Kompaktkurs im GRK 2075 Prof. Dr. Hermann G. Matthies The

Graduiertenkolleg 2075
Modelle für die Beschreibung der Zustandsänderung
bei Alterung von Baustoffen
Kompaktkurs im GRK 2075
Prof. Dr. Hermann G. Matthies
Institut für Wissenschaftliches Rechnen
The fusion of computational models and data
Donnerstag, 17.03.2016, 10.00 - 11.30 Uhr und 13.30 - 15.00 Uhr
Institut für Wissenschaftliches Rechnen, Raum 812
Mühlenpfordtstrasse 23, 38106 Braunschweig
When considering some existing system, both data from sensors, measurements, or
observations, and mathematical / computational models are forms of knowledge about the
system. Their fusion can allow one to exploit this information to increase or sharpen the
knowledge about the system. This is particularly true for complex systems, where often
the quantities of interest are not directly observable. Here the computational model can
act like an extension or additional measurement instrument. This has many different
potential areas of application, such as monitoring, state tracking and identification,
quantification and inverse identification of parameters and quantities of interest,
assessments after accidents or damage, etc.
One way to achieve this is the Bayesian approach to inverse problems, the basic ideas of
which will be presented. Traditionally often deemed too computationally demanding for
complex mathematical models, relatively recent numerical techniques for uncertainty
quantification - the forward problem - offer a realistic possibility for this task, considerably
increasing its range of practical applicability, and also benefit the inverse problem.
As these recent numerical techniques accelerate the forward problem, they also have
direct bearing - independent of the inverse problem - on the efficient uncertainty
quantification (UQ) of system states and other quantities dependent on it due to
uncertainties in system properties and / or system loading / excitations. This can then be
used in traditional UQ tasks like determining the probability of occurrence of certain
system states, as it is used for example in quantitative risk assessment.