Multiparametric Vibration Diagnostics of Machine Tools Within a Digital Twin Framework Using Machine Learning
Andrey Kurkin, Yuri Kabaldin, Maksim Zhelonkin, Sergey Mancerov, Maksim Anosov, Dmitriy Shatagin
In the context of the digital transformation of industrial production, the need for intelligent maintenance and repair systems capable of ensuring reliable operation of machine-tool equipment without operator involvement is growing. This present study reviews the current state and future development of diagnostic and condition-monitoring systems for metalworking machine tools. A review of international standards and existing solutions from domestic and international vendors in vibration diagnostics has been conducted. Particular attention is paid to non-intrusive vibration diagnostics, digital twins, multiparametric analysis methods, and neural network approaches to failure prediction. The architecture of the developed system is presented. The concept of the system is developed in full compliance with Russian and international standards of vibration diagnostics. At its core, the comprehensive digital twin relies on machine learning methods. The proposed architecture is a predictive-maintenance system built on interconnected digital twin realizations: the dynamic machine passport of a unit, operational data, and a comprehensive digital twin of the machine-tool equipment. The potential of neuromorphic computing on a hardware platform is being considered as a promising element for local-condition classification and emergency protection. At the current development stage, the operating principle has been demonstrated along with the integration into the control loop. The system is now at the beginning of laboratory testing. It demonstrates capabilities for comprehensive assessment of the equipment’s technical condition based on multiparametric data, short-term vibration trend forecasting using a Long Short-Term Memory network, and state classification using a Multilayer Perceptron model. The results of the system’s testing on a turning machining center have been analyzed.