中文摘要：

水质分类的信息条件无法控制无量纲化造成的信息失真，所以，选择距离作为相似性度量的模糊识别不宜用于水质评价；隶属度作为样本点与各类间的相似性度量成为水质评价的最佳选择。但是，基于隶属度的模糊综合评判在算法上存在如下问题：①用“超标权”实现隶属度转换是错误的；②模糊隶属函数不规范。本文的改进方法是：①规范模糊隶属函数的拓扑空间结构、代数性质和标准化方法；②隶属度转换是一个数学计算过程，指标区分权重由指标值隶属度的熵确定。改进的模糊综合评判模型只在构造模糊隶属函数时用到先验知识。比较同案例的评价结果。显现改进前后的不同应用效果。

英文摘要：

Information will be faulty because water quality classification information conditions can＇ t control non-dimension. Thus, theoretically, fuzzy recognition that uses ＇distance＇ as similarity measurement between sampling points and classes isn＇ t suitable for evaluating water quality. The best choice for water quality evaluation is to use membership as similarity measurement between sampling points and classes. However, fuzzy evaluation using membership has some problems： ①Using super standard weight to realize the conversion of index value membership to sampling point membership is faulty;②Fuzzy membership function is not normal. These problems cause undesirable results. Our improved methods are： ①Normalizing topological space structure, algebraic properties and normalized method of the membership function;② Pointing out that the conversion of index value membership to sampling point membership is a pure mathe-matical problem and index classification weight is decided by index value membership＇s entropy and hasn＇ t domain features. Priori knowledge is only needed when the fuzzy membership function is constructed in the improved fuzzy evaluation. Comparing with the results for same cases, we can observe the difference between the original method and the improved method.