Journal Article

FZJ-2023-00625

http://join2-wiki.gsi.de/foswiki/pub/Main/Artwork/join2_logo100x88.png Applying Physics-Informed Enhanced Super-Resolution Generative Adversarial Networks to Large-Eddy Simulations of ECN Spray C

Bode, M. (Corresponding author)FZJ*

2022
Soc. Warrendale, Pa.

This record in other databases:

Please use a persistent id in citations: http://hdl.handle.net/2128/33608  doi:10.4271/2022-01-0503

Abstract: Large-eddy simulation (LES) is an important tool to understand and analyze sprays, such as those found in engines. Subfilter models are crucial for the accuracy of spray-LES, thereby signifying the importance of their development for predictive spray-LES. Recently, new subfilter models based on physics-informed generative adversarial networks (GANs) were developed, known as physics-informed enhanced super-resolution GANs (PIESRGANs). These models were successfully applied to the Spray A case defined by the Engine Combustion Network (ECN). This work presents technical details of this novel method, which are relevant for the modeling of spray combustion, and applies PIESRGANs to the ECN Spray C case. The results are validated against experimental data, and computational challenges and advantages are particularly emphasized compared to classical simulation approaches.

---

Contributing Institute(s):

1. Jülich Supercomputing Center (JSC)

Research Program(s):

1. 5112 - Cross-Domain Algorithms, Tools, Methods Labs (ATMLs) and Research Groups (POF4-511) (POF4-511)
2. CoEC - Center of Excellence in Combustion (952181) (952181)
3. Using deep learning to predict statistics of turbulent flows at high Reynolds numbers (jhpc55_20190501) (jhpc55_20190501)

Appears in the scientific report 2022

---

Database coverage:
; ; ; SCOPUS

---