Response surface methods have been widely used in robust design for reducing turn-around time and improving quality. That is, from a given set of CAE data (design-of-experiments results), many different robust optimization studies can be performed with different constraints and objectives without large, recurring, computation costs. However, due to the highly nonlinear and non-convex nature of occupant injury responses, it is difficult to generate high quality response surface models from them. In this paper, we apply a cross validation technique to estimate the accuracy of response surface models, particularly in the context of robustness assessment. We then decompose selected occupant injury responses into more fundamental signals before fitting surfaces to improve the predictivity of the response surface models. Real-world case studies on an occupant restraint system robust design problem are used to demonstrate the methodology.
Improving Robustness Assessment Quality Via Response Decomposition
Sae Technical Papers
SAE 2006 World Congress & Exhibition ; 2006
03.04.2006
Aufsatz (Konferenz)
Englisch
Improving robustness assessment quality via response decomposition
Kraftfahrwesen | 2006
|2006-01-0760 Improving Robustness Assessment Quality Via Response Decomposition
British Library Conference Proceedings | 2006
|Sequential Subspace Robustness Assessment
AIAA | 2016
|IMPROVING ROBUSTNESS FOR REAL-TIME VEHICLE EGOMOTION ESTIMATION
British Library Conference Proceedings | 2016
|