稀疏规则库条件下的模糊推理方法

Fuzzy reasoning method based on sparse fuzzy rule base

当模糊规则库是稀疏型时,利用Kóczy线性插值推理方法不能保证推理结论的正规性和凸性,为了解决这一问题,石岩曾提出了插值推理方法的推理条件,当满足这些条件时利用Kóczy线性插值推理方法得到的推理结论也满足正规性和凸性;但是这些条件却限制了模糊推理系统的应用,而且如果多次推理中在同一输入点遇到稀疏情况,必须进行相同的计算才能得到正确的推理结果,这样增加了系统的计算量,降低了系统的速度和效率。因此提出了一种新的稀疏模糊推理方法,不仅能够简单的给出正确的推理结果,还能在相应的位置增加规则,提高规则库的紧密程度。

When fuzzy rule base is sparse, by Kóczy＇s linear interpolative reasoning method, the reasoning conclusions are not always normal and convex fuzzy sets. Then, Shi Yan have proposed some conditions for this method, however, these conditions limit the application of fuzzy reasoning system. Also, the more times of fuzzy reasoning, at the same rule the more sparse＇s states, the repeat calculations are needed, so it must increases the complexity of calculation and decrease the reasoning speed and efficiency of the fuzzy reasoning system. In this paper, a new fuzzy reasoning method based on the sparse fuzzy rule bases is proposed, by this reasoning method, the reasoning consequence can be simply obtained, a new rule can be generated and added to the rule base, and the fuzzy rule base＇s compactibility is increased.