Theory And Application Of Fault Diagnosis Based On Analysis Of Vibration Signals

With the development of oilfield equipment and the needs of sustained and stable production of crude oil, the importance for the fault diagnosis technology of beam pumping unit is more and more obvious, particularly for the vibration signals fault diagnosis of the beam pumping unit. For the beam pumping unit equipment with multi-source, complex and sudden fault, the traditional diagnosis technology and theoretical methods often show many limitations. In order to overcome these shortcomings, this thesis focused on the application of Recurrent Neural Network (RNN) theory in the pumping vibration fault diagnosis technology and conducted a systematic and thorough research. The main research works are as follows:Firstly, based on the study of lots of literature and previous research results, this paper proposed a kind of the pumping unit vibration fault diagnosis method based on RNN. The paper analyzed the deficiencies of traditional fault diagnosis methods of pumping unit, proposed a kind of method of fault monitoring and diagnosis based on the vibration signals of motor and reducer by making use of neural network, which could be a useful complement to traditional methods of pumping unit fault diagnosis. It also pointed out the recurrent neural network theory used in pumping unit vibration fault diagnosis has a great value and development potential.Secondly, with the pumping unit mechanical vibration signal data acquisition and processing technology designed in this paper, four major characteristics of the vibration frequency vector of the motor and reducer are extracted. Using the basic internal structure of the recurrent neural network, the learning efficiency of the algorithm is greatly improved. The paper has studied the means of pumping unit vibration fault diagnosis technology which uses recurrent neural network theory, and pointed out that it is the new idea using recurrent neural network theory in pumping unit fault diagnosis technology.Finally, the thesis has designed a method of neural network learning and diagnosis applied to the pumping unit vibration fault diagnosis, and programmed procedures using MATLAB6.5 programming language, achieved the fault diagnosis processing for the motor and reducer in the beam pumping unit. The results show that neural network method not only remedied the disadvantages of traditional pumping unit fault diagnosis method, but also improved the efficiency of pumping unit fault diagnosis.