An On-line Fault Diagnosis Method for Gas Engine Using AP Clustering

The purpose of the paper is to implement an onlinefault diagnosis system for gas engine based on its datastream. The method proposed consists of three steps. First, weextract the features in a sliding window and reduce thedimension by PCA (Principal Components Analysis), second, therepresentative features are obtained by an AP (AffinityPropagation) clustering algorithm, finally, the fault patternrecognition is accomplished according to the distance betweenfeature samples. The diagnosis results for three common faultsshow that this method achieves better performance than existingmethods based on different clustering algorithms.