Tools for Multi-Physics Simulation

1
Tools for Multi-Physics Simulation

• Hans Petter Langtangen

Simula Research Laboratory Oslo,
Norway Department of Informatics, University of
Oslo
2
History of the group
Future ideas/visions
multi-physics
PDE components
Python
F77
C
Python
3
1990s Diffpack applications
I/O
Grid
FDM
Field
FEM
Matrix
Vector
Axb
4
A Typical Diffpack PDE Solver
Class P1 Grid grid Field p LinEqAdm
Axb integrands () K()
Evaluate integrands in weak form Evaluate
variable coefficient
5
Principal Diffpack collaborators
Key names Are Magnus Bruaset, Xing Cai, Hans
Petter Langtangen, Aslak Tveito,
6
Diffpack distribution

• 1995 1997 Open source versions
• 1997-2003 Numerical Objects
• 2003-? inuTech
• Free test version (size limit)
• Simula collaborators have free access
• Customers include Cornell, Stanford, LLNL, Intel,
  NASA, Shell, IFP, DaimlerCrysler, Mitsubishi, ...
• gt200 customers in gt30 countries

7
Long-term basic research is needed to develop
core technology
1990 -
8
Some Diffpack features

• Linear solvers and preconditioners
• Grids and scalar/vector fields
• Biased towards FEM, support for FDM
• Mixed FEM, block systems
• Systems of PDEs
• Stochastic PDEs
• Multilevel solution
• Adaptive mesh (h-adaptivity)
• Parallel computing
• Problem solving environment

9
We have always been interested in core
technologies, but these also need application
outlets
PDEs
FEM
OOP
Python
C
Axb
10
Some Diffpack applications

• Heat and phase transfer
• Viscous laminar and turbulent flow
• Thermo-elastic-plastic deformation
• Electromagnetics
• Chemical reactors
• Design of quantum computers
• Stochastic porous media flow
• Electrical activity in the heart
• Tsunami simulation
• ...

11
Simula is a new government lab for advanced ICT
research
High quality research Educate graduate
university students Prepare for research- based
business
Startup Jan 1, 2001
12
Research groups were selected by competition
11 Applicants
Software Engineering
Scientific Computing
Networks and Distributed Systems
13
Simula emphasizes large projects
Inverse problems
Computational geosciences
Software for PDEs
Cardiac Computing
14
Computing the electrical activity in the heart
15
This is a computationally intensive problem
16
Computational geosciences

• Goal find more oil!
• Compute geological evolution
• Multi-resolution visualization
• PDEs for deposition/erosion
• Flow, heat, deformation
• Software integration
• Advanced tools for tracking geological events

17
(No Transcript)
18
The Storegga Slide (8150 yrsBP)
Headwall 300 km Run-out ? 800 km Volume ?
5.600 km3 Area ? 34.000 km2
19
(No Transcript)
20
The Mjølnir asteroide impact
21
Barents Sea, 142 mill. years ago
22
The Mjølnir event was 1000 times stronger than
the Dec 26, 2004 event!
23
The Mjølnir tsunami consisted of a series of
individual waves
24
The evolution of the front of the tsunami
25
The tragic Dec 26, 2004 tsunami

• How was the earthquake-induced bottom lift?
• Tsunami simulations yield important constraints

26
Welding
Courtesy of University of Ålborg and Odense Steel
Shipyard
27
The Silent Wings soaring simulator uses advanced
terrain visualization
28
(No Transcript)
29
We have made international impact on scientific
software
The scientific computing group is concerned with
applying modern software engineering practices to
scientific software. This group has achieved
international distinction by calling attention to
the feasibility and desirability of subjecting
scientific software a niche area to this
discipline. DIFFPACK (now commercialized) has
wielded an enormous influence on other scientific
software around the world. The publication
record of this group is impressive, and includes
internationally published books that go a long
way to defining scientific software
engineering, - Evaluation of ICT groups at
Norwegian Universities (2002)
30
Higher-level tools encourage flexible software
integration
SWIG
3rd party
Python
Diffpack
F2PY
Simula code
31
SWIG
GiNaC
Python
C/F77
SWIG/F2PY
FAMMS