|
|
|
|
NTNU is responsible for one out of three national supercomputing projects in Norway. The center for computer services at NTNU (ITEA) operates a CRAY T3E scalable parallel computer and a CRAY J90 vector-parallel computer. These computers will be used in the CSE project. Five persons are associated with the supercomputing project. Two of these have PhD in scientific computing, and they will be among the key personell in the CSE project.
There exist several courses at Department for Mathematical Sciences (MS) which are related to scientific computing. At the undergraduate level there are basic introductory courses in numerical analysis and mathematical modelling and analysis. Four PhD courses are given, three of them within numerical solution of differential equations and numerical linear algebra. Of particular interest for the CSE project is perhaps the course Introduction to the Use of Supercomputers, that has been given for some years, but which is subjected to continuing change to account for the rapid development in hardware and algorithms. The course focusses on numerical algorithms for parallel- and vector- based computers.
The numerics group at MS has a long tradition of research within numerical methods for ordinary differential equations (ODE) on an international level, and maintains contacts with researchers from several countries throughout the world. There is an ongoing NFR sponsored project, SYNODE, where the main objective is to study numerical solution techniques for differential equations with known underlying structures to be inherited by the numerical approximation.
Department of of Computer and Information Science, NTNU has been an active partner in the following activities related to object oriented program design:
Work is going on to establish a well equiped VR laboratory within NTNU. Several departments have already announced their interest. This laboratory and the activities around it are considered important for introducing VR in several contexts within NTNU.
The ultimate goal for this project is to make industry capable of designing fluid-exposed structures more accurately and efficiently by using high performance modelling and simulation techniques. In particular, the project will emphasize the development of a coupling module (FSI-Coupler) for integration of Computational Fluid Dynamics (CFD) codes with Computational Structural Design (CSD) codes. The FSI-Coupler may be utilised within the CSE-SUP for coupling of the developed CFD-code with existing CSD-codes (as USFOS, RIFLEX etc.) to enable coupled fluid-structure simulations of real world problems. Furthermore a first step in the direction of implementing algorithms for adaptive mesh refinements for real world Navier-Stokes problems is taken. An object-oriented program module denoted PAM: Parallel Adaptive Module is under development by SINTEF Applied Mathematics. This program module may be used as a starting point for the work on error estimation and adaptivity within the present CSE-SUP.
SINTEF Applied Mathematics has also a Strategic Institute Program (SIP) within scientific computing called Numerical Computations in Applied Mathematics (NAM). This is a joint SUP/SIP and is described below.
SINTEF Telecom and Informatics has in cooperation with Department of of Computer and Information Science, NTNU been involved in the following major research program related to object oriented program design:
In the past, research and PhD projects related to viscous flow problems conducted at Department of Marine Hydrodynamics includes; vortex tracking methods and separated flow around marine structures, study of sloshing in tanks and development of a finite element Navier-Stokes code for two-dimensional separated flow around marine structures. At the moment, topics for PhD student projects in progress involving viscous computation are; flow around two-dimensional moving ship sections with free surface, implementing free surface capability in existing Navier-Stokes code, flow through three-dimensional net structures and sloshing in tanks with nonlinear free surface. One PhD student has established equipment (PIV) for detailed measurements of fluid particle flow properties in breaking waves. Extension of the method to cases with fixed or moving two-dimensional bodies interacting with waves may give experimental data useful for validation and verification of two-dimensional CFD calculations.
Together with Viewtech A/S and Det Norske Veritas, the Department of Marine Hydrodynamics has participated in planning a project with the aim to utilise modern software and hardware visualisation technology to develop a "hydrodynamic workbench". A minor activity at the Department is related to three-dimensional grid generation dedicated to ship hull forms, motivated mainly because of the difficulties in utilising existing CAD tools directly in the preprocessing phase of CFD calculations.
At the Department of Structural Engineering fluid flow about bluff bodies has during the last 20 years been a major research topic. The main thrust has been on experimental work on pipelines, tethers, risers and on sediment transport, but during the last 12 years also a concentrated effort on the FEM solution of Navier-Stokes equations (with turbulence terms) has been carried out using a workforce of typically 2 to 3 persons full time. Several types of turbulence modelling has been attempted, often good results were obtained with simple models. Altogether in excess of 10 doctoral theses have been produced on some aspect of bluff body flow.
The Department of Marine Hydrodynamics is involved in the strategic institute program (SIP) SKIPRO-2001 started in 1997 at MARINTEK. This program focusses on resistance and propulsion and a minor part is preliminary tests with a Navier-Stokes code. One study may be calculation of flow around a submarine hull with sail, wings and control surfaces and comparisons with the substantial amount of experimental data available in the literature (SUBOFF). Initial tests with application of Navier-Stokes codes are also conducted in other groups.
Deep water analysis tools (Deeper) is a joint industry project with 40% funding from NTNU. It is a cooperative effort between DNV, MARINTEK, and NTNU with project manager at DNV. One of the projects herein is 'Deep water marine risers' which is of special interest in the present context. It has funding for one doctoral student who will have Geir Moe and Odd Faltinsen as thesis advisors. There is also some funding for comparative studies between results from computer codes developed by respectively DNV, Dr.ing Torbjørn Utnes (NTNU Faculty of Civil and Environmental Engineering), and Dr.ing Kjell Herfjord (Norsk Hydro). Finally, there will be a project in which information on flow-induced vibrations will be evaluated to produce an empirical model. (Professors Larsen and Moe.)
In Oslo a consortium was formed under the title: Numerical Computations in Applied Mathematics (NAM). Members of the consortium are: the SINTEF Applied Mathematics, The Norwegian Computing Center, and the departments of Informatics and Mathematics at the University of Oslo. Both the CSE and NAM initiatives strive towards an increased level of competence in generic disciplines. An important element in the NAM project has been the development and use of the object oriented software package DIFFPACK. The main difference between the NAM project and the CSE proposal is that the latter focusses on a particular area of application and incorporates a close collaboration with researchers from engineering departments. Thus, the CSE initiative is complementary to NAM, but will at the same time exploit and develop further the available technology from the NAM project.
The High Performance Computing Centre at the University of Bergen has recently been awarded a SUP called Analysis and Implementation of Parallel Algorithms for Scientific Computing. Its main focus is on parallel algorithms, in particular domain decomposition algorithms and implementation of parallel, portable software. In contrast to the Bergen initiative, the present proposal will focus on algorithms for time integration and adaptive model refinement tailored for parallel machine architectures. Furthermore, we will develop next generation scientific visualization techniques suitable for handling data from large scale simulations.
 |
|
|
|
Editor in Charge: Manager of CSE, Professor Syvert P. Nørsett
Snorre Boasson Last modified: Thu Apr 10 18:51:57 MEST 2003 |