fvc::div(phi,rho) vs fvc::div(phi)

nimasam · February 4, 2011, 09:29

hello
could you tell me whats the difference between
fvc::div(phi,rho) vs fvc::div(phi) ?

thank in advance

maysmech · February 4, 2011, 10:54

Quote:

Originally Posted by nimasam

hello
could you tell me whats the difference between
fvc::div(phi,rho) vs fvc::div(phi) ?

thank in advance

Hi Nima,
phi is rho*U so div(phi)=divergence of ((rho)U) and div (phi,U)=divergence of ((rho)UU)
I haven't seen div(phi,rho) but it should be divergence of ((rho)(rho)U).
Read user guide for more information:
http://www.openfoam.com/docs/user/fv...20-1120004.4.5

Regards,

haze_1986 · October 17, 2012, 03:43

Hi, may I also ask, in icoFoam I noticed that
fvm::ddt(U) + fvm::div(phi, U) - fvm::laplacian(nu, U)
why is phi used in convective div() term when nu (I suppose kinematic viscosity) is used in the laplacian term? shouldn't left hand side be divided by rho?

nimasam · October 17, 2012, 04:16

phi = Uf & Sf, which Sf is face area , and Uf is velocity at face center
so its a volume flux {m.s}
and
nu = mu/rho

haze_1986 · October 17, 2012, 04:42

Please correct me if I am wrong.

ddt(U) [m/s2]

laplacian(nu,U) [m/s2] I assume nu here = mu/rho

grad(p) [m/s2] I assume pressure here has been divided by rho as well?

However, div(phi, U)

If phi = rho * U, then the unit will be different?

meindert · October 17, 2012, 05:59

Hi,

You should realise that the definition of phi is different for an incompressible flow solver (e.g. icoFoam) than for an compressible flow solver.
For an incompressible flow solver phi = U, for a compressible flow solver phi = rho*U. Thus, the face flux for an incompressible flow solver is (S_f · phi_f) = (S_f · U_f), which is exactly what Nima was trying to tell you.

To be honest, the notation in OpenFOAM is a bit sloppy. The face flux is given the name phi, which is confusing if you write

fvm::ddt(U) + fvm::div(phi, U) - fvm::laplacian(nu, U).

I suggest you have a look at section 2.4 of the Programmer's Guide, which is located in the $WM_PROJECT_DIR/doc/Guides-a4 directory.

camille131 · October 20, 2012, 10:20

Hello and thank you for the precision .
I am actually trying to run a poiseuille case in channel with cyclic conditions.
I don't know If I have to choose icoFoam or channelFoam solver.

Actually in the channelFoam code I don't know what represents
sgsModel->divDevBeff(U) in the equation? see http://foam.sourceforge.net/docs/cpp/a02641_source.html on line 63

Thank tou for your answer.

Camille

meindert · October 20, 2012, 11:40

Hi Camille,

First, I think it would have been more appropriate to open a new thread for this. My guess is that a little searching on the forum would have gotten you a long way as well.

To answer your question, go with the channelFoam solver. It has been developed to do exactly what you want. Have a look at the chan395 tutorial.
The definition of divDevBeff depends on the particular subgrid-scale model that you use. For most subgrid-scale models this term is defined in GenEddyVisc.C.

http://foam.sourceforge.net/docs/cpp...ce.html#l00096

Use Doxygen to find out more about the different subgrid-scale models. Doxygen is there for a reason.

camille131 · October 20, 2012, 12:53

Thank you for your answer actually I also have to kearn how to search correctly on the help guide ^^

but so , what s the difference between nuEff and nu? and what does the temperature T have to do here? it because the problem is coupled? in laminar case we don't care?

thank you
sorry it s not so easy to come from nowhere ..

I have then introduced a new subject for other questions about Poiseuille. here: http://www.cfd-online.com/Forums/ope...tml#post387650

chaolian · November 9, 2012, 00:49

If alpha = 0.5*Foam::erf(4.0*(T-Tmelt)/(Tl-Ts))+scalar(0.5);
Are fvc::div(phi,alpha)
and 4.0*exp(-pow(4.0*(T-Tmelt)/(Tl-Ts),2))/Foam::sqrt(pi)/(Tl-Ts)*(U & fvc::grad(T))
the same?
From the results, they are not quite the same. Why? Is there anything different other than a simple algebraic transformation?

meindert · November 9, 2012, 10:04

Hi Chad,

No, those are not the same. The most important difference is the location at which the expressions are evaluated.

If fvc::div(phi,alpha) is used, alpha is evaluated at the cell face according to equation (2.16) in the Programmer's Guide. For a central difference scheme this value is obtained by linear interpolation of the cell center values.

If 4.0*exp(-pow(4.0*(T-Tmelt)/(Tl-Ts),2))/Foam::sqrt(pi)/(Tl-Ts)*(U & fvc::grad(T)) is used, the whole expression is evaluated at the cell center.

You could write out the seperate discretisations and you will see that they are not the same. I would recommend you to use fvc::div(phi,alpha), since it maintains the conservative form of the convection term.

chaolian · November 9, 2012, 13:25

Quote:

Originally Posted by meindert

Hi Chad,

No, those are not the same. The most important difference is the location at which the expressions are evaluated.

If fvc::div(phi,alpha) is used, alpha is evaluated at the cell face according to equation (2.16) in the Programmer's Guide. For a central difference scheme this value is obtained by linear interpolation of the cell center values.

If 4.0*exp(-pow(4.0*(T-Tmelt)/(Tl-Ts),2))/Foam::sqrt(pi)/(Tl-Ts)*(U & fvc::grad(T)) is used, the whole expression is evaluated at the cell center.

You could write out the seperate discretisations and you will see that they are not the same. I would recommend you to use fvc::div(phi,alpha), since it maintains the conservative form of the convection term.

Hello, Meindert,Thanks for the reply. I agree with you about the conservative aspect. But I am confused since the results are not correct. Say, a 2D cavity is initially solid and starts melt from the left side. The initial internal temperature is 301.45, and BC is left:311.15 and right: 301.45, up and down insulated. melting temperature is 302.
But at t=120s, temperature field is: (
309.228
306.009
302.865
303.179
.
.
.
)
As you can see, the temperature field is not monotone which is not correct.

meindert · November 9, 2012, 13:48

Hi Chad,

It is hard to tell what the source of your problem is, because you are not giving any details about the simulation itself. I suggest you open a new thread about your problem, so we stay on-topic here.

Try to give as much details about your simulation as possible. You could think of boundary conditions, grid resolution, numerical schemes, etc. Perhaps it is a good idea to post a part of the solver output aswell.

woodwick · March 12, 2013, 04:59

Maysam,
Do you know if does exist a tutorial or guide that list side to side the mathematical equations on one side and on the other side how they are written in the openFoam code. I have found something here http://www.foamcfd.org/Nabla/guides/UserGuidese14.html, but I was searching for something with some practical examples for a newbie.

February 4, 2011, 09:29	fvc::div(phi,rho) vs fvc::div(phi)	#1
nimasam Senior Member Nima Samkhaniani Join Date: Sep 2009 Location: Tehran, Iran Posts: 1,267 Blog Entries: 1 Rep Power: 25	hello could you tell me whats the difference between fvc::div(phi,rho) vs fvc::div(phi) ? thank in advance

October 17, 2012, 04:16		#4
nimasam Senior Member Nima Samkhaniani Join Date: Sep 2009 Location: Tehran, Iran Posts: 1,267 Blog Entries: 1 Rep Power: 25	phi = Uf & Sf, which Sf is face area , and Uf is velocity at face center so its a volume flux {m.s} and nu = mu/rho randolph likes this.

October 17, 2012, 04:42		#5
haze_1986 Senior Member Join Date: Jul 2011 Posts: 120 Rep Power: 15	Please correct me if I am wrong. ddt(U) [m/s2] laplacian(nu,U) [m/s2] I assume nu here = mu/rho grad(p) [m/s2] I assume pressure here has been divided by rho as well? However, div(phi, U) If phi = rho * U, then the unit will be different? pacopuma likes this.

October 17, 2012, 05:59		#6
meindert Member Meindert de Groot Join Date: Jun 2012 Location: Netherlands Posts: 34 Rep Power: 14	Hi, You should realise that the definition of phi is different for an incompressible flow solver (e.g. icoFoam) than for an compressible flow solver. For an incompressible flow solver phi = U, for a compressible flow solver phi = rhoU. Thus, the face flux for an incompressible flow solver is (S_f · phi_f) = (S_f · U_f), which is exactly what Nima was trying to tell you. To be honest, the notation in OpenFOAM is a bit sloppy. The face flux is given the name phi, which is confusing if you write fvm::ddt(U) + fvm::div(phi, U) - fvm::laplacian(nu, U). I suggest you have a look at section 2.4 of the Programmer's Guide, which is located in the $WM_PROJECT_DIR/doc/Guides-a4 directory. hadisafaei, GerhardHolzinger, sharonyue and 5 others like this. Last edited by meindert; October 17, 2012 at 06:19.*

October 17, 2012, 03:43		#3
haze_1986 Senior Member Join Date: Jul 2011 Posts: 120 Rep Power: 15	Hi, may I also ask, in icoFoam I noticed that fvm::ddt(U) + fvm::div(phi, U) - fvm::laplacian(nu, U) why is phi used in convective div() term when nu (I suppose kinematic viscosity) is used in the laplacian term? shouldn't left hand side be divided by rho?

October 20, 2012, 10:20		#7
camille131 Member Camille Join Date: Oct 2012 Posts: 54 Rep Power: 14	Hello and thank you for the precision . I am actually trying to run a poiseuille case in channel with cyclic conditions. I don't know If I have to choose icoFoam or channelFoam solver. Actually in the channelFoam code I don't know what represents sgsModel->divDevBeff(U) in the equation? see http://foam.sourceforge.net/docs/cpp/a02641_source.html on line 63 Thank tou for your answer. Camille

October 20, 2012, 11:40		#8
meindert Member Meindert de Groot Join Date: Jun 2012 Location: Netherlands Posts: 34 Rep Power: 14	Hi Camille, First, I think it would have been more appropriate to open a new thread for this. My guess is that a little searching on the forum would have gotten you a long way as well. To answer your question, go with the channelFoam solver. It has been developed to do exactly what you want. Have a look at the chan395 tutorial. The definition of divDevBeff depends on the particular subgrid-scale model that you use. For most subgrid-scale models this term is defined in GenEddyVisc.C. http://foam.sourceforge.net/docs/cpp...ce.html#l00096 Use Doxygen to find out more about the different subgrid-scale models. Doxygen is there for a reason.

October 20, 2012, 12:53		#9
camille131 Member Camille Join Date: Oct 2012 Posts: 54 Rep Power: 14	Thank you for your answer actually I also have to kearn how to search correctly on the help guide ^^ but so , what s the difference between nuEff and nu? and what does the temperature T have to do here? it because the problem is coupled? in laminar case we don't care? thank you sorry it s not so easy to come from nowhere .. I have then introduced a new subject for other questions about Poiseuille. here: http://www.cfd-online.com/Forums/ope...tml#post387650

November 9, 2012, 00:49		#10
chaolian New Member Chad Join Date: Sep 2011 Posts: 16 Rep Power: 15	If alpha = 0.5Foam::erf(4.0(T-Tmelt)/(Tl-Ts))+scalar(0.5); Are fvc::div(phi,alpha) and 4.0exp(-pow(4.0(T-Tmelt)/(Tl-Ts),2))/Foam::sqrt(pi)/(Tl-Ts)*(U & fvc::grad(T)) the same? From the results, they are not quite the same. Why? Is there anything different other than a simple algebraic transformation?

November 9, 2012, 10:04		#11
meindert Member Meindert de Groot Join Date: Jun 2012 Location: Netherlands Posts: 34 Rep Power: 14	Hi Chad, No, those are not the same. The most important difference is the location at which the expressions are evaluated. If fvc::div(phi,alpha) is used, alpha is evaluated at the cell face according to equation (2.16) in the Programmer's Guide. For a central difference scheme this value is obtained by linear interpolation of the cell center values. If 4.0exp(-pow(4.0(T-Tmelt)/(Tl-Ts),2))/Foam::sqrt(pi)/(Tl-Ts)*(U & fvc::grad(T)) is used, the whole expression is evaluated at the cell center. You could write out the seperate discretisations and you will see that they are not the same. I would recommend you to use fvc::div(phi,alpha), since it maintains the conservative form of the convection term.

November 9, 2012, 13:48		#13
meindert Member Meindert de Groot Join Date: Jun 2012 Location: Netherlands Posts: 34 Rep Power: 14	Hi Chad, It is hard to tell what the source of your problem is, because you are not giving any details about the simulation itself. I suggest you open a new thread about your problem, so we stay on-topic here. Try to give as much details about your simulation as possible. You could think of boundary conditions, grid resolution, numerical schemes, etc. Perhaps it is a good idea to post a part of the solver output aswell.

March 12, 2013, 04:59		#14
woodwick New Member Alessandro Join Date: Feb 2013 Location: Italia Posts: 15 Rep Power: 13	Maysam, Do you know if does exist a tutorial or guide that list side to side the mathematical equations on one side and on the other side how they are written in the openFoam code. I have found something here http://www.foamcfd.org/Nabla/guides/UserGuidese14.html, but I was searching for something with some practical examples for a newbie.