Coupling of Deep Learning on Graphs and Model Order Reduction for Efficient Preliminary Sizing of Mechanical Structures in Aircraft Crash Simulations

This poster presents work focused on combining two disparate methodologies: deep learning on graphs and model order reduction. The former demonstrates promising generalization capabilities and efficient execution times, while the latter relies on physics-based resolution of the governing mechanical equations and has been gaining traction in industrial applications. In this work, conducted in collaboration with Safran Tech, the targeted application is the mechanical sizing of aircraft seats in airplane crash simulations.