Engineering Research Express, Journal Year: 2024, Volume and Issue: 6(3), P. 035538 - 035538
Published: Aug. 2, 2024
Abstract Pressure pipelines are prone to leakage under harsh working condition for a long time, and the detection reaches unsatisfactory performance due influence of background noise insufficient sample acoustic signals. Therefore, signals adversarial augmentation method is proposed pressure pipeline based on reduction generation. By deeply connecting with generative network (GAN), denoising autoencoder (DAE) residual (ResNet), generation model (ADGM) established reduce signal. In addition, trained DAE ADGM applied augment samples, thereby completing signal, which significant detection. Besides, experiment implemented validate generation, can reach accuracy 93.02% augmented Further, effectiveness superiority tested by ablation experiments comparative methods.
Language: Английский