整合式與FUN式案例基推理於鋼結構大樓初步斷面設計模式之發展

Abstract

案例基推理是人工智慧領域的一種新觀念，開始發展於1980年代初期，而在1990年被應用在工程設計的領域內，因為它可累積專家的經驗，並且以案例的方式存放在案例庫中，以便在需要時，能快速的取出並加以應用，對於注重以經驗來解決問題的工程設計領域，為一最佳的輔助設計工具，且因又是一個很容易建立的系統，對使用者來說，可以在很短的時間內，即找到需要的答案。但目前已發展完成的案例庫系統，絕大部份僅能找出最近案例，無法在找到最近案例後，再將取出案例作調整，因此，無法完成整個系統的完整流程。本文即針對此項缺點，嘗試整合目前的已發展成功的案例基推理套裝程式Esteem，與由模糊組合型式發展出之案例基推理模式，結合而成一整合式之案例基推理模型，並以鋼結構初步斷面設計為例，檢驗此系統在鋼結構初步斷面設計之適用性。最後將同樣問題，取類神經綱路下的非監督式學習系統UFN（此系統亦為在案例基推理觀念下發出之程式）再作一次，將其結果與由整合式案例基推理所得結果作比對，以了解兩系統之優劣點跟實用性。並以系統誤差量的大小，來判斷案例基推理模式在輔助鋼結構初步斷面設計工作的可靠度，最後發現以此兩種模式求得結果對應用於工程設計問題上，是可以被接受的。

Case-based reasoning (CBR) is the one of means of facilitating the development of computer program that attempts to solve problems by directly accessing the case base. The approach relies on the explicit symbolic representation of a case base based on experience. With a given case base, case-based reasoning uses a representation involving specific episodes of problem solving not only to solve a new problem, but also to learn how to solve the new problem. Based on the approach of case-based reasoning, several related research involved engineering problem solving have been studied. However, most CBR systems can only retrieve the nearest cases and can not adapt these cases to generate solution, so it is not really finish the episodes of the whole system float of CBR. In this work, a novel model is developed via integrating Esteem CBR with fuzzy synthesis approach and applied to the problem of steel structural preliminary design. Finally, a neural network based CBR, UFN, is also used to solve the problem. The results shown that the solutions generated through the novel CBR and UFN system are acceptable in engineering view.