| 摘要: | 在現今資訊化的時代下，資料品質之良窳對於企業營運具有重大且關鍵的影響，在工業領域中，工業感測設備串流資料的品質所扮演的角色亦是如此。然而，現有品質控制之常見方式，主要著重於錯誤偵測與遺漏值偵測，即僅針對資料產出的結果做考慮，卻忽略了在即時串流資料上可能也會引發相異的品質問題。因此，本研究旨在以串流資料的性質以及工業感測設備應用的訴求之下，提出一全面性的即時串流資料品質綜效架構(a Multi-dimensional Data Quality framework for online Data Stream, 簡稱MDQDS)。同時，透過這一架構所建立的三個品質構面指標—準確性、完整性、一致性，以及其具體測試規則協助設備操作人員能即時地從不同面向來掌握設備資料的各種品質。在實務應用上，本研究依據所提出的架構，實作出一個應用程式MDQDSS(MDQDS System)以幫助設備操作人員能適時且有效的提升與維護工業感測設備資料之品質，並以工業蒸汽鍋爐感測設備的串流資料為例。並且，為了驗證本研究提出的架構與所實作出之系統的可行性，本研究亦將MDQDSS應用於紡織工業蒸汽鍋爐之感測設備資料中，來展示系統功能與使用情形最後，針對現有的運用與研究限制進行討論，並提出未來研究的可能方向。;In this information era, the quality of data is significant to enterprises. In industry, the quality of streaming data of industrial Sensing Equipment also plays an important role in decision-making. However, the existing methods of streaming data quality control focus on error detection and missing values detection, which aim at neglect the quality problems arises from the data processes. Therefore, this research aims to propose a comprehensive framework of the quality of streaming data of industrial Sensing Equipment, termed Multi-dimensional Data Quality framework for online Data Stream (MDQDS). This framework is based on online state, especially focusing on the demand of the system characteristics of industrial sensing equipment and streaming data. Simultaneously, by the three quality dimensions—accuracy, completeness, and consistency—built by this framework, MDQDS can help observers from different views to handle a variety of online streaming data qualities. In the practical application, according to the proposed framework, this research implements an application, MDQDSS (MDQDS System), to help observers to increase and maintain the quality of streaming data of industrial sensing equipment. Moreover, to verify the feasibility of the proposed framework and the implemented system, this research applies MDQDS to the industrial steam boiler sensing equipment in H textile factory to demonstrate the proposed system and the test situation. Finally, discussion and suggestions are presented for the existing application and this research proposes the probable direction of the future work. |
