This series of jupyter notebooks demonstrates features of the Chevron Optimization Framework for Imaging and Inversion
.
- The notebooks are intended to be run in order.
- Downstream notebooks may depend on artifacts created in previous notebooks. For example the Marmousi dataset is downloaded and processed in the series
20_marmousi_model_setup
, and used in subsequent notebooks. - Some notebooks may require reasonable HPC resources to run. For example the FWI and RTM examples can run for more than one hour, depending on what hardware you execute them on.
00_add_packages
adds and precompiles all packages used in these demos.10_jets_basics
introduction toJets
andDistributedJets
.20_marmousi_model_setup
download the Marmousi model.- The Marmousi 2 model is provided by the Allied Geophysical Laboratory of the University of Houston, license and more information at the SEG wiki entry AGL Elastic Marmousi.
- If you run these notebooks you will have a copy of the license to review in the directory
20_marmousi_model_setup
.
30_forward_modeling
static and dynamic scheduled modeling.40_sensitivity
generation of FWI sensitivity kernels, single trace and wavefield separation examples.50_fwi
01_fwi_L2.ipynb
contains a brute force Marmousi time domain FWI example using theLBFGS
algorithm fromOptim.jl
, includes upsampling and downsampling models, data analysis, illumination compensation, very simple box constraints, and nonlinear optimization usingOptim.jl
.02_fwi_L2_dynamic.ipynb
contains a "cloud native" implementation of the previous notebook on the Azure cloud.10_add_slim_packages.ipynb, 11_constrained_fwi_pqn.ipynb, 12_constrained_fwi_spg.ipynb
contain constrained FWI examples that demonstrate interoperability with Georgia Tech SLIM group's julia software.
60_rtm
brute force RTM of the Marmous FWI results, including data processing like applying a temporal mute, and image processing like a Laplacian filter to remove backscattered noise. Both static and dynamic scheduled examples are provided.