=93Why Mathematicians Are Still
Necessary to Industry Since
Engineers Can Do Everything=94
Gundolf Haase
Johannes Kepler University, Linz, (Austria)
Wednesday, April 7, 1999 at 3:30 p.m.
(Refreshments at 3:00 p.m.)
327 McVey Hall
ABSTRACT: This statement is true 98% of the time but the
remaining 2% determines the success of an enterprise in relation to the
competition. Therefore mathematicians and other scientists have to
play around with new ideas and their transfer to applications. Most
ideas for advanced iterative solvers were born on the unit square, e.g.,
the classical multigrid idea, adaptive solvers, or domain decomposition
techniques. The analysis is often performed on simple domains and
will be extended to more challenging examples. The development of
code starts in the same way. A new type of problem starts as soon
as the new solution techniques meet practical requirements. Many
unpredictable changes in the code are necessary. Potentially we
even have to change the data representation. Additionally, the new
solvers detect modeling inaccuracies and we have to adapt the
model. And suddenly, we are faced with some mathematical problem
that was not in the scope of our initial investigation. Certain
points of this painful but satisfying path are presented by the
development of parallel solvers, by the mass minimization of an injection
moulding machine, by magnetic field calculations, and some minor
projects.