Abstract:Distributed acoustic sensing (DAS) is a new technology for acquiring seismic data with high spatial resolution at low cost. Furthermore, in real downhole seismic exploration, DAS can receive some weak signals reflected from deep and thin layers. Unfortunately, some real downhole seismic data received by DAS are often characterized by low quality. Specifically, in real DAS records, desired signals with weak energy are often contaminated by some new noise not presented in seismic data received by conventional electronic geophones. Due to the characteristics of seismic data such as frequency band aliasing, low S/N, and complex noise wavefield, existing linear or nonlinear denoising methods based on information processing theories cannot effectively eliminate this complex and multi-type noise. Recently, the deep-learning method is regarded as a powerful tool for background noise attenuation in seismic data. Most of the existing deep-learning methods are concerned with local features, and ignore the global features which can be utilized to enhance their performance further. To simultaneously extract global and local features, we design a novel complete perception self-attention network (CP-SANet) based on the Transformer framework and apply it to the denoising of downhole DAS records. The network embeds a Transformer module into multi-level encoder-decoder framework. Depth-wise convolution is applied to enhance the local perception capability. Given Transformers requirement for large amounts of data, we specifically designed abundant seismic data samples using formation models with different parameters. The noise in the datasets is obtained from actual field DAS data. The effectiveness and feasibility of CP-SANet were verified on both synthetic and field DAS records. All the experimental results prove its satisfactory performance compared to some classical and network methods.

Self-Supervised Pre-Training Transformer for Seismic Data Denoising

Self-Supervised Pretraining Transformer for Seismic Data Denoising

Residual Learning of Deep Convolutional Neural Network for Seismic Random Noise Attenuation

Swin Transformer for Seismic Denoising

Enhancing Seismic Data Denoising Through Multiscale Analysis Across Transformer and GAN

Unsupervised Deep Learning for Seismic Data Reconstruction

A Denoising Method of Seismic Data Based on Self-Supervised Learning

Improving Seismic Fault Recognition with Self-Supervised Pre-Training: A Study of 3D Transformer-Based with Multi-Scale Decoding and Fusion

Self-Supervised Seismic Random Noise Attenuation With Spatial Attention From a Single Section

Transformer for seismic image super-resolution

Optimizing a Transformer-based network for a deep learning seismic processing workflow

ME-TransNet: A Modal-Enhanced Transformer Denoising Network for Attenuating Low-Frequency Swell Noise

Seismic random noise suppression by using deep residual U-Net

A Natural Images Pre-Trained Deep Learning Method for Seismic Random Noise Attenuation

Irregularly Spatial Seismic Missing Data Reconstruction Using Transformer With Periodic Skip Connection

Fault Detection via 2.5D Transformer U-Net with Seismic Data Pre-Processing.

Transformer and Convolutional Hybrid Neural Network for Seismic Impedance Inversion

Unpaired training: Optimize the seismic data denoising model without paired training data

Complete perception self-attention network (CP-SANet) for weak seismic signal recovery in DAS VSP data

Seismic Shot Gather Denoising by Using a Supervised-Deep-Learning Method with Weak Dependence on Real Noise Data: A Solution to the Lack of Real Noise Data