Research On Tool Condition Monitoring And Life Prediction Under Complex Milling Trajectory

Effective tool wear and remaining useful life(RUL)prediction have important research value which can greatly improve the quality of workpiece and the efficiency of production.On-line prediction of tool wear is a pattern recognition process,which includes signal processing,feature extraction,and establishing a nonlinear relationship between features and tool wear.Currently,most research of tool wear prediction is being carried out only for straight cutting process,while the complex tool trajectory are widely used in the real machining process.When the trajectory of the tool changes,the sensor signal and the features are affected by both the tool trajectory and the tool wear,which poses a huge challenge for tool wear value prediction.Therefore,tool wear prediction under complex tool trajectories has become an urgent problem to be solved.What’s more,tool life prediction based on particle filter algorithm has the problem that the state equation cannot be adaptively adjusted,which makes the prediction of future tool wear produce deviation with true tool wear when the tool wear trend changes,and reduces accuracy of the tool life prediction.The traditional features from the cutting force are affected by the changing tool trajectories.So it cannot predict tool wear effectively.In this paper,the working conditions of complex milling trajectories are divided into two categories.The first is non-repetitive complex milling trajectories,and the second is repetitive complex milling trajectories.Signal rate change features and periodic features are carried out for two cutting conditions of complex trajectories.To verify the effectiveness of the new features for tool wear prediction under complex tool trajectory,two tool life cycle cutting experiment were carried out and cutting force were collected.Then the traditional features and new features are input BP neural network to predict the tool wear respectively.The results show that the prediction accurancy of tool wear based on new features is higher than the traditional features.What’s more,the traditional particle filter algorithm cannot adaptively adjust the state equation.A hybrid trend particle filter algorithm is proposed to solve this problem and improve the accuracy of life prediction.The research results show that the BP neural network based on the features proposed in the paper has a higher prediction accurancy under the complex tool trajectory.In addition,the proposed hybrid trend particle filter algorithm can accurately predict the tool life when the tool wear trend changes.