Low rank tensor approximation in IgA matrix generation

Dr. Angelos Mantzaflaris

April 14, 2015, 3:30 p.m. S2 059

Isogeometric discretizations have several similarities with the
isoparametric concept in classical finite element methods. In the latter
methods, a local geometry map is used to forward reference elements to
the computational domain. In NURBS-based isogeometric analysis the
reference element is substituted by a structured collection of
parametric elements and the local geometry map is replaced by a global
one, acting on all parametric elements. In this talk we will show how we
can take advantage of this new paradigm in order to speed up
significantly the isogeometric matrix generation task. The main
computational tool which we are going to use is low rank tensor
approximation.