Technology Category
Search and Optimization: Mixed Discrete/Continuous SearchComputer Vision: Generative Adversarial Networks (GANs) for VisionReasoning under Uncertainty: Stochastic Optimization
Leveraging generative adversarial networks with spatially adaptive denormalization for multivariate stochastic seismic data inversion
🤖 AI Summary
Joint inversion of facies (discrete) and rock/elastic properties (continuous) in multivariate stochastic seismic inversion remains challenging due to strong coupling, while existing deep learning methods suffer from excessive model complexity and training instability. Method: This paper proposes SPADE-GANInv—a hybrid framework integrating Spatially-Adaptive Denormalization GAN (SPADE-GAN) with geostatistical simulation. Leveraging a pre-trained SPADE-GAN as a structural prior, it iteratively generates subsurface realizations, matches synthetic seismic responses, and incorporates well-log constraints to achieve phase-conditioned, multivariate joint inversion. Contribution/Results: The SPADE mechanism explicitly models spatial dependencies between facies and petrophysical parameters, substantially reducing reliance on initial probabilistic models and enhancing prior adaptability and inversion stability. Tested on 2D synthetic and field data, SPADE-GANInv simultaneously reconstructs facies, porosity, and acoustic impedance with high accuracy, achieving an optimal balance between geological plausibility and seismic-data consistency.
Technology Category
Application Category
📝 Abstract
Probabilistic seismic inverse modeling often requires the prediction of both spatially correlated geological heterogeneities (e.g., facies) and continuous parameters (e.g., rock and elastic properties). Generative adversarial networks (GANs) provide an efficient training-image-based simulation framework capable of reproducing complex geological models with high accuracy and comparably low generative cost. However, their application in stochastic geophysical inversion for multivariate property prediction is limited, as representing multiple coupled properties requires large and unstable networks with high memory and training demands. A more recent variant of GANs with spatially adaptive denormalization (SPADE-GAN) enables the direct conditioning of facies spatial distributions on local probability maps. Leveraging on such features, an iterative geostatistical inversion algorithm is proposed, SPADE-GANInv, integrating a pre-trained SPADE-GAN with geostatistical simulation, for the prediction of facies and multiple correlated continuous properties from seismic data. The SPADE-GAN is trained to reproduce realistic facies geometries, while sequential stochastic co-simulation predicts the spatial variability of the facies-dependent continuous properties. At each iteration, a set of subsurface realizations is generated and used to compute synthetic seismic data. The realizations providing the highest similarity coefficient to the observed data are used to update the subsurface probability models in the next iteration. The method is demonstrated on both 2-D synthetic scenarios and field data, targeting the prediction of facies, porosity, and acoustic impedance from full-stack seismic data. Results show that the algorithm enables accurate multivariate prediction, mitigates the impact of biased prior data, and accommodates additional local conditioning such as well logs.
Problem
Research questions and friction points this paper is trying to address.
Predict facies and continuous properties from seismic data
Overcome GAN limitations in multivariate stochastic inversion
Integrate SPADE-GAN with geostatistical simulation for accurate predictions
Innovation
Methods, ideas, or system contributions that make the work stand out.
SPADE-GAN integrates spatially adaptive denormalization for facies conditioning
Iterative algorithm combines pre-trained GAN with geostatistical co-simulation
Method predicts multiple correlated properties from seismic data accurately
🔎 Similar Papers
No similar papers found.
