how to determine delta t for a transient run?

[pei-ying.hsieh]

Dear Autodesk simulation CFD experts,

I am running a transient conjugate heat transfer case.  There is a fan in my model, hence, with force convection.  I solved the flow field without thermal first.  Then, turn flow off and turn heat transfer on.  I set iteration of each time step to 2.  How to determine what  the apporiate delta t for each step should be?

Thanks!

Pei-Ying

[apolo_vanderberg]

Here's a SimTV where we talk about selecting appropriate timestep sizes for a flow solution

Basic Transient

From what you outlined you are leveraging a steady state flow solution and doing transient just for thermal

In these situations typically the timestep size would be based the thermal setup.

Variations in any of the boundary conditions, such that if Heat Generation was varying with respect to time we need a timestep small enough to capture this variation.

Or the progression of how quickly the temperature may change over time (this would be specific to material assignments) however in most cases we can get by with a timestep that is on the order of seconds large ie10sec.

In some models if the thermal solution is expected to take a notably longer time we can step up to a timestep that might be minutes large as well.

[pei-ying.hsieh]

Thanks a lot Apolo!

In my model, I do have thermoelectric cooling material.  The heating/cooling depends on temperature difference between the hot side and cold side.  Currently, I used 1 second for delta t and solution seemed OK.  I am planning to go up to 2 seconds for delta t.  Total time of the run is 2 hours.  So, I probably do not have to go up to 1 min for delta t.

Pei-Ying

[OmkarJ]

I would suggest doing a sensitivity analysis on timestep by creating monitor points at critical areas.   The adequate timestep would be the one that gives you flatter Temperature plots (and other critical variables) at the end of each timestep (which consists of several inner iterations) for all these monitor points. This way you have ensured you have resolved adequately in time.

[pei-ying.hsieh]

Hi,

Thanks for the suggestion.  Yes, it will be the right way to do it.  But, this run takes 4 days to complete.  This will be a luxury that I probably do not have.

One question about your suggestion:  In the tutorial of forced convection case, it mentions that only 2 iterations is needed at each time step because solution is linear. Is there any reason why you suggestion several inner iterations at each time step?

Pei-Ying