Application Research Of Wavelet Analysis And Deep Learning In Pipeline Defect Recognition

As the service life of oil and gas pipelines increases,the risks existing in the operation of oil and gas pipelines also gradually increase.Regular inspection of oil and gas pipelines is required to ensure the stable operation of oil and gas pipelines.In this paper,signal noise reduction,feature extraction,and defect signal diagnosis are carried out on the magnetic eddy current testing data of oil and gas pipeline defects.The specific research contents are as follows:1.The on-site noise will interfere with the subsequent signal analysis.In view of the problem that the traditional noise reduction method has no obvious effect on the noise reduction of the oil and gas pipeline magnetic eddy current detection signal,the Generative Adversarial Networks(GAN)and the wavelet multi-scale decomposition(Combined with Wavelet Multiscale Analysis,WMA),a GAN-WMA noise reduction algorithm is proposed.The GAN-WMA noise reduction algorithm is verified by examples.Compared with the traditional noise reduction algorithm,the signal to noise of the GAN-WMA algorithm is improved by 4-7 percentage points,and the signal to noise ratio is improved to 20.8560 d B.Preserve the local features of the signal.2.Most of the existing methods for extracting magnetic eddy current defect signal features of oil and gas pipelines are based on time-frequency domain information,which cannot fully exploit the core features of magnetic eddy current defect signals of oil and gas pipelines.To solve this problem,an automatic feature extraction using wavelet scattering network is proposed.The automatically extracted features are brought into the K-means clustering model,and the clustering accuracy rate is about 75% after testing.The results show that the features automatically extracted by the wavelet scattering network can realize the clustering of different oil and gas pipeline defect signals,which is the basis for subsequent oil and gas pipelines.Magnetic eddy current defect signal diagnosis provides reliable eigenvalues.3.To solve the diagnosis problem of oil and gas pipeline defective signal,a magnetic eddy current signal diagnosis model for oil and gas pipeline defects based on Bayesian optimization(BO)Long Short-Term Memory(LSTM)is proposed.The parameters and structures in the LSTM network were adaptively selected by the Bayesian optimizer,and the optimized model was used to diagnose the magnetic eddy current data set of oil and gas pipeline defects.The average diagnostic accuracy of the optimized model defect signals reached 94%,which is more accurate The accurate rate is higher than that of the traditional LSTM model,which verifies the superiority of the BO-LSTM model.4.In order to improve the use efficiency of the model proposed in this paper and simplify the difficulty of using the model,a set of magnetic eddy current signal analysis software for oil and gas pipelines is designed based on the above deep learning algorithm.The software can perform noise reduction processing on the magnetic eddy current signal of oil and gas pipelines,extract defect signal features and realize defect signal diagnosis.