CSAR Seminar

SPEAKER: Jie Wan, Rensselaer Polytechnic Institute

TITLE: Adaptive Mesh Control for Evolving Geometry Problems

DATE: Wednesday, May 3, 2006
TIME: 12:00 Noon
PLACE: 2240 DCL
1304 W. Springfield Ave., Urbana, IL

ABSTRACT

In 3D simulations of large deformations, the geometries of deformable objects evolve to represent material flow. The meshes of the deforming objects typically must be updated frequently to continue the analysis due to large mesh-associated errors, including mesh discretization errors, geometric approximation errors, and element distortions. Historically, the approach taken to address the need for updating once a mesh becomes unsatisfactory is to perform a full remeshing of the entire domain even though the mesh may contain only a limited number of distorted elements. More importantly, remeshing process imposes the need for transferring history-dependent solution data globally from old mesh to new mesh, which tends to diffuse the transferred solution variables numerically. To achieve satisfactory solution accuracy and computational efficiency, an automated simulation process based on simulation monitoring and adaptive mesh control executed through the application of local mesh modifications is developed. The adaptive mesh control procedure is coupled with FEM engines for automated finite element simulations of various types of 3D problems with evolving geometry.

The decision on the need for adaptive mesh enrichment is based on monitoring the errors associated with the mesh during the analysis increments. When the errors become too large, the analysis is stopped and adaptive mesh control is triggered to perform the needed localized mesh improvements and incremental solution transfer. The topological description of the object boundary is properly updated to reflect the current deformation configuration and geometry features before the mesh enrichment process is applied. A posteriori error indicators based on plastic deformation related quantities are used to control the mesh discretization errors through an h-adaptive procedure. The mesh size field is further processed to control the geometric approximation errors in the critical portions of the object boundary. The mesh of the deformed object is then modified through controlled application of local mesh modifications to meet both the needs of mesh discretization and element quality. The history dependent solution fields are locally and incrementally transferred from the old mesh to the updated mesh as mesh modifications are applied, based on local mesh modification driven solution transfer operators. A set of industrially relevant problems is investigated to demonstrate the capabilities developed.

BIOGRAPHY

After graduating with B.E. in mechanical engineering from National University of Defense Technology in 1989 and M.E. in computer engineeing from Chinese Academy of Space Technology (CAST) in 1992, Jie Wan worked as a project manager and a senior CAE engineer in CAST for more than 4 years. He then received his M.S. from University of Massachusetts at Amherst and Ph.D. from Rensselaer Polytechnic Institute, both in mechanical engineering. His current research interests include finite element methods, automatic mesh generation, adaptive mesh control, large deformation simulation automation, advanced geometric modeling, computer aided design, interactive 3D computer graphics, and reverse engineering.