Research On Sample Completion Method Of End Milling Tool Condition Monitoring Based On Milling Process Simulation

The tool is one of the most critical parts in the milling process,and it is also the most vulnerable part.Its.wear state will directly affect the surface quality of the processed parts and the normal operation of the machine tool.Therefore,it is particularly important to obtain timely and accurate tool wear status information and establish an effective tool condition monitoring(TCM)method.Artificial intelligence(AI)diagnosis models such as support vector machine(SVM),random forest(RF),decision tree(DT),artificial neural networks(ANN),ect..have beenwidely utilized in TCM system.However,due to the complexity and dynamic randomness of the machining process,as well as the expensive experimental cost and long experimental time,the tool wear data set obtained is prone to sample missing and sample insufficient,which has a direct impact on the model diagnosis results.In order to solve the above-mentioned problems,the end milling tool in the CNC milling machine is taken as the research object in this paper,which carries out the research of the sample completion method based on the end milling process simulation and the generative adversarial network(GAN).Firstly,the milling process is simulated based on the finite element simulation software Deform.The material model,constitutive model,friction model,heat conduction model and other key technologies in the simulation of milling process are studied to ensure the accuracy of simulation.Secondly,a milling experiment based on cutting force was set up,and the correlation between the cutting force data of the simulation model and the experimental data under the normal state of the tool was compared.The results show that the simulation data matches the experimental data relatively well.Aiming at the problem of sample missing in milling experiments,a TCM method based on finite element simulation is proposed.A low-cost,easy-to-implement finite element simulation method is adopted to obtain a complete simulation milling tool wear sample.According to the measured wear value data in the experiment,the missing tool wear state value is complemented by interpolation to make up for the missing wear state sample in the experimental sample.The simulation signal under different tool wear is compared with the experimental signal,which proves the availability of the simulation signal.Finally,support vector machines,random forests,decision trees and generalized regression neural networks are used to test the experimental data and simulation data to verify the usability and effectiveness of the simulation samples.Aiming at the problem of sample insufficient in milling experiments,the 'infinite generation' capability of the GAN is used to expand the simulation and experimental samples to generate a large number of synthetic samples.Combine the experimental and the simulation data to obtain the complete tool wear status sample.Finally,the complete samples are tested on different diagnostic models.The results show that the method proposed in this paper can significantly improve the classification accuracy of TCM.