<html>
  <head>
    <meta content="text/html; charset=ISO-8859-1"
      http-equiv="Content-Type">
  </head>
  <body bgcolor="#FFFFFF" text="#000000">
    Hi Matt,<br>
    I see that you found a similar approach by manually fixing the GLL
    points. Do you have experience with the Moab interface in Nek? It
    seems that the geometry of spectral elements is supported. Then
    probably, the manual fix in nek could be replaced by  the readin of
    a curved mesh format via moab. What do you think?<br>
    <br>
    Florian<br>
    <br>
    <div class="moz-cite-prefix">Am 28.04.2014 12:16, schrieb
      <a class="moz-txt-link-abbreviated" href="mailto:nek5000-users@lists.mcs.anl.gov">nek5000-users@lists.mcs.anl.gov</a>:<br>
    </div>
    <blockquote
      cite="mid:mailman.7271.1398680182.3880.nek5000-users@lists.mcs.anl.gov"
      type="cite">
      <div dir="ltr">
        <div>
          <div>
            <div>
              <div>
                <div>
                  <div>
                    <div>
                      <div>Hi Wei and Florian,<br>
                        <br>
                      </div>
                      In general you can first generate a 2D grid for
                      flow past a wing if it is a straight wing (of if
                      you represent the 3 dimensionality by a forcing in
                      the z direction to indicate a fixed sweep angle).
                      This is not optimal though for a larger problem
                      even if you have periodic boundary conditions
                      because your dz spacing will be fixed at a small
                      value by the BL region requirements. <br>
                      <br>
                    </div>
                    For DNS of the boundary layer region close to the
                    wing you generally want <br>
                    dx+~10<br>
                  </div>
                  dy+ ~ 0.5 next to the wall, less than dz+ away from
                  the wall<br>
                </div>
                dz+ ~ dx+/2<br>
                <br>
              </div>
              <div>This gives a maximum aspect ratio of about 20 near
                the wall (dx/dz). <br>
              </div>
              <div><br>
              </div>
              In the wake / separated flow region you generally want
              max(dx,dy,dz)/eta < 4 where eta is the kolmogorov
              lengthscale. <br>
              <br>
              The problem is that with a wake calculation such as an
              airofil you generally also want the boundaries to be far
              away. You can do this by growing dx and dy away from the
              airfoil. However, you will eventually reach a point where
              your aspect ratio is 10-1,000 far away from the wing as
              well since dz is small. The problem with doing this is
              that when you are far away from the wing then dz will
              still be very small. With a structured mesh there appears
              to be no way to fix this issue without a multiblock
              method. For an unstructured mesh there is a possibility of
              coarsening more in all 3 directions but this is quite
              challenging to generate.<br>
              <br>
              Also, it is not as simple as just thinking about the
              maximum value of (dx,dy,dz) divided by the minimum value
              of (dx,dy,dz). The relative ratios of dx,dy,dz all matter
              and with a wake you generally have regions where you have
              all 6 cases of <br>
              dx < dy < dz<br>
              dx < dz < dy,  <br>
              dy < dx < dz<br>
              dy < dz < dx<br>
              dz < dx < dy<br>
              dz < dy < dx<br>
              <br>
            </div>
            <div>All these areas create problems for an iterative
              solver. You can write a more robust / complex solver (such
              as a semi-coarsening multigrid algorithm) to handle these
              different aspect ratio regions but then much more work is
              required per iteration. It is a trade-off between # of
              iterations required and computing time per iteration.<br>
            </div>
            <div><br>
            </div>
            For generating the mesh itself.<br>
          </div>
          We have an inhouse generated method based on gridgen-c. <a
            moz-do-not-send="true"
            href="https://code.google.com/p/gridgen-c/">https://code.google.com/p/gridgen-c/</a>
          . It is not ideal but it can be made to work. We also looked
          into using cubit for unstructured meshes but we are still
          testing that. With cubit you can either use moab to load in
          the mesh to nek but then you have some limitations. For both
          options you can write your own converter as well... we do this
          for now. <br>
          <br>
          To curve the boundary layer elements we first generate the
          element locations themselves and then the GLL points are
          created on straight line segments. We go back in and manually
          correct the location of the GLL points located closest to the
          wall using a spline of the wing. Then we solve a laplace
          equation to smooth out the GLL points in the rest of the
          domain. <br>
          <br>
          Matt<br>
        </div>
      </div>
      <div class="gmail_extra"><br>
        <br>
        <div class="gmail_quote">On Fri, Apr 25, 2014 at 10:17 PM, <span
            dir="ltr"><<a moz-do-not-send="true"
              href="mailto:nek5000-users@lists.mcs.anl.gov"
              target="_blank">nek5000-users@lists.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 dir="auto">
              <div>Hi matt,</div>
              <div>Sorry that I cannot fully answer your question, I
                know that at least the smallest edge length  in the mesh
                is a measure for the stiffness of the full problem, so
                maybe you should avoid too small element heights in the
                boundary layer. </div>
              <div>However, I also would like to know  how you are
                generating the airfoil mesh, since the mesh has to be
                coarser than a standard meshes and the boundary layer
                elements need to have curved boundaries, no? Which mesh
                generator you use and how do you convert the mesh to Nek
                format? </div>
              <div>The 3d problem should boil down to a 2d problem,
                since I assume that you want simulate a small part of
                the wing with periodic boundary conditions in spanwise
                direction... But wei, for the 2d mesh, did you resolve
                the issue to curve the boundary layer elements?</div>
              <div><br>
              </div>
              <div>Florian</div>
              <div><br>
                Am 25.04.2014 um 18:02 schrieb <a
                  moz-do-not-send="true"
                  href="mailto:nek5000-users@lists.mcs.anl.gov"
                  target="_blank">nek5000-users@lists.mcs.anl.gov</a>:<br>
                <br>
              </div>
              <div>
                <div class="h5">
                  <blockquote type="cite">
                    <div>
                      <div dir="ltr">·HI Matt,
                        <div><br>
                        </div>
                        <div>Till now I have no experiments on 3D
                          problem, what I am interested in is how you
                          generate the 3D or 2D airfoil mesh for
                          nek5000? I spend 2 weeks in generated a 2d
                          airfoil flow mesh without any good results.
                          would you like tell me some informations?
                          thank you a lot!</div>
                        <div><br>
                        </div>
                        <div>Wei<br>
                          <div class="gmail_extra"><br>
                            <br>
                            <div class="gmail_quote">2014-04-25 17:00
                              GMT+02:00 <span dir="ltr"><<a
                                  moz-do-not-send="true"
                                  href="mailto:nek5000-users@lists.mcs.anl.gov"
                                  target="_blank">nek5000-users@lists.mcs.anl.gov</a>></span>:<br>
                              <blockquote class="gmail_quote"
                                style="margin:0 0 0 .8ex;border-left:1px
                                #ccc solid;padding-left:1ex">
                                <div dir="ltr">
                                  <div>
                                    <div>
                                      <div>
                                        <div>Hello,<br>
                                          <br>
                                          I am looking to do simulations
                                          of flow past a wing in 3D
                                          using nek5000 and I have been
                                          thinking more about potential
                                          issues with high aspect ratio
                                          elements. In general we have
                                          very fine resolution near the
                                          wing and then as we get
                                          further away the wall normal
                                          and wall parallel spacing
                                          increases. As a first try we
                                          will extend the domain in the
                                          cross stream direction which
                                          will result in small dz
                                          values. I know that in general
                                          the best performance is
                                          obtained with elements having
                                          dx=dy=dz and that as the
                                          aspect ratio increases the
                                          performance will degrade. <br>
                                          <br>
                                          I'm wondering if there are
                                          general rules of thumb for the
                                          performance degradation with
                                          increased aspect ratio. For
                                          example, is an aspect ratio of
                                          10 ok but an aspect ratio of
                                          100 unacceptable? Is this even
                                          something we can estimate in
                                          general or does it vary so
                                          much problem to problem that
                                          no general estimate is
                                          possible?<br>
                                          <br>
                                        </div>
                                        I saw an earlier post that
                                        referred to the paper "An
                                        Overlapping Schwarz Method for
                                        Spectral Element Solution of the
                                        Incompressible Navier-Stokes
                                        Equations", P. Fischer JCP 1997.
                                        From the paper I see two general
                                        strategies. <br>
                                      </div>
                                      1. limit the maximum aspect ratio
                                      to a critical value<br>
                                    </div>
                                    2. design a grid for our case, run
                                    it for a short time and then
                                    iteratively add more grid points to
                                    decrease the aspect ratio until
                                    optimal performance is achieved. <br>
                                    <br>
                                  </div>
                                  Does anyone have a general or specific
                                  suggestion regarding how we should
                                  handle the grid generation in terms of
                                  selecting the largest aspect ratio
                                  possible with low computational cost?<br>
                                  <br>
                                  Thanks,<br>
                                  <br>
                                  Matt<br>
                                </div>
                                <br>
_______________________________________________<br>
                                Nek5000-users mailing list<br>
                                <a moz-do-not-send="true"
                                  href="mailto:Nek5000-users@lists.mcs.anl.gov"
                                  target="_blank">Nek5000-users@lists.mcs.anl.gov</a><br>
                                <a moz-do-not-send="true"
                                  href="https://lists.mcs.anl.gov/mailman/listinfo/nek5000-users"
                                  target="_blank">https://lists.mcs.anl.gov/mailman/listinfo/nek5000-users</a><br>
                                <br>
                              </blockquote>
                            </div>
                            <br>
                            <br>
                          </div>
                        </div>
                      </div>
                    </div>
                  </blockquote>
                  <blockquote type="cite">
                    <div><span>_______________________________________________</span><br>
                      <span>Nek5000-users mailing list</span><br>
                      <span><a moz-do-not-send="true"
                          href="mailto:Nek5000-users@lists.mcs.anl.gov"
                          target="_blank">Nek5000-users@lists.mcs.anl.gov</a></span><br>
                      <span><a moz-do-not-send="true"
                          href="https://lists.mcs.anl.gov/mailman/listinfo/nek5000-users"
                          target="_blank">https://lists.mcs.anl.gov/mailman/listinfo/nek5000-users</a></span><br>
                    </div>
                  </blockquote>
                </div>
              </div>
            </div>
            <br>
            _______________________________________________<br>
            Nek5000-users mailing list<br>
            <a moz-do-not-send="true"
              href="mailto:Nek5000-users@lists.mcs.anl.gov">Nek5000-users@lists.mcs.anl.gov</a><br>
            <a moz-do-not-send="true"
              href="https://lists.mcs.anl.gov/mailman/listinfo/nek5000-users"
              target="_blank">https://lists.mcs.anl.gov/mailman/listinfo/nek5000-users</a><br>
            <br>
          </blockquote>
        </div>
        <br>
      </div>
      <br>
      <fieldset class="mimeAttachmentHeader"></fieldset>
      <br>
      <pre wrap="">_______________________________________________
Nek5000-users mailing list
<a class="moz-txt-link-abbreviated" href="mailto:Nek5000-users@lists.mcs.anl.gov">Nek5000-users@lists.mcs.anl.gov</a>
<a class="moz-txt-link-freetext" href="https://lists.mcs.anl.gov/mailman/listinfo/nek5000-users">https://lists.mcs.anl.gov/mailman/listinfo/nek5000-users</a>
</pre>
    </blockquote>
    <br>
    <pre class="moz-signature" cols="72">-- 
-----------------------------------------------------------------------------
Dipl. Ing. Florian Hindenlang
Institut fuer Aerodynamik und Gasdynamik 
Phone: 0049 (0)711-685 63413
office 1.14
Pfaffenwaldring 21
70569 Stuttgart 
E-Mail: <a class="moz-txt-link-abbreviated" href="mailto:hindenlang@iag.uni-stuttgart.de">hindenlang@iag.uni-stuttgart.de</a>
---------------------------------------------------------------------------
</pre>
  </body>
</html>