Data Study Group Final Report: The University of Sheffield Advanced Manufacturing Research Centre

Multi-sensor based intelligent machining process monitoring


With the advent of the new industrial era, modern manufacturing equipment is expected to be more flexible, sustainable and operative with minimum human interference which requires the supporting process monitoring system to be smarter and more intelligent, e.g., carrying out automated machine health checks using embedded sensors. This challenge focused on the investigation of using modern data science and AI techniques to analyse the multiple sensor measurements to monitor the status of the machining process.

Process monitoring allows the integrity of a machining operation to be gauged through the sensor measurement. This methodology can help to identify issues with the component, cutting tool, or machine tool before the component undergoes final inspection. The benefits of identifying issues in-process, rather than a final inspection, include limiting further damage to the component or machine tool and preventing additional components from being machined before identification of the issue.

Installing such a system can therefore provide significant savings to a manufacturer in terms of scrap, machine tool maintenance, and downtime. Techniques for process monitoring of machining operations through sensor signals are well-established in the literature. Many solutions utilising such techniques have now been commercialised and are available to the industry. However, most of the commercial systems are often based on static limits or signal trending.

To gain further insight into the machining process, the nature of the signal beyond simple level/amplitude is expected to be examined through time-series analysis techniques, such as fast Fourier transform, and spectral entropy calculations. The use of sensor fusion, i.e., evaluating combinations of sensor signals rather than in isolation, has also been demonstrated to be able to achieve more robust monitoring and fault identification. Finally, rather than simple trending or setting static limits, machine learning techniques could be employed in determining the optimal indicators and setting tolerances.

Citation information

Data Study Group team. (2022, September 13). Data Study Group Final Report: The University of Sheffield Advanced Manufacturing Research Centre. Zenodo.