Numerics and Simulation
Overview
The current era of humankind is haunted by challenges such as climate change, increasing energy demands, and the ever growing desire for mobility. This results in the demand for more efficient lightweight and durable structures while featuring low Noise-Vibration-Harshness (NVH) for applications including wind turbines, electric cars or airplanes. It is crucial to reduce their vibrations to a minimum to obtain the desired acoustic behavior, efficiency and high endurance strengths.
As the development cycles become shorter and the structures more complex, numerical models are used to predict the dynamic behavior of the structures. For example one can observe more and more organic looking structures in architecture and aerospace applications, that are highly optimized and unimaginable without modern numerical methods. These methods are only usefull, if they are able to describe real-world problems as accurate as the problems require. This works well for simple materials and objects, but as soon as contacts, different materials and many parts are involved the required accuracy pushes up computational costs. Although the computer power has increased drastically over the last decades, it can take several days to compute results from numerical models.
One main research goal at the chair of Applied Mechanics is to reduce computation time for such simulations. We develop algorithms that help to exploit the parallel structure of today's computer architectures, i.e., algorithms that can be run on several cores or even High-Performance-Computers. Additionally, we develop model reduction methods that can reduce the computational effort to obtain simulation results by approximating large models with the help of mathematical methods.
Besides several scientific publications in literature, we also publish the software that we have developed to achieve the above mentioned goals. Currently, there are two main software packages available:
- AMfe: an open source finite element code that is able to simulate the dynamics of geometric nonlinear structures and also contains nonlinear model reduction methods
- pyFETI: an open source package that implements Finite Element Tearing and Interconnecting methods to parallelize the computation of the dynamics of structures with the help of substructuring techniques