The aerospace industry and TU Delft have set stringent climate goals for themselves and with each passing year, these goals seem all the more implausible. While sustainability in the aerospace industry is envisioned as futuristic, circular, net-zero aircraft, the change rarely comes from the grassroots. Although these technological solutions are undoubtedly our liberators from long-term climate change, there is relatively less focus on the education of engineers who grow up to design these machines. This thesis follows a bottom-up approach which can stimulate pro-environmental attitudes among future aerospace engineers to give ourselves a fighting chance to mitigate climate change.

Due to an exponential growth in the number of shipments of Unmanned Aerial Systems (UAS), the amount of these devices operating in the sky has increased remarkably over the last few years. This led to an increasing number of proximity incidents with manned aircraft. Since these devices share certain airspace with rotorcraft, the question arises how much damage a helicopter could sustain after an impact with a UAS. Within this thesis, a risk assessment was completed initially to determine which collision in terms of type of UAS and helicopter impact location poses the highest risk to the operator of the helicopter. Subsequently a validated model of a DJI Phantom III was developed and impacted onto a rotorcraft windshield in explicit Finite Element software. The sustained damage was compared with a simulated bird strike event to determine whether the prevailing certification requirements would suffice to guarantee safety of the crew.

Using a data-driven approach, a heterogeneous populated agent-based model of an airport security checkpoint is calibrated and validated. It is found that average performance using the model can be replicated to a sufficient degree. The heterogenous population could become useful when different passenger fractions - with different performance per passenger type - are analysed for different security checkpoint configurations. Furthermore, time-dependent agent performance could improve the models validity.