西安石油大学学报（自然科学版）2025，Vol.40Issue(3)：1-11,11.DOI:10.3969/j.issn.1673-064X.2025.03.001
基于VAE-EGAN架构的地震脉冲干扰异常检测
Detection of Seismic Pulse Interference Anomaly Based on VAE-EGAN Architecture
摘要
Abstract
In seismic exploration acquisition sites,pulse signals,as a form of interference,seriously affect the quality of seismic acquisi-tion records and are a key monitored interference object in seismic acquisition sites.To accurately detect the pulse signals and reduce the impact of seismic pulse signals on the processing and interpretation of subsequent seismic data,a anomaly detection method based on VAE-EGAN architecture is proposed.This method combines the generation stability of the variational autoencoder VAE with the dis-criminative ability of the generative adversarial network GAN,and reduces the possibility of model overfitting through weight attenuation and spectral normalization techniques.The newly designed loss function combines the unique structure of multiple discriminators to en-hance the competitiveness of GAN in anomaly capture tasks.The experimental results of actual seismic data in a western work area show that this method achieves the anomaly detection accuracy of 93.75％ and F1 value of 96.77％,and the anomaly localization accuracy of 89.82％ and F1 value of 92.73％.The experimental results have verified the effectiveness of this method in improving the accuracy of pulse signal anomaly detection and reducing the complexity of pulse signal detection in seismic data processing,helping to ensure the accuracy of seismic data.关键词
地震脉冲/异常检测/生成对抗网络/变分自编码器Key words
seismic pulse/anomaly detection/generative adversarial network/variational autoencoder分类
地球科学引用本文复制引用
严英殊,余贞侠,文晓涛,王秋成,文武..基于VAE-EGAN架构的地震脉冲干扰异常检测[J].西安石油大学学报（自然科学版）,2025,40(3):1-11,11.基金项目
四川省自然科学基金项目"四川盆地碳酸盐岩储层的脉冲神经网络识别理论及方法研究"(2023NSFSC0258) （2023NSFSC0258）
四川省中央引导地方科技发展专项"岩石物理驱动的非常规油气地震多参数预测理论与方法研究"(2023ZYD0158) （2023ZYD0158）
