Leak Identification In The Presence Of Fluid Noise Inside Pipeline And Distributed Monitor System

Pipeline leakage is one of the major factors that causes water wastage.Leakage not only means a significant loss of water,but increases the risk of the water pipes being contaminated by bacteria and pollutants,which seriously threatens the public's safety and health.Therefore,it is important for reducing economic losses and improving the quality of people's life to detect the pipe leakage in time.Leak detection technologies based on acoustic signal processing are simple and efficient,easy to operate,and are widely used in practices.However,the interference around the pipeline will severely affect the accuracy of the leak identification result.In all kinds of noises,the fluid noise inside the pipeline caused by the change of the pipe structure hasn't been studied clearly in the mechanism of production,propagation rules,etc.To improve the accuracy of leak identification,this thesis studies the leakage identification method under the existence of the noise inside the pipeline.(1)Analysis the generation mechanism of the noise inside pipeline.We investigate the fluid field in the mutation position of the pipeline by computational fluid dynamics(CFD).The results show that there are some regular low pressure regions existing in the mutation position of the pipeline and the noise is generated mainly from the cavitation sound and turbulence in this place.The noise inside pipeline also shows a characteristic with regularity under the influence of regular regions inside noise pipe,which offers theoretical support for features extracting and leak identification.(2)Research on leak identification technology under the influence of noise inside pipeline.The researched result of the noise generation mechanism shows that the difference in regularity between the leakage signal and the noise inside pipeline is obvious.And the regularity degree represents the randomness of the signal.Due to the influence of random turbulence and vapor,the leak signal shows more randomness than noise inside pipeline.According to the generation mechanism of the noise signal,we adopt the autocorrelation function to characterize the regularity of these two signals.In order to make the regularity of the signal more discrimination,the envelope of the autocorrelation function is extracted as our analysis objects and then decomposed by SVD to get the quantized value of signal regularity.At last,the BP neural network approach is developed as a classifier,which uses the extracted regularity feature as the network inputs,then we can identify pipeline leaks through the trained BP neural.(3)Development of the distributed leak detection system.Combining our previous experience in leak detection with the actual working mode of the equipment,we use STM32F103 microprocessor as control unit.For low power consumption,portability,wireless transmission,and controllable parameters,we design the system that can monitor pipe leakage in the field for a long time.Then we develop the PC software of signal processing system based on C# and MATLAB mixed programming and uses the former leak identification technology.Finally we test the whole system in field to prove that this system has a good low power consumption performance and high leak detection accuracy of 86%.