These slides were presented at the NAFEMS World Congress 2025, held in Salzburg, Austria from May 19–22, 2025.

Abstract

The complexity of vehicle development is increasing due to diverse power sources and advanced technologies, requiring efficient and flexible systems. The V-model-based target cascading method has limitations in early-stage design due to broad performance requirements and high design variability. This study aims to probabilistically quantify uncertainties in design variables and vehicle performance early in the development process to ensure robustness and reliability. A probabilistic analytical target cascading method using a machine learning model predicts R&H performance variability. In this study, we proposed a target cascading process for stochastic analysis using machine learning models to strengthen R&H performance in the architectural stage. Based on the above process, the correlation between handling performance, K&C characteristics, and design variables was quantitatively identified between finished cars, systems, and parts in the V model. In addition, handling performance and parts were quantitatively identified according to changes in hardpoint and bush rigidity. The uncertainty in the performance of the vehicle was quantified by checking the distribution of K&C characteristics. Based on the above results, the range of front and rear wheel K&C characteristics and the range of design variations of hard point and bush stiffness to improve the stability and responsiveness of the vehicle were determined by applying the optimization technique based on reliability analysis. As the probabilistic analysis target cascading technique was applied to the R&H performance development when developing a vehicle in the architectural stage, it was confirmed that the following expected effects were found. Through reliability-based optimization analysis, we were able to derive an optimal design that could be realized by presenting the distribution range of system characteristics and vehicle performance. In addition, we were able to secure the R&H performance reliability of the platform by identifying the performance distribution of various product groups within the same platform at an early stage. Finally, By quantifying the correlation between system characteristics and vehicle performance and predicting and managing dispersion, data-based efficient decision-making was possible for frequent design changes. Based on this research case, if the probabilistic analysis target cascading process presented in the paper is extended to other performance fields besides R&H performance, it is expected that the efficiency of development work can be increased by strengthening vehicle performance in the early stages of vehicle development.