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

Abstract

Set-based design methodologies have demonstrated a significant promise in offering systems engineers a broader range of choices earlier in the design process than conventional point-based design. This approach allows engineers to explore multiple alternatives simultaneously, making it particularly advantageous in dynamic and rapidly evolving fields. However, the dynamics in automotive design are evolving at an unprecedented pace. The myriad market and regulatory factors contribute to increased levels of complexity that threaten to overwhelm traditional time- and resource-intensive SBD processes. Challenges include integrating new technologies, meeting stringent emissions standards, and accommodating diverse consumer preferences, which collectively strain conventional design frameworks. In this talk, we will explore how advanced, data-efficient Machine Learning techniques can deliver a step change in system design optimization, pushing set-based design to new and essential levels of efficiency and impact. These ML techniques facilitate more comprehensive analysis through a highly data efficient approach, to inform decision-making, thus overcoming limitations of traditional methods. Through practical examples of real-world design challenges, we will demonstrate how ML helps engineers identify and rapidly evaluate better choices across vastly more complex design spaces. Additionally, we will discuss how these techniques can reduce dependencies on expensive simulations by predicting outcomes more accurately with fewer computational resources. This ability minimises schedule risks and cost overruns by adaptively learning and refining processes as new data becomes available. In particular, we will illustrate how set-based design frameworks can be used to decouple requirements across a multi-dimensional parameter space, enabling simultaneous consideration of multiple disciplines and teams. This aspect highlights its power in collaborative environments where interdisciplinary integration is crucial. Attendees will leave this talk with a deeper understanding of how augmenting set-based design with Machine Learning can significantly enhance its capabilities, particularly for complex products or systems involving multifaceted trade-offs. This approach not only augments human expertise and improves efficiency, but also uncovers novel, data-driven insights, making set-based design a more robust and powerful tool in the engineer'™s toolbox. By embracing these innovations, organisations can drive forward more effective, sustainable, and innovative automotive design processes.