Abstract
建構在QSIM的定性推理架構之上, 我們提出一種新的定性推理表達方法. 可以減化在定性模擬中出現的無限分枝問題. 我們以QSIM的限制條件將物 理上的直覺及知識陳述出來. 三種方法如下:時間反轉推理是一種具有物 理上意義的後向推理方法. 由時間反轉觀念, 模擬可以是正向的或後向 的. 錯誤的分枝可以減少.為處理沒有外界能量流入的系統, 我們提出能 量損失比率觀念. 由能量損失比率觀念可以決定何種行為可較可能.為處 理有外界能量流入的系統, 我們提出主動模型轉換觀念. 由一組定性微分 方程式轉換到另一組. 主動模型轉換觀念可以幫助推理何種行為的確可 能. Based on the qualitative simulation under QSIM's framework, we propose a new formulation to reduce the intractable branching problem, and enhance the capability of qualitative simulation. We formulate the physical intuition and knowledge in terms of QSIM's constraints. Three methods are proposed. Time-reverse reasoning is a backward reasoning with physical significance. With the time-reverse concept, simulation can be performed backward in addition to forward, so the number of spurious branches could be reduced. In problems of dealing with systems without energy coming from outside world, we propose a concept of energy losing rate, to decide which behavior is preferred. This helps to reason about which behavior is the most likely outcome. In problems of dealing with systems with energy coming from outside world. We propose a concept of enforced model transition from one set of QDEs to another. This helps to reason about how some behavior can become possible in the systems.