|
|
Development of numerical methods
Computational fluid dynamics (CFD)
Numrax's primary field of expertise is fluid dynamics, mainly the development of numerical methods for computational fluid dynamics (CFD) for a broad range of applications. These include incompressible, weakly compressible, subsonic, transonic and supersonic flows with or without extensions like chemical reactants (flames, detonations), suspensions
and aerosols or material discontinuities. Numrax will analyse a customers problem and develop specific solution methods for arbitrary physics involved.
Concerning the demand for computational resources, most applications in CFD are "insatiable". This means that the accuracy and eventually the value of the simulations is mostly restricted by the computational resources available and/or affordable. Jointly with our partner Engits GmbH, to push the limits, we develop the new CFD tool DrNUM (dual resolution numerics) featuring:
- Consistent use of GPUs: Compared to CPUs, GPUs feature far supperior computational performances in affordalbe hardware units, the latter considering investments as well as power consumption.
- Dual resolution approach: Optimized grid system for maximum numerical efficiency, minimized memory requirements and optimized memory access.
At present (early 2019) DrNUM is not yet available for all simulation problems in CFD. Where applicable, huge compuations that would else require large data centres, can be performed on simple gaming PCs at extremely low costs.
Other simulations methods
In our work, we often encounter simulation problems, for which no suitable, readily available software can be found. In such cases, we may thoroughly analyse the given problem. In many cases we can offer specific or highly customized numerical methods to solve the tasks being.
Artificial intelligence (AI)
Based on our long-time comptences in computational intense simulation problems on HPC architectures, we currently focus on developments in the broad field of artificial intelligence, including deep learning. As at present this field is still mostly restricted to science, it is our goal to offer efficient methods and solutions for a broad range of future applications.
Please do not hesitate to contact us with your most demanding numerical and simulation problems for which no suitable or fully satisfying solution could yet be found.
|
|
