A Study On Fault Intelligent Diagnosis And Applications Of Rod Pumping System

In recent years, rod pumping oil dominates the main position in the oil industry. The fault rate is very high because of complex conditions and bad environment under the pumping oil. Although at present appeared the synthesis diagnosis system, it can’t complete the diagnosis efficiently because they only depends on the expert system or the neural network, So this paper introduces a fault diagnosis method based on the wavelet packet and neural nerwork.This article regards the rod pumping oil well in fault diagnosis system as the research object, and analyses the working principle of rod punmping oil in-depth. Firstly, it simulates vibration signals of gear base on the theory of digital signal analysis and wavelet packet decomposition, to verify the feasibility of using the wavelet packet to extrat the energy eigenvector. Secondly, it normalized pretreatment the collect data by using poor regular method, and extrated the energy characteristic eigenvector by using wavelet packet decomposition methods. Thirdly, the article introduces several models of neural network and compares the performance of RBF network with BP network in function into fields, and the result indicates that RBF network has higher accuracy in function approximation ability. Finally, it validates the failures diagnosis system of pumping well. It validates the established model of RBF network by combining wavelet packet decomposition and RBF network, and takes the measurement damping-displacement figure data in certain oilfield as an example to analyze the six fault type, and it obtains good results.Based on the theoretical research effort, the fault diagnosis system of pumping oil well is designed, and it combines ACCESS database with MATLAB to realize effcient diagnosis. The experiment shows that the method combined wavelet packet with RBF network is feasible, and the results are correct and reliable.