Research On Embedded Condition Monitoring System Of Precision Machine Tool

The wide application of large-size aspheric optical elements in large-scale optical astronomical telescopes,space-borne cameras,laser fusion and other fields puts forward stricter requirements on their processing efficiency and processing quality.Ultra-precision grinding machine tools,as an important processing method for largesize aspheric optical components,are vital to the efficient and high-quality processing of components.Condition monitoring and fault diagnosis of ultra-precision grinding machine tools are important guarantees for its long-term stable and efficient machining.The condition of the machine tool,its grinding wheels,guide rails and other parts has a significant impact on ultra-precision machining.Online monitoring of these parts to ensure that the grinding wheels are in good condition is an important guarantee for the quality and stability of ultra-precision machining.Therefore,this article builds a set of machine tool intelligent monitoring system based on the embedded system around the needs of precision machine tool processing and monitoring,which monitors and collects the machine tool’s acoustic emission,vibration,temperature and other signals.In response to the needs of state recognition of parts such as grinding wheels,signal processing and analysis methods and pattern recognition methods are studied.The specific work is as follows:(1)Aiming at the monitoring requirements and characteristics of ultra-precision grinding machine tools,based on the NI sbRIO hardware platform,an embedded monitoring system is designed for condition monitoring and data acquisition of ultraprecision grinding machines.The sensors and acquisition modules of the system were selected according to the collection and monitoring requirements,and then the placement of the sensors was selected according to the structure and design characteristics of the machine tool.According to system requirements and functional planning,FPGA software modules for realizing signal acquisition and monitoring of vibration,temperature,and acoustic emission,and real-time processing program modules for real-time acquisition,analysis and diagnosis have been written to provide software support for pattern recognition algorithms.Corresponding real-time software modules and diagnostic server scripts are written for the deployment of algorithms such as data-driven diagnostic models.(2)Research on signal processing analysis and pattern recognition algorithms,and propose a diagnosis method combining simple feature parameter indicators to identify simple faults in real time and data-driven neural network algorithms to further identify complex states.Pattern recognition algorithms such as multi-layer neural network and cyclic neural network and their implementation are studied.(3)A monitoring system test experiment was carried out on a precision grinding machine tool,and signals such as vibration and acoustic emission were collected to verify the rationality and correctness of the embedded monitoring system design.Through the analysis of the experimental data,the time domain parameter index and frequency domain frequency point of the vibration signal to distinguish the fault are determined.(4)The wavelet packet energy coefficient method is used to extract the sensitive frequency band characteristics of the acoustic emission signal with respect to the wear of the grinding wheel,and use it as an input to train the model to identify the wear state of the grinding wheel,and compare the LSTM network under different hyperparameters with the traditional Recognition effect of feedforward neural network to obtain the best model.The results show that the network model based on LSTM can accurately identify the wear state of the grinding wheel,and the error is reduced by 80%compared with the traditional multi-layer neural network model.