The Virtual Microscope - Visualization and Inspection of the Geometry of Simulated Particle Mixtures

SWZ-funded research project

Partner: Prof.Dr. Michael Kolonko, TU Clausthal

Lots of materials such as concrete and tablets are made of particles and some properties of those materials are strongly determined by the geometric characteristics of the particle mixtures. For example, for concrete, the space filling rate of the dry mixture, i.e. the ratio of the container size to the volume of the particles contained, is decisive for the strength of the concrete after curing. In other applications, such as in the production of foams, the distribution and the shape of the spaces between the 'particles', which are cavities in this case, play a critical role for the properties of the material.

The right choice of the mixture composition or grain size distribution is therefore crucial in the development of particle-based materials with predetermined properties. Up to now, however, expensive laboratory experiments are still required, but there are successful approaches to simulate the geometry of the mixtures on the computer. In AG Kolonko, a program system "RASIM" is developed for several years that the random tight disposition (pack) simulates a mixture of spherical particles having a predetermined particle size distribution (PSD). At present, the main application area is the concrete research, where high space filling rate is pursued under given PSD. As a result, the simulation currently provides the achieved space filling rate as a percentage number and creates static images of simulated packs with standard software.

To use the simulation in a wider field of applications, such as the production of filters and membranes of particles, and also for the production of mixtures for foundry molds or 3D printing, more qualitative properties of the packings have to be determined, e.g. the local interaction of the particles. This is particularly necessary when non-spherical particles are simulated. Here one would like to recognize the relative position of the (differently shaped) particles to each other and to judge the mixture. This could also be used to study the effects of individual parameter settings of the simulation, which in turn correspond to real conditions such as pressure or duration of the mixing process.

In the case of real mixtures, it is possible, for example, under a microscope, e.g. the disposition and the shape of the interstices or possible (undesirable) sortings of the particles to detect. Within the scope of this project, a flexible visualization tool is to be developed that enables a qualitative analysis of a simulated mixture as a kind of virtual microscope. This would expand the application spectrum of existing simulation systems, and at the same time, this would be an important tool for the further development and improvement of the simulation.

Interacting in Photorealistic Augmented Reality (IPAR)

German Research Foundation (DFG)

Partner: Prof. Dr. Raimund Dachselt, TU Dresden


 Interactive Near-field Illumination for Photorealistic Augmented Reality on Mobile Devices
Kai Rohmer, Wolfgang Büschel, Raimund Dachselt and Thorsten Grosch
IEEE International Symposium on Mixed and Augmented Reality (ISMAR 2014)
Best Full Paper Award

Tiled Frustum Culling for Differential Rendering on Mobile Devices
Kai Rohmer and Thorsten Grosch
Conditionally accepted at IEEE International Symposium on Mixed and Augmented Reality (ISMAR 2015)

Interactive Out-of-Core Global Illumination

German Research Foundation (DFG)

Nowadays, global illumination simulations are possible with a high quality. Often, for large production scenes the available memory, especially on GPU, is not sufficient. Furthermore, large scenes are often illuminated with a single light bounce only due to time restrictions.

GPUs are highly parallel devices and well-suited for rendering approaches. Hence, there is a need for global illumination algorithms on GPU which can handle out-of-core scene sizes. This is difficult due to the global influence of light into the visible range. It is not possible to load only the visible part to get correct results. Additionally, the algorithms require a ray cast operation which badly scales on the GPU. Nowadays, the GPU ray-tracer just reached the performance of their CPU opponents, despite the theoretical higher power.

We are interested in a physically correct (unbiased) solution as well as the option to smoothly introduce a bias to get a high performance gain. We target an interactive GPU-based solution for static scenes.


 Distributed Out-of-Core Stochastic Progressive Photon Mapping
Tobias Günther and Thorsten Grosch
Computer Graphics Forum, Volume 33, Issue 6, September 2014

 Illumination-driven Mesh Reduction for Accelerating Light Transport SimulationsVideo]
Andreas Reich, Tobias Günther and Thorsten Grosch
Conditionally accepted at Eurographics Symposium on Rendering (EGSR 2015)