Tool Condition Monitoring System Based On Simulation Samples Of Main Cutting Force

Tool condition monitoring is an imperative part of intelligent manufacturing.At present,there are two key problems in this field,which are difficult to solve: the economy of monitoring model and the ability to deal with large data.In this paper,main cutting force models and machine learning technologies are introduced into the field of tool condition monitoring.The calculation of the main cutting force of a specific tool is deduced by instantaneous chip thickness.The simulation samples are generated based on theoretical main cutting force.The classification of samples is trained by using dimensionality reduction and kernel SVM.This paper offers a new idea for tool condition monitoring.The main works are as follows:(1)Models of main cutting force of large-feedrate milling tool and translocation turning tool are established.The results are compared with experimental data.The generation method of the simulation samples of the main cutting force is studied based on the possible occurrence of cutting force signals in time domain.The study of visualization and dimensionality reduction of the simulation sample set is carried out.On this basis,kernel SVM is trained and the evaluation method of the results of test set is studied.(2)The wear tests of large feed milling and cylindrical turning of TC4 titanium alloy are carried out.The wear laws are analyzed.The relationship between the main cutting force and the VB value of flank wear is established.The thresholds of the main cutting force of tools are determined.On this basis,the simulation samples are generated.Polynomial kernel SVM is trained.The ROC curve and AUC value are used to evaluate the performance of the classification models.The cutting force test samples are pretreated and classified.The results are evaluated by four indicators: Accuracy,Precision,Recall and F1-Measure.(3)The main platform is based on Matlab.LabVIEW is used to read the real-time cutting force data transmitted by the NI data acquisition system.Two main functions of model training and real-time data acquisition are compiled and the main interface of tool condition monitoring is designed.It integrates the key technologies in this paper,such as generation of simulation samples,visualization of data set,evaluation of models and so on.The friendliness for the users is enhanced.