Learning Forum

[Juan Martin]

Hi all, I have a query that you might be able to help me with. If I wanted to perform a direct optimisation of a multivariable problem to get the best candidate point, would there be a way to include a material as an input parameter?

Although this is not my case, I hope that giving an example will make the type of problem more understandable: if I have a liquid flowing through two pipes that go to a tank, which has a drain and I want to find the minimum pressure gradient or any objective function you imagine from some input parameters such as the inlet velocity in each pipe or the temperature in each one, could I somehow introduce a new fluid to evaluate different combinations of velocity/temperature/fluid to obtain the best possible combination?

It should be noted that each fluid would have its respective temperature-dependent cp, density, conductivity and viscosity conditions and they should act as a block (1 fluid with its respective properties, it could not be with mix of properties of different fluids).

Thank you very much in advance, hope I can get some insight.

Cheers.

[Rob]

I think you can set a material property (eg density) as a parameter but not a material type. For optimisation it's common to adjust boundary conditions and geometry shape, not so much material types in Fluids. 

[Pedro Quintanilla]

It is true that for user-defined materials, it would be really interesting the use of optimization for each one presenting the responses for the best material in the determined problem. I think, if you want the best candidate point for some inlet conditions, editing boundary parameters is not a good idea because you will have a mix of parameters and it has no value. 

If someone think it would be possible to assign the material as a input parameter in direct optimization, I would be happy to see how I could do that.

[Rob]

Assigning a material property may work, but not a material name. For solids the Granta MDS tools may be an option once you've done the calculation and want to see what other materials have similar properties. 

[Pedro Quintanilla]

Would it be possible to leave the material name whatever and yet assign an auxiliary variable to use a conditional that picks a value for density, conductivity, cp, and viscosity? This would have to be done with UDF, which is not my field. Do you think it would be possible this way?

 

To explain myself better: I create an Expression named aux with default value 1.

In the UDF code, I import the value of that variable and place a conditional saying if aux==1, rho= T*T+1 (INVENTED EXAMPLE), elseif aux==2, rho=T*T+2.. ..same for conductivity, viscosity and cp.

In this way, for the optimization problem, we could use only the auxiliary variable as input parameter with integer values. Do you think it can work?

[Rob]

Theoretically; parameters are available to UDFs and with the extra parametric options in Fluent at 2023R1 it may work. Feel free to post an example code snippet! 

[Pedro Quintanilla]

 

When I tried it, I had problems with the input of the auxilar parameter un the UDF. Could you help with the code line to fix it? I do not have a lot of experience in this. I tried with “real” declarison of var_aux but I had same results that with double. When I initialize the program, I get the text I put for wrong input value of var_aux (“Error: var_aux must be a integer value between 1-3" in spanish).

 

 

#include “udf.h”

 

DEFINE_PROPERTY(density, cell, thread)

{

    double var_aux = RP_Get_Double(“var_aux”);

    real rho;

 

    if (var_aux == 1)

    {

        rho = 1.055 – 0.660 * C_T(cell, thread);

    }

    else if (var_aux == 2)

    {

        rho = 1108 – 0.666 * C_T(cell, thread);

    }

    else if (var_aux == 3)

    {

        rho = 1185000 – 0.7333 * C_T(cell, thread);

    }

    else

    {

        Message(“Error: var_aux debe ser un valor entero entre 1 y 3.\n”);

        rho = 0.0;

    }

 

    return rho;

}

 

[Rob]

Have a look in the documentation for DEFINE_REPORT_DEFINITION_FN   I have no idea why you're using an RP_ command, and as it's not documented won't be able to tell you anything about it. 

[Pedro Quintanilla]

 

Dear Bob,

I think it was fixed importing with the next line:

real inlet_velocity = Get_Input_Parameter(“var_aux”); 

but now I have a problem with the function. Why do you think it happens? Maybe, it is due to I declared as real or it could be because I should not use the C_T(cell, thread) parameter. 

 

I also tryen declaring the imported parameter as double because the expression value is recognized in this way but there is no difference in the TUI failure return

 

Best regards

Thanks in advance. It is carrying me more time I expected…

 

[Rob]

C_T ought to be fine, if that was causing a more general problem we'd know as it's about the most called value in Fluent UDFs. Programming question (I don't program much), how close to being an integer is the  RP_Get_Double(“var_aux”) value? I was always warned about setting 

value = other value 

as in computational terms there's aways room for variation at the 16th decimal place. 

[Pedro Quintanilla]

It is totally integer. The idea is that the values will be 1,2 or 3 and depending on the value, it is selected one of the density equations.

I selected double declarison because when you write a number in the Expressions window, it is saved as double. In the code, I tryed with var_aux==1.0... but it does not fix the problem. I tryed with real/float/double but it did not change so I guess it is not the problem

Programming is not usual for me so, I try to fix with simple syntaxis but it is possible this is a wrong code...

[Rob]

Hopefully the Community can jump in - UDFs are getting beyond where Staff can help.