Uprime2Mean and R (LES)

cjc96 · February 13, 2020, 08:17

Hey everyone,

To repeat how 90% of posts here go, I'm relatively new to OpenFOAM, and I am having an issue processing the Reynolds Stress Tensor for my SA-based DES case.

My understanding is that 'Uprime2Mean' provides the Reynolds Stress Tensor of the resolved flow, while 'R' provides the sub-grid (modelled) tensor. With these two values it should be possible to produce a field that displays the percentage of resolved TKE as a method of mesh validation.

I'm seeking some guidance on how to either obtain an average of 'R' to use with 'Uprime2Mean' obtained through the fieldAverage function or an instantaneous value of 'Uprime2Mean' to use with 'R' obtained through the eponymous Post-Processing function.

I half expect that this is a horribly inefficient method of obtaining the desired data, and any other suggestions would be equally welcome!

Thanks,

cjc96 · February 13, 2020, 12:13

Quote:

Originally Posted by cjc96

Hey everyone,

To repeat how 90% of posts here go, I'm relatively new to OpenFOAM, and I am having an issue processing the Reynolds Stress Tensor for my SA-based DES case.

My understanding is that 'Uprime2Mean' provides the Reynolds Stress Tensor of the resolved flow, while 'R' provides the sub-grid (modelled) tensor. With these two values it should be possible to produce a field that displays the percentage of resolved TKE as a method of mesh validation.

I'm seeking some guidance on how to either obtain an average of 'R' to use with 'Uprime2Mean' obtained through the fieldAverage function or an instantaneous value of 'Uprime2Mean' to use with 'R' obtained through the eponymous post-processing function.

I half expect that this is a horribly inefficient method of obtaining the desired data, and any other suggestions would be equally welcome!

Thanks,

As an update, I have managed to produce time-averages of both 'Uprime2Mean' and 'R' by modifying the 'turbulenceFields' function to execute each time-step, thus allowing the 'fieldAverage' function to produce an average value for 'R'

I'm now hoping to produce the field:

Code:

(1/2*tr(Uprime2Mean))/(1/2*tr(R + Uprime2Mean))

as a post-processing step, but this is where my limited knowledge of C++ is starting to become a hindrance!

Any guidance would be appreciated,

HPE · February 13, 2020, 17:36

Might misunderstand you, but here we go: your understanding on these objects is correct.

You can obtain 'RMean' by the same FO 'fieldAverage', and use it with 'UPrime2Mean' at the end of a sim?

cjc96 · February 14, 2020, 05:59

Quote:

Originally Posted by HPE

Might misunderstand you, but here we go: your understanding on these objects is correct.

You can obtain 'RMean' by the same FO 'fieldAverage', and use it with 'UPrime2Mean' at the end of a sim?

Hey HPE,
Thanks for your reply!

I actually made a bit of progress last night, but my post got caught by the new user moderator checks so didn't come through until after your response.

If you read that post you'll see I am now trying to form a new field, produced from:

Code:

(1/2*tr(Uprime2Mean))/(1/2*tr(RMean + Uprime2Mean))

HPE · February 14, 2020, 16:22

codedFunctionObjects can do it easily. Have a go for a review, and see by urself. I can provide an example after sitting in front of a PC unless you or other helpmates do earlier.

cjc96 · February 17, 2020, 06:03

Quote:

Originally Posted by HPE

codedFunctionObjects can do it easily. Have a go for a review, and see by urself. I can provide an example after sitting in front of a PC unless you or other helpmates do earlier.

Hello again HPE,
Thanks once more for your help.

I've had a look into codedFunctionObjects. I was able to get a basic function to work (simply reading and printing out a field value) but I'm hitting some errors when I try to include the functionality to calculate the ratio I need. Unfortunately, my current lack of experience with C++ means most of the errors are beyond my understanding.

Code:

resolvedTKE
{
	type			coded;
	libs			("libutilityFunctionObjects.so");
	name			resolvedTKE;
	executeControl	timeStep
	executeInterval	1;
	writeControl	writeTime;
//	timeStart		0.5;
//	timeEnd			1;
	enabled			false;

	codeOptions
	#{
		-I$(LIB_SRC)/meshTools/lnInclude
	#};

	codeExecute
	#{
		const volSymmTensorField& RMean = mesh().lookupObject<volSymmTensorField>("RMean");
		const volSymmTensorField& UPrime2Mean = mesh().lookupObject<volSymmTensorField>("UPrime2Mean");

		Info << "Min RMean = " << min(RMean) << endl;
		Info << "Min UPrime2Mean = " << min(UPrime2Mean) << endl;

		// volScalarField resolvedTKE
		// (
		// 	IOobject
		// 	(
		// 		"resolvedTKE",
		// 		runTime.timeName(),
		// 		mesh,
		// 		IOobject::NO_READ,
		// 		IOobject::AUTO_WRITE
		// 	),
		// 	mesh,
		// 	dimensionedScalar(dimless, 0)
		// );
		//
		// resolvedTKE = ((1.0 / 2.0) * tr(UPrime2Mean)) / ((1.0 / 2.0) * tr(RMean + UPrime2Mean));
		// resolvedTKE.write();
	#};
}

If you view the above code, the function works as presented, but if I include the commented section it breaks.

I'd again appreciate any advice you can provide!

cjc96 · February 17, 2020, 09:38

The following code appears to mostly work:

Code:

resolvedTKE
{
	type			coded;
	libs			("libutilityFunctionObjects.so");
	name			resolvedTKE;
	writeControl	runTime;
	writeInterval	0.005;
	// timeStart		0.5;
	// timeEnd			1;
	enabled			true;

	codeExecute
	#{
		const volSymmTensorField& RMean = mesh().lookupObject<volSymmTensorField>("RMean");
		const volSymmTensorField& UPrime2Mean = mesh().lookupObject<volSymmTensorField>("UPrime2Mean");

		volScalarField resolvedTKE
		(
			IOobject
			(
				"resolvedTKE",
				mesh().time().timeName(),
				mesh(),
				IOobject::NO_READ,
				IOobject::AUTO_WRITE
			),
			mesh(),
			dimensionedScalar(dimless, 0)
		);

		resolvedTKE = ((1.0 / 2.0) * tr(UPrime2Mean)) / ((1.0 / 2.0) * tr(RMean + UPrime2Mean));
		resolvedTKE.write();
	#};
}

However, the function ignores any writeControl or executeControl statements I include at the start, instead writing every time-step which is somewhat undesirable.

Due to this being a 'time' related problem, I'm assuming the issue is a result of one of (or a combination of) these lines:

Code:

mesh().time().timeName(),

IOobject::AUTO_WRITE

resolvedTKE.write();

But playing around with them seems to only result in no writes whatsoever.

HPE · February 17, 2020, 13:18

I haven't compiled it below, because I don't have a test-case, might need to add a `codeOptions`.

Let me know if it would be any help. I'm not sure since I'm bloody getting old. :]

Code:

resolvedTKE
{
    functionObjectLibs  (utilityFunctionObjects);
    type                coded;
    name              resolvedTKE;
    writeControl    writeTime;
    codeExecute
    #{
        static autoPtr<volScalarField> tkeRatio;
        if
        (
            mesh().time().timeIndex() == 1
            ||
            mesh().time().startTimeIndex() == mesh().time().timeIndex() - 1
        )
        {
            Info<< "    # First time-step or restart:" << nl;
            tkeRatio.set
            (
                new volScalarField
                (
                    IOobject
                    (
                        "tkeRatio",
                         mesh().time().timeName(),
                         mesh(),
                         IOobject::NO_READ,
                         IOobject::AUTO_WRITE
                     ),
                     mesh(),
                     dimless
                 )
             );
         }
         if (mesh().time().timeIndex() != 1)
         {
             const volSymmTensorField& RMean =
                mesh().lookupObjectRef<volSymmTensorField>("RMean");
             const volSymmTensorField& UPrime2Mean =
                mesh().lookupObjectRef<volSymmTensorField>("UPrime2Mean");
             volScalarField& tkeRatio = 
                mesh().lookupObjectRef<volScalarField>("tkeRatio");

            Info << "Min RMean = " << min(RMean) << endl;
            Info << "Min UPrime2Mean = " << min(UPrime2Mean) << endl;

            tkeRatio =
                (0.5*tr(UPrime2Mean))/(0.5*tr(RMean + UPrime2Mean) + SMALL);

            if (mesh().time().outputTime())
            {
                Info<< "    ### Writing out tkeRatio field" << nl << nl;
                tkeRatio.write();
            }
        }
    #};
}

FYI: I'm using v1912.

FYI: You need to have `RMean` and `UPrime2Mean` beforehand this FO, e.g. by using `fieldAverage`.

cjc96 · February 18, 2020, 10:03

Hello again HPE,
Thanks again for your help!

I've got the function mostly working now thanks to your input, and as a form of practice merged your code with my own as opposed to simply taking your code in its entirety.

Hopefully these will be my last questions for you and then you can go onto more important things!

Firstly, using:

Code:

if
(
	mesh().time().timeIndex() == 1
	||
	mesh().time().startTimeIndex() == mesh().time().timeIndex() - 1
)

is there a method to have the field initialise in the 0 directory, as opposed to during the first time step? In this case it doesn't actually matter, but could be useful in the future.

-------------------

Secondly, in your code:

Code:

tkeRatio = (0.5 * tr(UPrime2Mean)) / (0.5 * tr(RMean + UPrime2Mean) + SMALL);

I'm failing to see what the '+ SMALL' addition denotes, I'm guessing it is something left over from one of your own functions?

-------------------

Finally, I'd like the function to only write during a set time interval, such as can be done with the 'fieldAverage' function by denoting 'timeStart' and 'timeEnd'

My best guess thus far has been:

Code:

if
(
	mesh().time().writeTime()
	&&
	mesh().time().value() >= $timeStart
)

however, the compiler states that 'timeStart' hasn't been properly declared, leading me to presume it is included in another function somewhere (to be added using #include) but I am unsure which one.

-------------------

Thanks again for your time, I really am grateful

HPE · February 18, 2020, 14:27

>> is there a method to have the field initialise in the 0 directory, as opposed to during the first time step? In this case it doesn't actually matter, but could be useful in the future.

OpenFOAM learned `setExprFields`, link, which you can use it to initialise fields like by using swak4Foam.

>> I'm failing to see what the '+ SMALL' addition denotes, I'm guessing it is something left over from one of your own functions?

Not sure throws error when compiled (because I haven't compiled it), but `SMALL` is an OpenFOAM object which is equal to the smallest floating-point, for my double-precision compilation SMALL=1e-15. SMALL is there in the denominator to avoid zero-division error, which might be annoying if you are not familiar with.

>> Finally, I'd like the function to only write during a set time interval, such as can be done with the 'fieldAverage' function by denoting 'timeStart' and 'timeEnd'

You can use these keywords in codedFOs as well. But they are not for writing, but execution. Outside of their interval, the FO will not be executed.

>> however, the compiler states that 'timeStart' hasn't been properly declared, leading me to presume it is included in another function somewhere (to be added using #include) but I am unsure which one.

`timeStart` is not known locally. If you want it, you need to construct it like

`const scalar timeStart = 10;`

Or you need to turn-on the generic `timeStart` keyword up there somewhere. Otherwise, compiler will not recognise it since C++ statically defined language (your object type must be known there).

Hope it helps.

Andrea1984 · February 20, 2020, 06:11

Hi Conor,

an alternative method could be to hardcode resolved TKE, epsilon and all the other variables you need in a modified version of the solver. I gave it a go here: pimpleTKEBudgetFoam, which also includes a LESResolutionIndex = k_res/(k_modelled + k_res)

Andrea

cjc96 · February 26, 2020, 07:08

Quote:

Originally Posted by Andrea1984

Hi Conor,

an alternative method could be to hardcode resolved TKE, epsilon and all the other variables you need in a modified version of the solver. I gave it a go here: pimpleTKEBudgetFoam, which also includes a LESResolutionIndex = k_res/(k_modelled + k_res)

Andrea

Hey Andrea,
Thanks for replying!

Had a look through your repository and while I think I'll stick with the function for the time being, as I think it offers a bit more flexibility in what I am trying to do at the moment, it was very useful and I did pinch a few aspects of your code!

One thing I did disagree with however, and this may in actuality just be poor understanding on my part, is your calculation of 'kRes'

Code:

volVectorField UPrime = (U - db.lookupObject<volVectorField>("UMean"));

volScalarField kRes = 0.5*(UPrime & UPrime);

I understand that you have avoided needing to actually build the Reynolds Stress Tensor, seeing as we are only interested in the leading diagonal. However, am I not right in saying that:

$R_{ij} \equiv \overline{u'_{i}u'_{j}}$

And that you are in fact calculating:

$Andrea_{ij} \equiv {u'_{i}u'_{j}}$

Again, forgive me if I have totally misinterpreted either your code, or the underlying maths but I am eager to learn either way!

cjc96 · February 26, 2020, 07:14

Also,
I just thought I'd include the function I settled on (for now) for reference, in case someone is working on a similar problem in the future:

Code:

LESIndex
{
	type			coded;
	libs			("libutilityFunctionObjects.so");
	name			LESIndex;
	executeControl	writeTime;
	writeControl	writeTime;
	// timeStart		0;
	// timeEnd		0;
	enabled			true;

/*---------------------------------------------------------------------------*\

	LES Quality Evaluation
		** Requires fieldAverage Function to Obtain UPrime2Mean**
			** Resolved Reynolds Stress Tensor
		** Requires turbulenceFields Function to Obtain R**
			** Subgrid Reynolds Stress Tensor

\*---------------------------------------------------------------------------*/

	codeExecute
	#{
		static autoPtr<volScalarField> LESIndex;

		if
		(
			mesh().foundObject<volSymmTensorField>("UPrime2Mean")
			&&
			mesh().foundObject<volSymmTensorField>("R")
			&&
			mesh().foundObject<volScalarField>("LESIndex") == 0
		)
		{
			Info << "LES Quality Index:" << endl;
			Info << "	Initialising" << endl;
			Info << "	Calculating" << nl << endl;

			LESIndex.set
			(
				new volScalarField
				(
					IOobject
					(
						"LESIndex",
						mesh().time().timeName(),
						mesh(),
						IOobject::NO_READ,
						IOobject::AUTO_WRITE
					),
					mesh(),
					dimensionedScalar
					(
						"LESIndex",
						dimless,
						0
					)
				)
			);

			const volSymmTensorField& R = mesh().lookupObjectRef<volSymmTensorField>("R");
			const volSymmTensorField& UPrime2Mean = mesh().lookupObjectRef<volSymmTensorField>("UPrime2Mean");

			volScalarField& LESIndex = mesh().lookupObjectRef<volScalarField>("LESIndex");
			LESIndex = (0.5 * tr(UPrime2Mean)) / (0.5 * tr(R + UPrime2Mean));
			LESIndex.write();
		}

		else if
		(
			mesh().foundObject<volSymmTensorField>("UPrime2Mean")
			&&
			mesh().foundObject<volSymmTensorField>("R")
			&&
			mesh().foundObject<volScalarField>("LESIndex")
		)
		{
			Info << "LES Quality Index:" << endl;
			Info << "	Calculating" << nl << endl;

			const volSymmTensorField& R = mesh().lookupObjectRef<volSymmTensorField>("R");
			const volSymmTensorField& UPrime2Mean = mesh().lookupObjectRef<volSymmTensorField>("UPrime2Mean");

			volScalarField& LESIndex = mesh().lookupObjectRef<volScalarField>("LESIndex");
			LESIndex = (0.5 * tr(UPrime2Mean)) / (0.5 * tr(R + UPrime2Mean));
			LESIndex.write();
		}

		else
		{
			Info << "LES Quality Index:" << endl;
			Warning << endl
					<< "	Unable to Calculate LES Quality Index" << endl
					<< "	UPrime2Mean and/or R Unavailable" << endl
					<< "	Enable fieldAverage and turbulenceFields Functions" << nl << endl;
		}
	#};
}

The final function takes elements from my original attempt, HPE's function and Andrea's solver

Andrea1984 · February 26, 2020, 07:21

Hi Conor,

in a previous version of the code (see first commit) I had something like:

Code:

RRes = symm((UPrime*UPrime)); //Instantaneous resolved Reynolds stress tensor
kRes = 0.5*tr(RRes);

Than I realised that I was not interested in the non-diagonal components of R, so I though I would avoid the calculation of R altogether and just use

Code:

volScalarField kRes = 0.5*(UPrime & UPrime);

which, if I am not mistaken, is essentially the same, i.e.:

$k=\frac{1}{2}\overline{u_i u_i}=\frac{1}{2} \left (\overline{u_1^2}+\overline{u_2^2}+\overline{u_3^2} \right)$

Hope this clarifies it

Andrea

cjc96 · February 26, 2020, 08:57

Hey Andrea,

Yes, I totally agree that we aren't interested in the entire Reynolds Stress Tensor so may as well simplify the calculation.

However, unless I am misinterpreting your code, which as I said at the start of this thread, is entirely possible (I am very much new to this) you are calculating:

$k = \frac{1}{2} (u'u') = \frac{1}{2} (u'^{2}_{i}+u'^{2}_{j}+u'^{2}_{k})$

as opposed to:

$k = \frac{1}{2} (\overline{u'u'}) = \frac{1}{2} (\overline{u'^{2}_{i}}+\overline{u'^{2}_{j}}+\overline{u'^{2}_{k}})$

Again, I do apologise if I am just being a massive idiot =D

Andrea1984 · February 26, 2020, 09:08

Lol no need to apologies, I am not 100% sure as well. Actually I believe I was wrong in my previous post to include the overbar in my definition of kRes.

It is true I am calculating the "instantaneous" value of kRes - I figured that I can then obtain the time-averaged kRes by just adding kRes (instantaneous) to the fieldAverage utility, which would output kResMean, i.e. including the overbar.

I am not entirely sure if this is mathematically correct, it was leading to reasonable results in a test I did in a simple channel flow though. See snapshot below:

https://imgur.com/yM6fNQx

joshm · February 27, 2020, 11:03

Quote:

Originally Posted by Andrea1984

Lol no need to apologies, I am not 100% sure as well. Actually I believe I was wrong in my previous post to include the overbar in my definition of kRes.

It is true I am calculating the "instantaneous" value of kRes - I figured that I can then obtain the time-averaged kRes by just adding kRes (instantaneous) to the fieldAverage utility, which would output kResMean, i.e. including the overbar.

I am not entirely sure if this is mathematically correct, it was leading to reasonable results in a test I did in a simple channel flow though. See snapshot below:

https://imgur.com/yM6fNQx

If you forget to do this at runtime, as long as you do have the mean U and nut fields, you can just change the name of UMean to U and nutMean to nut and run the R utility.

ari003 · April 17, 2020, 13:10

Quote:

Originally Posted by cjc96

As an update, I have managed to produce time-averages of both 'Uprime2Mean' and 'R' by modifying the 'turbulenceFields' function to execute each time-step, thus allowing the 'fieldAverage' function to produce an average value for 'R'

I'm now hoping to produce the field:

Code:

(1/2*tr(Uprime2Mean))/(1/2*tr(R + Uprime2Mean))

as a post-processing step, but this is where my limited knowledge of C++ is starting to become a hindrance!

Any guidance would be appreciated,

I can use that equation to find the value of resolved KE in spreadsheet. But I m not getting what does that tr denotes. Please can anyone say that?

cjc96 · April 17, 2020, 13:13

Quote:

Originally Posted by ari003

I can use that equation to find the value of resolved KE in spreadsheet. But I m not getting what does that tr denotes. Please can anyone say that?

Hey bud,

The symbol 'tr' denotes the trace of a matrix

Hope that helps!

lizzy · August 25, 2020, 05:50

Hello Conor,
Thanks for your share. I'm wandering if the "R" in your final coded function denotes the averaged value of R? Or just an instaneous tensor value?

Quote:

Originally Posted by cjc96

Also,
I just thought I'd include the function I settled on (for now) for reference, in case someone is working on a similar problem in the future:

Code:

LESIndex
{
	type			coded;
	libs			("libutilityFunctionObjects.so");
	name			LESIndex;
	executeControl	writeTime;
	writeControl	writeTime;
	// timeStart		0;
	// timeEnd		0;
	enabled			true;

/*---------------------------------------------------------------------------*\

	LES Quality Evaluation
		** Requires fieldAverage Function to Obtain UPrime2Mean**
			** Resolved Reynolds Stress Tensor
		** Requires turbulenceFields Function to Obtain R**
			** Subgrid Reynolds Stress Tensor

\*---------------------------------------------------------------------------*/

	codeExecute
	#{
		static autoPtr<volScalarField> LESIndex;

		if
		(
			mesh().foundObject<volSymmTensorField>("UPrime2Mean")
			&&
			mesh().foundObject<volSymmTensorField>("R")
			&&
			mesh().foundObject<volScalarField>("LESIndex") == 0
		)
		{
			Info << "LES Quality Index:" << endl;
			Info << "	Initialising" << endl;
			Info << "	Calculating" << nl << endl;

			LESIndex.set
			(
				new volScalarField
				(
					IOobject
					(
						"LESIndex",
						mesh().time().timeName(),
						mesh(),
						IOobject::NO_READ,
						IOobject::AUTO_WRITE
					),
					mesh(),
					dimensionedScalar
					(
						"LESIndex",
						dimless,
						0
					)
				)
			);

			const volSymmTensorField& R = mesh().lookupObjectRef<volSymmTensorField>("R");
			const volSymmTensorField& UPrime2Mean = mesh().lookupObjectRef<volSymmTensorField>("UPrime2Mean");

			volScalarField& LESIndex = mesh().lookupObjectRef<volScalarField>("LESIndex");
			LESIndex = (0.5 * tr(UPrime2Mean)) / (0.5 * tr(R + UPrime2Mean));
			LESIndex.write();
		}

		else if
		(
			mesh().foundObject<volSymmTensorField>("UPrime2Mean")
			&&
			mesh().foundObject<volSymmTensorField>("R")
			&&
			mesh().foundObject<volScalarField>("LESIndex")
		)
		{
			Info << "LES Quality Index:" << endl;
			Info << "	Calculating" << nl << endl;

			const volSymmTensorField& R = mesh().lookupObjectRef<volSymmTensorField>("R");
			const volSymmTensorField& UPrime2Mean = mesh().lookupObjectRef<volSymmTensorField>("UPrime2Mean");

			volScalarField& LESIndex = mesh().lookupObjectRef<volScalarField>("LESIndex");
			LESIndex = (0.5 * tr(UPrime2Mean)) / (0.5 * tr(R + UPrime2Mean));
			LESIndex.write();
		}

		else
		{
			Info << "LES Quality Index:" << endl;
			Warning << endl
					<< "	Unable to Calculate LES Quality Index" << endl
					<< "	UPrime2Mean and/or R Unavailable" << endl
					<< "	Enable fieldAverage and turbulenceFields Functions" << nl << endl;
		}
	#};
}

The final function takes elements from my original attempt, HPE's function and Andrea's solver

February 13, 2020, 08:17	Uprime2Mean and R (LES)	#1
cjc96 Member Conor Crickmore Join Date: Jan 2020 Location: Leicestershire, United Kingdom Posts: 36 Rep Power: 6	Hey everyone, To repeat how 90% of posts here go, I'm relatively new to OpenFOAM, and I am having an issue processing the Reynolds Stress Tensor for my SA-based DES case. My understanding is that 'Uprime2Mean' provides the Reynolds Stress Tensor of the resolved flow, while 'R' provides the sub-grid (modelled) tensor. With these two values it should be possible to produce a field that displays the percentage of resolved TKE as a method of mesh validation. I'm seeking some guidance on how to either obtain an average of 'R' to use with 'Uprime2Mean' obtained through the fieldAverage function or an instantaneous value of 'Uprime2Mean' to use with 'R' obtained through the eponymous Post-Processing function. I half expect that this is a horribly inefficient method of obtaining the desired data, and any other suggestions would be equally welcome! Thanks, __________________ Conor Crickmore PhD Researcher in Automotive Aerodynamics Aeronautical and Automotive Engineering Loughborough University LE11 3TU

February 13, 2020, 17:36		#3
HPE Senior Member Herpes Free Engineer Join Date: Sep 2019 Location: The Home Under The Ground with the Lost Boys Posts: 931 Rep Power: 13	Might misunderstand you, but here we go: your understanding on these objects is correct. You can obtain 'RMean' by the same FO 'fieldAverage', and use it with 'UPrime2Mean' at the end of a sim? cjc96 likes this. __________________ The OpenFOAM community is the biggest contributor to OpenFOAM: User guide/Wiki-1/Wiki-2/Code guide/Code Wiki/Journal Nilsson/Guerrero/Holzinger/Holzmann/Nagy/Santos/Nozaki/Jasak/Primer Governance Bugs/Features: OpenFOAM (ESI-OpenCFD-Trademark) Bugs/Features: FOAM-Extend (Wikki-FSB) Bugs: OpenFOAM.org How to create a MWE New: Forkable OpenFOAM mirror

February 14, 2020, 16:22		#5
HPE Senior Member Herpes Free Engineer Join Date: Sep 2019 Location: The Home Under The Ground with the Lost Boys Posts: 931 Rep Power: 13	codedFunctionObjects can do it easily. Have a go for a review, and see by urself. I can provide an example after sitting in front of a PC unless you or other helpmates do earlier. cjc96 likes this. __________________ The OpenFOAM community is the biggest contributor to OpenFOAM: User guide/Wiki-1/Wiki-2/Code guide/Code Wiki/Journal Nilsson/Guerrero/Holzinger/Holzmann/Nagy/Santos/Nozaki/Jasak/Primer Governance Bugs/Features: OpenFOAM (ESI-OpenCFD-Trademark) Bugs/Features: FOAM-Extend (Wikki-FSB) Bugs: OpenFOAM.org How to create a MWE New: Forkable OpenFOAM mirror

February 17, 2020, 09:38		#7
cjc96 Member Conor Crickmore Join Date: Jan 2020 Location: Leicestershire, United Kingdom Posts: 36 Rep Power: 6	The following code appears to mostly work: Code: resolvedTKE { type coded; libs ("libutilityFunctionObjects.so"); name resolvedTKE; writeControl runTime; writeInterval 0.005; // timeStart 0.5; // timeEnd 1; enabled true; codeExecute #{ const volSymmTensorField& RMean = mesh().lookupObject<volSymmTensorField>("RMean"); const volSymmTensorField& UPrime2Mean = mesh().lookupObject<volSymmTensorField>("UPrime2Mean"); volScalarField resolvedTKE ( IOobject ( "resolvedTKE", mesh().time().timeName(), mesh(), IOobject::NO_READ, IOobject::AUTO_WRITE ), mesh(), dimensionedScalar(dimless, 0) ); resolvedTKE = ((1.0 / 2.0) * tr(UPrime2Mean)) / ((1.0 / 2.0) * tr(RMean + UPrime2Mean)); resolvedTKE.write(); #}; } However, the function ignores any writeControl or executeControl statements I include at the start, instead writing every time-step which is somewhat undesirable. Due to this being a 'time' related problem, I'm assuming the issue is a result of one of (or a combination of) these lines: Code: mesh().time().timeName(), IOobject::AUTO_WRITE resolvedTKE.write(); But playing around with them seems to only result in no writes whatsoever. __________________ Conor Crickmore PhD Researcher in Automotive Aerodynamics Aeronautical and Automotive Engineering Loughborough University LE11 3TU

February 18, 2020, 10:03		#9
cjc96 Member Conor Crickmore Join Date: Jan 2020 Location: Leicestershire, United Kingdom Posts: 36 Rep Power: 6	Hello again HPE, Thanks again for your help! I've got the function mostly working now thanks to your input, and as a form of practice merged your code with my own as opposed to simply taking your code in its entirety. Hopefully these will be my last questions for you and then you can go onto more important things! Firstly, using: Code: if ( mesh().time().timeIndex() == 1 \|\| mesh().time().startTimeIndex() == mesh().time().timeIndex() - 1 ) is there a method to have the field initialise in the 0 directory, as opposed to during the first time step? In this case it doesn't actually matter, but could be useful in the future. ------------------- Secondly, in your code: Code: tkeRatio = (0.5 * tr(UPrime2Mean)) / (0.5 * tr(RMean + UPrime2Mean) + SMALL); I'm failing to see what the '+ SMALL' addition denotes, I'm guessing it is something left over from one of your own functions? ------------------- Finally, I'd like the function to only write during a set time interval, such as can be done with the 'fieldAverage' function by denoting 'timeStart' and 'timeEnd' My best guess thus far has been: Code: if ( mesh().time().writeTime() && mesh().time().value() >= $timeStart ) however, the compiler states that 'timeStart' hasn't been properly declared, leading me to presume it is included in another function somewhere (to be added using #include) but I am unsure which one. ------------------- Thanks again for your time, I really am grateful __________________ Conor Crickmore PhD Researcher in Automotive Aerodynamics Aeronautical and Automotive Engineering Loughborough University LE11 3TU Last edited by cjc96; February 18, 2020 at 11:07.

February 18, 2020, 14:27		#10
HPE Senior Member Herpes Free Engineer Join Date: Sep 2019 Location: The Home Under The Ground with the Lost Boys Posts: 931 Rep Power: 13	>> is there a method to have the field initialise in the 0 directory, as opposed to during the first time step? In this case it doesn't actually matter, but could be useful in the future. OpenFOAM learned `setExprFields`, link, which you can use it to initialise fields like by using swak4Foam. >> I'm failing to see what the '+ SMALL' addition denotes, I'm guessing it is something left over from one of your own functions? Not sure throws error when compiled (because I haven't compiled it), but `SMALL` is an OpenFOAM object which is equal to the smallest floating-point, for my double-precision compilation SMALL=1e-15. SMALL is there in the denominator to avoid zero-division error, which might be annoying if you are not familiar with. >> Finally, I'd like the function to only write during a set time interval, such as can be done with the 'fieldAverage' function by denoting 'timeStart' and 'timeEnd' You can use these keywords in codedFOs as well. But they are not for writing, but execution. Outside of their interval, the FO will not be executed. >> however, the compiler states that 'timeStart' hasn't been properly declared, leading me to presume it is included in another function somewhere (to be added using #include) but I am unsure which one. `timeStart` is not known locally. If you want it, you need to construct it like `const scalar timeStart = 10;` Or you need to turn-on the generic `timeStart` keyword up there somewhere. Otherwise, compiler will not recognise it since C++ statically defined language (your object type must be known there). Hope it helps. cjc96 likes this. __________________ The OpenFOAM community is the biggest contributor to OpenFOAM: User guide/Wiki-1/Wiki-2/Code guide/Code Wiki/Journal Nilsson/Guerrero/Holzinger/Holzmann/Nagy/Santos/Nozaki/Jasak/Primer Governance Bugs/Features: OpenFOAM (ESI-OpenCFD-Trademark) Bugs/Features: FOAM-Extend (Wikki-FSB) Bugs: OpenFOAM.org How to create a MWE New: Forkable OpenFOAM mirror

February 20, 2020, 06:11		#11
Andrea1984 Senior Member Andrea Join Date: Feb 2012 Location: Leeds, UK Posts: 179 Rep Power: 16	Hi Conor, an alternative method could be to hardcode resolved TKE, epsilon and all the other variables you need in a modified version of the solver. I gave it a go here: pimpleTKEBudgetFoam, which also includes a LESResolutionIndex = k_res/(k_modelled + k_res) Andrea acgnipper, HPE, cjc96 and 1 others like this.

February 26, 2020, 07:21		#14
Andrea1984 Senior Member Andrea Join Date: Feb 2012 Location: Leeds, UK Posts: 179 Rep Power: 16	Hi Conor, in a previous version of the code (see first commit) I had something like: Code: RRes = symm((UPrimeUPrime)); //Instantaneous resolved Reynolds stress tensor kRes = 0.5tr(RRes); Than I realised that I was not interested in the non-diagonal components of R, so I though I would avoid the calculation of R altogether and just use Code: volScalarField kRes = 0.5*(UPrime & UPrime); which, if I am not mistaken, is essentially the same, i.e.: $k=\frac{1}{2}\overline{u_i u_i}=\frac{1}{2} \left (\overline{u_1^2}+\overline{u_2^2}+\overline{u_3^2} \right)$ Hope this clarifies it Andrea cjc96 likes this.

February 26, 2020, 08:57		#15
cjc96 Member Conor Crickmore Join Date: Jan 2020 Location: Leicestershire, United Kingdom Posts: 36 Rep Power: 6	Hey Andrea, Yes, I totally agree that we aren't interested in the entire Reynolds Stress Tensor so may as well simplify the calculation. However, unless I am misinterpreting your code, which as I said at the start of this thread, is entirely possible (I am very much new to this) you are calculating: $k = \frac{1}{2} (u'u') = \frac{1}{2} (u'^{2}_{i}+u'^{2}_{j}+u'^{2}_{k})$ as opposed to: $k = \frac{1}{2} (\overline{u'u'}) = \frac{1}{2} (\overline{u'^{2}_{i}}+\overline{u'^{2}_{j}}+\overline{u'^{2}_{k}})$ Again, I do apologise if I am just being a massive idiot =D __________________ Conor Crickmore PhD Researcher in Automotive Aerodynamics Aeronautical and Automotive Engineering Loughborough University LE11 3TU

February 26, 2020, 09:08		#16
Andrea1984 Senior Member Andrea Join Date: Feb 2012 Location: Leeds, UK Posts: 179 Rep Power: 16	Lol no need to apologies, I am not 100% sure as well. Actually I believe I was wrong in my previous post to include the overbar in my definition of kRes. It is true I am calculating the "instantaneous" value of kRes - I figured that I can then obtain the time-averaged kRes by just adding kRes (instantaneous) to the fieldAverage utility, which would output kResMean, i.e. including the overbar. I am not entirely sure if this is mathematically correct, it was leading to reasonable results in a test I did in a simple channel flow though. See snapshot below: https://imgur.com/yM6fNQx cjc96 likes this.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
LES best practice: resolved and modelled turbulence kinetic energy	em17	OpenFOAM Running, Solving & CFD	22	April 17, 2020 14:01
Relation between k and UPrime2Mean etc in LES	Tarak	OpenFOAM Running, Solving & CFD	49	August 10, 2018 14:10
Worse accuracy in LES models than in RANS models	dferrando	OpenFOAM Running, Solving & CFD	0	August 30, 2017 05:56
Resolved KE versus SGS KE - LES	smog	OpenFOAM	0	February 11, 2016 19:01
Bug in Dynamic 1Eq Eddy Viscosity LES Model Incompressible flow	harishg	OpenFOAM Bugs	5	December 2, 2008 19:18