On Mon, 18 Jul 2016, Xujun Zhao wrote: > Yes, the points are not quad-points but arbitrary points, which are moving > particles in the domain. It mainly happened when the points are out of the > element.
Oh, yes, looking for points far outside an element can give you inverse map failures. If you imagine extending the master->physical element map for, say, a QUAD4 trapezoid with one very narrow side on top, you can see how the map stops being invertible a little way above that top. We avoid ever hitting many such cases because we precede the inverse_map with a bounding box test, but that's a performance optimization, not a guarantee of an invertible map. We pass secure=false to inverse_map to, in theory, avoid screaming and dying when handling a case that might be expected to fail. Looks like there's a bug in the 3D case, though? We rely on TypeTensor::solve() in the Newton iteration in 3D, but we *don't* warn solve() (or have a way to warn solve()?) that in the secure=false case we don't necessarily expect a solution to exist. So then I guess the reason you're getting a zero determinant is that, for people who hit a positive determinant the iteration tends to work, for people who hit a negative determinant the iteration diverges and contains_point() reports false, but you managed to test a point right on the cusp? --- Roy ------------------------------------------------------------------------------ What NetFlow Analyzer can do for you? Monitors network bandwidth and traffic patterns at an interface-level. Reveals which users, apps, and protocols are consuming the most bandwidth. Provides multi-vendor support for NetFlow, J-Flow, sFlow and other flows. Make informed decisions using capacity planning reports.http://sdm.link/zohodev2dev _______________________________________________ Libmesh-users mailing list Libmesh-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/libmesh-users
