Asteroid simulation shader

[jgalazm]

From @jgalazm on July 3, 2018 22:5

According to this reference

http://www.geosci.usyd.edu.au/users/prey/Teaching/Geos-2111GIS/Tsunami/Ward-Icarus00-ImpactTsumani.pdf

The initial condition should be given by equation (2), whose parameters can be obtained and aproximated in different ways.

The simplest is that of equation (16) with the following parameters:

(user modifiable)

• Dc (cavity depth)
• dc ( 2Rc, cavity diameter)
• rho_I : impactor (asteroid) mass density
• V_I: impactor speed

(non user-modifiable)

• rho_w : water mass density
• g: gravity acceleration
• epsilon: ammount of energy converted into tsunami waves. According to lab experiments this is about 15% (paragraph after equation (15) )

@spereirag @kamakshidasan

Copied from original issue: Inria-Chile/tsunami-lab#149

[jgalazm]

From @kamakshidasan on July 4, 2018 8:57

Figure 14 looks like something we can replicate as well. Do you think we could have something similar to Figure 5 and 6 from the main data?

[jgalazm]

Figure 14 should be easy to do. We can get it changing the output parameter.

Figures 5 and 6 I'm not sure. We would have to confirm from the vtk files (in Paraview) how to extract profiles and time series, but this may require an important amount of time depending on the experience with this data and model (the multi physics one). I have only used Paraview/vtk files a few times before.

[jgalazm]

From @kamakshidasan on July 4, 2018 19:41

I have lived on Paraview/VTK for the last two years and I continue to do so. I sort of ran few experiments on the raw Tsunami data on Paraview, and have some results.

On Wed, Jul 4, 2018 at 9:35 PM, José Galaz Mora notifications@github.com wrote:

Figure 14 should be easy to do. We can get it changing the output parameter.

Figures 5 and 6 I'm not sure. We would have to confirm from the vtk files (in Paraview) how to extract profiles and time series, but this may require an important amount of time depending on the experience with this data and model (the multi physics one). At least I only have used Paraview/vtk files a few times before.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/Inria-Chile/tsunami-lab/issues/149#issuecomment-402548055, or mute the thread https://github.com/notifications/unsubscribe-auth/AI7akKJ3YcZxMZAuLl2WgXXxPQh60EcQks5uDRjmgaJpZM4VBsK4 .

[jgalazm]

Cool. Can you try to extract some isosurfaces of water fraction around 1?that should be good for estimating the free surface. What do you think?

We could use that as initial condition in Nami and also extract profiles of the free surface.

On Wed, Jul 4, 2018, 15:41 Kamakshidasan notifications@github.com wrote:

I have lived on Paraview/VTK for the last two years and I continue to do so. I sort of ran few experiments on the raw Tsunami data on Paraview, and have some results.

On Wed, Jul 4, 2018 at 9:35 PM, José Galaz Mora notifications@github.com wrote:

Figure 14 should be easy to do. We can get it changing the output parameter.

Figures 5 and 6 I'm not sure. We would have to confirm from the vtk files (in Paraview) how to extract profiles and time series, but this may require an important amount of time depending on the experience with this data and model (the multi physics one). At least I only have used Paraview/vtk files a few times before.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub < https://github.com/Inria-Chile/tsunami-lab/issues/149#issuecomment-402548055 , or mute the thread < https://github.com/notifications/unsubscribe-auth/AI7akKJ3YcZxMZAuLl2WgXXxPQh60EcQks5uDRjmgaJpZM4VBsK4

.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/Inria-Chile/tsunami-lab/issues/149#issuecomment-402548777, or mute the thread https://github.com/notifications/unsubscribe-auth/ADQF-e6wMxdOuFI86fPWWO6CqOcXmU6Oks5uDRpggaJpZM4VBsK4 .

[jgalazm]

From @kamakshidasan on July 5, 2018 16:15

Look at the screenshot.

This is isovalue = 1 for the last time step. The colour is the magnitude of velocity. Is this similar to what you had in mind?

[jgalazm]

Yes. Definitely. After that we should interpolate these values in a 2D grid. Can you do it? Also, can you see if it changes too much with other values, such as 0.99 and 0.98 for example?

José Galaz Estudiante Ingeniería Civil Industrial Matemática Pontificia Universidad Católica de Chile https://jgalazm.github.io/

2018-07-05 12:15 GMT-04:00 Kamakshidasan notifications@github.com:

Look at the screenshot.

This is isovalue = 1 for the last time step. The colour is the magnitude of velocity. Is this similar to what you had in mind?

[image: tsunami-isosurface] https://user-images.githubusercontent.com/9362064/42335043-56da470c-807f-11e8-8594-61aef6c38900.png

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/Inria-Chile/tsunami-lab/issues/149#issuecomment-402775946, or mute the thread https://github.com/notifications/unsubscribe-auth/ADQF-UY6shgtNdrrOKEJL20OyfeMJXp0ks5uDjutgaJpZM4VBsK4 .

[jgalazm]

From @kamakshidasan on July 5, 2018 18:38

These values are already in a 2D grid. Why do we want to interpolate them?

Okay, I will look at the changes. Which time-steps should we be looking at? Before/during/after collision?

[jgalazm]

Because there may be more than one value for a single grid point (since it is a 3D surface), like "C" shaped profiles.

With Seba, we think that, if we use not only the free surface info but also the vertical average of the horizontal speeds (from the bottom to the surface) then we could take the last frame of the simulation.

Otherwise, if we only have the free surface elevation I think we should use the frame closest in time to when the maximum elevation happens and assume zero velocity from there.

*Edit: @kamakshidasan, wrote some clarifications

2018-07-05 14:38 GMT-04:00 Kamakshidasan notifications@github.com:

These values are already in a 2D grid. Why do we want to interpolate them?

Okay, I will look at the changes. Which time-steps should we be looking at? Before/during/after collision?

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/Inria-Chile/tsunami-lab/issues/149#issuecomment-402815402, or mute the thread https://github.com/notifications/unsubscribe-auth/ADQF-RwFF_7KL4gSYy3zC399GtUp3r-nks5uDl0jgaJpZM4VBsK4 .

[jgalazm]

From @kamakshidasan on July 6, 2018 12:21

Yikes - I barely understood those sentences. We should get on a call sometime and I'll screencast our Paraview experiments.

[jgalazm]

For the first implementation I'm using a "mockup model", that is, just put some dummy formulas (exponential initial condition) into the shader but the right coordinate transformation and uniforms and consts, just to see if data is given right. see commit e0113d2348e4b00f88c65b5daf0eb09fee808158

[jgalazm]

I just got the mockup asteroid working. (gaussian shape)

Next: formula that uses the data.

[jgalazm]

Just realized I got the formulas wrong:

The input parameters are

• center east,north coordinates: ce,cn
• Vi: impactor speed
• Ri: impactor radius
• rho_i: impactor mass density
• epsilon: fraction of energy transferred to the tsunami ~ 0.15

So with (Vi, rho_i, epsilon) we obtain Q. With Q and R_i we obtain Dc( With Dc we obtain dc and Rc=dc/2. Also Rd = Rc sqrt(2)

So this is pluggable into:

[jgalazm]

From @kamakshidasan on July 12, 2018 7:52

Wow! The mockup simulation is amazing! :)

[jgalazm]

Sorry, won't be able to connect again until 1 hour I think :-( (and the university wont allow me to log into my inria email either ... )

Just to finish the point, I wanted to say is that to get the surface as a function of X and Y only, we should interpolate it on a uniform (X,Y) grid later.

I can do that if you send me the (x,y,z) data of the "relevant" isosurface we discussed in a csv for example. For now it can be for any threshold around 1.

Lets discuss that later.

Thanks!

[jgalazm]

From @kamakshidasan on July 12, 2018 16:54

No issues!

As I see it, I have following tasks ahead of me. Correct me if you disagree with any of the below.

• We will have to find out at which timestep the collision of the asteroid takes place with the water surface. Look at Figure 5 from here https://sciviscontest2018.org/wp-content/uploads/sites/19/2017/09/DeepWaterImpactEnsembleDataSet_Revision1.pdf. The way to figure this out will be by looking at all timesteps and collecting essential data. Once after data collection, we will have to analyse which of the timesteps corresponds to collision - this could be done using plots of amplitude/wavelength/intensity. For instance, Figure 4 from here http://www.geosci.usyd.edu.au/users/prey/Teaching/Geos-2111GIS/Tsunami/Ward-Icarus00-ImpactTsumani.pdf initially starts from impact collision, whereas our data has "asteroid approach" as well. Till then, I will extract the water fraction from all frames, ready for analysis. We will require a domain scientists' expertise/prior understanding to know what plots we should look at from the extracted data -- who can we ask/where should we look?

• Once we have found the a "good" timestep, we will have to find an isosurface with maximum features for you. This part will have computational topology and should be simple from my end.

• This iso-surface should be interpolated to 2D grid. This also fine. :)

-- The first step will take some amount of time for data processing and cleaning. Till then, I will give you a sample 2D grid to play around with. What format do you want to be in? -- What exactly do you need this 2D grid for? What features/parameters in the equation will you be using it for?

On Thu, Jul 12, 2018 at 4:32 PM, José Galaz Mora notifications@github.com wrote:

Sorry, won't be able to connect again until 1 hour I think :-( (and the university wont allow me to log into my inria email either ... )

Just to finish the point, I wanted to say is that to get the surface as a function of X and Y only, we should interpolate it on a uniform grid later.

I can do that if you send me the data of the "relevant" isosurface we discussed (with some threshold).

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/Inria-Chile/tsunami-lab/issues/149#issuecomment-404533180, or mute the thread https://github.com/notifications/unsubscribe-auth/AI7akHzyicw3r9sjEiyfJQrCKl2Mf4Duks5uF136gaJpZM4VBsK4 .

[jgalazm]

From @kamakshidasan on July 12, 2018 17:29

-- The first step will take some amount of time for data processing and cleaning. Till then, I will give you a sample 2D grid to play around with. What format do you want to be in? -- What exactly do you need this 2D grid for? What features/parameters in the equation will you be using it for?

I'm attaching a sample 2D grid. Our end result will look similar to this.

On Thu, Jul 12, 2018 at 6:54 PM, Adhitya Kamakshidasan < adhitya.kamakshidasan@gmail.com> wrote:

No issues!

As I see it, I have following tasks ahead of me. Correct me if you disagree with any of the below.

• We will have to find out at which timestep the collision of the asteroid takes place with the water surface. Look at Figure 5 from here https://sciviscontest2018.org/wp-content/uploads/sites/19/2017/09/DeepWaterImpactEnsembleDataSet_Revision1.pdf. The way to figure this out will be by looking at all timesteps and collecting essential data. Once after data collection, we will have to analyse which of the timesteps corresponds to collision - this could be done using plots of amplitude/wavelength/intensity. For instance, Figure 4 from here http://www.geosci.usyd.edu.au/users/prey/Teaching/Geos-2111GIS/Tsunami/Ward-Icarus00-ImpactTsumani.pdf initially starts from impact collision, whereas our data has "asteroid approach" as well. Till then, I will extract the water fraction from all frames, ready for analysis. We will require a domain scientists' expertise/prior understanding to know what plots we should look at from the extracted data -- who can we ask/where should we look?

• Once we have found the a "good" timestep, we will have to find an isosurface with maximum features for you. This part will have computational topology and should be simple from my end.

• This iso-surface should be interpolated to 2D grid. This also fine. :)

-- The first step will take some amount of time for data processing and cleaning. Till then, I will give you a sample 2D grid to play around with. What format do you want to be in? -- What exactly do you need this 2D grid for? What features/parameters in the equation will you be using it for?

On Thu, Jul 12, 2018 at 4:32 PM, José Galaz Mora <notifications@github.com

wrote:

Sorry, won't be able to connect again until 1 hour I think :-( (and the university wont allow me to log into my inria email either ... )

Just to finish the point, I wanted to say is that to get the surface as a function of X and Y only, we should interpolate it on a uniform grid later.

I can do that if you send me the data of the "relevant" isosurface we discussed (with some threshold).

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/Inria-Chile/tsunami-lab/issues/149#issuecomment-404533180, or mute the thread https://github.com/notifications/unsubscribe-auth/AI7akHzyicw3r9sjEiyfJQrCKl2Mf4Duks5uF136gaJpZM4VBsK4 .

[jgalazm]

From @kamakshidasan on July 12, 2018 17:30

sample-slice.csv.zip

[jgalazm]

Lets continue this discussion on the gitlab issue please https://gitlab.inria.fr/epietrig/seawall/issues/1.

Just to keep this one for tracking changes on the asteroid shader, please.