Application of machine learning techniques in conceptual design

Abstract

設計中開發計算理論的一個挑戰是必須能支持計算機制的有效運用 ,這一機制允許從設計專家那兒或設計樣例中取得產生 ,累加和轉換的設計知識。而其中的一個方法是把機器學習機制綜合成基于知識的支持系統 ,以模擬設計過程初級階段 ,使設計成為一個增加和誘導學習的過程。模擬的需要產生于在不同的提取階段獲取 ,提煉和轉移設計知識的需求 ,從而使得能輕而易舉的熟練操作。在設計中 ,現有的知識產生于過去的設計解決方案 ,而過去的解決方案提供的反饋信息能更新和提高設計理論知識基礎。但是 ,沒有學習接受能力 ,設計系統不能反映設計家們在這一領域的成長經歷 ,也不能反映設計家們從以往設計案例中提取知識的能力。在此提出了方案設計和效力評價中的三種方法 One of the challenges of developing a computational theory of the design process is to support learning by the use of computational mechanisms that allow for the generation, accumulation and transformation of design knowledge learned from design experts, or design examples. One of the ways in which machine learning techniques can be integrated in a knowledge based design support system is to model the early stage of the design process as an incremental and inductive learning of design problem structures. The need for such a model arises from the need of capturing, refining, and transferring design knowledge at different levels of abstraction so that it can be manipulated easily. In design, the knowledge generated from past design solutions provides a feedback for modifying and enhancing the design knowledge base. Without a learning capability a design system is unable to reflect the designer’s growing experience in the field and designer’s ability to use knowledge abstracted from past design cases. In this paper, we present three applications of machine learning techniques in conceptual design and evaluate their effectiveness.