A Fault Diagnosis System For The Pump Oil Wells Based On Neural Network

In the opening up oilfields, it's an important task in petroleum industry to predict and diagnose the fault under the oilfields, to understand and master working-pump's condition of oilfields, and to realize automatic monitor and scientific management in an exploit oil well. Because it is difficult that the faults are diagnosed automatically, a fault diagnosis expert system is designed to utilize pattern recognition of neural networks.First, the working principle of the pumping wells is discussed. The main fault of pumping wells and the reasons are discussed. Because much information is included in the working-pump's graph of oil well, so it focuses study the concept, the principle and the type of the oil wells working-pump's graph and extracts the geometric parameters as a neural network input signal.Second, the neural network is discussed, and its fault diagnosis in the application is introduced. BP network, as the most commonly used neural network, is discussed. Its topology and algorithm are detailed analysis, and address its deficiencies, given additional momentum method and adaptive learning method which both improved algorithm.Finally, the improved algorithm for BP pumping wells fault diagnosis. MATLAB using a neural network based on the pumping wells fault diagnosis system, and the system design and implementation process.Theoretical analysis and experiments show the improved BP algorithm for the pumping wells fault diagnosis system's samples has a better ability to identify with a certain practicality.