Study On Third Party Damage Detection Of Underground Gas Pipelines Based On Acoustic Emission Signals

In recent years,with the increase of natural gas production and consumption,the length of gas pipelines in China is increasing,and underground gas pipelines are one of the important components.The safety of gas pipelines directly affects the use of gas and public safety.The third-party damage is the main cause of gas pipeline damage.The existing pipeline leak detection technology has some problems,such as poor timeliness of detection information and so on.In addition,the distributed optical fiber technology is not suitable for underground pipelines.Therefore,the acoustic emission detection of pipelines is used as the research object in this paper.The acoustic emission signal generated by a variety of the thirdparty damage on the pipeline is collected.More efficient third-party damage detection is achieved through signal decomposition,noise reduction and classification.The main work is as follows:(1)The theory of acoustic emission generation and propagation is studied.Collected acoustic emission signals generated by third-party damage on buried gas pipelines with detection equipment.According to the requirements of signal processing,the signal is cut and the experimental data set is formed.(2)The acoustic emission signal is easily disturbed and the signal-to-noise ratio is low,so the Empirical Mode Decomposition(EMD)and Variational Mode Decomposition(VMD)theories are studied.The best parameters of VMD are not easy to find,so the EMD method based on Whale Optimization Algorithm(WOA)is proposed.Compared with EMD and VMD,the signal-to-noise ratio of the acoustic emission signal processed by this method is higher.The 1D-CNN model is used to classify the signal.The processed signals were better classified compared to the data without noise reduction.(3)In some cases,the number of acoustic emission signals collected from the pipeline is small.Therefore,taking the acoustic emission signal generated by cutting the pipeline as the research object,the classification method of the unbalanced data set is studied.Using the theory of Wasserstein Generative Adversarial Network with Gradient Penalty(WGAN-GP)to increase the number of acoustic emission signals,the effectiveness of WGAN-GP is verified by 1D-CNN model,and the classification accuracy of fewer categories of acoustic emission signals is improved.