Seminars Graduate School SimTech: Winter Term 2010/11

Finite Element Technology

Prof. Dr.-Ing. Manfred Bischoff

- finite elements for thin-walled structures and continua

- spurious stiffening phenomena, locking

- advanced finite element methods: reduced integration, mixed and hybrid mixed finite elements, collocation methods

- finite elements using b-splines, isogeometric approach

- lecture notes "Advanced Computational Structural Mechanics"

- O.C. Zienkiewicz, R.L. Taylor, The Finite Element Method (several issues since the 80s).

- T.J.R.Hughes, The Finite Element Method - Linear Static and Dynamic Finite Element Analysis, Dover 1987 (new edition 2000).

- C. Felippa, Introduction to Finite Element Methods, lecture notes, available online: http://www.colorado.edu/engineering/CAS/courses.d/IFEM.d/Home.html

Numerical Methods and Their Application in Molecular Simulation on New Computer Architectures

Judith Rommel, Jun.-Prof. Johannes Kästner, Prof. Helmut Harbrecht

To extend our interdisciplinary communication, algorithms and numerical solvers are discussed. As simulations are often based on numerical libraries, we will focus on the concepts behind them. Themes reach from eigenvalue to partial differential equation solver in numerical libraries: e.g. the very common BLAS, LAPACK libraries, the sparse solver package PARDISO, or special solvers constructed for parallel platforms. You can choose something connected with your own work or your special interest. Your talk can be given in German or English.

Numerical simulation of multiphase flow

Prof. Bernhard Weigand, Prof. Claus-Dieter Munz, Prof. Christian Rohde

multiphase flow, modelling, numerical methods, simulation, experimental methods, jet breakup, drop dynamics, phase change

Physics and computer applications

Prof. Christian Holm, Prof. Rudolf Hilfer und Jun.-Prof. Axel Arnold

- computer simulations in physics

- soft matter and granular media

- state-of-the-art methods in simulations

Stochastic and Statistical Topics in Simulation Technology: Responsibility of Developers, Simulators, and Best Practice

Jun.-Prof. Dr.-Ing. Wolfgang Nowak

(partially adaptive, depending on mixture and preferences of participants)

Many technical, economical, ecological and political are based on simulation results. Highly developed simulation technology may be misleading, making the end user think that simulation is reality. Of course, we know better. But what exactly is the responsibility of model/code developers, software distributors? How to make sure that the involved models, algorithms, software meets quality requirements? What are these quality requirements? How to communicate approximations, limitations, model and parameter uncertainties to end users and to the public?

Within this general set of questions, we will discuss, learn from each other, and try to find answers:

* What are sources of error in the chain of conceptual model development, mathematical/numerical implementation, software engineering, model calibration, simulation, visualization, interpretation of results and final decision making?

* What exactly is the responsibility of model/code developers?

* How to communicate the above errors and uncertainties to the end user, the decision maker, and the public? How do we do this without undermining the value and esteem of simulation?

* Can we develop a "Best Practice" of how we should work so that we stand up to our responsibility? What approaches for "best practice" and "quality assessment/assurance" exist in other fields, and which of them make sense for simulation technology? In specific:

* Should there be public benchmarks for each discipline? Should a certain set of benchmarks (accuracy, performance,...) be part of all code manuals?

* Should simulation software contain warning systems to prevent critical application?

* What does method and code documentation have to be like, so that the final user understands all of this and can use the software truthfully?

* Does the "Workflow" concept pose additional problems, due to its "plug and play" character?

* Is there a useful concept for built-in uncertainty quantification and visualization?

* Can a computer experiment replace a real-world experiment? Is simulation theory or model? How far may we claim to approximate "reality"?

...any any other topic that fits into this spectrum.

selected key papers, if appropriate, will be found during the seminar.