Application Of Improved VMD And Bubble Entropy In Pipeline Leakage Detection

As main transportation mode of natural gas,pipelines are easy to cause pipeline leakage due to long-term corrosion of pipeline materials,man-made damage,theft,and geological disasters.If there are false or false alarms in pipeline leakage detection,and the leakage is not repaired in time,it will cause harm to the surrounding environment,and cause explosions,fires and other vicious events,threatening life safety.Therefore,it is imminent to solve the problems in pipeline leak detection.In this paper,the relevant theories and technologies of pipeline leakage detection are studied in depth,including signal processing,intelligent optimization algorithm,entropy theory and neural network,etc.,and a set of natural gas pipeline leakage detection methods are designed through steps such as signal preprocessing,feature extraction and working condition identification.The main work is as follows:Firstly,in-depth study of the basic principles of Variational mode decomposition(VMD),the feasibility and superiority of the VMD algorithm in the denoising of simulated data are verified through the simulation analysis of noisy signal and sawtooth signal.The influence of preset scale K,penalty factor and artificial selection of effective mode on VMD effect is analyzed.Introduce the pipeline experiment platform,conduct experiments on the data collected by the platform,and verify that VMD is more suitable for pipeline data denoising than Empirical mode decomposition(EMD).Secondly,in view of the difficulty in selecting the effective mode and noise mode after VMD,it is proposed to use Itakura-Saito distance(ISD)to select effective mode,and use Wavelet transform(WT)to remove the noise mode after screening the high-frequency noise is reconstructed from the ISD selection mode and the filtered noise mode to obtain the filtered signal.Simulation signal and pipeline leakage signal experiments show that compared with VMD-CC,VMD-EV and WT,the denoising effect is better,and It can be more effectively applied to the denoising processing of natural gas pipeline leakage signals Thirdly,aiming at the difficulty of selecting VMD parameter[,],Seagull optimization algorithm(SOA)is proposed to optimize VMD parameters,and combined with ISD to select effective modes to obtain filtered signals.The simulation signal and pipeline leakage signal experiments show that compared with other optimization algorithms,SOA-VMD optimizes VMD,has faster convergence speed under the premise of the minimum fitness value,and has better signal decomposition effect.have a more pronounced inhibitory.Finally,under complex working conditions and strong background noise interference,in view of the difficulty of feature extraction caused by weak pipeline signals and the selection of Extreme learning machine(ELM)parameters will affect the classification accuracy,a pipeline signal condition identification method combining Bubble entropy and SOA-ELM is proposed.Calculate the Bubble entropy of the filtered pipeline signal,construct the fault feature vector,optimize the optimal input weight and bias of ELM through SOA,input the fault feature vector into the optimized ELM for working condition identification,and apply its classification model in pipeline leak detection,the accuracy rate of recognition results reached98.33%.