Physical state guided diffusion improves subsurface velocity imaging accuracy
Physical-State-Guided Diffusion Sampling for Full-Waveform Inversion
Machine Learning
Summary
Understanding underground structures by analyzing seismic waves is very difficult because the process depends a lot on how you start and the information you have. The authors developed a new method that uses a learned geological model together with physical laws to better guide this imaging. Their approach separately improves the physical model and the learned model steps, making the final results more reliable and detailed. They demonstrated this method works well on different simulated geological data and can handle incomplete or noisy measurements.
What this means in practice
- •For geophysical exploration teams: Improve underground velocity maps for oil, gas, and mineral exploration using more accurate seismic inversion methods that combine learned geological models and physical laws.
- •For earthquake hazard modelers: Refine subsurface velocity structures to better understand seismic wave propagation and improve earthquake risk assessments under noisy data conditions.
Authors
Chen Min, Haowen Jiang, Zheng Ma, Xiongbin Yan
Abstract
Full waveform inversion (FWI) estimates subsurface velocity from seismic recordings, but its ill-posedness and nonlinearity make accurate reconstruction strongly dependent on initialization and prior information. Diffusion posterior sampling provides a learned geological prior, yet directly coupling its denoiser to the nonlinear wave solver can yield unreliable physical guidance. We propose Physical-State-Guided Diffusion Sampling (PSG), which couples a persistent physical velocity to the diffusion prior through a Gaussian bridge. The physical state is refined by waveform fitting regularized by the denoised velocity, and in turn guides the reverse diffusion process. This formulation separates the wave-equation and denoiser gradients while preserving conventional FWI initialization and accumulated optimization history. On four OpenFWI families, PSG's terminal denoised estimates outperform classical and diffusion-based baselines under clean and missing-trace acquisitions and maintain strong structural recovery under measurement noise. Repeated stochastic runs preserve the dominant geological structures, with ensemble variability concentrated near geological interfaces and positively associated with local inversion error. A frozen OpenFWI-trained prior further supports inversion of the larger Marmousi, Overthrust, and BP2004 Salt models, recovering complex geological structures without retraining.