CFD Online Discussion Forums - Turbulence model implementation errors

CFD Online Discussion Forums (https://www.cfd-online.com/Forums/)

- OpenFOAM Programming & Development (https://www.cfd-online.com/Forums/openfoam-programming-development/)

- - Turbulence model implementation errors (https://www.cfd-online.com/Forums/openfoam-programming-development/81895-turbulence-model-implementation-errors.html)

elbertj

November 10, 2010 01:48

Turbulence model implementation errors

2 Attachment(s)

Quote:

BSLEARSM.C:254: instantiated from here
/home/ejp/codes/OpenFOAM/OpenFOAM-1.7.x/src/OpenFOAM/lnInclude/FieldFunctions.C:695: error: no match for ‘operator&&’ in ‘*(f2P ++) && *(f3P ++)’
/home/ejp/codes/OpenFOAM/OpenFOAM-1.7.x/src/OpenFOAM/lnInclude/FieldFunctions.C:695: note: candidates are: operator&&(bool, bool) <built-in>
/home/ejp/codes/OpenFOAM/OpenFOAM-1.7.x/src/OpenFOAM/lnInclude/dimensionSet.H:285: note: Foam::dimensionSet Foam::operator&&(const Foam::dimensionSet&, const Foam::dimensionSet&)
make: *** [Make/linuxGccDPOpt/BSLEARSM.o] Error 1

Seems like I'm getting an error due to wrong use of double inner product operator (&&) and consequently with dimensions. However, I am unable to spot the error with the formula implementation. Kindly let me know if you know the source of error

Thanks,
Elbert

niklas

November 10, 2010 06:26

Is IV_ a scalar or tensor?

IV_( (S_&&W_)&&W_),

If it is a scalar there should be a normal matrix multiplication somewhere

either
IV_( (S_&W_)&&W_)
or
IV_( S_&& (W_ & W_) )

niklas

November 10, 2010 07:28

some other issues...

C1prime_ is defined as a volScalarField, but its just a constant, so I would change it to dimensionedScalar, otherwise you need to supply size information et.c....

Beta2_ is a volScalarField, but you cant just initialize it with a value it needs more information than
that. (same error as C1prime_)
To get it to compile i just set it to
Beta2_( 0.*N_), // why use a volScalarField for a zero?

but Im guessing that it really should be something else, otherwise why have it :)

aij_ is a symmetric tensor
W_ is a tensor and
W_ & W_ is just a tensor multiplication, this does not imply that the result is a symmetric tensor.
sure in this example it will be, but the code sees this as any other tensor multiplication

so you need to explicitly extract the symmetric part.

aij_
(
Beta1_ *S_
+ Beta2_ * ( (S_ & S_)-(1./3.)*(IIs_ * I ))
+ Beta3_ * (symm(W_ & W_)-(1./3.)*( IIo_*I)) // why should a tensor multiplication yield a symm tensor?
+ Beta4_*symm((S_ & W_)-(W_ & S_))
+ Beta6_*(symm(((S_ & W_) & W_) + ((W_ & W_) & S_)) - (2./3.)*IV_*I - IIo_*S_)
+ Beta9_*(symm(((W_ & S_) & (W_ & W_)) - ((W_ & W_) & (S_ & W_)) + 0.5*IIo_*((S_ & W_) - (W_ & S_))))
),

elbertj

November 11, 2010 16:03

1 Attachment(s)

Hi Niklas,
Thanks, after making changes you suggested, it is working! I did this for aij_ ,

Quote:

aij_ = symm(
Beta1_*S_
+ Beta2_*((S_ & S_)-(1./3.)*IIs_*I)
+ Beta3_*((W_ & W_)-(1./3.)*IIo_*I)
+ Beta4_*((S_ & W_)-(W_ & S_))
+ Beta6_*(((S_ & W_) & W_) + ((W_ & W_) & S_) - (2./3.)*IV_*I - IIo_*S_)
+ Beta9_*(((W_ & S_) & (W_ & W_)) - ((W_ & W_) & (S_ & W_)) + 0.5*IIo_*((S_ & W_) - (W_ & S_)))
);

Are you saying that though A_ = W_ & W_ is symmetric it has to be explicitly said, as
A_=symm(W_&W_) ?

In the constructor portion of my BSLEARSM.C (attached) , can you please tell me how I can include this code:

Quote:

forAll(N_,cellI)
{
if(P2_[cellI] >= 0)
{
N_[cellI]= (C1prime_.value()/3.)+pow((P1_[cellI]+sqrt(P2_[cellI])),(1./3.))
+sign(P1_[cellI]-sqrt(P2_[cellI]))*pow(mag(P1_[cellI]-sqrt(P2_[cellI])),(1./3.));
}
else
{
N_[cellI]= (C1prime_.value()/3.)+2.*pow((sqr(P1_[cellI])-P2_[cellI]),(1./6.))
*cos((1./3.)*acos(P1_[cellI]/(sqrt(sqr(P1_[cellI])-P2_[cellI]))));
}
}

-Elbert

niklas

November 12, 2010 02:02

Quote:

Originally Posted by elbertj (Post 283137)

Are you saying that though A_ = W_ & W_ is symmetric it has to be explicitly said, as
A_=symm(W_&W_) ?

Yes, because
in the tensor class the multiplication operation between two tensors returns another tensor,
not a symmetric tensor.

Quote:

In the constructor portion of my BSLEARSM.C (attached) , can you please tell me how I can include this code:

N_(scalar(0.)*IIs_),
This is a good start if the dimensions of IIs_ and N_ are the same otherwise use another field
which has the same dimensions as N_ or a constructor like the one for nut_.
For that kind of complicated expressions you have to place it in here.

Code:

{

    bound(omega_, omega0_);



    nut_ = a1_*k_/max(a1_*omega_, F2()*mag(symm(fvc::grad(U_))));

//    nut_ = k_/(omega_+smallOmega_);

    nut_.correctBoundaryConditions();



    // calculation of N_ goes here ....



    printCoeffs();

}

maolongliu

December 21, 2010 04:51

Hi, I am also using this model.
Which reference did you use to write this model.
Because your correlations are a little different with mine.

Quote:

Originally Posted by elbertj (Post 282837)

Dear Foamers,
I have implemented a non-linear eddy viscosity model, by modifying the kOmegaSST and including additional non-linear terms (code attached). I get a number of errors as:
no matching function to call to 'Foam:: .....'
final error as:

Seems like I'm getting an error due to wrong use of double inner product operator (&&) and consequently with dimensions. However, I am unable to spot the error with the formula implementation. Kindly let me know if you know the source of error

Thanks,
Elbert

elbertj

December 21, 2010 08:25

I used the formulation described here:
"Explicit Algebraic Reynolds Stress Models for Anisotropic Wall-Bounded Flows", F.R. Menter, A.V. Garbaruk, and Y. Egorov.
EUCASS – 3rd European Conference for Aero-Space Sciences, July 6-9th 2009 Versailles

maolongliu

December 22, 2010 03:34

Does this model work well?
I use the formula of " Wallin, S. and A. Johansson (2000). "An explicit algebraic Reynolds stress model for incompressible and compressible turbulent flows." Journal of Fluid Mechanics 403: 89-132 ", it does not work quite well.
I want to try this one you use. Would you mind to send me the reference you use because I cannot find the file.
my mail: maolongliu@gmail.com
Thank you.

Quote:

Originally Posted by elbertj (Post 287996)

cfdmarkus

December 26, 2010 13:47

Hi Elbert,

I am also using the WJ model. The reference you are using seem much newer.
Would it be possible for you to send me copy as well please: mw405@soton.ac.uk

Thanks, Markus.

florian_krause

June 17, 2011 11:07

Quote:

Originally Posted by niklas (Post 283182)

Code:

{

    bound(omega_, omega0_);



    nut_ = a1_*k_/max(a1_*omega_, F2()*mag(symm(fvc::grad(U_))));

//    nut_ = k_/(omega_+smallOmega_);

    nut_.correctBoundaryConditions();



    // calculation of N_ goes here ....



    printCoeffs();

}

Hi Niklas,
I recently had a similar issue when implementing Wilcox 1998 k-omega model (http://turbmodels.larc.nasa.gov/wilcox.html), which is different than the k-omega model implemented in OF (Wilcox 1988 k-omega model).

I have the following expression

Code:

    forAll(fBetaStar_, cellI)

    {

    if(xiK_[cellI] <= scalar(0))

        fBetaStar_[cellI] = scalar(1);

    else

        fBetaStar_[cellI] = (scalar(1)+680*sqr(xiK_[cellI]))/(scalar(1)+400*sqr(xiK_[cellI]));

    }

and I was wondering where to place it in the constructor. Well, I placed it where you suggested above and it works.

Why should it be placed there and how is that related to the complexity of the expression?

Thanks and best regards,
Florian

khedar

November 29, 2016 08:28

Hi all, I know this thread is a bit old but I have a question regarding the implementation of elbertj(thanks for the code), I have been playing with this formulation and my question is the following:

In the kOmega_LowRel.H file to update the omega and G field:

Code:

forAll(curPatch, facei)

            {

                label faceCelli = curPatch.faceCells()[facei];

                // For corner cells (with two boundary or more faces),

                // omega and G in the near-wall cell are calculated

                // as an average

                cellBoundaryFaceCount[faceCelli]++;

                omega_[faceCelli] += scalar(60)*nuw[facei]

                    /(beta1_.value()*sqr(y_[faceCelli]));

                G[faceCelli] +=

                    (nutw[facei] + nuw[facei])*magFaceGradU[facei]

                    *Cmu25*sqrt(k_[faceCelli])/(kappa_.value()*y_[faceCelli]);

            }

Does one need to update the G field as well? G in this case is derived from the equilibrium assumption and hence it might not be a good idea to use it.

All times are GMT -4. The time now is 21:05.