A Physics-guided Generative AI Toolkit for Geophysical Monitoring

TL;DR

This paper introduces EdGeo, a physics-guided diffusion model toolkit that enhances geophysical imaging by generating high-quality data to improve machine learning model accuracy in subsurface velocity mapping, especially with limited data.

Contribution

The EdGeo toolkit uniquely combines physics-based diffusion modeling with ML fine-tuning to address data scarcity and improve geophysical imaging accuracy.

Findings

Significant SSIM score improvements across pruning ratios

Reduction in MAE and MSE in velocity map predictions

Outperforms existing methods in representing unprivileged features

Abstract

Full-waveform inversion (FWI) plays a vital role in geoscience to explore the subsurface. It utilizes the seismic wave to image the subsurface velocity map. As the machine learning (ML) technique evolves, the data-driven approaches using ML for FWI tasks have emerged, offering enhanced accuracy and reduced computational cost compared to traditional physics-based methods. However, a common challenge in geoscience, the unprivileged data, severely limits ML effectiveness. The issue becomes even worse during model pruning, a step essential in geoscience due to environmental complexities. To tackle this, we introduce the EdGeo toolkit, which employs a diffusion-based model guided by physics principles to generate high-fidelity velocity maps. The toolkit uses the acoustic wave equation to generate corresponding seismic waveform data, facilitating the fine-tuning of pruned ML models. Our results demonstrate significant improvements in SSIM scores and reduction in both MAE and MSE across various pruning ratios. Notably, the ML model fine-tuned using data generated by EdGeo yields superior quality of velocity maps, especially in representing unprivileged features, outperforming other existing methods.