Physics-informed neural network for seismic data reconstruction of compressed sensing
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
Seismic exploration represents one of the most critical methodologies and highest-cost expenditures in the pre-exploration phases. The main cost of seismic exploration is seismic data acquisition, which can be significantly reduced through compressed sensing (CS) techniques. This abstract proposes a novel physics-informed neural network (PINN) framework for reconstructing seismic data acquired via down-sampling from Ocean Bottom Seismometer (OBS) observation systems. The acquisition system involves two types of sparsity: random missing entire source lines and random missing source points. The proposed method utilized moveout (MO) transformations with multiple constant velocities to address aliasing artifacts and enhance reconstruction accuracy. Then, a 2D pre-interpolation process is employed for the MO-transformed seismic data groups. Besides, an overlay effect evaluation mechanism dynamically assigns weights to each MO dataset, generating optimized pre-interpolated seismic profiles. Finally, the PINN architecture integrates physical constraints to refine the reconstructed data. Experimental results demonstrate the superior reconstruction performance and computational efficiency of the proposed method compared to conventional approaches.
Fields of Study & Subjects
Seismic Imaging and Inversion Techniques
Model Reduction and Neural Networks
Seismic Waves and Analysis
Publication details
- Journal / source
- Second International Meeting for Applied Geoscience & Energy
- Volume
- 1
- Pages
- 2210–2214
- Publication date
- 10 February 2026
- Language
- EN