Fuzzy knowledge representation and processing is a very important topic in artificial intelligence research, the key of the fuzzy knowledge inference is the knowledge matching. Via our investigation on kinds of conventional methods in fuzzy knowledge matching, we propose a knowledge matching method based on variable weights IDM (Inscribed Diameter Matching). Through our comparative analysis, Variable weights IDM is able to overcome some drawbacks exist in the traditional methods. We also give a fusion model which can improve efficiency and accuracy of knowledge inference. The experimental results verify the practicability and availability of our method.