Unplanned maintenance activities are one of the biggest cost drivers within manufacturing. Late maintenance can cause damage to the products or production components, or even downtime. To prevent this, manufacturers are developing strategies for timely replacement of components or maintenance operations of (crucial) parts in their manufacturing environment.
One of the ways to prevent unplanned maintenance activities is to utilise machine sensor data for continuously monitoring the machine status and health. By timely failure anticipation, downtime can be reduced, and maintenance activities can be optimised and scheduled at a convenient time. However, some critical components require external sensors to monitor their health status for maintenance activities. This is the core of the VITALS project, which is an acronym for Visualisation of Information Trends and AnaLytics of data for production Systems. This is conducted by utilising external sensors and advanced analytics tools for decision-making, as a standalone monitoring system. This AMP-subsided project serves as an initial step towards predictive maintenance for the involved companies.
The VITALS consortium consists of Watson Marlow Bredel from Delden, CM Data from Oldenzaal, and High Tech Maintenance from Hengelo, under the support of the FIP-AM@UT. Watson Marlow Bredel’s core activity is manufacturing hose pumps for thick liquids, such as concrete and oil. These products are mainly applied in the mining and construction industry and food production. The production of the housing of these pump systems involves rotary machining operations with higher cutting forces. Rolling element-bearing, housed in a rotary machining table, plays a crucial role in handling such greater forces that require smart maintenance practices to avoid critical failure. CM Data, specialised in sensor technology and IoT platforms, plays an active role in sensor integration and data acquisition from the production component. High Tech Maintenance is an organisation that specialises in maintenance operations in the heavy-duty industry.
The project aims to successfully link sensor data to failure modes and failure mechanisms, to monitor the rolling element-bearing health status. Using data analytics techniques, amongst other things, a machine learning algorithm will be developed to detect the healthy and failure modes. The final objective is to build a demonstrator that showcases the successful extraction of sensor data from a production machine, to implement a machine learning algorithm to monitor the status and events of the machine and visualise this process on a dashboard. This can lead to extended research toward accurate prediction of a Remaining Useful Life (RUL). Eventually, this method will be extended to other engineering systems.
