We are studying the effects of oblique flow on variable pitch propellers, and we did wind tunnel experiments for these. Now I am running CFD simulations to study the flowfield.
This is the domain. There is a cuboid of dimensions 2.5 x 2.5 x 2.5 m, within which I have 2 cylindrical halves for each blade. Within them I have another cylinder which actually encloses the blade. The middle and the inner domain have a conformal mesh. I did this so that I can quickly rotate the inner section to simulate different pitches.
The strategy I was following is to first run the steady state simulation for 500 iterations, with the transition-SST solver (other transition solvers are crashing) and SIMPLE coupling, and then switch to NITA FSM for transient because it's faster, and I assumed that the solution will be unsteady at higher oblique angles.
While this strategy worked for the no-wind simulations (CFD results close to experimental,) I am getting divergence in the continuity equations when I give a velocity input of 10 m/s. This happens in the transient part, static is able to converge well till 1e-3 for continuity. On observing the velocity contours I observe spurious sections with very high speeds, near the inlet (>200 m/s).
Some sections of the mesh could do with some refinement, so I enabled automatic adaption for volume change and mass imbalance, every 10 iterations. I also reduced the under relaxation parameter to 0.1 for the momentum eq. I am still observing divergence. I ran the simulations with a more refined mesh, but I am still observing the same issues.
As a stopgap solution I am now running the simulations with SIMPLE pressure coupling because it works, though it takes a lot of time. But I still want to know why this problem is happening. Could it be because the compressibility condition is not able to be satisfied, acc to this (
Monitoring residuals , is it safe ?