The CRF team explains in this short interview a few key concepts for the Work Package 2: AI based designs for aerodynamics.
The main focus of this work package is the definition of a framework addressed to aerothermal simulations (external aerodynamics, cooling flows, thermal protection) based on model order reduction.
The main picture shows the other partners involved on this work package.
The importance of the aerodynamic performance is a growing issue in the modern automotive industry. As leaders of the Work Package 2: AI based designs for aerodynamics, how AI can help improve the design performance?
Since AI is spreading exponentially in all technological fields we took the opportunity given in this project to test potential approaches for aerodynamics, and especially CFD, integration. Affordable and reliable applications in this field are still limited, but we saw an opportunity for the development of prospectively cutting-edge techniques for results analysis and data-driven shape optimization methodologies. Final objective, something that was a dream just a decade ago, could be the availability of a real-time tool for the prediction of flow fields around a detailed full vehicle. Nevertheless, the road ahead is still a lot.
Why have you decided to start the input modelling with 2D shapes?
As you may reasonably question, 2D CFD is certainly not the state-of-the-art for ground vehicle external aerodynamic simulation, and at the same time is not the goal of this project; however, since the AI-based workflow is on early stage a bottom-up approach has been chosen in the setup phase of the activity. This could guarantee several benefits: quick turn-around times for preliminary tests and debugging, draft assessment of data requirements, reasonable projections on final cost of the procedure. At the same time, it is considered a definitely non-trivial application from a physical point of view and quite easily extendable, when the maturity of the tool will be considered acceptable, to three dimensions. Of course, the goal will be to develop a procedure that can be applied on 3D shapes significant from an industrial perspective.
Which are the requirements considered for aerothermal simulations reduced order model?
Generally speaking, all virtual methodologies related to complex physical phenomena such as fluid dynamics need to keep into account feasibility in terms of computational cost. The aim of the requirements definition phase is to identify the best compromise between accuracy/quantity of extracted data and cost. For sure, developed methodologies need to guarantee a reliable trend assessment for shape modifications in order to correctly drive the style development phase of the vehicle; a faithful description of vehicle flow fields in terms of velocity and pressure will be also a goal of this project. Main constraint, of course, would be the usability of the procedures on current industrial cluster in order to actually exploit the findings of the projects to develop our vehicles.