Many tribosystems are designed for long-lasting operation. Therefore, a reliable automated condition monitoring is necessary to identify critical operation in time. On the example of porous journal bearings, experts from AC²T have developed a new Machine Learning approach for the classification of a tribosystem’s operational states, which was recently published in the article “Classification of operational states in porous journal bearings using a semi-supervised multi-sensor Machine Learning approach” in the peer-reviewed journal Tribology International.

The presented Machine Learning algorithm is able to differentiate between the four states ‘run-in’, ‘steady’, ‘vibration’ and ‘critical’. Critical operation indicating an impending catastrophic failure was predicted up to 50 hours before a conventional, threshold-based alarm was raised. Such an algorithm may assist a predictive maintenance strategy, as it can be integrated into an online monitoring tool that allows the classification of the operational state in real time.

For the first time at AC²T, a 3D laser Doppler vibrometer was applied to reveal the main vibration modes of the test setup. Detailed knowledge of the dynamic behavior is crucial for a deeper understanding of the origin of a tribosystem’s operational states.

For more information see https://doi.org/10.1016/j.triboint.2023.108464

Free download until May 25, 2023: https://authors.elsevier.com/c/1gsmY,6t04LOoL