Study On Intelligent Leakage Aperage Identification For Natural Gas Pipeline In Compressed Sensing Domain

The natural gas industry plays an important role in the development of the national economy and people’s lives.Pipeline transmission is the main mode of transportation.Therefore,attention has been paid to the research on the real-time monitoring of natural gas pipeline leakage and the identification of leakage apertures.As an important application of fault diagnosis theory,this paper aims at the new problems and new challenges in the leakage monitoring of natural gas pipelines.Compressed sensing and deep learning theories proposed in recent years have been introduced into the collection and analysis of leakage signals,providing new ideas for the research of identification on pipeline leakage apertures.The main research contents of this article include:First of all,it deeply understands the development and current status of pipeline leakage monitoring,and has a deep understanding of the new problems encountered in the large amount of data,acquisition and transmission of leakage monitoring.Aiming at the problems of traditional leakage signal collection redundancy and high degree of subjective dependence of the diagnosis process,the theory of compressed sensing is proposed to achieve high-dimensional signal compression and acquisition with less redundancy while ensuring that important signal information is not lost.This paper introduces deep neural network to analyze compressive sensing domain signals,learns hidden features in complex leaky signals through deep hierarchical intelligence and adaptively,and extracts important information hidden in the data,effectively solving the problem of identifying pipeline leaks.Second,for the serious dependence on prior knowledge and diagnostic experience of the traditional methods,based on the theory of compressed sensing and deep learning,a leakage aperture identification method based on compressed acquisition and sparse filtering deep neural network is proposed.According to the sparse characteristic of the leakage signal,the method obtains information in the compressed sensing domain through the measurement matrix.The obtained small amount of observation signal can be regarded as an adaptively extracted features,this signal through the sparse filtering neural network achieving further streamlining of the acquired signal features by setting a reasonable number of feature outputs.Finally,the high-precision identification of leakage aperture was realized using Softmax regression classifier.The important influencing factors of this method are also analyzed,and a large number of experiments verify the effectiveness of this method.In the end,aiming at the poor robustness of features and overfitting in training in the traditional self-encoding method,a leaky aperture identification method based on compressed acquisition and denoising self-encoding deep neural network is proposed.After the leakage signal is compressed and collected,a certain proportion of random noise is added,self-encoding method is used to adaptively learn the robustness feature,and the over-fitting correction method is used to randomly process the characteristics of the hidden layer,effectively preventing Overfitting in the training process;Finally,based on the Softmax classifier,high-precision identification of leakage apertures was achieved,and the effectiveness of the proposed method was verified through extensive experiments.