Study Of Monitoring Methods For End Milling Tool Breakage Based On Multi-sensor Fusion

In the milling process, tool breakage has become the main reason of tool failure due to high hardness and of great brittleness of the brittle material. As a result, it causes low surface processing quality and production efficiency, high machine downtime and increase production cost. Therefore, milling breakage monitoring has become the key issue to be solved urgently in production. With the development of data process technology and monitoring techniques based on multi-sensor, the accuracy of tool breakage detection has improved. But both of these technologies make the tool condition system more complicated and dimension of feature extraction increased. How to reduce the dimensions caused by the utilizing of multi-sensor and select a best feature subset must be solved in tool condition monitoring based on multi-sensor fusion.Signals in the end milling tool breakage monitoring process extraction methods were studied. The power spectrum analysis and wavelet packet decomposition method were used to analyze the monitoring signals in the frequency domain and the time-frequency domain respectively. Then source signals in the frequency domain and the time-frequency domain respectively for feature extraction were selected. The features were extracted from the selected source signals as the features for tool breakage detection. An approach which is the GA-SVM Wrapper method was proposed to features selection extracted in the frequency domain and the time-frequency domain. The features selection method was adopted by the wrapper method combined feature selection based on genetic algorithm with parameters optimization of the support vector machine(SVM).A HASS VF-2 CNC was used as the test bed and a n end milling tool breakage monitoring system was developed by monitoring the vibration signals in two mutually perpendicular direction and the machine sound signal. A 33 orthogonal experiment under diverse cutting conditions including spindle speed, feed rate and axial depth of cut and the tool breakage experiment were designed for acquiring sensor signals while the new tool and the tool with breakage cutting. Based on the monitoring signals, the statistical features of the spindle vibration signals and the machine sound signal. Then, the features among the statistical features were selected based on the GA-SVM Wrapper method to get the best feature subset and the best parameters of the SVM. At last, another orthogonal experiment under other cutting parameter levels in the range of the former one to verify the classification accuracy of the tool breakage detection model built based on the best feature set and the best parameters of SVM.An end milling tool breakage monitoring system based on multi-sensor fusion was designed and developed using Lab VIEW and MATLAB mixed programming because of the advantages of Lab VIEW which is easy to be developed the program graphical interface and MATLAB which is power for math calculation. This system can display the real-time signals, transmits the data, save the data and the tool breakage detection. The platform can be used in the end milling tool breakage monitoring process based on the multi-sensor fusion.