It is an SA turbulence model and the discrete adjoint computed exactly with AD. Certainly the grids are highly stretched in the BL since the grids are resolving the viscous sublayer (y+ < 1) and the Reynolds numbers are on the order of 10's of millions. I tend only to see this behaviour at higher mach numbers when stronger shocks start to appear. For example, the adjoint system may solve fine at M=0.80, and fail to converge at M=0.85. For these RANS cases, the non-linear solution is solved using only RK with multigrid. <div>
<br></div><div>It is entirely possible a different preconditioner may not help at all, but there's not much else you can do. </div><div><br></div><div>Gaetan<br><br><div class="gmail_quote">On Mon, Apr 29, 2013 at 10:31 AM, Jed Brown <span dir="ltr"><<a href="mailto:jedbrown@mcs.anl.gov" target="_blank">jedbrown@mcs.anl.gov</a>></span> wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div class="im">Gaetan Kenway <<a href="mailto:gaetank@gmail.com">gaetank@gmail.com</a>> writes:<br>
<br>
> For the forward solve I use ASM+ILU in the same manner as for the adjoint<br>
> problem.<br>
> The ASM not a bottleneck per se. Typically we see the adjoint problem<br>
> taking the same amount of time as the non-linear problem for well-behaved<br>
> flows, and the adjoint is shorter for less well-behaved flows.<br>
<br>
</div>Sounds reasonable.<br>
<div class="im"><br>
> The real problem I am having is for certain RANS cases, the<br>
> frozen turbulence adjoint is extremely difficult to solve --- requiring<br>
> GMRES subspace sizes on the order of 400-500 to converge.<br>
<br>
</div>Hmm, which turbulence model are you using? Is it related to stretched<br>
grids? Continuous or discrete adjoint?<br>
</blockquote></div><br></div>