SimpleFoam case not converging for second order scheme
 

Hello everyone,
I am trying to develop a numerical model of flow past serpentine microchannel for Reynolds numbers 10, 50, 100, 250, 500, 750 & 1000.
I want to run it for incompressible, steady state and laminar flow. Hence I have chosen the simpleFoam solver in OpenFOAM v2012.
The issue that I am encountering is that the solution is not converging for Re 500, even though the solution converged for previous Re Nos: 10, 50, 100 & 250.
For the Re 500 case - I ran the case with first-order scheme (Gauss upwind) and it converged. But it isn't converging with second-order scheme (bounded Gauss linearUpwind grad(U)). I tried lowering relaxation factors or adding non-orthogonal correctors but didn't work much. The residual stays at a constant value of around 1e-5 for Ux forever.

I have attached the checkMesh output, fvSolution and fvScheme file for reference.

checkMesh Output

Edit: I increase Reynolds numbers by increasing the velocity only. The working fluid is constant throughout the cases, which is water. The inlet width of the channel is 300 micrometres.

Hello,
I had this problem before, It usually happens actually. This can be due to the simulation is transient in nature. It's actually a common thing when using simpleFoam.

What I might suggest is to track the residuals (usually it's fluctuating). After tracking the residuals and getting almost stable residuals (i.e., fluctuating around a mean value), you can compare the first order vs. the second order solutions. Maybe it's in acceptable range, the error between the two is not much.

Lastly, you can use mapFields utility to map the solution of the first-order to the case with second-order scheme, maybe it helps. But I suggest you check the behaviour of the solution first.

Remember not converging to the desired tolerance doesn't mean the solution is wrong, it might be within acceptable range of accuracy.

Hey,
Thanks for the insights.
I tried running my case with first order for around 20k iterations (I had changed the residuals to 1e-10 just to let it run) and then took the same case and ran for another 20k iterations with second order discretization.
However, I see notable differences in the Velocity profile ( Velocity magnitude vs the inlet width of the channel) and the pressure gradient graph (Pressure Gradient vs length of the channel)
I am attaching the above mentioned plots and the residual plot as well below:
https://ibb.co/k1bFVTg
https://ibb.co/fdLQ1K1
https://ibb.co/3BKWH09