Learning Forum

[suraj9735]

I am solving a pure conduction problem in solid, And I test the results in both ANSYS Mechanical and Fluent too.

I have to share something.

ANSYS Mechanical-

1) Gives very good/appropriate/relevant result in automatic mode.
2) I am not familiar with scripting/command in the workbench Mechanical (WM). So I switch to Fluent and easy to use UDF for my model.
3)Very less access of solver through GUI in workbench Mechanical. And I also don't know how to check convergence plot (like Fluent shows at every iteration), How to put Boundary condition varying with solver variable, etc. in WM.


Fluent-
1)It's very hard to let the solution converged at every time step in long heating time (1 hour) and reports don't change much with iterations. So it gives bad result comparing to WM.
2)Fluent has very much access to solver through GUI and can set so many things easily about convergence, reports, etc.
3) I think UDF in Fluent is relatively easier than workbench Mechanical scripting or APDL command.

[suraj9735]

I have two pictures of results given by Mechanical and Fluent of the same case for comparison but have too much difference, Even energy equation is converged till e-14 in Fluent.....

How can I make my result better in Fluent?
How can I study quickly to write command in the workbench Mechanical?

I confused what to do now?, Should I start to learn scripting/command/APDL-command in workbench Mechanical or try to improve my result in Fluent?

 

 

[Keyur Kanade]

for converging it in every time step, please use fine mesh. please reduce the time step size. 

fluent uses k as unit for temperature. you may to change it. 

fluent uses cell center values. then they are recalculated for nodes and used for showing node vales in fluent post processing. 

if this helps you, please mark this as 'is solution' to help others on forum. 

[suraj9735]

Thanks, Kkanade for your reply!

What should be the strategy to do the simulation for long heating (1 hour) in Fluent? Only solve energy equation for pure conduction problem.

1) For an accurate result, What should be the number of elements in Mesh?

2) Solution method? (SIMPLE-SIMPLEC-PISO-COUPLED)

3) In my case, the heating time of plate is near about one hour, But I am interested in a solution in between suppose 30-50 minutes. Not care about the beginning solution.

   a) What time step should I choose at the beginning, in between 30-50 minutes, and at last?

   b) What about "Adaptive Time-step or Fixed Time-step"?

   c) I read some post that suggests doing steady-state solution before transient run for better convergence/for saving time. But I am still unable to figure it out, How can I apply it in my model?  While I need data of my model when it crosses the target temperature. And It seems me, I have to do transient run from the beginning (time=0 second) to get the desired data at the desired time.

[Amine Ben Hadj Ali]

Use Biot and Fourier number number analogy to accrlerarte the runand then scale the transient behavior back. That a compromise for engineering complicated cases.

Regarding your questions

1/carry out mesh dependency study

2/for simple condition all numerical methods are equivalent 

3/time scale is a portion of Lsolid^2/thermal diffuivity. Lsolid is the cubic root of the solid volume for example 

[suraj9735]

Thanks a lot!