Unsupervised clustering for fault diagnosis

Abstract

We develop an unsupervised clustering method for the classification of transient data. A fuzzy-based technique is employed to measure the similarity among the transients; a spectral clustering technique, embedding the unsupervised Fuzzy C-Means (FMC) algorithm, is applied to the matrix of similarity values so that the clusters are formed by patterns most similar to each other. The performance of the proposed technique is tested with respect to a case study with data artificially generated.