Search for a command to run...
This is the final report GIS 2 work package 2 on injury prediction (LBMC of Université Gustave Eiffel). The activities are now completed although some scientific publication activities may still take place.The main objective aim of the research was to study the risk of skull fracture in case of impact with small drones, and consider its transfer towards a method to assess this risk in testing with actual drones. Based on the low number of fracture cases in the literature, it was assumed that the risk to estimate would be low and could not be based on field observations. Most knowledge regarding skull fractures resulting from impact is derived from the automotive safety field in which impact velocities are typically lower. Therefore, an approach essentially based on detailed human body models that can describe bone deformation was proposed. The hypothesis was that using a model able to describe the fracture mechanisms closer to the material level would be more robust to changes of conditions related to velocity, stiffness or surface characteristics. The idea was to develop risk curves for the model including large amount of data including high speed impact to increase to confidence in the risk prediction and then to use this risk curve as a reference. The data would be complemented by new high-speed tests on PMHS to also help with the model validation. The core of the project on risk prediction was be the topic of a PhD thesis.After a first exploratory study (section 2, Pozzi et al., 2022) conducted with simplified impactors showing the importance of contact pressure and impact velocity for the occurrence of fracture simulated by element elimination in the model, work focused on the assessment of the model biofidelity and the development of a bone strain based risk curve (section 3, Pozzi et al., 2025). While significant limitations were observed in terms in model biofidelity (force prediction, reference paper reproduction), risk curves could be developed based on principal strains in the skull cortical bone using a large number of impact cases to increase the confidence at low risks (n=156). The best curves were found to predict well the risk of fracture (rating good in the sense of ISO TS18506:2014 for risks of 50%, 25%, and just below at 10%). No major discrepancies were observed between low and high-speed conditions within the limitation of the number cases available (only 19 drones including 5 drones).Efforts to complement the experimental datasets (section 4) to support model improvements in terms of biofidelity and injury prediction (section 5) were initiated in parallel. A protocol was developed and two PMHS tested, with injuries fully in line with the risk curve predictions (18 tests, 3 locations, 2 injuries). Observations from the tests and previous simulations drove the improvement efforts. While changes performed on the scalp in particular improved the biofidelity, the overall quantitative gain was lower than hoped for and the improvement in risk prediction was marginal. Further adjustments of the model may be required.The risk transfer was therefore developed (section 6) using the published model and risk curve (Pozzi et al., 2025). Based on paired simulations between the human model and test devices in 165 conditions each, peak acceleration filtered 1000Hz measured on human surrogates covered by a deformable skin ranked first (e.g. Hybrid III). Although requiring more validation, further improvements seem possible if accounting for the contact surface area which was larger for the drones. The results were compared to existing literature, past drone tests and possible test speeds discussed. Results can be used to support the development of test protocols.Based on all the results from the simulation, testing and literature, the general discussion tries to examine regulatory guidelines (FAA and EASA). The comparison suggests high difference between the two approaches, with potential issues regarding the 80 J of energy transmitted to the head for the EASA (questionable severity ranking, robustness to evolution of drone properties). The discussion concludes about the perspectives with propositions and steps to derive improved test protocols from the risk transfer.