Aims

On a worldwide scale efforts targeted at Monte Carlo based reactor dynamics codes are about to leave scouting phase and turn towards bringing proof-of-concept schemes closer to realistic industrial applications. The development of GUARDYAN approaches this topic by focusing on massively parallel architectures such as General Purpose GPU’s. Recently commercially available GPU memory sizes reached 8 Gbyte breaking through the barriers associated with memory needs of an industrial application. For reference, a fully detailed geometry and cross section model of a commercial VVER-440 power plant geometry was created in GUARDYAN necessitating 5 Gbyte of memory including cross sections preprocessed for fast Monte Carlo sampling. Hardware developers recently announced GPU memory capacities of 24 Gbyte and up, finally clearing all the obstacles standing in the way of industrial size applications. GPU implementation is expected to offer at least a factor 5 speedup compared to similarly priced CPU’s (Central Processing Unit) if the GPU calculation flow is optimized to use the hardware capacities. This latter condition is not to be overlooked: a poorly demised parallelization scheme can very easily underperform the CPU implementations of decades of careful optimization work.

Implementation of GUARDYAN started in 2015 in the frame of VKSZ_14-1-2015-0021 Hungarian project supported by the (Hungarian) National Research, Development and Innovation Fund (NRDI). The code is expected to include full physics modeling capabilities, kinetic calculation capabilities without thermal feedback and rudimentary thermohydraulic modeling with feedback at the project end of 2019.

next: Guardyan Features

VVER-440 type power plant geometry modelled by GUARDYAN

• Aims
• Examples
• Funding
• PhysModel
• Publ
• VarRedu
• features

---