A5028205120- Seismic Imaging and Inversion Techniques
- Seismic Waves and Analysis
- Seismology and Earthquake Studies
- Underwater Acoustics Research
- Reservoir Engineering and Simulation Methods
- Energy Efficient Wireless Sensor Networks
- Geophysical Methods and Applications
- Hydraulic Fracturing and Reservoir Analysis
- Network Time Synchronization Technologies
- Age of Information Optimization
Kootenay Association for Science & Technology
2020-2024
King Abdullah University of Science and Technology
2019-2024
Building realistic and reliable models of the subsurface is primary goal seismic imaging. We have constructed an ensemble convolutional neural networks (CNNs) to build velocity directly from data. Most other approaches attempt map full data into 2D labels. exploit regularity acquisition train CNNs gathers neighboring common midpoints (CMPs) vertical 1D logs. This allows us integrate well-log inversion, simplify mapping by using labels, accommodate larger dips relative single CMP inputs....
Low-frequency data are essential to constrain the low-wavenumber model components in seismic full-waveform inversion (FWI). However, due acquisition limitations and ambient noise it is often unavailable. Deep learning (DL) can learn map from high frequency updates of elastic FWI a update, producing an initial estimation as if was available low-frequency data. We train FusionNET-based convolutional neural network (CNN) on synthetic dataset produce set produced by with missing low frequencies....
Strong near-surface heterogeneity poses a major challenge for seismic imaging of deep targets in arid environments. Inspired by the initial success learning applications to inverse problems, we investigate possibility building nearsurface models directly from raw elastic data including surface and body waves conditions. Namely, train convolutional neural network map into model supervised way on part SEAM Arid synthetic dataset evaluate its performance different same dataset. The main feature...
Summary Building realistic and reliable models of the subsurface is primary goal seismic imaging. By employing an ensemble convolutional neural networks (CNNs), we build velocity directly from pre-stack data quantify model uncertainties by analyzing all results. Most attempts are made to infer as a whole. Here, CNNs trained map subsets into 1D vertical logs. This allows us integrate well inversion simplify mapping using regularity active acquisition geometries. The presented approach uses...
SummaryConventional seismic data are naturally mainly sensitive to the very smooth velocity variations that alter transmission traveltimes (low-model wavenumbers) and abrupt discontinuities cause reflections (high-model wavenumbers). Full-waveform inversion (FWI) of inherits this lack middle model wavenumber illumination, which results into ringy artifacts in gradients. Multiple methods have been suggested overcome issue. Here we tackle problem missing wavenumbers with a deep-learning...
We solve the 3D acoustic wave equation using finite-difference time-domain (FDTD) formulation in both first and second order. The FDTD approach is expressed as a stencil-based computational scheme with long-range discretization, i.e., 8th order space 2nd time, which routinely used oil gas industry environmental geophysics for high subsurface imaging fidelity purposes. Absorbing Boundary Conditions (ABCs) are employed to attenuate reflections from artificial boundaries. discretization...
Summary Seismic full-waveform inversion (FWI) as a non-linear, iterative optimization benefits from low-frequency data to constrain low-wavenumber model updates and improve convergence. However, is often limited in active seismic acquisitions. Using model-domain approach, we attempt generate existing gradients at higher frequencies within deep learning framework. Namely, train convolutional neural network (CNN) provide missing FWI associated with frequency updates. We test this technique on...
Low-frequency data is crucial for successful retrieval of low-wavenumber model component in seismic full-waveform inversion (FWI), yet it often limited by hardware. Deep learning (DL) can fuse early high-wavenumber updates elastic FWI and map them into desired that would be available from low-frequency data. FusionNET-based convolutional neural network (CNN) trained on a synthetic dataset produces meaningful models taking initial iterations field as inputs. Elastic initiated "DL-fused" shows...