- 學位論文
設計並建構一個以知識本體論為基之模糊推論系統 -以半導體製程診斷為例
Design and Implementation of Ontology-Based Fuzzy Inference System: An Application Case of Semiconductor Manufacturing Process Diagnosis
若您是本文的作者，可授權文章由華藝線上圖書館中協助推廣。
摘要
在這篇論文中，提出了一個以知識本體論為基礎的模糊推論（Ontology-based fuzzy inference）架構，簡稱為「OFI」，並以半導體製程診斷作為應用問題。其設計目標在於期望將所有的知識含蓋於OFI 運行流程之中：著眼於不斷的演進具彈性、可擴充性；其他系統可以分享與重複使用；對於不同的應用，可依需求隨時自行組裝使用。 OFI架構分為兩個部份：integrated OFI ontology與configurable OFI system。integrated OFI ontology編輯可分享的Web Ontology Language（OWL），來描述包含於OFI process中的Knowledge。相較於其他的ontology設計，integrated OFI ontology不僅僅只設計了事實與規則的ontology，同時也設計了流程與方法的ontology來表示推論的步驟。此外，integrated OFI ontology還具備了表達模糊理論的能力，並能有效的表示不確定性的知識。 configurable OFI system是用JAVA程式語言所開發而成的。在系統的運作階段（run-time phase），實際操作與控制integrated OFI ontology的運行。 借由OWL概念與JAVA物件之間的轉換設計（OWL-concept-to-JAVA-object transfer），在系統的運作階段，integrated OFI ontology中所描述之各個不同的Knowledge與JAVA物件之間可以被動態的轉換與存取。 基於OFI架構，本論文設計實作了一具max-min fuzzy logic運算的OFI系統，並於文中描述其詳細的過程與步驟，並且以半導體的製程診斷為例，示範OFI系統的運作，此外，也引用了兩個評估的案例來驗證OFI架構的設計目標。
並列摘要
This thesis proposes an ontology-based fuzzy inference (OFI) architecture using the semiconductor manufacturing process diagnosis as a driving problem. The design objective is to make all the knowledge involved in OFI process: flexible and scalable with respect to continuous evolutions, shareable and reusable by other systems, and configurable for adapting to various applications. The OFI architecture is composed of two parts: integrated OFI ontology and configurable OFI system. The integrated OFI ontology adopts the sharable Web Ontology Language (OWL) to describe all the knowledge involved in OFI processes. As compared to other designs of ontology, the integrated OFI ontology not only designs fact and rules ontology, but also designs flow ontology and function ontology to present inference procedure. Moreover, the integrated OFI ontology has the capability to model fuzzy logic theory and present vagueness knowledge. The configurable OFI system is constructed by JAVA programming to manipulate the integrated OFI ontology at the system run-time phase. An OWL-concept-to-JAVA-object transformer is designed and implemented so that it can access and dynamically transfer various knowledge sources described in integrated OFI ontology to corresponding JAVA objects during the run-time phase. On top of the OFI architecture, the procedures to design and implement a real OFI system with the mechanism of max-min fuzzy logic inference are proposed and demonstrated by its applications to semiconductor manufacturing process diagnoses. Two evolutions of the max-min fuzzy logic inference mechanism are then designed and implemented to validate the original design objectives.
參考文獻
1. 關銘，”以OWL DL 及SWRL 為基礎建置推論雛型系統-以大學排課問題為
2. CLIPSTab Documentation, <
3. Fuzzy logic Toolbox,
401-456. 1991
7. Baader F., McGuinness D.L., Nardi D., “The Description Logic Handbook: Theory,