Negative values in scalarTransportFoam with upwind discretization

G_German · June 6, 2017, 06:18

I tried to set up the well known test case for the 1D Advection/Diffusion equation with scalarTransportFoam. With a linear discretization of the advection term the results look as expected. However, also with an upwind discretization the result becomes negative for large Peclet numbers, which should not be the case, right? Is there something wrong with my setup (or my understanding of what should happen) or is this an OpenFOAM problem?

GerhardHolzinger · June 7, 2017, 07:58

I made the following changes:

change the convection scheme to bounded Gauss upwind to remove warnings
change the laplacian scheme to orthogonal, since there is no non-orthogonality
change the outlet BC of T to inletOutlet
change the solver for the linear system to smoothSolver

The last two points are the ones that do the trick.

G_German · June 8, 2017, 05:06

Hi Gerhard,
thanks for your reply.
With the zeroGradient boundary condition imposed by the inletOutlet setting the solution is no longer negative but T=0 over the whole domain. This is the correct solution for the zero Gradient BC, but unfortunately not the solution for the problem I want to solve (with the boundary condition T(L)=1).

I already had checked and looked at the warning (suggesting the bounded Gauss method) and other solvers, but I think these settings do not cause/solve the problem.

ngj · July 25, 2017, 15:15

FYI:

I have just noticed that this issue has been reported as a bug here (not by myself):

https://develop.openfoam.com/Develop...lus/issues/545

Kind regards,

Niels

FlyBob91 · August 21, 2017, 15:57

I have the same problem. Does someone have a solution?

santiagomarquezd · August 25, 2017, 20:33

Hello, there is no native solution for this problem in FOAM. I commented on the topic in the FOAM+ bugtracker.

Regards.

G_German · August 30, 2017, 10:46

Thanks for this hint.
Were the boundary conditions suggested in the earlier discussion (Wrong fvm::div assembling) actually added to -extend?

G_German · August 30, 2017, 12:22

Another thing:
In the earlier thread only the fixedValue BC was discussed. However, for me the same thing happens with the fixedGradient BC.

G_German · August 30, 2017, 13:24

With another look I am not quite sure that this is the same problem.
As I understand the earlier discussion, the solution oscillated for the upwind scheme as it did for the linear scheme.
In my case the linear scheme oscillates at high Pe but the Upwind (but also the Gamma-scheme) simply give a negative solution.

July 25, 2017, 15:15		#4
ngj Senior Member Niels Gjoel Jacobsen Join Date: Mar 2009 Location: Copenhagen, Denmark Posts: 1,900 Rep Power: 37	FYI: I have just noticed that this issue has been reported as a bug here (not by myself): https://develop.openfoam.com/Develop...lus/issues/545 Kind regards, Niels __________________ Please note that I do not use the Friend-feature, so do not be offended, if I do not accept a request.

August 25, 2017, 20:33		#6
santiagomarquezd Senior Member Santiago Marquez Damian Join Date: Aug 2009 Location: Santa Fe, Santa Fe, Argentina Posts: 452 Rep Power: 23	Hello, there is no native solution for this problem in FOAM. I commented on the topic in the FOAM+ bugtracker. Regards. __________________ Santiago MÁRQUEZ DAMIÁN, Ph.D. Research Scientist Research Center for Computational Methods (CIMEC) - CONICET/UNL Tel: 54-342-4511594 Int. 7032 Colectora Ruta Nac. 168 / Paraje El Pozo (3000) Santa Fe - Argentina. http://www.cimec.org.ar

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Avoid negative scalar values (UDS)	Bruno Machado	Fluent UDF and Scheme Programming	8	November 8, 2023 20:19
Negative value in scalarTransportFoam	waiter120	OpenFOAM Running, Solving & CFD	3	May 31, 2021 12:16
[blockMesh] edges not aligned with or perpendicular to non-empty directions	ynos	OpenFOAM Meshing & Mesh Conversion	6	March 26, 2020 15:02
negative values of additional variable	cosine	CFX	0	November 15, 2011 04:42
exact face values	RubenG	Main CFD Forum	0	June 22, 2009 11:09

June 8, 2017, 05:06		#3
G_German New Member Gerd Join Date: Nov 2015 Posts: 10 Rep Power: 10	Hi Gerhard, thanks for your reply. With the zeroGradient boundary condition imposed by the inletOutlet setting the solution is no longer negative but T=0 over the whole domain. This is the correct solution for the zero Gradient BC, but unfortunately not the solution for the problem I want to solve (with the boundary condition T(L)=1). I already had checked and looked at the warning (suggesting the bounded Gauss method) and other solvers, but I think these settings do not cause/solve the problem.

August 21, 2017, 15:57		#5
FlyBob91 Senior Member Join Date: Mar 2016 Location: Bergamo Posts: 157 Rep Power: 10	I have the same problem. Does someone have a solution?

August 30, 2017, 10:46		#7
G_German New Member Gerd Join Date: Nov 2015 Posts: 10 Rep Power: 10	Thanks for this hint. Were the boundary conditions suggested in the earlier discussion (Wrong fvm::div assembling) actually added to -extend?

August 30, 2017, 12:22		#8
G_German New Member Gerd Join Date: Nov 2015 Posts: 10 Rep Power: 10	Another thing: In the earlier thread only the fixedValue BC was discussed. However, for me the same thing happens with the fixedGradient BC.

August 30, 2017, 13:24		#9
G_German New Member Gerd Join Date: Nov 2015 Posts: 10 Rep Power: 10	With another look I am not quite sure that this is the same problem. As I understand the earlier discussion, the solution oscillated for the upwind scheme as it did for the linear scheme. In my case the linear scheme oscillates at high Pe but the Upwind (but also the Gamma-scheme) simply give a negative solution.