The integration of artificial intelligence into engineering processes has become a major transformation lever for the automotive industry, which is facing increasing complexity, stricter regulatory constraints, and ever-shorter development cycles. In this context, numerical simulation is undergoing a profound evolution, shifting from traditional approaches to hybrid workflows that combine high-performance computing with AI-based models. This talk will share concrete industrial feedback on this ongoing transformation.
Two key pillars will be highlighted. The first focuses on accelerating simulation loops through deep learning surrogate models, enabling a drastic reduction in computation time, from several hours on HPC systems to a few seconds on GPUs. The second addresses improved predictivity through hybrid twin approaches, combining physics-based models with experimental data to enhance robustness and result reliability.
These evolutions are driving a fundamental shift in how computing resources are used: increased reliance on HPC upstream for model training, followed by much more agile and distributed usage downstream, empowering engineers in their day-to-day work.
Finally, we will discuss the emergence of foundation models for engineering, paving the way for shared training efforts and enabling scalable industrial deployment of AI in CAE processes. This evolution marks a major transition—from heavy, sequential simulation workflows to augmented engineering that is faster, more predictive, and more collaborative. |