ABSTRACT: To solve the problem that traditional Kernel Fuzzy C-Means Algorithm (KFCM) is very sensitive to outliers and noises in the training set, a novel Kernel fuzzy C-Means Algorithm based on distribution density around samples and maximum variance between clusters method is proposed in this paper. In the proposed method, the value of distribution density around samples is used as weight values according to the feature of sample distributing to overcome the shortcomings of KFCM’s sensitivity to noises and outliers. The maximum variance between clusters methods is applied to segment the sample’s distribution density vector, whose segmentation results are used to as the initial centers of the proposed KFCM algorithm, which overcomes the problems of KFCM sensitivity to initial values. The proposed method can not only solve the problems of traditional KFCM’s sensitivity to noises and outliers and sensitivity to initial values, but also can be applied to analyze samples’ contribution to clustering performance. The experimental results with various real data sets illustrate the effectiveness of the proposed algorithm. The proposed method is applied to fault diagnosis field which outperforms traditional cluster methods.