Study Of Pumping Well Failure Diagnosis Based On RBFNN

Mechanical oil is widely used in various methods of crude oil production,and sucker rodoil Pumping system dominates,thus sucker rod oil Pumping occpy an important position inthe whole crude oil production.but in practice oil wells could malfunction with the deepeningof oil jobs.that is straightly related to the normal operation of oil extraction,and causes oilcompanies great economic loss.on that account,this paper studies the condition monitoringand fault diagnosis.the main content of this paper is monitoring system、the feature vectors、model.The main contents are as follows.(1) Build wells pumping condition monitoring system. Describes the realization ofcondition monitoring system consists of various functional modules. For this topic researchthe fault diagnosis of pumping unit well done well pumping unit indicator diagram, such asdata collection Thereby establishing a network diagnostics used behind signs-fault sample.(2) For pumping wells dynamometer feature vector extraction. Fourier descriptorsselected contour shape has a strong ability to identify, The traditional shape of the Fouriertransform in order to ensure accuracy,many sampling points, leading to a large amount ofcomputation disadvantage, Presented to the straight line segment approximation continuouslychanging the line to replace the original Fourier transform of a discrete Fourier transform,This method not only reduces the distance between the boundary curve discrete errors caused,but also greatly reduces the amount of computation of Fourier transform, The entire pumpingwells diagnostic systems become more accurate, more efficient.(3) Through comparative analysis of RBF network based on gradient descent andgenetic characteristics of each algorithm, RBFNN proposed layered loop learning algorithmbased Fourier descriptors, through the simulation of nonlinear function approximation provesthat the proposed algorithm is accurate and effective.(4) Verify pumping fault diagnosis system.through a large number of field measurementindicator diagram of RBF network model the results indicate is validate,the RBF networkcan establish a good type of failure to make accurate judgments.