Analysis And Research On Cutting Characteristics Of CNC Machine Tools Based On Machine Learning

According to the spirit of "Made in China 2025",basic manufacturing equipment and aviation equipment are the pillar industries of the country.And the technical level of high-end CNC machine tool equipment can reflect the industrialization level of a country.Therefore,the real-time monitoring is very necessary for CNC machine tool.It is wildly used that by comparing the original health data and real-time monitoring data in the database.Through this method,the machine tool can give warning clearly before failure,and improve production efficiency and processing quality.This research taking a certain type of CNC milling machine as the research target,the real-time data acquisition and analysis of the spindle system working characteristics are carried out.The relationship model between different cutting workpiece-materials and spindle system characteristic data in a healthy state is established for future research and application.The main contents of this paper are as follows:1.By analyzing the theory of cutting period,the dynamic model of cutting system is established,and the factors influence the cutting period is analyzed.Four cutting elements including cutting speed,feed speed,back feed,radial depth and the cutting system influence the vibration the most.Based on the cutting theory,the cutting experiment is established.2.By acquiring the spindle vibration signal of cutting carbon steel,6061-T6 aluminum alloy and HT250 gray cast.And the spindle vibration of no-load state of CNC machine tool is acquired too.In order to improve the signal to noise ratio,by using wavelet de-noising,the signal to noise ratio is improved.After using WMPE and t-sne the Intrinsic Mode Function can used as eigenvector.3.In order to build the relationship model between spindle cutting vibration characteristics and various materials.Three optimization algorithms including Adaboost,PSO,GWO are used to optimize BP neural network and SVM.The result shows that the GWO algorithm has higher recognition rate than Ada Boost algorithm and PSO algorithm.4.Through the experimental analysis,the relationship model between the vibration characteristic information and the material type of the machine tool in the healthy state is established with high accuracy,which provides certain basic data and basic model for the follow-up research of machine tool health diagnosis.