This conference paper was submitted for presentation at the NAFEMS World Congress 2025, held in Salzburg, Austria from May 19–22, 2025.

Abstract

The automotive design process is inherently complex and time-consuming, involving numerous specialists who must collaborate closely to predict and optimize every aspect of a vehicle'™s performance. Engineers face the challenge of meeting strict targets and regulations while adapting to rapidly changing market demands, all without sacrificing quality or increasing development costs. Traditional approaches often rely on repeated trial-and-error, focusing on only a small subset of all possible design configurations. This not only slows down the process but also risks overlooking potentially better solutions that could enhance product performance or reduce overall costs. To address these challenges, integrating Machine Learning (ML) into Computer-Aided Engineering (CAE) has emerged as a promising strategy. By leveraging ML-based methods, engineers can quickly evaluate a wide range of design variables and performance criteria without running excessively detailed and time-consuming simulations. Instead of limiting exploration to a handful of configurations, they can now consider a much broader portion of the design space, identifying optimal choices more efficiently. This early insight reduces the likelihood of late-stage design changes, cutting down on both expenses and delays. SEAT S.A. has developed ARCO (Advanced Real-time Car-body Optimization) as a physics-informed ML technique specifically aimed at improving the noise and vibration (NVH) performance of car-body structures. NVH behavior is highly sensitive to changes in global static and dynamic stiffness, making it a critical aspect of vehicle design. By applying the Proper Generalized Decomposition (PGD) method, ARCO can efficiently handle parametric analyses of static and dynamic behaviors. This approach enables a single offline computation to capture a wide range of design possibilities, effectively reducing model complexity and highlighting key structural patterns. Previous studies have already demonstrated the potential of ARCO in addressing NVH challenges. The next step involves adapting ARCO to handle full-scale industrial models and ensuring seamless integration with commercial engineering tools. By streamlining the design process and making advanced ML approaches more accessible, ARCO represents a significant step forward in achieving improved, efficient, and reliable automotive development.