Machine learning for predictive maintenance diagnosis with motor fault spectrum [應 用 機 器 學 習 於預 測 維 護 診 斷 之馬 達 故 障 頻 譜 研究]
Journal
Journal of Taiwan Society of Naval Architects and Marine Engineers
Journal Volume
38
Journal Issue
3月4日
Pages
157-163
Date Issued
2020
Author(s)
Abstract
In recent years, the increase of green energy awareness has risen with the policy to “20% of green electricity in 2025.” Many green energy companies have sprung up. The establishment of the Taiwan Wind Industry Association in 2019 further highlights the development potential and importance of offshore wind energy in Taiwan. As the maintenance cost of equipment in the offshore wind farm is higher than that of the onshore wind farm, the fault prediction and health management of offshore wind turbines have become important research. In recent years, Artificial Intelligence (AI) has been widely used in Prognostic and Health Management (PHM). By equipment monitoring with the Industrial Internet of Things (IIoT) architecture, huge amounts of data collected can be transmitted through Supervisory Control and Data Acquisition (SCADA) system. How to apply this big data for PHM has a decisive influence on the maintenance cost of equipment in the offshore wind farm. The goal of our study is to import and test some machine learning algorithms, including decision tree classifier, K-nearest neighbor classifier, logistic regression classifier, and support vector machine, for Predictive Maintenance (PdM). The experimental results show that the accuracy is over 95%. It provides a preliminary discussion on machine learning algorithms for a possible reference to PdM analysis of wind turbines in the future. ? 2019 Taiwan Society of Naval Architects and Marine Engineers. All rights reserved.
Subjects
Costs; Data acquisition; Decision trees; Electric utilities; Industrial internet of things (IIoT); Learning algorithms; Learning systems; Logistic regression; Nearest neighbor search; Offshore oil well production; Offshore power plants; Onshore wind farms; Predictive analytics; Predictive maintenance; Support vector machines; Support vector regression; Wind turbines; Decision tree classifiers; Development potential; Equipment monitoring; K-nearest neighbor classifier; Logistic regression classifier; Off-shore wind energy; Prognostic and health management; Supervisory control and dataacquisition systems (SCADA); Offshore wind farms
Type
journal article