| description abstract | Assessing vehicle safety is a challenging, yet fundamental task. In the early phase of development, car manufacturers need to ensure the compliance with strict safety requirements. An interesting task to automate these early-stage operations is to harness information from already developed products. Established designs are largely accessible, with abundant data; novel designs' data are scarce. While established and novel designs are (by definition) different, it is expected nonetheless that there is a degree of correlation between them. Thus, the established design could be regarded as a low-fidelity (LF) model of the novel design, in the sense that it may provide an approximation of the behavior of the novel design. In turn, the novel design could be regarded as a high-fidelity (HF) model, as it represents the true product being designed. This bifidelity character of the problem stands at the basis of this paper. This work explores the application of control variates (CV) to a crashworthiness analysis scenario. Control variates is a variance reduction technique that exploits the low-fidelity information to improve the accuracy of the response statistics of the high-fidelity model. Such an approach could be most useful for industrial applications. Therefore, we apply control variates to a crash box example and compare its performance to its plain Monte Carlo (MC) counterpart. The results of this paper show the benefits of this bifidelity approach, resulting in control variates being a powerful technique to extract valuable information from limited data sets. Indeed, control variates can serve as an innovative solution to support car manufacturers in the early phase of vehicle development and thus improve the performance in crashworthiness scenarios. | |
