Abstract: An anthropopathic clustering algorithm based on the data set's intensity and neighborhood was proposed in this paper. The algorithm imitates human's thinking procedure of classification, and it obtains the given number of centers and accomplishes the clustering by searching the whole data set globally and restrictedly, incorporating and optimizing the initial centers. The experiment using the data from the hardware emulational platform of PHM indicates that the algorithm has the accuracy on local and secondary clustering about 5 times larger than Fuzzy C Means algorithm.