Neighborhood grid clustering and its application in fault diagnosis of satellite power system

Data-driven fault diagnosis, known to be simple and convenient, is more suitable for diagnosing the complicated spacecraft systems, e.g. the satellite power system. Nevertheless, it is difficult to extract the rules for diagnosing from unlabeled data. In this paper, a clustering approach based on neighborhood relationship and spatial grid partition is proposed to compensate for the above deficiency. In order to deal with the data-driven fault diagnosis issue, a diagnostic strategy is designed, which is a combination of the proposed clustering method and the entropy weight. Finally, multiple experiments, consisting of the artificial data clustering, comparison experiments on satellite data mining, and a case of fault diagnosis on satellite power system, are carried out to illustrate the versatility and superiority of the proposed method.