Study On Pipeline Leak Detection Method Based On Fuzzy Neural Network

As the transport tools for conveying important materials such as oil, natural gas and water etc, long distance pipelines have become absolutely necessary parts of current transportation industry. However, the economic loss and environments pollution are very serious caused by pipeline leakage frequently occurring which is generated by inevitable aging, corrosion and artificial damage. So the real-time detection for pipeline leakage has become the worldwide important research subject. There are a lot of methods until now. In general, three kinds can be classified that is the method based on information, the method based on mathematical model and the method based on knowledge. Among them, the method based on knowledge is universally focused mainly in the use of neural network and expert systems in resent years. But these three methods have some certain deficiency. In this paper, fuzzy-neural network is attempted to be used in the detection for pipeline leakage. The main research works are as following:1. By studying on mechanism of pipeline leakage, the pressure and flow signal inside pipelines are surely chosen as the basis for analyzing. At the same time, by studying the relevant documents and using the relevant experiments data for reference to establish fuzzy-diagnosis-rules and learning samples.2. Using improved BP algorithm to train the network parameters. Simulation experiments are done to test the results of network-diagnosis reliability. The experiments results also testify that the method is effective.3. According to the disadvantages that BP algorithm is easy to be influenced by initial weights to fall into local minimum, slow convergence speed and causing oscillating effect and so on, genetic algorithm is added to optimize the initial weights of BP algorithm. Simulation results show by combining the two kinds of algorithm, the accuracy of detection and estimation for pipeline leakage is improved.4. Considering that the false symptom has influence on the fuzzy-diagnosis-rules, the guiding plan for network-tailoring is proposed to improve the rapidity of the leakage detection.5. Conbining the characteristics of several finished leakage-detection and algorithm modules of location method, the systematic software using to detect the pipeline leakage is realized generally and the simplified statement of sever methods to detect the leakage is given.