Browse Prior Art Database

Method For Processing Waveform Data

IP.com Disclosure Number: IPCOM000241973D
Publication Date: 2015-Jun-11

Publishing Venue

The IP.com Prior Art Database

Abstract

The disclosure is directed to a method for processing waveform data. The method includes selecting an initial model and a computational domain for computing waveform data. The method further includes splitting the computational domain into a number of subdomains. A Stieltjes continued fraction (S-fraction) is determined for each subdomain using a reducedorder model. Then, each subdomain is joined together to generate compressed waveform data.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 21% of the total text.

Page 01 of 10

METHOD FOR PROCESSING WAVEFORM DATA

TECHNICAL FIELD


[0001]This disclosure relates to inversion of data, and more particularly to inversion of waveform data.

BACKGROUND


[0002]Full waveform inversion of data (e.g., acoustic and seismic data) uses significant computational resources due to multiple solutions for a forward problem. After spatial discretization, inversion yields dynamical systems with up to billions of unknowns. Inverse problems are typically solved using high performance computing (HPC) platforms.


[0003]Forward modeling schemes and, particularly, discretization techniques have been developed to more efficiently solve inverse problems. For example, conventional finite- difference and finite-element methods with low order accuracy have small stencil, which is favorable for efficient communications on HPC platforms. However, such methods require a large number of unknowns for accurate results [1].


[0004]High-order discretizations, such as high-order finite-difference methods, high- order and spectral finite elements methods [2], and spectral multi-scale methods [3], not only suffer from large stencil, but also have poor Courant-Friedrichs-Lewy (CFL) conditions. Moreover, for realistic medium structure, high-order discretizations may require using complicated gridding techniques, which are not feasible for iterative inversion algorithms with model updates.


[0005]Another approach to obtain spectrally accurate discretization is via an optimal grids method (also known as spectrally-matched grids and finite-difference Gaussian rules)
[4]. The optimal grids method has the same stencil as a regular second-order finite-difference scheme. However, optimal grids can only be constructed for simple mediums.

1


Page 02 of 10

SUMMARY


[0006]The disclosure is directed to a method for processing waveform data. The method includes selecting an initial model and a computational domain for computing waveform data. The method further includes splitting the computational domain into a number of subdomains. A Stieltjes continued fraction (S-fraction) is determined for each subdomain using a reduced- order model. Then, each subdomain is joined together to generate compressed waveform data.


[0007]Interactions between adjacent subdomains are represented using a Neumann-to- Dirichlet (NtD) map. Each S-fraction is determined by reducing the NtD map using the reduced-order model.


[0008]The compressed waveform data is compared to measured data and the model is revised based upon the comparison. The method is repeated until the compressed waveform data converges with the measured data.

BRIEF DESCRIPTION OF THE DRAWINGS


[0009]Fig. 1 shows a method for processing waveform data;


[0010]Fig. 2 shows a computational domain for waveform data that is split into a discretization grid with 25 subdomains; and


[0011]Fig. 3 shows a system for obtaining seismic waveform data.

DESCRIPTION


[0012]The disclosure is directed to a method for processing wavefo...