Feature engineering –based machine learning models for operational state recognition of rotating machines

作者

Jukka Junttila¹ , Ville Lämsä¹ , Leonardo Espinosa Leal² , Anssi Sillanpää³

1. VTT Technical Research Centre of Finland
2. Arcada University of Applied Sciences
3. Wärtsilä Finland Oy

会议/活动

AI DAY 2022, November 2023 (Espoo, Finland)

海报摘要

This poster sums up the work done in a master’s thesis and two subsequent publications. Our aim is to provide data-based models for operational state recognition and detection of abnormal operation of a gas engine generating set in near real-time. Our idea is to use computationally light features extracted from measured mechanical vibration data that are sensitive to the changes in the operational state to build classification and novelty detection models. The classifier model groups different states of normal operation, different power output levels in our case, and the novelty detector can detect abnormal operation at a specific power output level. As a result, we managed to build fast and accurate two-step state recognition model by combining the classification and novelty detection models. We also studied simulated (FEM) mechanical vibration responses. A comparative validation showed that the simulated responses resemble the measured ones statistically but revealed significant absolute differences.

关键词

Operational state recognition, Feature engineering, Internal combustion engine, Mechanical vibration, Simulation, Classification

研究领域

Mechanical Engineering, Computer and Information Science

参考文献

暂无数据

基金

暂无数据

补充材料

1. Masters thesis related to poster   Download

附加信息

利益冲突

No competing interests were disclosed.

数据可用性声明

Data sharing not applicable to this poster as no datasets were generated or analyzed during the current study.

知识共享许可协议

Copyright © 2023 Junttila et al. This is an open access work distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.