High aspect ratio problem....
 

I am currently solving Euler eqn from my compressible SIMPLE algorithm using structured grids. On some high-aspect ratio grids, convergence is slow and it sometimes fails. I found that happened when the gradient of variable( u v p ...) is large where the aspect ratio of grid is also large. It's just a small region, but make the convergence slow. then i changed second order upwind scheme to first upwind , everything became ok. I don't known why? Does second order scheme makes that happened? and how should i do to solve this problem if i want to get second order accuracy?:(

reduce the CLF number