Research On The Fault Diagnosis Of The Submersible Pump Units Based On Analysis Of Vibration Signals

Electric submersible pump (ESP) unit is the main device to extract oil in the world presently. Its security and efficiency is important to ensure the high and stable yields of the oil field. ESP is sealed, the structure of which is relatively complex. ESP goes wrong easily in operation, and many faults of ESP will lead to fierce vibration. By means of analyzing the vibration signals of ESP and collecting the correlative characteristics of the fault, the type of some faults can be determined preliminarily.By gathering and analyzing the well-head vibration signals of ESP test wells, ESP's vibration feature can be extracted,and some faults of ESP can be diagnosed. The main content of the thesis is concluded:1. To detect the vibration of ESP, a vibration signal acquisition system, which is based on USB7360AF (digital acquisition module) and piezoelectric acceleration sensor, is devised to collect the ESP's well-head vibration acceleration signals. The particular charge amplifier is designed independently, with which the weak charge signal can be amplified efficiently.2. Make a research on the wavelet analysis theory and wavelet transform in the vibration signal processing applications. Wavelet analysis is a kind of the time-frequency analysis, which is provided with the property of the time-frequency localization, has the characteristics of the multiresolution analysis and overcomes the limitations of Fourier transform which is lacking in the time-frequency local analysis ability. Through the wavelet transform and the frequency-domain analysis of the ESP's vibration signals, the fault traits can be extracted. Many groups of detected vibration signals are analyzed in many manners, including the time-domain analysis, the wavelet analysis and the frequency-domain analysis, thereby, the effective method of extracting the features is received about the impeller eccentric wear, the impeller tip screenout and the bearing eccentric wear by comparison and contrast.3. Have a research on the application of the neural network in the fault diagnosis. Directed to the polymorphism and the multilevel of the ESP's fault, the backpropagation neural network is created on the basis of the Gaussian-Newton method. By training and studying the designed neural network, three kinds of ordinary faults are diagnosed effectively.In conclusion, it is feasible and effective to extract fault feature of ESP, by analyzing the vibration acceleration signals of ESP's well-head, extracting the fault features and building nonlinear model to carry out fault diagnosis. And the work of the paper provides the helpful thinking for the fault diagnosis technology of ESP.