Research On The Classification And Identification Of Indicator Diagram Based On Freeman Chain-Code Eigenvalues

Oil exploration is such a high-risk industry that the production and development of oil industry presupposes the production safety on oil field. Oil well is usually divided into the flowing well and non-flowing well. However, the flowing well may also stop flowing because of long-term oil exploitation. To ensure the oil output, the non-flowing wells and flowing wells which lose the flowing capacity should employ the mechanical oil extraction method. The current commonly used extraction method at home and abroad is based on sucker-rod pumping system.Analysis and diagnosis of the oil pumping machine indicator diagram is a main way to know the working condition of sucker-rod pumping system directly and rapidly. The traditional indicator diagram diagnosis is mainly by manual operation, the diagnosis of which is easily affected by objective and subjective factors, such as the experience and technological level of oil staff. Nowadays more and more oil fields begin to employ the computer intelligent diagnosis technology.The classification and identification based on artificial neural network requires the accurate extraction of eigenvalue for the indicator diagram, the recognition efficiency and reliability of which has the direct bearing on the quality of eigenvalue. Many of the indicator diagram eigenvalue extraction method needs a large amount of calculation, which is inconsistent with the real-time requirements of oil field. In order to solve this problem, this thesis proposes Freeman chain code to express the features of indicator diagram and further researches into its recognition. Firstly, the thesis introduces the structure of sucker-rod pumping system and the principle of the indicator diagram, and describes the forming process of the indicator diagram. Then the thesis provides an introduction of the related concepts of Freeman chain code and the extraction methods of Freeman chain code eigenvalue, including the selection method of indicator diagram boundary points and the pretreatment method of chain code eigenvalue. Finally, the thesis plays emphasis on the introduction of the concept of artificial neural networks, especially the algorithm and the structure of BP network. According to the actual needs of the current project, the thesis designs a BP neural network model from the perspective of the input layer, the output layer and the hidden layer, and finally conducts the simulation verification amid MATLAB. The experimental results show that Freeman chain code eigenvalue can effectively classify indicator diagram for various typical working conditions and this method has a certain theoretical and practical significance.