Daniel Grealy

In 2023, roughly 1.2 million people died in road traffic accidents. Crashworthiness design involves selecting a set of design parameters - which can be shape parameters, size parameters, material selection, etc. - such that in a crash scenario, the vehicle absorbs the impact energy instead of passengers and pedestrians. Reducing vehicle mass increases energy efficiency and is therefore a secondary goal of crashworthiness design. However, simulating a crash using finite element analysis (FEA) is computationally expensive, especially if complex human body models are used to evaluate passenger or pedestrian injuries.

Crashworthiness optimization aims to find the optimal combination of design parameters with as few simulations as possible by creating and optimizing a surrogate model of the crash response. Multi-fidelity optimization is a special case where simplified models of the system or component being optimized are available. Simplified models trade a decrease in accuracy for a decrease in computational cost. Multi-fidelity modelling aims to combine results from models of varying complexity to retain or increase surrogate model accuracy, while reducing overall computational cost. This is commonly achieved using various extensions of a surrogate modelling approach called Gaussian Processes.