SC²S Colloquium - February 21, 2018
|Date:||February 21, 2018|
|Time:||2:00 pm, s.t.|
Piet Jarmatz: Parallel Noise Reduction for Transient Molecular-Continuum Coupled Flow Simulations
Molecular-Continuum flow simulations often suffer from high thermal noise, created by fluctuating hydrodynamics within the molecular dynamics simulation. Multi-instance MD computations are an approach to tackle this, but they require a huge amount of computational resources. Filtering particle-based simulation data, using signal processing methods to reduce numerical noise, can lead to promising results and reduced computational cost. In this Master's Thesis we will compare several noise reduction algorithms under the main aspects of parallelization and filtering efficiency. We plan to present an implementation of a massively parallel noise filter using the Macro-Micro-Coupling tool and validate the bidirectional transient coupling using a three-dimensional Couette Flow scenario. Finally we will conduct performance benchmarks and scaling tests on supercomputing platforms to demonstrate the feasibility and power of such a massively parallel efficient noise reduction approach.