| 摘要: | 最近幾年全球製造業興起推展工業4.0的變革，其核心願景乃是藉由導入先進資訊與智慧科技(例如物聯網、虛實整合系統、雲端運算、大數據分析等)來建立智慧工廠。在各種智慧技術中，預測保養(PdM, Predictive Maintenance)可監測機台之狀態及預測機台關鍵組件之剩餘壽命(RUL)，然後工程師可據以在機台失效前進行維修保養。因此，與傳統定期保養相比，預測保養可以為製造業帶來許多節省成本之效益(例如降低維修保養之頻率、減少因維修帶來的生產損失、減少備用零件等)。目前已經有很多預測保養的文獻，但主要著重在發展目標機台的預測保養演算法，缺乏一個系統化且方便的方式來建置可支援智慧製造之預測保養服務平台。因此，本論文參考雲製造的觀念，利用雲端運算技術與雲端機器學習工具，設計與實作一個基於虛實整合技術的預測保養雲端服務平台，可讓使用者方便地以系統化的方式為目標裝置或設備建置預測保養雲端服務。最後，本論文透過為一個太陽能廠機台關鍵組件建置預測保養雲端服務為研究案例，驗證本預測保養服務平台的便利性與可行性。本論文成果可作為發展支援智慧製造之預測保養雲端服務平台之參考。
In recent years, manufacturing enterprises around the world energetically promote the revolution of Industry 4.0 whose core vision is to build smart factories by leveraging advanced information and intelligent technologies, such as IoT (Internet of Things), CC (Cloud Computing), BDA (Big Data Analysis), and CPS (Cyber-Physical Systems). Among intelligent technologies, PdM (Predictive Maintenance) can monitor the states of production equipment and machines and further predict the RUL (Remaining Useful Life) of their key components so that engineers can accordingly conduct maintenance before the equipment and machines fail. Thus, PdM can bring many cost-effective benefits to manufacturing enterprises, such as reducing the frequency of maintenance, reducing production losses due to maintenance, reducing spare parts, etc., compared to traditional regular maintenance (i.e., preventive maintenance). There exist many PdM-related works, but most of them mainly focus on developing algorithms of prediction maintenance for key components of the targeted machines. It lacks a convenient and systematic approach to building a PdM service platform that can support the operation of intelligent manufacturing. Thus, by referring to the concept of cloud manufacturing, this thesis utilizes cloud computing technology and cloud machine learning tools to design and implement a predictive maintenance service platform based on cyber-physical technology (called PdMSP), which allows users to construct PdM cloud services for targeted devices or machines in a convenient and systematic manner. Finally, this study conducts an illustrative case study that constructs a PdM cloud service for a key component of a piece of solar cell manufacturing equipment to validate the convenience and feasibility of the proposed PdMSP. The results of this thesis can serve as a reference for building PdM cloud service platforms that can support intelligent manufacturing. |
