Research On Cutting Tool Wear Monitoring Technology Based On Two-stage Feature Fusion

In the process of metal cutting,tool wear is inevitable due to the friction and extrusion between workpiece,chip and tool.As the executive end of the machining field,the tool is an important part to improve the production efficiency and ensure the machining quality.The tool wear status is one of the key factors affecting the reliability and stability of the machining system.In the process of machining,the excessive wear of the tool will lead to a sharp increase in cutting force,vibration and even system noise,and the surface roughness of the parts will increase accordingly,which will ultimately affect the machining accuracy and service life.Premature tool change cannot make full use of tool service life and increase production costs;Too late tool change is not conducive to ensuring the processing quality and production efficiency,even causing losses.Therefore,it is very important to establish an efficient tool wear monitoring system to monitor the tool wear status in the machining process in real time and accurately.Turning is the most typical cutting method.Studying the changing law of cutting force and the mechanism of tool wear in the process of turning difficult-to-machine materials can lay the foundation for building the monitoring system of cutting tool wear.The results show that the cutting depth has the greatest influence on the three-dimensional cutting force,and the main cutting force and the cutting depth resistance are significantly affected.The main wear forms of tools are hard spot wear and adhesive wear,and hard spot wear runs through the whole process of tool wear;At different wear stages,the tool has different degrees of diffusion wear and oxidation wear;When the tool is damaged,the machine tool has obvious vibration,and the tool has the phenomena of edge collapse,coating peeling,crack damage and chip accumulation.In order to realize real-time and accurate monitoring of tool wear during milling,this paper selects cutting force and vibration acceleration as monitoring signals,and proposes a prediction model of milling tool wear based on least squares support vector machine(LSSVM).Because the optimization objectives involved in the research are related to complex multi-extremum problems,common optimization algorithms are prone to fall into local optimization and difficult to find the global optimal solution.In this paper,the PSO-ALS(particle swarm optimization with adaptive learning strategy)algorithm is selected to find the optimal LSSVM core parameters,simplify the optimization problem and strengthen the group diversity and global search ability.On the basis of common signal processing methods,this paper proposes a unique signal feature extraction and fusion method:multi-domain feature extraction,feature reduction based on principal component analysis(PCA),and feature dimension elevation based on stacked multilayer denoising autoencoders(SMDAE).In order to minimize the influence of redundant information in the original signal features and ensure the prediction accuracy of the model,a new PCA+SMDAE two-stage feature fusion method is proposed.In this paper,12 sets of milling experiments are carried out,and five models such as PSO-ALS LSSVM are used to predict milling cutter wear.The advantages of PSO-ALS algorithm are proved by comparing the prediction performance of different models on unknown samples.The research results show that the proposed PSO-ALS LSSVM model still has the best prediction performance when there are many noises in the original signal features,and the two-stage feature fusion method can effectively improve the prediction accuracy of the model.The research results of this paper provide theoretical guidance for monitoring tool wear in practical production.