摘要
以模糊c-means算法对实例初步分类,以分类正确率和隶属度构造两级目标函数对实例进行挖掘获得个体分类效果最佳的敏感信息源、主矢量、类别原型和距离计算方法。以模糊综合决策算法对实例敏感主矢量进一步挖掘获得信息融合分类效果最佳的主矢量集合。同时从实例库中挖掘出模糊综合决策权重分配系数。将c-means算法的意义扩展应用于信息融合诊断。另外还论述了两种综合诊断方法,模式识别与模糊综合决策联合的综合诊断、扩展的模糊c-means算法与综合决策联合的二级优化综合诊断。最后以列车滚动轴承故障实例库的挖掘为例阐述了原理的应用。
The fault cases are first clustered by c-means algorithm. Then two objective functions are built from clustering correct ratio to all fault cases and the grade of membership of all fault cases. The sensitive information sources, the vectors of principal components, clusters' prototypes and the corresponding method of calculating distance are gotten. By means of fuzzy synthetic decision, sensitive principal vectors are further mined and the set of principal vectors are obtained which do the best in the correct result of clustering fault cases with fusion of information sources. The distribution of weight coefficients is mined from fault cases by using optimization method of external penalty function. The c-means algorithm is developed and the developed c-means is used in fuzzy synthetic diagnosis of information fusion. There are two other methods of synthetic diagnosis supposed. One is the combination of pattern recognition and fuzzy synthetic decision. The other is the combination of developed c-means algorithm and fuzzy synthetic decision. It is demonstrated how to use the methods by mining the database of fault cases of rolling bearings of freight train.
出处
《模式识别与人工智能》
EI
CSCD
北大核心
2003年第4期486-493,共8页
Pattern Recognition and Artificial Intelligence
基金
国家"九五"攀登计划(No.PD9521908Z1)
河南省自然科学基金(No.994044000)
