CFD Online Discussion Forums

CFD Online Discussion Forums (https://www.cfd-online.com/Forums/)
-   Main CFD Forum (https://www.cfd-online.com/Forums/main/)
-   -   XFlow (https://www.cfd-online.com/Forums/main/91704-xflow.html)

Yaping August 19, 2011 14:14

XFlow
 
Hello,

Anybody has any experience with XFlow?

Thanks,

Yaping

Zep August 25, 2011 12:21

..................................

Saket.Chaudhary August 25, 2011 12:26

Hey,

I am using XFlow, How can I help?

Thanks

sail August 25, 2011 14:20

Quote:

Originally Posted by Saket.Chaudhary (Post 321600)
Hey,

I am using XFlow, How can I help?

Thanks

hi Saket.

Given your experience with Xflow I'd like to ask some questions about it:

1) could you point me to any paper or verification and validation for external , subsonic aerodynamics testcases in which Xflow is used?

2)could you compare the solving time, given the same accuracy, with fluent, starccm, etc...

3) just out of curiosity: how did you came using it? i tried to get a demo licence but the reply was negative... frankly speaking, their hope that i pay 50k€ for a software I've never seen before is wishful thinking. just wandering if other users have bought it because they were sure that it was the right tool for the job or... dunno.

thanks.

Saket.Chaudhary August 25, 2011 14:54

Hey Sail,

In the interest of full disclosure, I work for MSC in the XFlow team.
- Certainly have validation of external aerodynamics on academic problems that are well researched e.g ASMO car model and real industry problems with our customers. The customer data is sensitive as you can understand, I can certainly share other data with you. If you are interested to take a deeper look into XFlow we can talk about how we can accomplish this including trial.
- When you say comparing time, I assume you want to see how much time it takes from the time when one starts to do a simulation, this includes importing the geometry, cleaning up, meshing, setting up the analysis, run time and post processing. When one starts comparing this you will find that the time invested in bringing in CAD models, cleaning up and meshing is eliminated in XFlow...which is the bulk of the time in doing simulations with traditional CFD software's. Users can focus of doing more design iterations in their product development in the same time period.
- I agree with you, I would feel the same way that paying $$ for a software I have never tried doesn't sound like a good idea. We are not looking to get $$ this way either....we would engage with the company and show how XFlow will impact their product development by saving time, saving $$, increasing throughput.
- XFlow is a paradigm change in CFD, when someone says no meshing...first reaction is WOW and also skepticism. Our customers have seen the value it brings to their product development.
- We would love to engage more and show you more in depth into the technology. I have myself used several CFD software's and I still manage to amaze myself everyday running XFlow.
- I am saket.chaudhary@mscsoftware.com, send me note and we can dive deep into XFlow.

Check out the below
http://www.youtube.com/user/XFlowCFD

Thanks
Saket

sail August 25, 2011 15:21

Hi Saket.

Thank you for your prompt reply. Your software is interesting, and i'm writing you an email to have a private feedback.

best regards

leflix August 26, 2011 06:58

Hi Saket,

when looking at the simulation results provided by Xflow http://www.youtube.com/user/XFlowCFD
it's terricfic !!

I even don't think that there exist other softwares able to perform the kind of simulations Xflow did. I think you are 100 light years away from Fluent, Star CCM+ and others.

But....

How did you validated your results ? How could you assure that your computed results are good?

I guess you have performed some benchmarks on classical flows where other numerical or experimental results are available.
Did you communicate on this ? Where are your comparisons?

Have you also compared your results in the case of a high resolution DNS turbulent flow (channel flow for example) where huge databases (velocity field statistics) are available one very fine grids ( 4096^3 ~68 billions grid points ) ?



Quote:

Originally Posted by Saket.Chaudhary (Post 321617)
- When you say comparing time, I assume you want to see how much time it takes from the time when one starts to do a simulation, this includes importing the geometry, cleaning up, meshing, setting up the analysis, run time and post processing. When one starts comparing this you will find that the time invested in bringing in CAD models, cleaning up and meshing is eliminated in XFlow...which is the bulk of the time in doing simulations with traditional CFD software's. Users can focus of doing more design iterations in their product development in the same time period.

yes it will be interessting to know for a very complicated flow (complex geometry, complex physics ) how much time it takes from the starting point until post-processing?


But also the CPU time only required to compute the flow, for example during 10 physical seconds of simulations ( how many time steps, the magnitude of the time step...)

What is the overall precision of the method ? first order, second order, higher?

leflix August 26, 2011 07:10

Is XFlow based on lattice boltzman method like the code Powerflow?

cfdnewbie August 26, 2011 07:19

Quote:

Originally Posted by leflix (Post 321722)
Hi Saket,

when looking at the simulation results provided by Xflow http://www.youtube.com/user/XFlowCFD
it's terricfic !!

I even don't think that there exist other softwares able to perform the kind of simulations Xflow did. I think you are 100 light years away from Fluent, Star CCM+ and others.

I'm not familiar with the capabilities of engineering codes like Fluent and such, but there are many and more research codes that are able to perform this kind of simulations and do so routinely. However, these codes are generally not open to the public or are in some cases classified. These videos are certainly nice, but similar stuff can be seen at decent CFD conferences as well. I'm not saying that they are not impressive, but let's keep it real :)
[quote]
But....

How did you validated your results ? How could you assure that your computed results are good?

Quote:

I guess you have performed some benchmarks on classical flows where other numerical or experimental results are available.
Did you communicate on this ? Where are your comparisons?
Again, I haven't worked outside academia, so this is new to me as well. How in general are commercial CFD codes validated? Are there publications/papers about this kind of stuff? Nice and fast pictures are one thing, but there's no free lunch in CFD. If you want to be lightning fast, you have to sacrifice accuracy.... So anybody in the know, are there any publications about the accuracy and even consistency of commercial codes available?

Quote:

Have you also compared your results in the case of a high resolution DNS turbulent flow (channel flow for example) where huge databases (velocity field statistics) are available one very fine grids ( 4096^3 ~68 billions grid points ) ?
DNS codes are almost always of a spectral nature, and I sincerely doubt that XFlow has a spectral basis, so the comparison with a high RE DNS is not so relevant in my opinion. What I would like to see before I touch any piece of software not written by myself or proven is this: Does the solution of XFlow converge to the DNS solution for low/moderate Re? At what cost?

Is there any organization/publication that compares commercial codes in terms of accuracy and efficiency?







Quote:

What is the overall precision of the method ? first order, second order, higher?
Yes, interesting!

leflix August 26, 2011 08:49

Quote:

Originally Posted by cfdnewbie (Post 321728)
I'm not familiar with the capabilities of engineering codes like Fluent and such, but there are many and more research codes that are able to perform this kind of simulations and do so routinely.

If there are research codes able to perform this kind of simulations I would like to know which one?

On a very specific point you can find a research code able to do it. I mean if you take a code dedicated to free surface flow for example it will be perhaps able to perform one of the simulations seen... But this research code won't be able perhaps to manage in the same time
moving meshes or coupled FSI or heat transfer etc....

generally research codes are very specialized in certain tasks and not versatile like commercial codes.

Why I was so amazed by Xflow it's because of the performed simulations in a lot of very different areas ( automotive, aerospace, aerodynamic, free surface flows, FSI, heat transfer,..)
None of existing research codes are able to perform well in so different areas.
Just for full disclosure I'm from academic side and I do not sell any commercial code nor use none of them in my research activities.

cfdnewbie August 26, 2011 09:19

Quote:

Originally Posted by leflix (Post 321751)
If there are research codes able to perform this kind of simulations I would like to know which one?

On a very specific point you can find a research code able to do it. I mean if you take a code dedicated to free surface flow for example it will be perhaps able to perform one of the simulations seen... But this research code won't be able perhaps to manage in the same time
moving meshes or coupled FSI or heat transfer etc....

generally research codes are very specialized in certain tasks and not versatile like commercial codes.

ok, I agree with you on that point. There are some versatile research codes like the ones used i.e. by Onera or Airbus or academic research groups and such that can do compressible/incompr, FSI, heat transfer, but in general, you are right, research codes tend to be very specific in what they do.

Quote:

None of existing research codes are able to perform well in so different areas.
Maybe it's true, I don't claim to have an overview over all the research codes out there, I'm just saying that I'm not convinced of XFlow in terms of accuracy. We have no way of telling how "well" it actually performs. Take the flapping wings for example... it would look equally impressive even if the drag and lift were totally off the mark...


I agree with you that it's impressive to see how many codes/methods have been coupled in this approach, and nice to see that it works - at least on a qualitative level. However, without careful validation and verification, which has been published in reviewed journals, it remains - at least for me - pretty pictures....

Cheers

leflix August 26, 2011 10:39

Quote:

Originally Posted by cfdnewbie (Post 321754)
ok, I agree with you on that point. There are some versatile research codes like the ones used i.e. by Onera or Airbus or academic research groups and such that can do compressible/incompr, FSI, heat transfer, but in general, you are right, research codes tend to be very specific in what they do.

you are right and I have to admit that the code CEDRE from ONERA can do indeed a lot of stuffs even if it is very specialized in multiphasic reactive flows. But for example if you want to perform aeroacoustic you have to switch to another product of ONERA named SPACE.
It is in that sense that verstatility of research codes are weaker than commercial ones. And CEDRE is probably one of the most versatile research code.

Quote:

I'm just saying that I'm not convinced of XFlow in terms of accuracy. We have no way of telling how "well" it actually performs. Take the flapping wings for example... it would look equally impressive even if the drag and lift were totally off the mark...
[.....] However, without careful validation and verification, which has been published in reviewed journals, it remains - at least for me - pretty pictures....
You got it !!
Yes I totally agree with you and that was my point in my first post.
Without a tons of rigorous validations and the communication of the real cost (CPU time) for some given benchmarks attested by published papers in high level reviewed journals, these impressive animations will remain nice moving pictures.

The flapping wings is indeed a very impressive example, but who can say it is right and accurate?
We just need to be convinced to believe it.

Saket.Chaudhary August 26, 2011 15:50

Thanks Guys. So we agree that the technology is innovative and designed to help bring down the barriers from traditional CFD by giving power to the users in doing more design iterations in less time.
I agree with your points that users of the technology who invest $$ will have to see if the technology fits their application, solves problems in acceptable engineering time and accuracy.
If you are interested in using the technology send me a note and I can share validation examples and also have your own models run in XFlow to show the correlation. Totally on board with your thoughts that seeing your own products is best way to believe.

cfdnewbie August 26, 2011 15:55

Quote:

Originally Posted by Saket.Chaudhary (Post 321828)
Thanks Guys. So we agree that the technology is innovative and designed to help bring down the barriers from traditional CFD by giving power to the users in doing more design iterations in less time.
I agree with your points that users of the technology who invest $$ will have to see if the technology fits their application, solves problems in acceptable engineering time and accuracy.
If you are interested in using the technology send me a note and I can share validation examples and also have your own models run in XFlow to show the correlation. Totally on board with your thoughts that seeing your own products is best way to believe.


Thank you for your reply, Saket. Could you please comment on the question of validation and accuracy of XFlow? That would be highly appreciated! thank you!

Cheers!

f-w October 4, 2011 13:07

How about doing an analysis on the aircraft from the Drag Prediction Workshop:
http://aaac.larc.nasa.gov/tsab/cfdlarc/aiaa-dpw/

I would be interested in the results along with documented steps going from geometry to post-processing, including associated times.

Saket.Chaudhary October 5, 2011 10:00

Hey,

How about the high lift prediction workshop.
http://hiliftpw.larc.nasa.gov/

We have done the high lift simulations which shows the coreleation of XFlow results with the NASA test results of drag, lift, polar curve etc.

Please ping me at saket.chaudhary@mscsoftware.com or at +1 804-405-1242 and I can share the results with you.

Best,
Saket

f-w October 5, 2011 12:38

Quote:

Originally Posted by Saket.Chaudhary (Post 326772)
Hey,

How about the high lift prediction workshop.
http://hiliftpw.larc.nasa.gov/

We have done the high lift simulations which shows the coreleation of XFlow results with the NASA test results of drag, lift, polar curve etc.

Please ping me at saket.chaudhary@mscsoftware.com or at +1 804-405-1242 and I can share the results with you.

Best,
Saket

Saket, your results from the high lift workshop are of interest as well, so I'll send you an e-mail request.

However, I mainly conduct cruise analysis (as do many other external aero engineers), and the drag prediction workshop is of higher interest. Please add it to your company's to do list.

Thanks

mohammad.sarfaraz November 14, 2012 08:19

XFLOW userguide
 
I am a M.Sc. student.
Can anyone send me the userguide of the XFLOW?

Thanks in advanced;

sarfaraz@alum.sharif.edu

arjun November 14, 2012 17:34

Quote:

Originally Posted by leflix (Post 321751)
If there are research codes able to perform this kind of simulations I would like to know which one?
.


I could do that today in 2 codes and if I spent time may be around 2 more weeks I could do that in 4 codes.

Currently that could be done by:
1. The in house code that I wrote for my last company. We could do that for last 4 years. Validated the results against experimental data from wind tunnel too.
2. Inavier has immersed boundary so it can also do it.

I could do:
3. I have written 2D versions of lattice boltzman that support immersed boundary. If given time I could write down 3D versions too that would work just like xflow.

4. Now that I could program starccm too, I could add immersed boundary and could do that with it too. It won't take more than 2 weeks to do that.


Further I assume fluent had immersed boundary code long ago so that could have also done it. And Ansys CFX also has immersed solid , so i guess that could also do it.


Other than that literature is filled with cases where people have done this type of thing. From acadmics I can tell you that inhouse of Tohoko univ of Prof Nakahashi's group can also do it. Probably frontflow also can do it.

mecobio June 30, 2013 17:17

Quote:

Originally Posted by Saket.Chaudhary (Post 321617)
Hey Sail,

In the interest of full disclosure, I work for MSC in the XFlow team.
- Certainly have validation of external aerodynamics on academic problems that are well researched e.g ASMO car model and real industry problems with our customers. The customer data is sensitive as you can understand, I can certainly share other data with you. If you are interested to take a deeper look into XFlow we can talk about how we can accomplish this including trial.
- When you say comparing time, I assume you want to see how much time it takes from the time when one starts to do a simulation, this includes importing the geometry, cleaning up, meshing, setting up the analysis, run time and post processing. When one starts comparing this you will find that the time invested in bringing in CAD models, cleaning up and meshing is eliminated in XFlow...which is the bulk of the time in doing simulations with traditional CFD software's. Users can focus of doing more design iterations in their product development in the same time period.
- I agree with you, I would feel the same way that paying $$ for a software I have never tried doesn't sound like a good idea. We are not looking to get $$ this way either....we would engage with the company and show how XFlow will impact their product development by saving time, saving $$, increasing throughput.
- XFlow is a paradigm change in CFD, when someone says no meshing...first reaction is WOW and also skepticism. Our customers have seen the value it brings to their product development.
- We would love to engage more and show you more in depth into the technology. I have myself used several CFD software's and I still manage to amaze myself everyday running XFlow.
- I am saket.chaudhary@mscsoftware.com, send me note and we can dive deep into XFlow.

Check out the below
http://www.youtube.com/user/XFlowCFD

Thanks
Saket


Excuse me, with all respect, but LB need indeed mesh.
Some people use LB models based Cartesian, orther on spherical, etc.
Some people use refinned meshes in LB too.
It is at those mesh-nodes where collisions happen, usually modelled by the SRT BGK/Wallander model or MRT.
So, where the meshless-myth come from?!


All times are GMT -4. The time now is 14:55.