On the impact of geometric variance on the performance of formed parts: A probabilistic approach on the example of airbag pressure bins

TL;DR

This study quantifies how manufacturing-induced geometric variance translates into scatter in airbag pressure bin performance using a probabilistic framework. It models geometry as random variables (φ, h), uses Latin hypercube sampling to drive Abaqus FE analyses of a cold-forged airbag bin in 16MnCrS5 steel, and builds a Gaussian process surrogate to compute variance and Sobol' indices. The results reveal φ as the primary driver of stress scatter, with indentation playing a minor role, suggesting calibration and QA should target φ for lightweight design. The approach enables probabilistic performance prediction post-manufacture and informs quality assurance to reduce conservative safety factors while maintaining safety.

Abstract

Scatter in properties resulting from manufacturing is a great challenge in lightweight design, requiring consideration of not only the average mechanical performance but also the variance which is done e.g., by conservative safety factors. One contributor to this variance is the inherent geometric variability in the formed part. To isolate and quantify this effect, we present a probabilistic numerical study, aiming to assess the impact of geometric variance on the resulting part performance. By modelling geometric deviations stochastically, we aim to establish a correlation between the variance in geometry with the resulting variance in performance. The study is done on the example of an airbag pressure bin, where a better understanding of this correlation is crucial, as it allows for the design of a lighter part without changing the manufacturing process. Instead, we aim to implement more targeted and effective quality assurance, informed by the performance impact of geometric deviations.

Figures (3)

• Figure 1: Schematic view of the considered geometry and variation.
• Figure 2: FEM analysis; nominal geom. (a), perturbed geom. (b, sample 31).
• Figure 3: Sobol' indices resulting from the variations.