Abnormal events recognition and classification for pipeline monitoring systems based on vibration analysis and artificial neural networks

Abstract

Pipelines have become the principal means of oil and gas transportation. However, pipeline leakage takes place due to some natural or artificial damages, which may cause loss of life and properties along with the environmental pollutions. A new pipeline detection and pre-warning system based on distributed optical fiber sensor is proposed, and the hardware has been accomplished. Now, its following key problem is how to recognize and classify the abnormal events, such as oil stealing, construction, artificial excavation, motor work, and train passing. This paper involves a study on this and proposes a solution method. First, original vibration signal is pre-processed and segmented according to threshold of energy within a narrower bandwidth. Then, event features in time and frequency domain are analyzed through statistical analysis and short-time Fourier transform (STFT). The energy coefficients at some bandwidth can distinguish different type of abnormal events, which are chosen as feature vectors. At classification, abnormal events are first divided into discrete and continuous events with single classifier, which can decrease classified event sets and improve recognition accuracy. Then, BP artificial neural network is applied to identify the type of abnormal events. Finally, proposed method will be verified with actual collection data sets.