Tool Wear Condition Monitoring Of High Speed Milling Based On Sparse Decomposition And Support Vector Machine

In the recent years,high speed milling has been widely applied in many fields for its high efficiency,high precision and high surface quality.In high speed milling,the milling tool works discontinuously with an ultra-high speed,leading to its quick wear or damage,which affects the machining accuracy and product quality directly and will cause serious accident when the machine and the work-piece are damaged.Therefore,the online monitoring of tool wear condition in high speed milling is very important.In this paper,a new method of fault diagnosis is proposed by taking advantage of advanced sensor technology on the basis of sparse representation and pattern recognition to achieve real-time monitoring of tool wear state and improve the safety of production system as a result.The work of this paper mainly includes the following aspects:(1)Various scientific methods,research progress of scientific research institutions and scholars in the field of high-speed milling for tool wear condition monitoring are learned and summarized,and the mechanism of tool wear and its classification problems are introduced,which lay a solid theoretical foundation for the development of the subject.(2)Based on compression sensing and sparse representation theory,a dual basis pursuit optimization algorithm is constructed and solved combining morphological component analysis and Split Variable Augmented Lagrangian Shrinkage Algorithm.In this way,the impulse components and harmonic components of the signal can be separated successful.Then simulation analysis is conducted to analyze and verify the performance of this algorithm.(3)A high-speed milling experiment platform is established and its construction and the usage of each module is introduced.Various sensor signals are collected and stored.According to its sparsity in the frequency domain and other characteristics,vibration signal is decomposed into impulse components and harmonic components.Then two features,including the impulse density and the amplitude ratio of higher harmonic frequency and the basic frequency are constructed,extracted,and the correlation analysis between tool wear and the features are performed to explore the physical meaning and the practicality of the features.(4)A multi-class support vector machine classifier is constructed.Then the classifier is trained by inputting the feature samples into it,thus the classifier can identify tool wear state by multiple features.New experimental data arc input into the classifier to predict tool wear status,in order to verify the performance of the classifier.