Schedule

Schedule 2008-2009

Schedule 2009-2010

1st semester (Sept 2009-Jan 2010)

October, 8

15:45-16:45, Room H-327, Citadel Building

Speakers:

1. Philip Hölzenspies on Towards Platforms - How to run applications that do not yet exist

In Embedded Systems, power is a very scarce resource. Both the diversity of applications and the demand for integration into single appliances increase, despite of power problems. Power consumption can be reduced by specializing chips, whereas integration increases when generalizing them. This is why Multi-Processor System-on-Chips (MPSoCs) have become popular; flexibility and integration by offering several processors, all with their own specialization, in a single chip. We like to look at MPSoCs as platforms, but - especially when real-time constraints are involved - the assignment of tasks to processors is a design-time process. This means that two applications that require the same resource can not be run simultaneously and that when a resource becomes faulty, a set of applications can no longer run. This talk is about a way to loosen this constraint, by means of a heuristic approach to run-time resource assignment.

2. Denis Miretskiy

cancelled

Poster

November, 12

15:45-16:45, Room 2126, Zilverling Building

Speakers:

1. Didit Adytia on Variational Boussinesq Model for Simulation of Tsunamis and Coastal Waves

We describe the basic ideas of a so-called Variational Boussinesq Model (VBM) which is based on the Hamiltonian structure of gravity surface waves. By using a rather simple approach to prescribe the profile of the vertical fluid potential in the expression for the kinetic energy, we obtain a set of dynamic equations extended with one additional elliptic equation for the amplitude of the vertical profile. All expressions in the energy contain at most first order derivatives, which makes a numerical implementation with finite elements relatively easy. The applicability of the code is illustrated for two different applications. One application deals with tsunami simulation with study case a historical tsunami on July 2006 in Pangandaran, Indonesia. Another application deals with simulation of coastal waves entering the small harbour of Cilacap on the south-coast of Java, Indonesia. The approach is to investigate a resonance problem in the harbour.

2. Matthijs Bomhoff on Space-efficient Recognition of Perfect Elimination Bipartite Graphs

When applying Gaussian elimination to a sparse matrix, it is desirable to avoid turning zeros into non-zeros to preserve the sparsity. The class of perfect elimination bipartite graphs is closely related to square matrices that Gaussian elimination can be applied to without turning any zero into a non-zero in the process. Existing literature on the recognition of this class of graphs and matrices (and finding a suitable set of corresponding pivots) mainly focusses on the time complexity of this problem. However, when viewed from a practical perspective, the space complexity also deserves our attention: it may not be worthwhile to look for a suitable set of pivots for a sparse matrix if this requires a `dense' amount of space. This talk will focus on a new, more space-efficient algorithm for this problem.

Poster

December, 10

15:45-16:45, Room H-327, Citadel Building

Cancelled

January, 14

15:45-16:45, Room H-327, Citadel Building

Speakers:

1. Sid Visser on Modeling the Neocortex with Meso-scale Models and Population Models

Models of the brain yield new physiological insights as well as a better understanding of the brain and related disorders, e.g. epilepsy and Parkinson’s disease. A large model, having a detailed description of all neurons, is developed and numerically analyzed. Typical types of behavior, like oscillations and travelling waves, are observed. A different model, consisting of two delayed differential equations with two different delays, is derived that qualitatively captures the behavior of the detailed model. This model is then analyzed by construction of a bifurcation diagram that provides new insights in the dependence of both models on certain parameters.

2. Bob Peeters on A Conservative Numerical Scheme for Hydrostatic Flow in Isentropic Coordinates

Within this project we try to construct new numerical schemes that aim to improve current ensemble forecasting models for climate. Most contemporary climate models conserve energy and potential vor- ticity in the case of no forcing and dissipation. What’s missing in most cases, however, is a consistent discretization of the phase space structure. Constructing conservative Eulerian schemes is highly non- trivial. We take a different route and start with the equations for a hydrostatic atmosphere in isentropic coordinates. In absence of forcing and friction, entropy is conserved on infinitesimal fluid parcels. De- composing the atmosphere into a finite number of air particles is the key ingredient for a conservative numerical scheme.

Poster

2nd semester (Feb 2010-July 2010)

February, 11

15:45-16:45, Room H-327, Citadel Building

Speakers:

1. Peter Vanberkel on Modelling Hospital Patient Mix Decisions

Dutch hospitals are becoming more focused in the types of treatments they want to provide to patients. To achieve these targets for each treatment type, hospitals can choose to attract only certain patient classes to the hospital. The actual treatment types required by a patient class are subject to randomness. Furthermore, the number of patients in a patient class is unknown and evolves over time. In this presentation we discuss models to determine, based on given targets, which patient classes to attract to the hospital and when to do so.

2. Wenny Kristina on Wave Reflection over Slowly Varying Bathymetry modelled by Effective Boundary Conditions

In tsunami simulations, the wave height near the shore is the most important aspect scientists would like to calculate accurately. Unfortunately, the present-day simulation tools still cannot calculate the waveheight near the shore accurately enough. One source of inaccuracy is the interaction of the incoming waves with the reflected waves from the coast. Besides, computing the details of run-up and run-down of waves on the shore is computationally very demanding and expensive since closer to the shore a finer computational resolution will be needed. Moreover, the modelling of the physical processes is bound to be rather rudimentary because many aspects of tsunami propagation, e.g. nonlinearity, dispersion, friction, etc., have to be considered. In this talk, an analytical expression for the reflection of a wave that travels over slowly varying bathymetry is derived by using the Wentzel-Kramer-Brillouin (WKB) approximation of the linear Shallow Water Equations. This solution is then used to model the effective boundary condition to be imposed in a zone before the shoreline.

Poster

March, 11

15:45-16:45, Room H-327, Citadel Building

Speaker:

1. Hanumant Singh Shekhawat on Computation of Chopped System Norm

If system G represents the class of signals we are aiming for, then L²-norm of the optimal signal re-constructors over all samplers and holds require the norm computation of the system G in the frequency range [ω,∞) instead of usual (−∞,∞). The computation becomes more complicated if system G is allowed to have the poles not only in the left half complex plane but also on the imaginary axis of the complex plane. Hence a Lyapunov type answer is not applicable in this case. This talk is about a method for efficient computation of L²-norm of the system given in state space form with poles in the left half plane as well as on the imaginary axis. The relationship with L²-norm of the stable system and H²-norm of the system will also be discussed.

Poster

April, 8

15:45-16:45, Room H-327, Citadel Building

Speakers:

1. Daniël Reijsbergen on Rare Event Simulation for Model Checking Highly Dependable Systems

We are interested in estimating system failure probabilities in highly dependable systems, such as a telecommunications network or a nuclear power plant. Probabilities of interest could be system failure before some time bound or the time fraction that the system is down in equilibrium. Often, the models have state spaces that are too big to allow for iterative methods (e.g. Gauss-Seidel). Stochastic simulation is then typically used as the alternative. Obviously, in a highly dependable system, system failure is a rare event, so we need to apply efficient simulation techniques. We use Importance Sampling, i.e. we simulate under a new distribution which oversamples occurrence of the rare event. In this talk we explain why existing techniques fail in some situations and propose a new distribution for systems with high component repair rates.

2. Julia Mikhal on Prediction of Shear Stress in Cerebral Aneurysms

The prediction of blood flow inside aneurysms that can develop in a human brain is a field of intense research. It is important in relation to the growing medical need for an effective planning and execution of surgical intervention. The shear stress that occurs at the vessel walls of cerebral aneurysms is a key component for the prediction of the risk of their long-term rupture. The computational model for the simulation of blood flow through vessels in the human brain is formulated in terms of the incompressible Navier-Stokes equations in three-dimensional domains. We apply the immersed boundary method based on volume penalization to represent the complex shaped solid vessel walls of the aneurysm. In this presentation we focus on the flow near the boundaries, where the shear stress distribution appears as a reliable characteristic of the flow.

Poster

May, 6

15:45-16:45, Room H-327, Citadel Building

Cancelled

June, 10

15:45-16:45, Room H-327, Citadel Building

Speaker:

1. Lilya Ghazaryan on Simulation of Impaction Filtration of Aerosol Droplets in Porous Media

Driven by various applications, several approaches have been developed to determine the filtration efficiency of particulate filters. Our aim is to simulate the particle filtration efficiency of several basic structured porous media under a number of flow conditions and to quantify the role of the particle’s inertia on filtration. This is a steppingstone problem towards the analysis of realistic systems of porous filters.

Particle motion is governed by a variety of forces. Here, we focus on the dominant drag-force dynamics to quantify the filtration efficiency as a function of the droplet inertia, expressed in terms of its relaxation time. Filtration of a droplet is assumed to arise whenever this droplet hits the surface of the porous medium. The immersed boundary method is applied to capture the detailed particle motion in the gas flow through a complex porous medium. The flow field is calculated by solving the incompressible Navier-Stokes equations. For the particle phase the Lagrangian approach is used, in which we track the trajectories of a large number of particles embedded in the flow. The porous medium is formed by a staggered arrangement of square rods in 3D in order to develop and validate our approach.