Leakage Diagnosis And Neural Network-based Location Research Of Complex Pipe Networks

With the continuous development of the city,the scope of the centralized pipe network is gradually expanded,the structure of the pipe network tends to be complex,and the leakage situation is changeable,which puts forward higher requirements for leakage diagnosis and positioning technology.The leakage of the pipe network will not only cause a large amount of fluid loss in the pipe and waste resources,but also affect the normal life and safety of the surrounding residents.At a result,it is very important to quickly diagnose and locate the leakage of the pipeline network in real life.However,the existing pipe network leak diagnosis and location technology has low precision and cannot fully meet the needs.In order to improve the leak diagnosis and location accuracy of complex pipe network and reduce the economic loss of pipe network,this paper studies the leak diagnosis and location technology of pipe network.Firstly,a complex pipe network test bench was built to collect the leakage pressure signal of the pipe network.After performing wavelet threshold denoising and short-time Fourier transformation on the pressure signal,the pressure wave leakage transmission law was explored from the complex cepstrum,and the Based on this,a leakage identification method of complex pipeline network is raised.It can be shown that the leakage of the pipeline network identification rate reaches 93% by using this method.Meanwhile,this method is compared with methods such as empirical mode decomposition,which once again proves the superiority of this method,which can be further applied to the actual pipeline network.Secondly,on the basis of identifying the leakage in the pipeline network,the leakage point is further located based on the neural network technology.According to the experimental pipe network structure,the hydraulic model of pipe network leakage is established and checked.The results show that the error between the simulated value and the experimental value of the corresponding monitoring point is less than 1 m when the leakage point corresponding to the experiment leaks,which fully satisfies the check standard of the pipe network model.At the same time,according to this model,the pressure of the monitoring points under different leakage conditions and different leakage points were simulated,and the leakage eigenvalue library of the pipeline network were further established through the calculation of the relationship between the pressure difference before and after the leakage of the five monitoring points,and the leakage position and amount.Finally,the deep belief network model is trained by using 1305 sets of data in the pipeline network leakage eigenvalue database,and the initialization parameters,topology,iterative steps,learning rate of the DBN model are discussed to obtain the best model.Analysis of the prediction results of the DBN model shows that the prediction accuracy of the DBN leak location model has reached more than 96.2%.Compared with the other neural networks,the DBN model has the best prediction effect.In the meantime,The predicted positioning accuracy of the experimental pipe network reaches more than 95.35% by using the DBN model,which further proves that the method has a good prediction effect for the actual complex pipe network.