An ART2-based pumping unit fault diagnosis

Since the environments the pumping units working in, is always severe and the dowhole condition is very complex, so it´s difficult to diagnose the faults of the pumping units. The indicator diagram can be a good indication of the pumping unit´s work conditions, it is also the main basis of fault analysis. Judging the pumping unit´s faults using intelligent algorithms is also the demand of digital oilfield. The key is how to extract the indicator diagrams´ features and what kind of intelligent algorithm is applicable. In this article, we first extract seven moment invariants of the typical faults´ indicator diagrams, then we use ART2 neural network to classify these moment invariants. In which way, we can recognize pumping unit faults efficiently.