GitHub is home to over 40 million developers working together to host and review code, manage projects, and build software together. If nothing happens, download GitHub Desktop and try again.
If nothing happens, download Xcode and try again.
If nothing happens, download the GitHub extension for Visual Studio and try again. Matlab code used to produce the results in the commentary on "Application of the variable projection scheme for frequency-domain full-waveform inversion" by Li et al. Skip to content. Dismiss Join GitHub today GitHub is home to over 40 million developers working together to host and review code, manage projects, and build software together.
Sign up. Branch: master. Find file. Sign in Sign up. Go back. Launching Xcode If nothing happens, download Xcode and try again.
Latest commit Fetching latest commit…. Variable-Projection-for-FWI Matlab code used to produce the results in the commentary on "Application of the variable projection scheme for frequency-domain full-waveform inversion" by Li et al. Tristan van Leeuwen Aleksandr Y. Aravkin Felix J. You signed in with another tab or window. Reload to refresh your session. You signed out in another tab or window.A word in response to the corona virus crisis: Your print orders will be fulfilled, even in these challenging times.
Recent progress in numerical methods and computer science allows us today to simulate the propagation of seismic waves through realistically heterogeneous Earth models with unprecedented accuracy. Full waveform tomography is a tomographic technique that takes advantage of numerical solutions of the elastic wave equation. The accuracy of the numerical solutions and the exploitation of complete waveform information result in tomographic images that are both more realistic and better resolved.
This book develops and describes state of the art methodologies covering all aspects of full waveform tomography including methods for the numerical solution of the elastic wave equation, the adjoint method, the design of objective functionals and optimisation schemes. It provides a variety of case studies on all scales from local to global based on a large number of examples involving real data.
Gives a comprehensive account of full waveform tomography Develops and describes state-of-the-art methodologies covering all aspects of full waveform tomography Includes methods for the numerical solution of the elastic wave equation see more benefits. Buy eBook. Buy Hardcover. Buy Softcover. FAQ Policy. About this book Recent progress in numerical methods and computer science allows us today to simulate the propagation of seismic waves through realistically heterogeneous Earth models with unprecedented accuracy.
Earth Sciences Research Journal
Table of contents 15 chapters Table of contents 15 chapters Preliminaries Pages Fichtner, Andreas. Introduction Pages Fichtner, Andreas. Visco-elastic Dissipation Pages Fichtner, Andreas. Absorbing Boundaries Pages Fichtner, Andreas. Show next xx. Recommended for you. PAGE 1.With full-waveform inversion FWI solutions for every exploration, appraisal, or production environment, we can create highly detailed velocity models that honor the geologic structures in your reservoir.
Not only do our FWI algorithms work with all acquisition geometries, but they also complement the low frequencies inherent to broadband seismic data. Complying with the half-wavelength criteria is a fundamental challenge for FWI. When the criteria is violated, cycle skipping issues between the predicted and acquired data can cause the inversion to converge at a local minimum, resulting in an inaccurate model. Conventional least-squares FWI LsFWI mitigates this risk by using a highly accurate initial model, restricting its scope of use to basins with mature velocity models.
In locations where an accurate initial model is not available, our adjustive FWI AdFWI algorithm is designed to build the relationship between traveltime shift and model error to correct the erroneous background model while also mitigating cycle-skipping issues. The white dashed line on the inline section indicates the location of the corresponding depth slice. Deepwater subsalt prospects are difficult to image because of the complexities of the overburden and the limited penetration depth of the refraction energy needed for conventional FWI.
Initially, a Born modeling—based gradient kernel is implemented to directly compute the reflection-based low-wavenumber components of a conventional FWI gradient. Then, a robust kinematics-oriented objective function ensures that the low-wavenumber components update the model in the correct directions. Through their combined use, RFWI can derive more accurate velocity models at depths where traditional FWI is limited, even when starting with a smooth initial model.
RFWI improves the velocity model of the section deep below the salt. Move the slider to see the improved geologic detail between the legacy model and the model updated with RFWI.
Low-frequency transmitted seismic energy is crucial for the success of FWI to overcome sensitivity to starting velocity fields. To mitigate the impact of elastic effects, we include only the diving and postcritical early arrivals in the waveform inversion.
With the aid of HR-Radon preconditioning and a carefully designed workflow, acoustic FWI can derive a reliable high-resolution near-surface model that could not be otherwise recovered through traditional tomographic methods.
FWI applied to data acquired on land captures subsurface geologic detail that cannot be obtained through traditional tomography. The legacy tomography model left has been updated using FWI rightproviding significant uplift. Comprehensive array of algorithms and workflows to maximize the value of your multicomponent data. Improve data quality and fill in data gaps with our integrated geophysical reprocessing services and multiclient data.One of the first decisions in any pattern recognition system is the choice of what features to use: How exactly to represent the basic signal that is to be classified, in order to make the classification algorithm's job easiest.
Speech recognition is a typical example. Through more than 30 years of recognizer research, many different feature representations of the speech signal have been suggested and tried. PLP was originally proposed by Hynek Hermansky as a way of warping spectra to minimize the differences between speakers while preserving the important speech information [Herm90].
RASTA is a separate technique that applies a band-pass filter to the energy in each frequency subband in order to smooth over short-term noise variations and to remove any constant offset resulting from static spectral coloration in the speech channel e. In order to understand the algorithm, however, it's useful to have a simple implementation in Matlab.
Scientific Research Codes
This implementation offers only a few control parameters, namely a switch to select or disable rasta filtering, and an option to set the order of PLP modeling which disables PLP modeling when set to zero.
Other important options, such as the basic window and hop sizes, can easily be altered by editing the relevant routines, if desired. Since Mel-frequency Cepstral Coefficients, the other really popular speech feature, involve almost the same processing steps, I decided to make an implementation for them as well, using the same blocks as far as possible. See below. You can do this, crudely, by recovering the short-time magnitude spectrum implied by the cepstral coefficients, then imposing it on white noise.
For more details on reproducing and inverting cepstra from several common feature calculation programs, see the companion page on Reproducing Feature Outputs You can download the complete set of routines above as rastamat.
This example calculates 20th order MFCC features as close as I can get it to the features we distribute for the uspop Music IR dataset and then turns them back into audio - pretty weird sounding!
This version has been verified to give nearly identical results, but offers flexibility to adapt to different bandwidths, sampling rates, etc. If you use this code in your research and would like to acknowledge it and direct others to ityou could use a reference like this:.This paper investigates the capability of acoustic Full Waveform Inversion FWI in building Marmousi velocity model, in time and frequency domain.
FWI is an iterative minimization of misfit between observed and calculated data which is generally solved in three segments: forward modeling, which numerically solves the wave equation with an initial model, gradient computation of the objective function, and updating the model parameters, with a valid optimization method.
An initial model is obtained by smoothing the true model to initiate FWI procedure. Smoothing ensures an adequate starting model for FWI, as the FWI procedure is known to be sensitive on the starting model.
The final model is compared with the true model to review the number of recovered velocities. An initial model is derived from smoothing the true model to initiate FWI procedure. The final model is compared with the true model to review theamount of recovered velocities.
Broyden, C. The New Algorithm. Journal of the Institute of Mathematics and its Applications, 10, Chabert, A. Seismic imaging of the sedimentary and crustal structure of the hatton basin on the Irish Atlantic margin.
Thesis, University College Dublin, Ireland. Clayton, R. Bulletin of the Seismological Society of America, 67 6 Dai, Y.
A nonlinear conjugate gradient method with a strong global convergence property. Demanet, L. Waves and Imaging, Class Notes. Erlangga, Y. Archives of Computational Methods in Engineering, 15 1 Fletcher, R. Computer Journal, 13 3 Goldfarb, D.Deep-learning inversion: A next-generation seismic velocity model building method. Velocity model building by deep learning. Multi-CMP gathers are mapped into velocity logs. FBD: Multi-channel blind deconvolution with focusing constraints. Kronecker-product-based linear inversion under Gaussian and separability assumptions.
Documentation request. Describe your documentation request clearly bellow To write a wiki page with decription of the VFSA algorithm and provide references. Locate seismic events on a simplified 2-d model of the earth. This is a challenge problem and data set for PPLs Probabilistic Programming Languages or generative models in general. Probabilistic inference techniques can be tested and published on this challenge.
Add a description, image, and links to the seismic-inversion topic page so that developers can more easily learn about it. Curate this topic. To associate your repository with the seismic-inversion topic, visit your repo's landing page and select "manage topics.
Learn more. Skip to content. Here are 15 public repositories matching this topic Language: All Filter by language. Sort options. Star Code Issues Pull requests. Open hotfix master branch with accurate documentation regarging penobscot dataset - allows user to more carefully gauge the performance of the repo. Prior: ShowStopper Type: Documentation. Open automate update of pre-run notebooks linked from README before release - allows the user to view up to date pre-run notebooks faster repo exploration without running the repo.
Updated Aug 27, Python.Documentation Help Center. Generate linear, quadratic, and logarithmic chirps using chirp. Create square, rectangular, and triangular waves using squarerectpulsand sawtooth.
Signal Generation and Visualization. Generate periodic and aperiodic waveforms, sequences such as impulses, steps, and ramps, multichannel signals, pulse trains, sincs, and Dirichlet functions. Create Uniform and Nonuniform Time Vectors. Waveform Generation: Time Vectors and Sinusoids.
Impulse, Step, and Ramp Functions. Common Periodic Waveforms. Common Aperiodic Waveforms. Generate widely used aperiodic waveforms such as Gaussian-modulated sinusoidal pulses and linear, logarithmic, or quadratic chirps. The pulstran Function. The Sinc Function. The Dirichlet Function. Choose a web site to get translated content where available and see local events and offers.
Based on your location, we recommend that you select:. Select the China site in Chinese or English for best site performance. Other MathWorks country sites are not optimized for visits from your location. Toggle Main Navigation. Search Support Support MathWorks. Search MathWorks. Off-Canvas Navigation Menu Toggle. Functions expand all Waveforms.
Modulation and Quantization. Topics Signal Generation and Visualization Generate periodic and aperiodic waveforms, sequences such as impulses, steps, and ramps, multichannel signals, pulse trains, sincs, and Dirichlet functions. Create a sample signal consisting of two sinusoids.