Knowledge Representation and Reasoning Methodology based on CBR Algorithm for Modular Fixture Design

應用於模組化夾具設計之以案例為基礎推理法則內的知識表示與推理方法

孫書煌(Shu-Huang Sun) ； 陳家豪(Jahau-Lewis Chen)

《中國機械工程學刊》 28卷6期 (2007/12) Pp. 593-604

https://doi.org/10.29979/JCSME.200712.0006

摘要

比起以規則形式的知識庫系統而言，以案例為基礎的知識庫系統是一種更先進的人工智慧輔助系統。以案例為基礎的知識庫系統是透過尋找以往曾解決過的類似案例並加以修正，來對現有的問題提供解決方案。在本研究中即應用了以案例為基礎的法則提出一個著重在工件定位的模組化夾具設計方法。在進行設計時，本方法先以自由度(DOF)為依據，由案例庫中搜尋出與現有設計問題最類似的案例，然後再加以修正以求得最後結果。依據本方法，未來將可以建立一個電腦輔助模組化夾具設計系統。此系統將含有三個子系統，資料庫子系統中儲存多個模組化夾具模組，每個模組各對應其所具有之功能。知識庫子系統儲存工件的定位知識。案例庫子系統儲存以往的設計案例。本方法亦使用MOP知識表示法以便能將複雜的知識條理化地儲存進系統中，最後並以一個例子說明本方法的推理過程。 本方法主要著重在案例的評估及修正上，此部分是案例為基礎推理法則中最重要也困難的部分。在工件定位評估中，幾何外形辨識技術是主要的關鍵，由於這部分超出本研究的範圍，將不會在本文中作詳細討論，因此目前本方法僅能對外形較簡單的工件進行處理。

並列摘要

CBR algorithm provides a better knowledge transfer and explanation than rule-based inference. It solves new problems by adapting solutions that were used to solve old problems. Based on CBR algorithm, a methodology applied in modular fixture design and focus on workpiece locating is proposed in this study. A similar solution can be retrieved from past experiences. Evaluation is applied for this retrieved case by checking degrees of freedom (DOF) to determine whether it is satisfactory for a new problem and some components would be replaced if it is not. According to this methodology, a computer-aided modular fixture design system can be established in future. In the system, three sub-bases would be included. Data base stores many function structures that are assembled by modular components to complete some functions. Knowledge base stores the qualitative knowledge that is required in considering the location of the workpieces. Case base stores previous successful design cases that can be applied to develop a new solution. MOP-based memory technique is applied to organize these complex data, knowledge and case base. A demonstrated example is finally provided in this study to illustrate how this methodology works.
This methodology principally focuses on inference process of case evaluation and modification. This is the most important and difficult issue on CBR algorithm. In the evaluation of workpiece locating, geometry recognition play a critical role. Feature recognition is beyond this study and then too detail discussion about that would not be given here. For this reason, the methodology can handle simple geometry workpiece only presently.

並列關鍵字

CBR ； Modular fixture ； MOP ； Knowledge representation