Research On Fault Diagnosis Of Sucker Rod Pumping System Based On Analysis Of Indicator Diagram

In oil exploration, a larger number of production equipment works in hostile environment and the underground condition of the pumping system is complex, which results in the pumping going wrong frequently. What is worse, the pumping are geographically dispersed, it is difficult for people to detection in time, so oil production and efficiency is quite low. Therefore, it is an urgent task to understand and master the working conditions of production systems in time, so as to achieve automatic control and scientific management. Currently, no matter at home or abroad, almost each method of the oil field pumping diagnosis depends on indicator diagram. This approach is mainly dominated by experienced experts. They analysis the indicator diagram for themselves, which inevitable leads into a variety of confounding factors in the diagnosis and makes the results cannot guarantee the accuracy and stability. In addition, the actual production of oil wells pumping operation is in cluster situation, manual analysis is not only a waste of human material and financial resources, but also affects the efficiency of production. This article makes a research in rod pumping oil well fault diagnostic, which is based on indicator diagram analysis, and advanced artificial intelligence methods. In this article we use to computers instead of manual operations in order not only to improve the accuracy of fault diagnosis, but also improve the efficiency of field operations and automation level of production. Thesis is as follows:First, we introduce the principle of pumping well system and the development of fault diagnosis system. After that the main cause of failure are described. The indicator diagram of the pumping well includes a wealth of information about working conditions, so we focus on the concept and the formation of the indicator diagram.Second, by analyzing the dynamics of the sucker rod, we set up a mathematical model of the diagnosis of conditions. This fault diagnosis model is mainly based on Gibbs wave equation. After solving the model by using the method of separating variables, the ground indicator diagram has been converted to down-hole pump indicator diagram. Since we eliminate the affect of the sucker rod deformation, vibration load and inertia load, the down-hole pump indicator diagram can be more truly reflected the working status of the pump. This part makes a good foundation of the subsequent feature extraction. Third, the characteristic extracting is obtained based on the geometric characteristic of working-pump’s graph, and the calculated model that select feature of failure of the subsurface equipment is setup.Third, by studying the theoretical of the, as well as learning from the method of’four point’, which is commonly used by the experts in fields, we presents a new feature extraction methods, which is based on the curve invariant moment. This method consists of two steps. First, the pump indicator diagram is divided into four parts according the formation. Second, extract seven invariant moments in each curve. All of the twenty-eight characteristic parameters can be described of the pump working condition.Finally, we use the SVM as the classifier. When choosing the penalty factor C and the RBF parameter g, we make a comparison between CV and PSO. We have a simulation experiment by using the LibSVM, and the result shows that the PSO algorithm to select out of the C and g as the parameter of SVM can improve the accuracy of classification.Theoretical analysis and simulation results show that the proposed method is feasible. It has important theoretical and practical value in research in fault diagnosis of sucker rod pumping system.