References and case studies
In this part of the web site, you find a number of detailed examples of my work. If you like more detailed information, don't hesitate to ask me!
A very interesting and beneficial kind of application is the modeling and simulation of machines. Using simulation, it is possible to optimize machine parameters or to try out new configurations, without having to reconfigure or build the physical machine. An example for this is the simulation of an industrial food cutting machine. In another project, the geometric errors of a high-precision CNC machine were modelled mathematically and simulated, in order to derive these errors backwards from the errors of the products. Ultimately, this permitted to automatically correct those errors by adapting the numerical control parameters of the tool paths.
While in the projects just mentioned, an efficient implementation of the algorithmic parts is only one of several aspects determining the success of a solution, in other projects speeding up compute-intense software is at the heart of the matter, such as in optimizing a application for real-time signal processing using vectorization and improving data locality, or the parallelization of a fluid simulation code employing OpenMP.
Another group of case studies has its roots in the work I did for the @neurIST European project. @neurIST targetted the improvement of clinical management of brain aneurysms. An aneurysm — a local dilation of an artery — is utterly life-threatening if it ruptures, but which ones will rupture? This is an open question of utmost clinical relevance.
One path towards an answer to this question is the quest for physical characterisations of aneurysms suitable for predicting rupture risk. I was involved in the specification, design and implementation of a highly automated toolchain for computing blood flow, aneurysm wall stresses and shape descriptions, starting from medical images. I compiled and edited the projects final detailed modeling decisions and specifications, and set up a system for managing the simulation data to handle both provenance information (where did data come from?) and future variations (what else can we do with the data?).