Colloquium September 1
Date: Friday September 1
Time: 16:00 to 17:00
Place: 455 McBryde (Commons Room)
Speaker: Eric de Sturler of the Math Dept
Title: Large-scale numerical simulation for analysis and design
Party at University Club at 8p.m.
Abstract
Large-scale numerical simulation is becoming an indispensable tool for
understanding and analyzing processes in engineering, physics, chemistry,
biology, and medicine. In addition, simulation more and more helps us to
optimize complex processes and designs. The fast solution and optimization
of very large linear and nonlinear algebraic systems is an important
aspect of such simulations, since these usually requires by far the most
time and memory. Looking at a range of applications, such as designing
structures and advanced materials, tomography, computer graphics, and
simulations involving uncertainty quantification, we will introduce and
discuss some of the central issues of efficient and robust linear and
nonlinear solvers and optimization methods. Finally, we will discuss the
Quantum Monte Carlo method for computing materials properties. For these
problems the dimensionality of the problem precludes directly modeling
accurately all forces and possible configurations, and we need either
significant
simplifications or we have to resort to stochastic methods. We will see
that in such problems the methods discussed earlier (and other unsuspected
numerical analysis methods) can play a fundamental role.
Return to