From: Rafael V. <va...@im...> - 2016-06-20 08:39:37
|
Dear Hoang-Giang Bui, to be honest, I'm not an expert of patch tests (not a fan, either). I have tried your test, and asked a colleague to run the same with another code (igatools), and the results are identical. It seems that you only get sigma_xy=0 exactly in the case of an affine transformation. Even in a simple bilinear case (2x2 elements), equivalent to FEM, if you move the central control point the computed sigma_xy is not zero anymore (but it converges when you refine). So, I guess that this is not a problem of GeoPDEs, but about the understanding of (isoparametric) quadrilateral elements. Best, Rafa |