I'm working on a time dependent convection-diffusion simulation. The Peclet number is 10^3 to 10^4. The maximum Reynolds number is ~100. The solution tends to diverge at a point where the input velocity increases. The solver starts lowering the time step as it should, but then gives up lowering the time step and quits with an error. I can get the solution to converge nicely by lowering the velocity while keeping the Peclet number constant (decreasing the diffusion coefficient). Or I can get convergence if I check the "exclude algebraic" in the error estimation under the advanced solver settings. The solution with the "exclude algebraic" setting looks good and shows some finite Re effects that I don't want to suppress by artificially lowering the Re number. However, I'm not really sure what the "exclude algebraic" setting does and I'm concerned I might be introducing some error into my solution. I've tried to find an explanation of this option in the documentation without much luck. Can anyone point me to some documentation or explain what this option does?
↧