Scientific Computing (SC) enables the simulation of phenomena, processes and systems that cannot be studied by real experiments for technical, financial, safety or ethical reasons. SC also allows for automatic design and optimization, through inverse computations. Many disciplines in science and engineering have their own computational branches now. With continuing growth in speed, memory and cost-effectiveness of computers and similar improvements in numerical mathematics, the future benefits of SC are enormous.

Providing better simulation tools for a wide range of different applications.

Within CASA’s SC group, we propose, analyze, develop and implement new numerical mathematics methods, particularly structure-preserving discretization methods for partial differential equations and numerical linear algebra methods for linear systems of equations. We carry over these methods to Computational Science and Engineering (CSE), for application to particularly fluid-structure-interaction and energy-conversion problems.

Traditionally, CSE is model-driven; based on mathematical models of first principles (physical laws for instance). Because of the immense growth in the availability of data, CSE is becoming data-driven as well, with an important role in this for neural networks. Within CASA’s SC group, neural-network technology in CSE is also studied. One of our research goals is to further develop neural networks in the context of CSE, combining data- and model-driven approaches, hand-in-hand with the development of more theory for a rational and trustworthy use of data and neural networks within CSE.