預測維護技術的實施方法、現況與發展趨勢

The Implementation Methods, Current Status, and Future Trends of Predictive Maintenance Techniques

吳鴻材 ； 偕睿仁 ； 王俊傑

《機械工業雜誌》 425期 (2018/08) Pp. 76-82

摘要

預測維護技術(Predictive Maintenance)因提供了設備維護所需的狀態診斷及狀態預測資訊，因此被視為是可取代傳統消耗維護(Run to Failure)、預防維護(Preventative Maintenance)模式的有效方法。然而，預測維護技術的開發，需透過訊號感測處理、設備聯網、設備健康狀態監測及預測、機器學習(Machine Learning)演算法、機械故障鑑別等跨領域技術的整合，因此大幅提高了業界瞭解技術、掌握技術、導入技術的困難度。有鑑於此，本文將從預測維護技術的實施方法出發，為讀者簡介整體技術的流程。而後將以工研院機械所所開發之預兆診斷系統(Prognosis Monitoring System, PMS)為例來說明技術現況，並針對其系統功能及相關技術做介紹。最後，則將盤點目前技術缺口及未來發展走向，期使讀者對預測維護技術能有基本的瞭解。

關鍵字

預測維護 ； 機器學習 ； 預兆診斷系統

並列摘要

Predictive maintenance techniques, which are designed to help determine the condition of equipment in order to predict when maintenance should be performed, are considered as the alternative of conventional run to failure and preventative maintenance methods. However, the successful development of such techniques heavily relies on multidisciplinary skills such as signal sensing, communication, prognostic and health management, machine learning, and machinery fault diagnostics; thus is considered burdensome for industrial implementations. This article aims to mitigate the gap by firstly giving a brief introduction about the whole implementation process, followed by the presentation of system function and capabilities of the prognosis monitoring system developed by ITRI. Finally, the insufficiencies and future development trends of predictive maintenance techniques are discussed, and therefore hope the readers can get an in-depth understanding of this useful and effective technique.

並列關鍵字

Predictive maintenance ； Machine learning ； Prognosis monitoring system