Research And Application Of On-line Monitoring Method Of Cutting Chatter On CNC Milling Machine

With the full implementation of "Made in China 2025",China’s industrial modernization process is advancing steadily,and intelligent manufacturing has become a current research hotspot.Cutting chatter during machining of machine tools is a key issue that restricts high-performance machining.It will cause poor surface finish of workpieces and accelerate tool wear,thereby reducing machine tool life,reliability and safety of machining operations,resulting in machining costs Improvement.In response to this problem,this paper proposes an online monitoring method for cutting chatter based on deep learning algorithms,and develops an embedded online monitoring system for cutting chatter based on a neural network processor.First,the principle of the online monitoring system for cutting chatter is studied.An online monitoring method for machine tool cutting chatter based on wavelet kernel convolutional neural network is proposed.This method uses a continuous wavelet convolutional layer to replace the first convolutional layer of the standard convolutional neural network,which can extract more effectively The pulse signal component related to flutter in the data.Compared with traditional deep learning methods,the recognition accuracy and speed of this algorithm are significantly improved.Then,according to the functional requirements of the online cutting chatter monitoring system,an overall software and hardware scheme was developed.The system uses Cortex-A series CPU as the core for modular hardware design,and is equipped with a neural network processor with a computing power of 3.0TOPS to accelerate calculations to improve the real-time performance of the entire system.The embedded Linux operating system is used as the core to carry out hierarchical embedded software design,and the upper computer software is written in the cloud to realize remote human-computer interaction.Finally,an experimental platform for milling chatter monitoring was constructed for experimental verification.A variable depth of cut experiment was designed to collect the vibration signals during the milling process,and build a milling chatter monitoring signal data set.Use this data set to train and test the convergence speed and accuracy of three sets of deep learning models of wavelet kernel convolutional neural network,standard CNN and Sin CNN in milling chatter monitoring.The wavelet kernel convolutional neural network has obvious advantages in the online monitoring application of milling chatter,with an average accuracy rate of 99.75% and an average recognition time of 0.35 seconds.The experimental results show that the embedded cutting chatter online monitoring system developed can well meet the current demand for online monitoring of machine tool cutting chatter.The upper computer software can view the status of machine tool chatter in real time,and the online cutting chatter monitoring system can realize remote control.Computer interaction makes highperformance intelligent manufacturing possible.