Refactor and better implementations of `fatiando.seismic.wavefd` by leouieda · Pull Request #52 · fatiando/fatiando

leouieda · 2013-08-26T18:59:35Z

I have changed and simplified a lot the interface for the finite difference simulation functions. They can now take care of recording seismograms at specified stations, calculating grid spacings, etc. They are still iterators but instead of yielding on every time step, a snapshot parameter controls how often the results are yielded.

The methods used for the finite differences have evolved from simple forward differences to staggared grid methods. For SH waves, I used the Equivalent Staggared Grid of Di Bartolo et al. (2012). This method can and should also be used for acoustic (scalar) waves. For P-SV waves I used the Parsimonious Staggared Grid of Luo and Schuster (1990). These methods have the advantages of being more stable to physical property gradients and high Poisson ratios. Because of these new algorithms, simulations now require the Lamé parameters and density instead of wave velocities.

Instead of 3, now only 2 grids are stored for each displacement. The t+1 grid can overwrite the t-1 grid because I only need it at i,j.

The absorbing boundaries have also improved. I still use a Gaussian tapering (damping) in extended borders of the grid. But now I do it right. Damping is applied to grids at times t + 1 and t for more efficiency. There are also special non-reflecting boundary conditions help reduce plain wave reflections.

For P-SV waves, an option was added to convert x and z displacements into representations of P and S waves. They are calculated from the divergence and curl of the displacement vector, respectively, using forth order centered differences. This way, only one wavefield panel can be plotted with more intuitive representations of the wave field.

I also added a new source function for simulating a blast in all directions. This function actually produces various point sources. This is useful because this source only produces P waves in P-SV simulations:

This is cools because its then easy to see that when the wave reaches an interface:

It splits into a P and an S wave that wasn't there before

Closes #4

… Still not working

…damping all time panels

…tead of wavelength

…orithm. Not working in water

…package

…coordinates and convert to indexes

(central diffs)

Conflicts: cookbook/seismic_wavefd_elastic_sh.py fatiando/io.py fatiando/mesher.py fatiando/seismic/wavefd.py

hbueno · 2013-08-26T19:36:27Z

\m/

eusoubrasileiro · 2013-08-27T20:05:37Z

Really cool master @leouieda . I will try to test the recipes and merge with mine to have the changes included.
I will read the papers to have a better understanding of all that. Too much for my head right now.
Again congrats, really cool. 👍

leouieda · 2013-09-02T14:49:06Z

@hbueno @eusoubrasileiro tudo bem se eu fechar esse PR? Objeções?

eusoubrasileiro · 2013-09-02T17:08:31Z

manda ver mestre @leouieda , vou testar as recepies ainda mas põem pra dentro! vou modificar meu código depois.
Achei que tu já tinha posto na verdade... fogo na 💣

Refactor and better implementations of `fatiando.seismic.wavefd`

eusoubrasileiro · 2013-09-18T02:49:36Z

velho tá a maior droga pra poder fazer merge, tem alguma ferramenta decente aew pra recomendar. Fudeu com tudo!
@hbueno alguma sugestão pra conflito com o git? valeus

hbueno · 2013-09-18T11:09:06Z

O que aconteceu?

Em 17 de setembro de 2013 23:49, André notifications@github.com escreveu:

velho tá a maior droga pra poder fazer merge, tem alguma ferramenta
decente aew pra recomendar. Fudeu com tudo!
@hbueno https://github.com/hbueno alguma sugestão pra conflito com o
git? valeus

—
Reply to this email directly or view it on GitHubhttps://github.com/leouieda/fatiando/pull/52#issuecomment-24637650
.

eusoubrasileiro · 2013-09-18T12:25:45Z

rsrs @hbueno então tenho que fazer merge do novo master com o meu. Tenho alguns conflitos vou acabar fazendo na mão a gambiarra do merge mas se vocês conhecerem algo legal aew de merge ui pra linux seria maneiros. valeus

leouieda · 2013-09-18T12:53:39Z

Eu uso um programa chamado Meld
Em 18/09/2013 09:25, "André" notifications@github.com escreveu:

rsrs @hbueno https://github.com/hbueno então tenho que fazer merge do
novo master com o meu. Tenho alguns conflitos vou acabar fazendo na mão a
gambiarra do merge mas se vocês conhecerem algo legal aew de merge ui pra
linux seria maneiros. valeus

—
Reply to this email directly or view it on GitHubhttps://github.com/leouieda/fatiando/pull/52#issuecomment-24660203
.

eusoubrasileiro · 2013-09-18T13:54:18Z

É to vendo o povo recomendando aqui... vou tentar rsrs cara mó trabalharei q droga uhaiahiua ⛵
Valeus @leouieda

leouieda · 2013-09-18T17:02:34Z

Cara, o que tinha tanto no seu branch que tá dando conflito?

eusoubrasileiro · 2013-09-18T17:06:39Z

ja to arrumando as chamadas pras implementacoes do python, o arquivo de implementacoes de python foi removido. Foi criado outro no lugar. Fora isso o merge automático nao funcionou no wavefd.py por conta das minhas mudanças . Mas ta sussa vou arrumar em mais uns 2 commits.
Sei que a PR já foi fechada.
Só uma sugestão a maneira como o sismograma é retornado poderia ser como a iteração para honrar o modelo de snapshot. Mas tanto faz. Funciona do mesmo jeito e o stations ficou maneiro.

eusoubrasileiro · 2013-09-18T18:27:53Z

Todos os exemplos que testei estão rodando (era problema com o backend que não estava instalado direito). Depois olha a pull request do scalar lá finalmente deixei usável.

eusoubrasileiro · 2013-10-12T21:39:30Z

@leouieda mestre... to dando uma estudada aqui. Realmente ESG ou no geral staggered grid's são muito mais fodas! Lembrando do curso de CFD que fiz lendo o artigo do Bartolo, tanto CFD qto Eletromag usam esse trem. Realmente deve ser bom! Vou tentar implementar um scalar com isso aew vamos ter duas versões de scalar. Um non-staggered e um staggered bonito o bicho. Valeus pela dica dos artigos.

hbueno · 2013-10-12T22:16:11Z

Achei que só eu estava nerdiando essa hora num sábado... =/

Em 12 de outubro de 2013 18:39, André notifications@github.com escreveu:

@leouieda https://github.com/leouieda mestre... to dando uma estudada
aqui. Realmente ESG ou no geral staggered grid's são muito mais fodas!
Lembrando do curso de CFD que fiz lendo o artigo do Bartolo, tanto CFD qto
Eletromag usam esse trem. Realmente deve ser bom! Vou tentar implementar um
scalar com isso aew vamos ter duas versões de scalar. Um non-staggered e um
staggered bonito o bicho. Valeus pela dica dos artigos.

—
Reply to this email directly or view it on GitHubhttps://github.com/leouieda/fatiando/pull/52#issuecomment-26206630
.

eusoubrasileiro · 2013-10-12T22:35:23Z

nada como a diversão rsrs

leouieda · 2013-10-16T11:44:02Z

@eusoubrasileiro prefeito! Não esqueça de fazer em outro branch e de avisar
na docstring do módulo que a versão staggered é melhor. Bom trabalho
nerdão!
Em 12/10/2013 19:35, "André" notifications@github.com escreveu:

nada como a diversão rsrs

—
Reply to this email directly or view it on GitHubhttps://github.com/leouieda/fatiando/pull/52#issuecomment-26207494
.

eusoubrasileiro · 2013-12-29T00:23:50Z

@leouieda agora entendi esse trem seu aiii... eu acho... rsrs
só fonte na componente z

https://gist.github.com/eusoubrasileiro/8165804

leouieda · 2013-12-29T00:25:50Z

Hahaha depois precisamos marcar um dia pra discutir a implementação OO
On Dec 28, 2013 10:23 PM, "André" notifications@github.com wrote:

@leouieda https://github.com/leouieda agora entendi esse trem seu
aiii... eu acho... rsrs

—
Reply to this email directly or view it on GitHubhttps://github.com/leouieda/fatiando/pull/52#issuecomment-31308170
.

eusoubrasileiro · 2013-12-29T00:27:06Z

A com certeza... vamos, pela alegria e felicidade de mundo. Até lá feliz ano novo!!! ✋

leouieda added 30 commits June 11, 2013 10:22

Started replacing wavefd algorithm with an equivalent staggered grid.…

1567442

… Still not working

Fixed new FD solution to SH waves and better absorbing boundaries by …

fe79309

…damping all time panels

Added 1D nonreflexive boundary conditions to SH wave FD

00c4b15

Redefined the Mexican hat source in wavefd to use the cutoff freq ins…

af78480

…tead of wavelength

Using 2 panels instead of 3 in wavefd SH simulation

97fa1bc

Added ignore rule for temp image files

8c02191

Replaced P-SV code in wavefd with the Parsimoneous Staggered Grid alg…

6ef8300

…orithm. Not working in water

Updated docstrings in wavefd

9fa265c

Small tweaks to wavefd notebooks

f367e89

Moved notebooks to their own dir

e9d5bfa

Outsourced SquareMesh.img2prop to io.fromimage

e9d0ca4

P-SV simulation unstable. probably error in derivatives of tensions

5b2f876

Updated rules in gitignore

08da159

Made lovewave notebook save snapshots instead of using the animation …

b065916

…package

Making the notebook of lovewaves nicer

4068edf

Replaced Cython ndarray declarations with memory views in wavefd

ef49561

Removed unused variables from seismic._wavefd

f87543a

Tried a different free surface bc for P-SV. Still unstable

ffb9167

Simplified interface to wavefd.elastic_sh. Sources now take the x, z …

987a6d1

…coordinates and convert to indexes

Converted p-sv simulation to new interface

4aed8e4

Moved time step to P-SV into single function

9ba2874

Added function to convert ux, uz into P and S waves. Added blast source

2c0d20e

PSV stable!!! Better free surf bc and fixed bug in PSV time step

f99aed8

(central diffs)

Cleaned up and updated docstrings in wavefd

c8008a1

Added option to convert output of wavefd elastic_psv to P and S

2b6c526

Calculating P and S from ux uz using 4th order centered diffs

1117cd3

Merge branch 'master' into wavefd

40eba2a

Conflicts: cookbook/seismic_wavefd_elastic_sh.py fatiando/io.py fatiando/mesher.py fatiando/seismic/wavefd.py

fixed the wavefd recipes

bf37d4a

Got rid of notebooks dir

6f2ced1

Added a theory section to wavefd describing equations used

74d04d0

leouieda added 2 commits August 26, 2013 15:25

Added function to calculate max dt for wavefd

9de3460

Try installing cython from pip for latest version

4d6eda0

leouieda added a commit that referenced this pull request Sep 2, 2013

Merge pull request #52 from leouieda/wavefd

54ac6a8

Refactor and better implementations of `fatiando.seismic.wavefd`

leouieda merged commit 54ac6a8 into master Sep 2, 2013

leouieda deleted the wavefd branch September 2, 2013 17:12

Uh oh!

Conversation

leouieda commented Aug 26, 2013

Uh oh!

hbueno commented Aug 26, 2013

Uh oh!

eusoubrasileiro commented Aug 27, 2013

Uh oh!

leouieda commented Sep 2, 2013

Uh oh!

eusoubrasileiro commented Sep 2, 2013

Uh oh!

eusoubrasileiro commented Sep 18, 2013

Uh oh!

hbueno commented Sep 18, 2013

Uh oh!

eusoubrasileiro commented Sep 18, 2013

Uh oh!

leouieda commented Sep 18, 2013

Uh oh!

eusoubrasileiro commented Sep 18, 2013

Uh oh!

leouieda commented Sep 18, 2013

Uh oh!

eusoubrasileiro commented Sep 18, 2013

Uh oh!

eusoubrasileiro commented Sep 18, 2013

Uh oh!

eusoubrasileiro commented Oct 12, 2013

Uh oh!

hbueno commented Oct 12, 2013

Uh oh!

eusoubrasileiro commented Oct 12, 2013

Uh oh!

leouieda commented Oct 16, 2013

Uh oh!

eusoubrasileiro commented Dec 29, 2013

Uh oh!

leouieda commented Dec 29, 2013

Uh oh!

eusoubrasileiro commented Dec 29, 2013

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants