Graduate Seminar

Scheduled Talk - October 28, 2009 - [ 12:00PM in DH 1064 ]

Sean Hardesty

"Optimization Problems for Naghdi Shells"

Abstract:
The Naghdi model for elastic shells accomplishes reduction of three-dimensional elasticity to a two-dimensional model by means of a kinematic assumption and integration across the thickness. It can achieve substantial computational savings, but discretizations may suffer from the "locking" phenomenon, in which the convergence is non-uniform in the thickness, and is prohibitively slow for practical problems. This phenomenon also has important ramifications for optimization: in the presence of locking, the computed derivatives and optimum may differ substantially from their true values. I will discuss the formulation of the Naghdi model and adjoint equations for derivative computations, and present some numerical examples involving optimization of roof shape so as to minimize deflection or strain energy under various loads.