Research On Acoustic Emission Detection And State Identification Method Of Leakage In Gas-liquid Two-phase Flow Pipeline

Gas-liquid two-phase flow pipeline is an important part of the oil and gas pipeline.It is very necessary to ensure the safe operation of the gas-liquid two-phase flow pipeline.Because the gas-liquid two-phase flow pipeline has complex and changeable flow conditions,this type of pipeline will have the characteristics of strong interference and strong vibration.The former will greatly increase the difficulty of leak detection,while the latter will make the pipeline more likely to leakage.Due to the difficulties mentioned above,it is difficult for the traditional detection technology to accurately detect the leakage of this kind of pipeline.Acoustic emission(AE)technology,as a common non-destructive testing technology in pipeline leakage detection,can sensitively detect the existence of leaks by receiving the acoustic emission signals generated by nearby leak sources.Therefore,it has a good theoretical feasibility in gas-liquid two-phase flow pipeline leakage detection.However,there are still many problems that are difficult to solve with traditional acoustic emission signal processing methods,such as noise interference and difficult to distinguish problems.Therefore,appropriate methods must be proposed to effectively improve the detection accuracy of gas-liquid two-phase pipeline leakage acoustic emission detection.Based on the above requirements,this article conducts related research on the acoustic emission detection and status recognition of gas-liquid two-phase flow pipeline leakage.(1)Experimental research on acoustic emission detection of gas-liquid two-phase flow pipeline leakage considering gas-liquid flow changes and various leakage conditionsBased on the theoretical basis of acoustic emission pipeline leakage detection technology,a gas-liquid two-phase pipeline leakage acoustic emission detection experimental system is designed,which can realize the simulation of leakage conditions considering the leakage type,leakage direction,leakage location and other parameters.A series of simulation experiments on pipeline leakage of gas-liquid two-phase flow were carried out,and acoustic emission signals were acquired in real time.Based on parameter analysis,time-frequency analysis and other methods,the difference between the acoustic emission signals under the state of no leakage and leaking of this type of pipeline was compared and analyzed,and the influence of various factors on the characteristics of acoustic emission signals mechanism was revealed.(2)Feature extraction and manual identification of acoustic emission signal of gas-liquid two-phase flow pipeline leakageThree different methods are used to extract the features of the signals and identify them artificially.The first category is the mean value of characteristic parameters,including amplitude,RMS and ASL.This feature is widely used in leak detection because it is simple and easy to implement,but it is poor in universality in this kind of pipeline.The second category is the energy proportion of the wavelet packet.The energy proportion of each frequency band obtained after four-layer wavelet decomposition is used as the characteristic to judge the leakage.This method has good universality,but the accuracy is not high.The third category is Mel-frequency cepstral coefficients(MFCC),which can extract two kinds of features,32×177and 96×177,but there are many features in this category and it is difficult to realize artificial recognition.(3)Research on leak identification model and generalization ability test of gas-liquid twophase flow pipelineBased on four machine learning models of support vector machine(SVM),BP neural network(BPNN),long and short time memory network(LSTM)and convolutional neural network(CNN)combined with the 9 kinds of extracted features in 3 categories,the intelligent identification algorithm of pipeline leakage in gas-liquid two-phase flow was researched.Finally,MFCC feature and LSTM network to realize leakage detection based on a lot of experimental data was raised.Further research on this method is carried out,including: hyper parameter optimization,K-fold cross validation,anti-noise interference ability test,unfamiliar situation recognition ability test.The results show that the verification accuracy of the model is extremely high,and the theoretical accuracy of the comprehensive recognition results after multiple recognition is as high as 100%.A series of studies in this paper have proved that MFCC characteristics and LSTM network can be used to achieve high precision leak detection of gas-liquid two-phase flow pipeline based on a large number of data,and to a certain extent,the practical engineering application ability of this method has been proved,providing support for the practical application of this method.This study provides a solution to the problem that gas-liquid two-phase flow is difficult to accurately detect,and provides a theoretical basis for the practical application of gas-liquid twophase flow pipeline leakage acoustic emission detection.