Self-Supervised Three-Dimensional Ocean Bottom Node Seismic Data Shear Wave Leakage Suppression Based on a Dual Encoder Network

Zhaolin Zhu ¹, Zhihao Chen ¹, Bangyu Wu ²

⁰Hainan Institute, Zhejiang University, Sanya 572025, China.

Sensors (basel, Switzerland) +

|

February 13, 2025

View abstract on PubMed

Summary

A novel self-supervised neural network effectively suppresses shear wave noise in 3D Ocean Bottom Node (OBN) seismic data. This method uses horizontal components to predict and remove noise from the vertical component without requiring clean data.

Area Of Science

Geophysics
Seismic Data Acquisition and Processing

Background

Ocean Bottom Node (OBN) seismic acquisition uses hydrophones and geophones.
Shear wave noise in the vertical component degrades dual-sensor seismic data processing.

Purpose Of The Study

Introduce a self-supervised neural network for shear wave noise suppression in 3D OBN data.
Replace traditional adaptive matching subtraction with a more efficient neural network approach.

Main Methods

A neural network takes horizontal seismic components as input to predict shear wave noise.
A loss function, incorporating cross-correlation regularization, trains the network using the noisy vertical component.
The method is self-supervised, eliminating the need for clean reference data.

Main Results

The proposed neural network effectively suppresses shear wave noise in 3D OBN seismic data.
Experiments on synthetic and field data validate the method's performance.
The approach balances signal preservation and noise reduction.

Conclusions

The self-supervised neural network offers an effective and efficient solution for shear wave noise suppression in 3D OBN seismic data.
This method simplifies data processing by removing the need for manual noise identification or clean data.

Keywords:

OBN U-Net deep learning noise suppression shear wave