Real-time Monitoring Of Cutting Tool Conditions In NC Machining Process Of Complex Structural Parts Based On Deep Learning

Real-time monitoring of tool conditions in CNC machining process plays an important role in ensuring the quality and efficiency of part processing and is a key technology in intelligent CNC machining.The tool condition monitoring based on signal like cutting force and vibration is an effective way in CNC machining.But the monitoring signal in NC machining process is not only related to the tool condition,but also affected by the geometry,process parameters and other factors of the workpiece.For complex structural parts with single-piece or small-batch production,the geometries and cutting parameters are continuously changing,which makes a great challenge for accurate recognition of cutting tool state.In order to address the issue mentioned above,this paper studies on the real-time monitoring of tool condition in CNC machining based on deep learning.The main innovations are as follows:(1)The influence mechanism of tool condition,part geometry and process on the tool condition monitoring signals is studied.The single factor experiment was designed to analyze the influence of tool condition,part geometry and process on the monitoring signals,and then the multi-factor experiment was designed to analyze the coupling relationship between different factors on monitoring signals.Experiments show that the influence of part geometry and process on the monitoring signals has a strong similarity with the influence of the tool condition.(2)Aiming at the problem that the traditional vector method destroys the high-dimensional structural relationship between geometry-process-monitoring information,a multi-dimensional input space expression based on tensor is proposed.A real-time correlation between geometry information,process information and monitoring information based on tool point coordinates is established,and then a high order tensor including information order,characteristic order and space order is constructed for the deep learning problem of tool condition monitoring.(3)Aiming at the problems of low level feature extraction and poor generalization ability in traditional machine learning in tool condition monitoring,a tool wear calculation model based on TDAE is studied.TAEs is stacked to extract the abstract features,while the deep learning model is established based on TDAE,which achieves the tool wear calculation.(4)Based on CATIA / CAA and LABVIEW platform,a real-time monitoring system of tool condition based on deep learning is developed and verified in the experiment.