It is possible the information provided by the discrete adjoint here is somewhat less meaningful, but I need to analyze them for off-design conditions for optimizations. I am using a centered discretrization plus scalar JST dissipation. I have not tried using LU on the subdomains, that is certainly something to try. <div>
<br></div><div>Thank you for your help,</div><div><br></div><div>Gaetan<br><br><div class="gmail_quote">On Mon, Apr 29, 2013 at 10:51 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>
> It is an SA turbulence model and the discrete adjoint computed exactly with<br>
> AD. Certainly the grids are highly stretched in the BL since the grids are<br>
> resolving the viscous sublayer (y+ < 1) and the Reynolds numbers are on<br>
> the order of 10's of millions. I tend only to see this behaviour at<br>
> higher mach numbers when stronger shocks start to appear. For example, the<br>
> adjoint system may solve fine at M=0.80, and fail to converge at M=0.85.<br>
<br>
</div>How meaningful is the information provided by the discrete adjoint here?<br>
Limiters and even just upwind discretizations on non-uniform grids lead<br>
to inconsistent discretizations of the adjoint equations. If the<br>
adjoint equation is full of numerical artifacts, it can cause the linear<br>
problem to lose structure, resulting in singular sub-problems, negative<br>
pivots, and other badness. What happens when you use a direct solve for<br>
subdomain problems (ASM+LU; use smaller subdomains if necessary)?<br>
</blockquote></div><br></div>