Event-driven tool condition monitoring methodology considering tool life prediction based on industrial internet

Tool condition monitoring (TCM) and remaining useful life (RUL) prediction is of great practical significance for any machining process to ensure machining quality and reduce the machine tool downtime. At the standpoint of workshop management, the current TCM has two drawbacks. (i) Continuously acquiring data without distinguishing the working states of the machine tool and the machining tasks will inevitably bring a large volume of unwanted signals, making difficulty for tool RUL prediction. (ii) The tool condition is independent of machining task, thus cannot provide further decision-making support for workshop scheduling and machining parameters optimization. Therefore, it is an important issue to consider various random events under the right machining tasks to trigger monitoring and RUL prediction just in time. This paper proposes an event-driven tool condition monitoring (EDTCM) methodology. The structure of EDTCM is designed based on the architecture of the Industrial Internet. Multi-source events are collected under the architecture, including MES events, machine tool events based on the OPC-UA (OPC Unified Architecture) standard, smart mobile terminal events, etc. The event-driven mode is designed to process these events such that the monitoring is triggered just in time. Then the Tool RUL is predicted online with the monitored sensor data based on the Bayesian method. A prototype system of EDTCM is developed and a case study is implemented to verify the feasibility of the proposed methodology. Our work promotes that the theories of TCM and tool RUL prediction deeply integrate with the real industrial practical applications.

Automated summary

Tool condition monitoring and remaining useful life prediction are crucial for ensuring machining quality and reducing machine downtime. However, current monitoring methods have limitations and require consideration of triggering monitoring and prediction under the right machining tasks. This paper proposes an event-driven tool condition monitoring method, combining multiple event sources and Bayesian methodology for online prediction of tool RUL, demonstrating the feasibility of the approach in a case study.