Computation

A wide variety of calculation methods are an integral part of the research and teaching activities at the Chair of Casting Technology (LGT). In addition to classic simulation methods such as computational fluid dynamics (CFD) and the finite element method (FEM), statistical calculation methods from the field of machine learning (ML) are also applied and further developed at the chair. Classic FEM, CFD, or coupled simulations are mainly calculated at the chair using COMSOL Multiphysics, ANSYS, and ABAQUS. In addition, numerous CFD applications are calculated in OpenFOAM, an open-source software. Proprietary software for pre- and post-processing as well as for calculation is mainly developed in Matlab, Python, or C++. Statistical calculation methods from the field of ML, such as neural networks, are primarily implemented and used at the LGT in Python. A wide variety of Python libraries, such as Pytorch, Keras, deepxde, etc., are used for this purpose.

Depending on the specific requirements of the respective application, the calculations themselves are performed using the chair’s hardware resources (see below). As part of FAU, the LGT also has access to the high-performance computing (HPC) resources of the NHR (https://hpc.fau.de/), which are particularly necessary for highly parallelized calculations (hyperparameter studies, generation of AI datasets, etc.).