FEATURE EXTRACTION AND ESTABLISHMENT BASED ON PUMPING UNIT WORKING CONDITIONS AND GLOBAL FAULT IDENTIFICATION

In response to the challenging task of fault feature classification in the current working conditions of pumping units，which results in poor adaptability and low recognition rate of the established diagnosis model. The dynamometer card was pretreated by mathematical morphology through the analysis of the motion state of the pumping unit valve and the sucker rod. Then, two methods of obtaining valve opening and closing points and load variation characteristics were proposed, and 54 new features of global faults of pumping units were extracted, and the characteristic database of working conditions of the pumping unit was established.Finally, the algorithm of decision tree, logistic regression and support vector machine was used to verify that the feature database has good classification effect under different working conditions. The characteristic indexes of different fault conditions were evaluated, and the private rule database of each working condition was obtained. The research results demonstrate that the proposed features in this study are capable of effectively identifying comprehensive faults in pumping units, exhibiting a high level of recognition accuracy.