Session Details
MS16-2: Reduced Order Models for Multiscale and Multiphysics Problems (Ganzes Minisymposium anzeigen)
Wednesday, 11. October 2017; 16:00 - 18:00 Uhr in Raum 7.02
Sitzungsleitung: Felix Fritzen
16:00
Reduced Order Modelling for the Simulation of Quenches in Superconducting Magnets (Keynote)
Sebastian Schöps (Technical University of Darmstadt), Idoia Cortes Garcia (Technical University of Darmstadt), Michał Maciejewski (CERN), Bernhard Auchmann (Paul Scherrer Institut)
Kurzfassung:
This contributions discusses the simulation of magnetothermal effects in superconducting magnets as used in particle accelerators. An iterative coupling scheme using reduced order models between a magnetothermal partial differential model and an electrical lumped-element circuit is demonstrated. The multiphysics, multirate and multiscale problem requires a consistent formulation and framework to tackle the challenging transient effects occurring at both system and device level.
Reduced Order Modelling for the Simulation of Quenches in Superconducting Magnets (Keynote)
Sebastian Schöps (Technical University of Darmstadt), Idoia Cortes Garcia (Technical University of Darmstadt), Michał Maciejewski (CERN), Bernhard Auchmann (Paul Scherrer Institut)
Kurzfassung:
This contributions discusses the simulation of magnetothermal effects in superconducting magnets as used in particle accelerators. An iterative coupling scheme using reduced order models between a magnetothermal partial differential model and an electrical lumped-element circuit is demonstrated. The multiphysics, multirate and multiscale problem requires a consistent formulation and framework to tackle the challenging transient effects occurring at both system and device level.
16:40
Computational homogenization and model order reduction of pressure diffusion in fractured rock
Ralf Jänicke (Chalmers University of Technology), Fredrik Larsson (Chalmers University of Technology), Kenneth Runesson (Chalmers University of Technology)
Kurzfassung:
Pressure diffusion in fracture networks is the dominating physical process that causes attenuation of elastic waves traveling trough fluid-saturated rock. We simulate this process in a multi-scale approach where pressure diffusion occurs on the sub-scale and the related seismic attenuation is observed on the macro-scale. We introduce a computational homogenization scheme and develop a NTFA-type model order reduction technique which allows to derive the macroscopic properties of fractured rock.
Computational homogenization and model order reduction of pressure diffusion in fractured rock
Ralf Jänicke (Chalmers University of Technology), Fredrik Larsson (Chalmers University of Technology), Kenneth Runesson (Chalmers University of Technology)
Kurzfassung:
Pressure diffusion in fracture networks is the dominating physical process that causes attenuation of elastic waves traveling trough fluid-saturated rock. We simulate this process in a multi-scale approach where pressure diffusion occurs on the sub-scale and the related seismic attenuation is observed on the macro-scale. We introduce a computational homogenization scheme and develop a NTFA-type model order reduction technique which allows to derive the macroscopic properties of fractured rock.
17:00
Numerical Model Reduction in Computational Homogenization of Transient Heat Flow
Fredrik Ekre (Chalmers University of Technology), Fredrik Larsson (Chalmers University of Technology), Kenneth Runesson (Chalmers University of Technology)
Kurzfassung:
We present a two-scale finite element (FE$^2$) formulation for transient linear heat flow. For the sub-scale problem, we use spectral decomposition in order to to establish a reduced basis. We discuss a few methods to estimate the error introduced by the reduction, and in particular we aim for explicit bounds on the error in (i) energy norm and (ii) an arbitrary ``quantity of interest''. Numerical results confirm the validity of the computed error bounds.
Numerical Model Reduction in Computational Homogenization of Transient Heat Flow
Fredrik Ekre (Chalmers University of Technology), Fredrik Larsson (Chalmers University of Technology), Kenneth Runesson (Chalmers University of Technology)
Kurzfassung:
We present a two-scale finite element (FE$^2$) formulation for transient linear heat flow. For the sub-scale problem, we use spectral decomposition in order to to establish a reduced basis. We discuss a few methods to estimate the error introduced by the reduction, and in particular we aim for explicit bounds on the error in (i) energy norm and (ii) an arbitrary ``quantity of interest''. Numerical results confirm the validity of the computed error bounds.
17:20
Substituting FE analysis of cyclic processes by a space-time reduced order model
Mohammad Reza Hassani (University of Stuttgart), Felix Fritzen (University of Stuttgart)
Kurzfassung:
The huge computational costs make classical Finite Element (FE) simulations of nonlinear structural problems subjected to long-term or e.g. cyclic loading a challenging task. One approach to tackle this is through Model Order Reduction (MOR). Space-time MOR leads to a low-dimensional nonlinear system of equations which is solved in coarse time intervals, e.g. for each load cycle. Thus, remarkable computational saving in terms of CPU time and memory space are attained.
Substituting FE analysis of cyclic processes by a space-time reduced order model
Mohammad Reza Hassani (University of Stuttgart), Felix Fritzen (University of Stuttgart)
Kurzfassung:
The huge computational costs make classical Finite Element (FE) simulations of nonlinear structural problems subjected to long-term or e.g. cyclic loading a challenging task. One approach to tackle this is through Model Order Reduction (MOR). Space-time MOR leads to a low-dimensional nonlinear system of equations which is solved in coarse time intervals, e.g. for each load cycle. Thus, remarkable computational saving in terms of CPU time and memory space are attained.