Proposal for a Strategic University Program
in Computational Science and Engineering at NTNU
3 Program elements
3.1 General philosophy
The project should be seen as a model project for Computational
Science and Engineering at NTNU. The core of the activity should be
within mathematical modelling, numerical analysis and computer
science, but the efforts should focus on one particular area of
application. The choice of application area is Marine Fluid
Mechanics. The criteria for choosing this area can be summarized
as follows.
- There are problems and challenges of national importance.
- It is appropriate for demonstrating the benefits of modern
methods for CSE.
- NTNU possesses expertise on a high international level in marine
fluid mechanics, and the representatives of the field at NTNU
have used scientific computing tools in their work and
acknowledge the need for a wider cooperation in the area of
CSE.The applied departments involved in the CSE-project will
be: Department of Marine Hydrodynamics, Faculty of Marine
Technology and Department of Structural Engineering, Faculty of
Civil and Environmental Engineering.
- The Department of Mathematical Sciences, Faculty of Physics,
Informatics and Mathematics possesses the necessary competence
in the areas of mathematical modelling and numerical analysis
that is required for solving these problems.
- The complexity of computer codes and output data from
simulations brings up the importance of using modern methods
for software design and modeling and for visualization. The
expertise in these areas is provided by the Department of
Computer and Information Science, Faculty of Physics,
Informatics and Mathematics.
Most of the activity should be conducted through PhD students that
will be supervised by academic staff at all 4 departments involved in
the project. Each PhD student will be assigned a topic and based at
one department. The flow of information between the research groups
at the various departments will be strongly emphasized. At least two
measures will be taken in order to ensure that a genuine collaboration
will take place
- Each PhD student will have two supervisors, when possible these
supervisors will belong to different research groups/
departments
- The course requirements for the PhD students will ensure that
students take courses at departments involved, but other than
where he/she is based. New courses will be tailored for the
project and required taken by most of the PhD students in the
project
- Seminars will be organized with participance from all involved
parties
3.2 Grand challenge problems
Computational tools have recently demonstrated the capability of
producing promising results of some complex turbulent flows of
practical interest in marine fluid mechanics. Two major requirements
for practical engineering applications are computational speed and
accuracy. One of the most challenging problems today in Computational
Fluid Dynamics (CFD) is the prediction of the forces and moments
acting on a realistic body at high Reynolds numbers (on the order of
10^6 to 10^9) with computational accuracy within few percent of the
measured data. To meet the stringent requirements of speed and
accuracy, sophisticated numerical techniques must be implemented and
algorithms that utilise modern parallel architectures must be used in
the simulations of such "grand challenge" problems. Efforts to expand
computational performance beyond conventional regimes and the need for
studying truly unsteady, high Reynolds number flows are increasing.
At NTNU we have the main necessary tools available for contributing to
the research activity, i.e. large parallel supercomputer CRAY T3E and
experimental facilities. The success of the CFD simulation however,
relies to a larger extent than before, on effective user interfaces
like preprocessor, grid generation tools and evaluation tools to
present the time-varying simulation results.
There is a strong tendency that commercial available Navier-Stokes
codes are tuned and marketed to the low capacity workstations or even
for PCs. Very few codes are optimised and available on modern
parallel hardware-platforms. This is a pure license and economic
optimisation from the software-vendors. It is therefore the
responsibility of the university research community to take a lead and
develop and demonstrate new possibilities in addition to transfer new
knowledge to the industrial engineering community.
Experience shows that using CFD and experimental facilities together
enhances the value of data derived from each source alone. CFD
techniques today cannot in general compute accurately enough the
highly complex, separated flow fields of a complete ship hull with
appendages in steady conditions, let alone unsteady ones. At the
moment we lack physical models on which to base the computations of
time-dependent, turbulent, separated flow fields. In developing such
models, flow field data from sufficiently realistic flows is a
necessity. One way to obtain such data is to conduct experiments.
3.3 Technical content
The technical content of the proposed Strategic University Program in
Computational Science and Engineering (CSE-SUP) may be characterized
by the following four key words:
- Parallel methods
- Object oriented philosophy
- Communication of results
- Marine fluid mechanics
Viewing the project from the broadest possible perspective, one may
divide the activities into a generic part and an application part. In
the generic part, development and analysis of parallel algorithms for
semilinear parabolic partial differential equations play a central
role. More specifically, the incompressible unsteady Navier-Stokes
equations will be subjected to a close investigation. Furthermore,
aspects of modelling and design of object-oriented software for
numerical computation as well as the development and implementation of
tools for communicating the results from the numerical computations,
will be treated in a generic setting, but again, never losing sight of
the proposed applications to marine fluid mechanics. The success of
the project rests heavily on the ability to transfer the developed
tools from the generic activities to the chosen applications, and as a
consequence of this, some of the activities based at the departments
of Marine Hydrodynamics and Structural Engineering will begin later
than those at the departments of Mathematical Sciences and Computer
and Information Science.
The PhD proposals at each of the 4
departments will be described briefly below. The project will comprise
9 PhD students distributed as follows:
| Department of Mathematical Sciences |
3 students |
| Department of Computer and Information Science |
2 students |
| Department of Marine Hydrodynamics |
2 students |
| Department of Structural Engineering |
2 students |
3.3.1 Activity based at the Department of Mathematical Sciences
Three generic concepts of great significance to computation within the chosen
application area are:
- Parallel methods
- Time integration
- Adaptivity
The topics of the PhD studies presented below are all to some extent
based on these three concepts. In particular the implementation which
is to be carried out in all three proposals is directed towards
parallel machine architectures. There are overlapping elements
involved in the three proposals, but they are emphasizing different
computational aspects. It is assumed that the three PhD students may
work independently over time periods of some months, but continuing
exchange of ideas and results between the students and their
supervisors is considered a vital part of the work.
- Parallel Algorithms for Time Dependent Partial Differential
Equations
Object-oriented modelling methods have proven their potential
in a diversity of applications. They are by themselves powerful
tools for systems work (e.g. virtual reality systems with
dynamically changing intelligent objects like living
creatures). Successful applications for reuse and engineering
for industrial purposes are demonstrated in projects like
REBOOT, PROTEUS and RENAISSANCE. The quality gains will be
metricated in SPIQ. Object-oriented reuse for distributed and
partly parallel systems is being worked on i CAGIS.
The proposed PhD projects aimes at extending the areas of
application of object-oriented modelling in CSE.
- Time Integration Schemes for ODE's Originating from
Semidiscretized Nonlinear PDE's
The semidiscretisation of the Navier-Stokes equations leads to
a huge system of nonlinear ODE's whose solution must be
approximated by means of a time integration scheme. It is
necessary for such schemes to possess the ability to take
advantage of parallel machine architectures. The objective of
this doctoral study is to design, analyse and implement such
time integration formulas, by taking into account the posed
requirements to the quality of the approximation, in particular
accuracy, stability and conservation properties. The schemes
should be validated by considering their suitability for
mathematical models arising from fluid-structure interaction,
like vortex induced vibrations of slender structures of risers
and pipelines.
- Parallel Adaptive Finite Element Methods for Time Dependent
Nonlinear Problems
The computation of numerical solution to time dependent
(nonlinear) problems as the Navier-Stokes equations is among
the most challenging problems within the field of scientific
computing. The resolution of the finite element model is
therefore usually limited by the capacity of the computer
hardware at disposal. Thus it is important to obtain
"close-to-optimal" distribution of elements throughout the
domain. The objective for this doctoral study is to address
this issue by developing error estimates and rezoning
techniques for adaptive finite element computations of time
dependent nonlinear problems. The developed techniques shall be
tested on fluid structure interaction phenomena as vortex
induced vibrations of slender structures as risers and
pipelines.
3.3.2 Activity based at the Department of Computer and Information
Science
Object-oriented modelling has been successfully applied for reuse and
reengineering in many industrial projects (REBOOT, PROTEUS,
RENAISSANCE). The quality gains will be metricated (SPIQ), and
object-oriented reuse is now being applied for distributed and partly
parallel systems (CAGIS).
- Object-oriented Modelling of Numerical Calculations
Object-oriented applications for e.g. discrete simulation,
computer graphics and distributed systems are well-known
success stories. The potential for reuse and engineering of
customized computer models is generally large, and may give
substantial quality and productivity benefits. Work has already
been done to organize numerical libraries in an object-oriented
framework. However, this is not yet well connected to
systematic reuse, use of graphics (VR), and formal product
models (e.g. STEP). In this PhD-study the focus will be on
linking object-oriented theory and methods, with emphasis on
abstraction and reuse, with numerical techniques and volume
visualization by VR for parallel systems.
- 3D Visualisation of Turbulent Flow
Volume visualisation has with considerable success been
utilised for 3D ultra sound images for medical
applications. The experiences will be transferred to this new
application. Virtual Reality (VR) techniques has also been used
to give users the opportunity to immerse into a synthetically
created "world". This is also techniques with rich
possibilities in fluid flow studies.
3.3.3 Activity based at the Department of Marine Hydrodynamics
In this research program we will focus on the general problem of flow
around a typical ship hull and use it to gain experience and increase
our competence in viscous flow hydrodynamic calculations. The
solution of the nonsymmetric problem with oblique flow around a ship
hull gives a solid foundation and experience for future extensions to
more complex hydrodynamic problems. The ship hull flow problem is
however a tremendous task where many research groups world-wide
already have spent several man-years to find solutions by calculation.
Reasons for choosing this application is first of all the fact that
tank tests to predict the maneuvering properties requires a lot of
time and efforts and some maneuvering motions cannot be well realised
in the experimental facilities. Secondly, to improve safety in
navigation and in order to cope with IMO's (International Maritime
Organization) regulation on ship maneuverability, a sophisticated
calculation method is highly requested. CFD can be a useful tool to
predict the effect of principal dimensions of a ship hull in an early
design phase and thereby incorporate the IMO maneuverability criteria
at initial stage. With concentration of both ph.d. students on the
same topic and through incorporation of experience from the other
research groups participating under the proposed computational science
and engineering program, we should be able to demonstrate results not
possible if working alone.
- Viscous Flow Around Ship Hull
The flow field around a ship moving at an angle of attack is
complicated even at low speed where no waves are
present. Crossflow separation forms the wake at low angles of
attack. At increasing angles of attack where the crossflow
separation eventually results in vortex roll-up which then
departs from the body. The location of separation and the
separation topology are highly dependent on the state of the
local boundary layer, that is whether it is laminar,
transitional or turbulent. The work will focus on modelling and
calculation of typical flow situations with a realistic ship
hull form moving in an infinite fluid domain. The distribution
of grid points and flexibility in redistribution (adaptivity)
will be investigated.
- Maneuvering of Ships in Restricted Waters
It is well known that maneuvering performance of ships may
differ to a great extent by even small differences in the hull
form, in particular in the aft part of the ship
hull. Theoretical prediction methods of hydrodynamic forces on
a maneuvering ship should be so improved that difference in
hydrodynamic characteristics caused by (small) change in hull
form may be appropriately evaluated. The need for knowledge
about the maneuvering performance of a ship is greatest in
restricted waters. So far much information needed is based on
empirical approaches, experience from full scale trials and
scaled model test data. We will extend available software or
develop new tools able to handle the effect of restricted
water, both finite width and finite depth, on the flow around a
ship hull.
3.3.4 Activity based at the Department of Structural Engineering
Turbulence is one of the main topics in applied fluid mechanics. The
focus of the proposed PhD studies below is on the modelling of
turbulent fluid flow, applied to relevant geophysical and engineering
studies. Although there are overlapping elements in the two
proposals, they are emphasizing somewhat different aspects. It is
assumed that the two PhD students may work independently for part of
the time, but a practical coordination and exchange of ideas is also
considered a vital part of the work.
- Turbulent Flows over Bottom Topography
Variable bottom topography, subsea structures, etc. tend to
organize the vorticity of the incoming boundary layer such that
intense fluctuations foreign may occur. The objective of this
doctoral study is to model coherent turbulent flows over such
bottom typography by use of large eddy simulations
(LES). Realistic simulation of the incoming turbulence and much
validation will be main themes.
- Large Eddy Simulation of Flows about Flexible Cylinders
Long and slender engineering structures are flexible and
therefore susceptible to vortex-induced vibrations. The
objective of this doctoral study will be to study and model the
turbulent flow about such structures by use of LES in a FEM
context.
