Journal Article/Contribution to a conference proceedings

FZJ-2024-06724

http://join2-wiki.gsi.de/foswiki/pub/Main/Artwork/join2_logo100x88.png Evaluation of GPU-based Conductive Heat Transfer Algorithms

Döhle, D. ; Börger, K.FZJ* ; Arnold, L. (Corresponding author)FZJ*

2024
IOP Publ. Bristol

This record in other databases:  

Please use a persistent id in citations: doi:10.1088/1742-6596/2885/1/012108  doi:10.34734/FZJ-2024-06724

Abstract: The one-dimensional heat transfer algorithm of the Fire Dynamics Simulator (FDS) is currently implemented to run on a CPU (Central Processing Unit). This study explores the potential advantages of adapting the algorithm for Graphics Processing Units (GPUs), which could o!er significant computational benefits. The motivation behind this work stems from the intention to speed up numerical fire simulations. Up to now, simplifications with regard to grid resolution and level of detail have been made, compromising accuracy for quicker results. Simulations, especially for heat transfer in solid objects such as walls, require computationally intensive resources. By leveraging the GPUs’ superior parallel processing capabilities, it is possible to conduct faster and more accurate simulations, avoiding these compromises. Both a CPU and a GPU algorithm for computing the 1D heat transfer are developed, and the computation time is compared against each other. Both implementations are validated against a simple FDS simulation with identical boundary conditions. The investigations show that the GPU algorithm is promising above a certain number of wall elements, depending on the employed hardware. The results show that this is generally the case from 2048 elements.

---

Contributing Institute(s):

1. Zivile Sicherheitsforschung (IAS-7)

Research Program(s):

1. 5111 - Domain-Specific Simulation & Data Life Cycle Labs (SDLs) and Research Groups (POF4-511) (POF4-511)

Appears in the scientific report 2024

---

Database coverage:
; ; ; Nationallizenz ; SCOPUS

---