Research On Detection And Location Of Leakage In Crude Oil Pipeline

The leakages of crude oil pipelines frequently happen as a result of natural aging corrosion and oil-stolen orifice. The research on pipeline leak detection has become an important subject in the pipeline transportation industry in the world. Although there have been many kinds of detecting methods, there are still some defects in the existing detection systems which result in low sensitivity, reliability and accuracy. It is necessary and difficult to find and confirm the accurate location of the leak point of crude oil pipelines.Through analyzing the stress-acoustic signals which can reflect the operating conditions of oil pipelines, the thesis illustrated the principles of the generation and transmission of leak signals, then analyzed the stress-acoustic signals both in time domain and frequency domain which are helpful to preliminarily understand the characteristics of the signals in the cases of normal, leakage and regulating pumps operating conditions. Then we chose the generalized RBF neural network to recognize the patterns of the three operating conditions. We use the normalized method to preprocess the signals, selected the characteristic parameters of signals to train the generalized RBF neural network, then applied the well trained neural network to recognize the patterns of signals. What is more, the signals are denoised by wavelet analysis. Base on the characteristics of the leakage signals and the theory of singularity, we use the wavelet transform modulus maximum method to get the singularity point of leak signals.The simulation results demonstrated that the generalized RBF neural network can recognize the signals of normal, leakage and regulating pumps operating conditions. The denoising method of wavelet analysis successfully improved the SNR of the signals. At the same time, we can get the singularity point of signal accurately by wavelet transform modulus maximum method with scale function attenuation.