| The urban water supply network is an important municipal area related to the health of the people and the national economy.However,at present,some domestic cities still have the problems of old water supply facilities,serious aging of the network,slow update rate,and even the location of the pipeline network has been lost.Leakage accidents occur from time to time,causing a large amount of clean water resources to be wasted,which means huge losses to the national economy and hidden safety hazards such as road surface collapse.At present,there are many types of leak detection methods used at home and abroad.Among them,the acoustic leak detection method represented by the listening method is more commonly used,but it relies heavily on the work experience of the leak detection staff and consumes a lot of manpower and material resources.This paper try to find a new leak location method based on collecting leaky acoustic signals on the ground,inputting the eigenvalues extracted from the signals into the BP neural network model to calculate the probability of leaking at the measuring point,and establishing the probability model of leaking and the positioning method of multiple measuring points to achieve the purpose of identifying the leaking.The causes of pipeline leakage and the principle of generating acoustic signals are analyzed,and the eigenvalues of leakage acoustic signals that can distinguish noise are screened out;the changes of six kinds of eigenvalues before and after the formation of stable erosion pits are recorded through experimental observations,it turns out that Power spectrum,EMD and MFCC values change a lot,but no significant effect on the three eigenvalues of Lempel-Ziv complexity(LZC),approximate entropy and short-term zero-crossing rate.The numerical simulation of the flow field of the eroded pits formed in the experiment was conducted by using the computational fluid dynamics software FLUENT,and the effects of different leak orientations and size on the shape,size and formation time of the eroded pits were compared.Three methods for identifying the eigenvalues of leaky acoustic signals have been analyzed and compared,and the BP neural network method was chosen to establish a prediction model for leakage probability.After processing the collected acoustic signal data,its acoustic signal eigenvalues are used as the model input,and the leakage probability is the model output.According to the test data,most of the judgment probability of the model on the leakage point can reach more than 80%,which has a certain effect.The analysis demonstrates that when the leak point of the pipeline is directly above,the measuring point has no obvious effect on the eigenvalues of the acoustic signal relative to the direction of the leak point.On this basis,Two kinds of optimization methods are analyzed and compared,and the interior point method algorithm based on penalty function is selected to explore the optimal arrangement of leak points in the process of multi-probe array leak point positioning in theory.The acoustic signal data collected in the design test was fitted by polynomial interpolation method,and the probability diffusion model of leakage point leakage under specific working conditions was established,reflecting the predicted leakage probability of the measurement points on the ground around the leakage point shows a downward trend like parabolic surface.Demonstrated and ruled out the possibility of locating leak points by two probes,and using the restrictions on the placement of the probes as constraint,determined the objective function of optimization from two directions(the distance difference between the model surface vertex and the model surface axis distance),abstracting the problem into a model that can be solved by using the interior point method algorithm based on the penalty function,and calculate the final objective by using MATLAB software programming. |
PDF Full Text Request