Research On Fault Diagnose Method Of Pipeline Leakage Based On Hybrid Model

With the rapid development of energy resources, the pipeline transportation has become widely used of oil and gas because of the security and economy. However, the leakage occurs frequently as the pipeline sabotage, natural aging equipment, geology, environment, natural disasters and other reasons. The pipeline leakage cause huge damage to the environment and people’s lives. Therefore,establishing an efficient, economic pipeline leak detection system is of great importance.At present, there are a lot of pipeline leak detection methods, with the neural network and expert system being used more and more widely in the leak detection. In this thesis, pipeline leak detection method based on hybrid model has been studied. The goal of the detection method is realizing the detection of pipeline leakage rapidly and accurately. The main research work as follows:Firstly, the typical pipeline leak detection methods have been discussed and their advantages and disadvantages are compared. Then, a two-stage detection method has been proposed. The two-stage detection method improves not only the detection sensitivity but also the response time. The method detects the abnormal signal of the pipeline based on neural network module for testing, then, the abnormal signal goes into the leakage detection module which based on the hybrid model. The high neural network and the accuracy of hybrid model both play important roles in the two-stage method.Secondly, the RBF neural network algorithm, the BP neural network algorithm, the retina neural network algorithm and the time sequence algorithm are studied respectively. And a large number of simulations have been done based on the actual data. Simulation results of the four methods have been compared. This thesis analyzes the response time and anti-jamming of each method. Then choose the retina neural network as the first level of abnormal signal detection algorithm.Finally, a hybrid model of the pipeline has been established.The hybrid model overcomes the difficulty to establish accurate mathematical model and the difficulty to overcome the restrictions on data model. Then, pipeline leak detection algorithm has been proposed based on hybrid model. The effectiveness of the hybrid model has been proved through the simulation.In this thesis, the theoretical research and simulation experiments has shown that the pipeline leak detection method based on hybrid model can diagnose the pipeline leakage fault timely and accurately, and has certain practical significance.This method will play a great role in pipeline leak detection with further improvement in theory and practice.