[Mina_Mbg]

I am doing a simulation for a gas quenching process in ansys Fluent 16.1.
the equations are:
Navier-Stokes Equation
SST k-omega
and Energy equations
which are solved in couple mode.
Applied gases are Nitrogen and helium in 8bar and 10bar respectively.inlet velocity in 35 m/s.
Inlet temperature(in steady) is 1143°K.
The gases density are temperature dependent(piecewise-linear)
I have 10 million elements.
in steady mode I have to consider 0.0027 for residual of continuity otherwise I will not get convergence.
where this high residual come from?
Maybe it is helpful to know that I use Polyhedral mesh.

[LuckyTran]

Is everything else okay other than continuity residual being high?

The continuity residuals are scaled based on the worst residual encountered during the first 5 iterations. If your worst residual during these 5 is not high, then you'll end up with relatively high residuals.

For example, if you initialize the flow with the (numerically) exact solution, your continuity residual will be stuck at 1 even though your solution is perfect.

[Mina_Mbg]

thanks for yourexplanation.
yes the others behave pretty good.Althogh omega is decreasing a bit slower than k.
I use hibrid initialization.You think I have to check the standard one?

can you lease explain this sentence by an excample?
" the worst residual encountered during the first 5 iterations. If your worst residual during these 5 is not high, then you'll end up relatively high residuals."

Originally Posted by LuckyTran

Is everything else okay other than continuity residual being high?

The continuity residuals are scaled based on the worst residual encountered during the first 5 iterations. If your worst residual during these 5 is not high, then you'll end up relatively high residuals.

For example, if you initialize the flow with the (numerically) exact solution, your continuity residual will be stuck at 1 even though your solution is perfect.

[LuckyTran]

The residuals you see outputted by Fluent are scaled/normalized.

Let's say you run 10 iterations and get the two following raw residuals for continuity.

0 50 150 10 5 4 3 2 1 1

The worst of the first five residuals is 150, therefore from iterations 6 onwards the residuals you see output by Fluent are 4/150, 3/150, 2/150, 1/150

Now say you have a different set of initial conditions and you get these residuals instead

0 50 50 10 5 4 3 2 2 2

The raw residual values for iterations 6 and onwards are greater than before, but because the worst residual is not 50 instead of 150, you will see scaled residuals output as 4/50, 3/50, 2/50, 2/50, 2/50.

Thus if you solution is well-behaved during the first five iterations, you can see smaller scaled continuity residuals even though the actual imbalance of mass is greater than the first case.

In the extreme example, if you initialize your simulation with the exact numerical solution you will get the following residuals:

1e-16 1e-16 1e-16 1e-16 1e-16 1e-16 1e-16 1e-16 1e-16 1e-16

Now the worst residual during the first five iterations is 1e-16 and scaled residuals from iterations 6 onwards are 1 1 1 1 1.

Now 1 is much larger than the previous residuals (1/150 or 2/50) however, the solution is indeed the exact numerical solution to the problem. If you are to judge convergence and accuracy by continuity residuals alone then you would erroneously conclude that the exact solution is NOT the solution you are after or that it is not yet converged.

If everything else in your simulation is okay except for the continuity residual, I am suggesting that this may be the cause. To verify if this is the case, reset your residual history and run your simulation for some more iterations. If the continuity residual is stuck at 1 then you have confirmed it.

Note that only the continuity residuals are scaled this way. The other quantities have much more meaningful scalings.

[Mina_Mbg]

very usefull information!
if I understand correctly I have to set my continiuity residual exactly the same as the number it has converged and re-do the calculation without initialization to check if this you have explained happens in my case?

Originally Posted by LuckyTran

The residuals you see outputted by Fluent are scaled/normalized.

Let's say you run 10 iterations and get the two following raw residuals for continuity.

0 50 150 10 5 4 3 2 1 1

The worst of the first five residuals is 150, therefore from iterations 6 onwards the residuals you see output by Fluent are 4/150, 3/150, 2/150, 1/150

Now say you have a different set of initial conditions and you get these residuals instead

0 50 50 10 5 4 3 2 2 2

The raw residual values for iterations 6 and onwards are greater than before, but because the worst residual is not 50 instead of 150, you will see scaled residuals output as 4/50, 3/50, 2/50, 2/50, 2/50.

Thus if you solution is well-behaved during the first five iterations, you can see smaller scaled continuity residuals even though the actual imbalance of mass is greater than the first case.

In the extreme example, if you initialize your simulation with the exact numerical solution you will get the following residuals:

1e-16 1e-16 1e-16 1e-16 1e-16 1e-16 1e-16 1e-16 1e-16 1e-16

Now the worst residual during the first five iterations is 1e-16 and scaled residuals from iterations 6 onwards are 1 1 1 1 1.

Now 1 is much larger than the previous residuals (1/150 or 2/50) however, the solution is indeed the exact numerical solution to the problem. If you are to judge convergence and accuracy by continuity residuals alone then you would erroneously conclude that the exact solution is NOT the solution you are after or that it is not yet converged.

If everything else in your simulation is okay except for the continuity residual, I am suggesting that this may be the cause. To verify if this is the case, reset your residual history and run your simulation for some more iterations. If the continuity residual is stuck at 1 then you have confirmed it.

Note that only the continuity residuals are scaled this way. The other quantities have much more meaningful scalings.

[ama294]

Hi Mina_Mbg

I have the same problem. Have you tried the suggested method? would you please share your experience if you have solved the continuity high residual convergence..

Thanks in advance
Aziz