The Pumping Unit Fault Diagnosis System Based On The Analysis Of BP Neural Network And The Indicator Card

The sucker-rod pumping accounts for a large proportion of the mechanical oil extraction. The diagnosis of sucker-rod pumping and fault treatment are delayed because of the oil wells’ dispersed location in the wild and the complicated circumstance. All these make the diagnosis of sucker-rod pumping being a difficult problem in the oil production field for a long time. Therefore, it is very important to diagnose the fault of the pumping units timely and provide operation advices reliably, which means lot to for improving the production efficiency and economical operation in oil field.The indicator card of pumping unit can be used to judge the work condition of the pumping unit, this paper discusses the technology of pumping unit diagnosis and the fault severity level judgment based on indicator card.First, the development of the diagnosis system and the operation principle of the pumping unit are introduced. And the generation of the indicator card, the sort of the indicator card and the characteristic of the indicator card are discussed particularly.Second, Thesis introduces artificial neural network in detail, especially the elementary theory and application of BP network, and researches on the construction principle of BP network architecture, the processing method and training way of network samples. Take the common failure of pumping units for instance, three-layer11-21-14network architecture for failure diagnosis is established. For the shortcomings of BP neural network, thesis discusses several kinds of improved BP algorithms, and analyzes the effects of these improved algorithms applied in failure diagnosis.Finally, pumping unit diagnosis system is established based on the technology of Labview and Matlab mixed programming. The whole system including indicator card identify module, fault alarm module and diagnosing module are totally introduced.Experiments prove that the diagnosis system is adaptive, stable and high reliability. And the judgment of the fault severity level needs further confirm in practical application.