2 Fundamental CFD Questions regarding convergence
Hello,
I have two questions regarding the rate of convergence in a problem. I am solving a fairly simple creeping flow problem in a square duct - zero applied pressure,slowly varying axial viscosit, and the flow is driven by slowly varying electroosmosis along the boundaries. I am using the SIMPLER algorithm, and solving for the steady state flow solution. Initially, I had used a channel of dimensions 1x1x20, and grid spacing of .05 in x,y, and z (20, 20, and 400 nodes). This allowed me to get very tight convergence criteria after around 500 iterations, with a pressure correction residual around 10^-10. Now, however, I have needed to lengthen my channel to 1x1x100, and I have changed my grid spacing to .10 in x,y, and z (10, 10, and 1000 nodes). Suddenly, I cannot get the pressure correction below approximately 10^-6, even after 2000 iterations. Once it gets in that vicinity, it changes VERY slowly and I don't think it will ever get smaller. Two fundamental questions- 1. Is this behavior expected? I was thinking that I have fewer total grid points (100,000 vs. 160,000), so I was expecting to reach tighter convergence more quickly. Is there a chance that I will never be able to reach that very low (10^-10) residual from my earlier runs, because now the grid spacing is more coarse? 2. If I run this simulation for an "infinite" amount of time (i.e. - 10,000 iterations), will the residuals always continue to get smaller, or do all residuals actually converge to a steady value, eventually slowing in their approach to the steady state value? Could this be why I am observing such a slow change in the pressure correction residual? Thoughts? Thanks, Jon |
All times are GMT -4. The time now is 16:37. |