CFD Online Discussion Forums (https://www.cfd-online.com/Forums/)
-   Main CFD Forum (https://www.cfd-online.com/Forums/main/)
-   -   Implementation of QUICK scheme (https://www.cfd-online.com/Forums/main/1734-implementation-quick-scheme.html)

 Romuald Skoda January 20, 2000 04:02

Implementation of QUICK scheme

Hallo,

I am implementing a QUICK convection scheme into a 2d pressure-correction (SIMPLE-like) RANS code for non-orthogonal grids. The code in formulated in cartesian co-ordinates. The time discretisation is fully-implicit. I intend to scale-up the code to 3D later.

1) In the open literature I found the QUICK formulation for non-uniform grids, not for non-orthogonal and body-fitted grids. So firstly I implemted the scheme proposed for non-uniform but orthogonal grids. Is there an advice or a reference what could be done better with non-orthogonal grids?

2) I neglected the transverse curvature terms. Does someone know some references about the implementation of the QUICK scheme including the transverse curvature contribution and/or its performance for highly convective / turbulent flow?

3) I suppose that the scheme works equally for steady-state and transient solutions, if I use an implicit time discretisation. Am I right? What about an explicit time-marching scheme? I intend to implement an explicit pressure-correction scheme. Can I expect that I do not have to modify the interpolation scheme?

Romuald

 Dr. Hrvoje Jasak January 20, 2000 05:16

Re: Implementation of QUICK scheme

Hi,

You haven't chosen my favourite scheme (you'll have boundedness problems later!), but I think I can help:

1) QUICK for non-uniform grids in fine and you don't need to modify it any further. All 'curvilinear' or 'non-orthogonal' corrections are unnecessary - they will just increase your computational molecule but won't give you any additional accuracy.

2) You did fine - forget about curvature.

3) The scheme is fine. You'll have to under-relax the system for pressure correction (as does everybody else); this kind of stuff is well tested and publicised with both SIMPLE and PISO.

The only trouble you can expect (to my knowledge is):

a) boundaries need special treatment as there's no "far upwind" node. b) the scheme is unbounded, i.e. for a variable which is laways positive (e.g. T or k) you can get negative face values, which causes trouble. However, there's a bunch of NVD-modifications to QUICK (called SMART of SHARP, I'm not quite sure) - you will need those when you start convecting scalars c) QUICK is a real bitch to implement on unstructured meshes, as you loose the directions. But that's another subject.....

Hrv

 Romuald Skoda January 20, 2000 06:39

Re: Implementation of QUICK scheme

Thank you very much for your quite encouraging advice. Actually, I know about the boundedness problem and intend to implement the SMART scheme, which includes the QUICK scheme. May I ask what is your favourite scheme you mentioned, which does not seem to suffer from boundedness problems, even if no flux limiter is used?

Thanks Romuald

 Dr. Hrvoje Jasak January 20, 2000 09:28

Re: Implementation of QUICK scheme

Hi,

Unfortunately, there's no linear scheme other than Upwind differencing that does not suffer from unboundedness, so I'm afraid that you'll always need a limiter. I have developed a scheme called Gamma (Int. J. Numer. Meth. Fluids,31:431-449, 1999) and it's been extensively tested on a lot of stuff, from combustion to two-phase and free surface flows. It is an NVD (limited!) scheme and its main advantages are i) it works on arbitrarily unstructured meshes (no need for "far upwind node") ii) compact support (blends UD and CD) iii) no time-step dependency and iv) (and most important) it's the simplest I could do! However, (I don't know the aim of your project), I think you made a decent choice, unless you ultimately want to make your code arbitrarily unstructured, so I think you can happily just QUICK away!

Have fun,

Hrv

 Seok Ki Choi January 28, 2000 20:43

Re: Implementation of QUICK scheme

Hi Dr.Hrvoje Jasak

You mentioned that you have tested your Gamma scheme on a lot of practical flow problems. How was accuracy of your Gamma scheme for practical flow calculations when compared with QUICK, SMART, HLPA or SOUCUP schemes ?

Seok Ki Choi

 technocrat.prakash March 23, 2009 00:18

Quote:
 Originally Posted by Romuald Skoda ;7215 Hallo, I am implementing a QUICK convection scheme into a 2d pressure-correction (SIMPLE-like) RANS code for non-orthogonal grids. The code in formulated in cartesian co-ordinates. The time discretisation is fully-implicit. I intend to scale-up the code to 3D later. 1) In the open literature I found the QUICK formulation for non-uniform grids, not for non-orthogonal and body-fitted grids. So firstly I implemted the scheme proposed for non-uniform but orthogonal grids. Is there an advice or a reference what could be done better with non-orthogonal grids? 2) I neglected the transverse curvature terms. Does someone know some references about the implementation of the QUICK scheme including the transverse curvature contribution and/or its performance for highly convective / turbulent flow? 3) I suppose that the scheme works equally for steady-state and transient solutions, if I use an implicit time discretisation. Am I right? What about an explicit time-marching scheme? I intend to implement an explicit pressure-correction scheme. Can I expect that I do not have to modify the interpolation scheme? Thanks a lot in advance! Romuald
Certainly, this is not a advice for you. Actual purpose is I need the open literature for the Quick Scheme. Can you send me to my mail technocrat.prakash@gmail.com

 ztdep March 23, 2009 04:03

Quote:
 Originally Posted by Dr. Hrvoje Jasak ;7220 Hi, Unfortunately, there's no linear scheme other than Upwind differencing that does not suffer from unboundedness, so I'm afraid that you'll always need a limiter. I have developed a scheme called Gamma (Int. J. Numer. Meth. Fluids,31:431-449, 1999) and it's been extensively tested on a lot of stuff, from combustion to two-phase and free surface flows. It is an NVD (limited!) scheme and its main advantages are i) it works on arbitrarily unstructured meshes (no need for "far upwind node") ii) compact support (blends UD and CD) iii) no time-step dependency and iv) (and most important) it's the simplest I could do! However, (I don't know the aim of your project), I think you made a decent choice, unless you ultimately want to make your code arbitrarily unstructured, so I think you can happily just QUICK away! Have fun, Hrv
Dear friends:
Regards

 Hochola March 23, 2009 04:11

Hi, if it's possible I would also like to get a copy of your paper about the GAMMA scheme. Could you e-mail it to me, if it's not a problem.

Thank you

 ziemowitzima July 14, 2010 15:10

Dear Dr HJ

You wrote that QUICK scheme is unbounded, but in OF-1.6 UserGuide in:
Table 4.10: Behaviour of interpolation schemes used in divSchemes.
It is written that QUICK is bounded.

could you tell me which statment is correct ?

Best,
ZM

 NickG August 3, 2010 11:50

There's a link to Prof.(!) Jasak's PhD at
http://openfoamwiki.net/index.php/IcoFoam

 ziemowitzima August 3, 2010 12:23

Thank You !
It is really great works !

 All times are GMT -4. The time now is 00:19.