In general skewed element dicrease a lot the accuracy of the scheme, if not sometimes prevent convergence with certain FV schemes. See work of Christophe Le Potier on google scholar.

Yes ... highly skewed cells can lead to convergence difficulties and inaccuracies in the numerical solution... its always the quality of grid that most critical for reliable solutions.

Is the problem in the solid body governed only by a linear heat transfer equation?

Try to mesh with no or minimum number highly skewed cells as it leads in deviating actual results

Filippo Maria Denaro actually it is heat conduction with temperature dependent thermal conductivity. No heat generation.

Well, I don't think that the conduction in the body, even if the solution is not of high accuracy, can alter the flow solution in the interior. The linear heat transfer, also on skweed grid, should be enough well solved. I would check the grid resolution in the thermal flow boundary layer, I will check the condition for the normal derivative of the temperature at the wall that must match the conduction condition.

However, if you have discrepancy in the flow solution compared to known data I think you should provide more details.

Filippo Maria Denaro

Thanks for your elaborate answer.

Here are few more details.

The problem is : The fluid is flowing through circular cavity having four layers of insulation. The inner surfaces of cavity has S2S radiation. Inside fluid there is very small radiating solid (internal heat generation) which gives some heat to fluid through convection and mostly to inner cavity walls through radiation.

The problem is as one of the insulation layer is too thin compared to others, I have only one element across that material's thickness, that too very skewed.

The outer layer of insulation has radiation and convection to ambient.

The result shows the some portion of solid touches the lower temperature limit (1K) and some becomes hotter than inner surface. Which is not possible.

The problem is transient. The flow is not so critical here as its role in overall temp distribution is very limited. Radiation and conduction are governing mechanisms.

The fluent shows warning of

"Temperature of x cells on solid 1 of zone y limited to 1.000000e+2 K"

I am not aware about this error, but I have modeled similar problem with very thin insulation layer.

According to interaction with Ansys Application Engineer, the cell zone in Fluent must contain at least three cells along the thickness of zone. I think you can try another mesh accordingly.

Cheers!

Well, your problem seems to show numerical wiggles that can be due to the onset of numerical instability. Try to regularize the grid skweness, add some more layer, use implicit method or explicit but at lower time step.

Simple answer: Yes, sure.

Thanks Pratik Joshi , Filippo Maria Denaro and Thomas Frank

Reducing under relaxation factor for Energy and time step size has resolved the problem.

Great!...@Hardik Mistry. But you have mentioned in the problem description that,

The problem is as one of the insulation layer is too thin compared to others, I have only one element across that material's thickness, that too very skewed.

And you got results, in your opinion, Are those results seem physical/correct?

Because, as discussed with Ansys people, they are suggesting me to go for at least three cells along the thickness of cell zone.

Hi Pratik Joshi

Yeah, the results seem physically correct. Not sure about the exact accuracy though.

Three to five elements are suggested across the thickness. But, I am not sure whether this holds true for solids as well.

In fluids, the velocities are zero at the walls and maximum at the center and this behavior might be difficult to capture by one element.

But, in solids the temperature distribution across the thickness can be accurately captured by a second order polynomial. So I guess one second order element should give the sufficiently accurate results.

My problem was with high URF the solution was diverging. I mean in few elements it was reaching to minimum value of 1 K. Which is not possible. Reducing the energy URF it has converged well. And with the comparison of earlier results and some experience with this type of systems, I can say it is fairly accurate.

Hardik