Design And Study Of Oilfield Electric Submersible Pump Fault Diagnosis System

Since the1970s, the theory study of equipment fault diagnosis technology has becomemature and perfect. The main study of this paper is the fault diagnosis method for oilfieldelectric submersible pump’s linear motor. The main contents are as follows:Study on a fault diagnosis method based on power spectrum of the energy coefficients inthe Wavelet Packet for oilfield electric submersible pump’s linear motor. SinceLow-resolution of Fourier transform and its disadvantage of losing information, and WaveletPacket analysis is just the opposite with it. We combine the Wavelet Packet analysis withFourier transform, propose the fault diagnosis method based on power spectrum of the energycoefficients in the Wavelet Packet for oilfield electric submersible pump’s linear motor. Whena fault occurs, different vibration signal means different vibration energy. This methoddiagnoses faults by comparing the various layers’ energy distribution.Study on a fault diagnosis method based on Wavelet-BP neural network for oilfieldelectric submersible pump’s linear motor. Since the real meaning and form features ofWavelet packet analysis results, and the requirements of the input signal for BP neuralnetwork analysis, design a fault diagnosis method based on Wavelet-BP neural network foroilfield electric submersible pump’s linear motor. This method extracts energy feature vectorsof vibration signal by using Wavelet Packet. Without a clear case of hidden layer nodes,determine it in a range, then train it one by one, finally determine the number of hidden layerby comparing output error and training time. Input the energy feature vectors to train and testnetwork.Study on a fault diagnosis method based on Wavelet-Genetic BP neural network foroilfield electric submersible pump’s linear motor. The effect of wavelet packet analysis andBP neural network’s combination is good, but the convergence is slow and easy to fall into alocal minimum. Combine the Wavelet Packet-BP neural network and Genetic algorithm,proposed a fault diagnosis method based on Wavelet-Genetic BP neural network for oilfieldelectric submersible pump’s linear motor. This method increases the genetic algorithm tooptimize the connection weights and thresholds in the network. It can improve the speed ofnetwork convergence. And further improve the accuracy of diagnosis.