作者: Luis Martí , Nayat Sanchez-Pi , José Manuel Molina , Ana Cristina Bicharra Garcia
DOI: 10.3390/S150202774
关键词:
摘要: Anomaly detection is the problem of finding patterns in data that do not conform to an a priori expected behavior. This related which some samples are distant, terms given metric, from rest dataset, where these anomalous indicated as outliers. has recently attracted attention research community, because its relevance real-world applications, like intrusion detection, fraud fault and system health monitoring, among many others. Anomalies themselves can have positive or negative nature, depending on their context interpretation. However, either case, it important for decision makers be able detect them order take appropriate actions. The petroleum industry one application contexts problems present. correct such types unusual information empowers maker with capacity act correctly avoid, react situations associated them. In context, heavy extraction machines pumping generation operations, turbomachines, intensively monitored by hundreds sensors each send measurements high frequency damage prevention. this paper, we propose combination yet another segmentation algorithm (YASA), novel fast quality algorithm, one-class support vector machine approach efficient anomaly turbomachines. proposal meant dealing aforementioned task cope lack labeled training data. As result, perform series empirical studies comparing our other methods applied benchmark real-life oil platform turbomachinery detection.