A numerical and parallelised method to describe a reacting packed bed in contact with a gas phase is being developed by linking material and computer sciences.
Packed bed processes are important engineering applications as encountered in renewable energy and hard metal production. The latter in particular is investigated for conversion of pulverised tungsten oxide to metallic tungsten in a packed bed. Reduction of tungsten oxide to tungsten in a hydrogen atmosphere is described by a reaction scheme, for which temperature and reaction progress is described by the Discrete Particle Method (DPM). Hydrogen as a gaseous phase streams over the packed bed of tungsten oxide particles in push-type furnace. The flow over and penetration of hydrogen into tungsten particles’ bed is represented by an advanced two-phase CFD-tool for porous media. These efforts will complement experimental investigations and provide a deeper insight into the process, because particle temperatures and interaction of particles with the fluid are inaccessible in a packed bed during experiments. Parametric studies will be carried out for improving the performance of the process.
As these predictions require significant amounts of CPU-time, parallelisation of the conversion module of the Discrete Particle Method is approached by an interdisciplinary concept as collaboration between engineering and computer science disciplines.
The workload for each processor is determined by the Orthogonal Recursive Bisection (ORB) algorithm that distributes the particles uniformly onto the processors available. MPI (Message Passing Interface) is employed for inter-processor communication. Additionally, the scalability of the approach on massively parallel machines is being investigated.
Add comment