results show that the WOA-SVM model performs better than the commonly used methods,such as PSO-SVM and GSA-SVM,in terms of time-consumption for parameter optimization and has comparable identification accuracy.In addition,the WOA-SVM model has higher prediction accuracy than some classical classification algorithms,such as k-NN,FFNN,LDA,QDA and CART.This study provides theoretical guidance and technical support for tool wear monitoring in real industrial settings.