AMDECC Project wins award at Teratec Forum

AMDECC stands forAdaptation de Maillage Dynamique et massivement parallèle pour la simulation aux grandes Echelles des Chambres de Combustion aéronautiques (massively-parallel dynamic mesh adaptation for large-scale simulation of combustion chambers for the aeronautics industry). The aim of this project is to design helicopter engines that afford better, and cleaner, performance. To achieve this, the team uses large-scale simulations of engine combustion chambers with a view to improving quality-to-cost ratio. Optimising the simulations is key to the project's success since this approach requires intensive use of high-performance computing.

The team gave two demonstrations which impressed the judges. The first entailed predicting fuel atomisation in an industrial fuel injection system using 4-micron resolution and dynamic monitoring. To achieve this, the team needed to simulate the liquid-gas interface using an extremely refined mesh. The computations required were performed in parallel using 18,000 processor cores, a world first in dynamic meshing.

The second demonstration involved the total, fully-automatic simulation of a combustion chamber, together with mesh convergence and post-processing performed during computation. With total computing time reduced from two weeks to just two days thanks to this simulation, industrial applications will be capable of exploring seven times more technological designs than is currently possible.

The solution proposed by AMDECC targets computational resources at specific physical zones of interest, which may change position as a function of time, and thus makes it possible to selectively increase the resolution without increasing the number of processor cores required.

The project initiators dedicated the trophy to Cécile Dobrzynski, who passed away on 23 May after battling against illness for 15 years. Cécile was a Lecturer at the Bordeaux INP ENSEIRB-MATMECA Engineering School and a member of the CARDAMOM team at Inria. Without her research on mesh generation and adaptation, the AMDECC Project would not be the success it is. Aware of the importance of conducting research that has an impact on industry, she chose to develop stable and long-lasting research software. Her PhD thesis was the foundation for the development of the open-source Mmg platform as well as for setting up the Mmg consortium, which is supported by its academic and industrial members.