[robingilbert]

Hello everyone,

I work on CFD simulations of data centers using OpenFoam. I am trying to validate the simulations with experimental data. I use transientSimpleFoam solver for my simulations. I place a sensor at the inlet of the cooling unit to see how the temperature varies over time. I perform the simulations using the same conditions but my results are way off. Although the steady state temperatures match, the path followed by simulations are not accurate. I am attaching the results below. The blue line is the experimental result while the red one is simulation result. Please give you suggestions to improve my simulation.

[gschaider]

Originally Posted by robingilbert

Hello everyone,

I work on CFD simulations of data centers using OpenFoam. I am trying to validate the simulations with experimental data. I use transientSimpleFoam solver for my simulations. I place a sensor at the inlet of the cooling unit to see how the temperature varies over time. I perform the simulations using the same conditions but my results are way off. Although the steady state temperatures match, the path followed by simulations are not accurate. I am attaching the results below. The blue line is the experimental result while the red one is simulation result. Please give you suggestions to improve my simulation.

You don't give us very much information to work with. For instance a description of the case would be nice to see whether the solver is appropriate.

About the solver: which OF-version are you using? Because the transientSimpleFoam that I found is quite old and doesn't solve the temperature equation (it's incompressible)

[robingilbert]

Hello Bernhard Gschaider,

Sorry about not giving much details... so here goes...

What I am trying to solve for is a data center room with so many servers in it and one cooling unit. The air flow through the servers and the cooling unit are constant and I modified the solver based on channelFoam to fix the flow rate (it works because I tested it with swak4Foam application- thanks to you!!). Now the heat added by the servers is modeled as a source term in the temperature equation. Here is the temperature equation that I solve. I added this to the transientSimpleFoam solver. I am using OF version 1.6.x.

fvScalarMatrix TEqn
        (
         fvm::ddt(T)
         + fvm::div(phi, T)
         - fvm::laplacian(DT, T)
    == Q/(rho0*Cp0)-
        );
    TEqn.solve();

Q is the source term which is active only where the servers are present. For cooling unit, the temperature follows a linear curve with respect to inlet temperature to the unit. so I fix the temperature of the air within the cooling unit for every iteration as follows:

T[cellI]= (0.21739*T[labelOfClosestCell]+225.739);

T[labelOfClosestCell] is the temperature at the inlet of cooling unit. Maybe this is where I am wrong? I am not sure about the above step that I took. Anyway, as shown in the plot above, the steady state temperatures look reasonable but the path it takes does not look good. Am I missing something? or is my set up wrong? I would like to fix this issue so please let me know if you need any more details.

Thank you so much

[gschaider]

Originally Posted by robingilbert

Hello Bernhard Gschaider,

Sorry about not giving much details... so here goes...

What I am trying to solve for is a data center room with so many servers in it and one cooling unit. The air flow through the servers and the cooling unit are constant and I modified the solver based on channelFoam to fix the flow rate (it works because I tested it with swak4Foam application- thanks to you!!). Now the heat added by the servers is modeled as a source term in the temperature equation. Here is the temperature equation that I solve. I added this to the transientSimpleFoam solver. I am using OF version 1.6.x.

fvScalarMatrix TEqn
        (
         fvm::ddt(T)
         + fvm::div(phi, T)
         - fvm::laplacian(DT, T)
    == Q/(rho0*Cp0)-
        );
    TEqn.solve();

Q is the source term which is active only where the servers are present. For cooling unit, the temperature follows a linear curve with respect to inlet temperature to the unit. so I fix the temperature of the air within the cooling unit for every iteration as follows:

T[cellI]= (0.21739*T[labelOfClosestCell]+225.739);

T[labelOfClosestCell] is the temperature at the inlet of cooling unit. Maybe this is where I am wrong? I am not sure about the above step that I took. Anyway, as shown in the plot above, the steady state temperatures look reasonable but the path it takes does not look good. Am I missing something? or is my set up wrong? I would like to fix this issue so please let me know if you need any more details.

Thank you so much

Don't know. Fixing the temperature that way IS kinda ugly and probably messes up conservation.

Anyway. One more thing that comes to mind: have you checked the VALUE of DT (because you probably wil have "dived" rho*cp out of that too).

[robingilbert]

Originally Posted by gschaider

Anyway. One more thing that comes to mind: have you checked the VALUE of DT (because you probably wil have "dived" rho*cp out of that too).

Yes the thermal diffusivity that I set was 0.00002 m^2/s for air.