Study Of Sucker Rod Pump Indicator Card Recognition Method Based On Biomimetic Pattern Recognition

At present, a great many oil fields in the world using rod pump oil production method for mining.The equipment of the method has simple structure, convenient operation and lower comprehensive cost. However, that the work site environment of oil field is changing and the underground condition is complex, can easily leads to failure of pumping wells. Timely and accurate fault diagnosis can increase oil production and oil field effectively. Indicator diagram is an important characterization of oil well working status, and can reflect the various anomalies pumping the oil pump. The correct analysis of the indicator diagram can diagnose the oil well failure effectively. Therefore, accurate analysis to identify indicator diagram is of great significance. In this paper, sucker rod pump dynamometer card has been the object of study. The main context is as follows,1. This paper analyzes the work principle of the rod pumping wells, and combines the work process of rod pumping to discuss the formation of the indicator diagram in detail. Meanwhile, a typical indicator diagram sample library is established. A detailed analysis of the features and causes of various types of fault indicator diagram is made. This provides the basis for a follow-up indicator diagram characteristics extraction and fault diagnosis.2. Feature is extracted for the indicator diagram. Two different characteristic of the indicator diagram feature extraction algorithm is selected in this paper:the algorithm based on moment feature and the algorithm of frequency spectrum analysis method. The method has the characteristic of rotation, translation and scale invariance based on the indicator diagram of the moment feature extraction algorithm, which effectively removes its recognition effect of the indicator diagram rotation deformation. The feature vector based on the indicator diagram feature extraction algorithm of frequency spectrum analysis method is not sensitive to the translation and scale transformation of the indicator diagram and rotation can lead to the high frequency region energy increase. The advantage of this method can effectively exclude the impact of the sucker rod vibration indicator diagram.3. Classification and identification for the indicator diagram. This paper studies that biomimetic pattern recognition is applied to identify the effect of the sucker rod pump indicator diagram, and lists theoretic principle of biomimetic pattern recognition. Biomimetic pattern recognition is applied to the implementation of the indicator diagram recognition, and two different feature vector dynamometer graph recognition effect. As a comparison, the pattern recognition theory of support vector machine is also introduced. LIBSVM that developed and designed by National Taiwan University Professor Lin Zhiren is adopted for recognition, and then compare with the recognition effect of the biomimetic recognition. The experimental result shows that the recognition results of two methods are similar based on feature vector of moment feature algorithm,the correct recognition rate is above98%. The performance of biomimetic pattern recognition is better using frequency spectrum analysis method extracting the feature vector, the correct recognition rate can achieve96.5%.