Application of fuzzy rough sets in patterns recognition for bearing faults diagnosis
Ren-fa Shen, Zheng Hai-qi, Yan‐Jie Qi, Kang Haiying · Zhendong yu chongji · 2010
A method of patterns recognition was presented based on fuzzy rough sets.A dynamic clustering algorithm and a anaiysis method of variance analysis were introduced to fuzzify the continuous condition attributes,and fuzzy membership functions were derived,avoiding information losing caused in the rough set discretization process.F test was introduced to judge the validity of clustering,which can overcome the disadvantage of determining artificially the class number of clustering.The fuzzy decision table obtained by use of attributes fuzzified was used to attributes reduction,and then clear and concise pattern rules were obtained.The applications in bearing faults detection show the proposed algorithm can raise the accuracy of recognition greatly in comparison with the use of normal rough sets.