Since 1908 Avio Aero design and produce the best-in-class aeronautical components that allow people to fly all around the world safely. To be the leading edge of the highly technological aeronautical industrial sector, the company has never stopped to innovate always providing to the market the prime of key aeronautical products.
In the last decade, more attention has been paid to the design of greener aero engines. One of the goals of Avio Aero is to reduce the impact of jet engines and thus contribute to re-build a greener world. To achieve this important objective, aeronautical companies are working to design newer key components leveraging advanced simulation processes that are powered with the ultimate HPC technology permit to produce more precise and reliable results. The ACROSS project gives to Avio Aero and partners the opportunity to exploit the most advanced technology in the HPC area and, thanks to it, overcome the limitation of the current design process bringing more accurate simulation.
Supported by Morfo, the University of Florence, and the University of Genova, Avio Aero provided two test cases to the ACROSS project. The first pilot proposes an innovative design process for low-pressure turbine blades. The goal is to develop an advanced AI-based design system aimed at reducing the time-to-design by 50% and better predicting turbine efficiency which is translated into a reduction of specific fuel consumption. This test case will take advantage of modern AI techniques and high-performance data analytics (HPDA). As a task of the simulation workflow, this pilot can leverage the exploitation of advanced artificial neural networks to build a solution space suited for low-pressure turbine blades that operate in a wide range of conditions. The new data-centric environment will be based on multi-fidelity CFD results (combining URANS and LES calculations) and it will be driven by AI to face a multidimensional, multi-objective, and highly constrained optimization problem. When the design system will be validated it will be very fast to obtain the geometry optimal candidate by querying the previously trained ANN.
The second pilot aims to introduce a novel approach for the design of combustion chamber cooling systems. For the correct prediction of the temperatures, an accurate solution of the velocity field, of the combustion reaction, of the heat exchanged by conduction and radiation is essential. To allow the correct modeling of the physical phenomena just introduced in a consistent way with the times dictated by the industrial sector it is necessary to adopt multi-physics and multi-scale approaches. The expected results are the correct prediction of metal component temperatures combined with a 30-50% reduction in design time.
This pilot consists of an advanced CFD tool developed within the Ansys fluent platform using user-defined functions and python scripts. A specific solver will be adopted for each previously introduced physical phenomenon to ensure that it is solved as efficiently as possible, this is the reason for which the greatest effort will be concentrated on the management of data exchange between the different solvers.
Both Avio Aero pilots are very hardware consuming but ACROSS project is providing really powerful infrastructure. The turbine use case is running on the Galileo100 cluster made available by CINECA that has a theoretical peak of 3.5 TFlops/s. The second pilot is taking advantage of two supercomputers offered by IT4I named Karolina and Barbora with a theoretical peak of 15.7 Pflops/s and 849 Tflops/s respectively.
Finally, the Experimental R&D cluster provided by Atos will be exploited to further improve the performance of the AI part by leveraging high-performing AI-acceleration devices like NNPs, VPUs, and even neuromorphic emulators/simulators.