Introduction

Traumatic brain injury (TBI) sustained by athletes occurs at the highest rate in American football (6.4 per 10,000 athletics exposure) with high concussion rates at the high school, collegiate10 and the professional levels of play25 and accounts for 47.1% of sport-related concussions.19 Considering head impacts in American football, the helmet is the primary energy absorbing component, with the aim to reduce head kinematics. Since a combination of linear and rotational kinematics12,16,17,21,26,29 has been hypothesized to cause concussive neurotrauma,40 head angular velocity and linear acceleration are commonly used to evaluate football helmet performance. The National Operating Committee on Standards for Athletic Equipment certification standard22,36 limits allowable linear accelerations in a drop test impact with a rigid neck. The Summation of Tests for the Analysis of Risk (STAR) rating system28 uses peak angular velocities to determine the helmet performance (1 to 5 stars) using the Hybrid III head and deformable neck with four head impact locations. In general, helmets with lower angular kinematics measured at the head center of gravity (COG) perform better on the STAR evaluation system. However, it has been suggested that test apparatus allowing head rotational displacements and including a broader combination of impact locations and speeds could improve helmet assessment.12 This is, in part, related to the finding that concussions are associated with specific impact directions and that different types of impact are associated with the varying player position.36 The linear impact test23 evaluates the helmet in a wide variety of impact locations and speeds, and the incorporation of a deformable Hybrid III head-and-neck assembly enables the assessment of angular kinematics. The linear impact test will be adopted by NOCSAE24 for helmet evaluation in 2019.

Finite element (FE) models have been used as a research tool to investigate the structural and kinematic response to impact, and to a lesser extent to evaluate and optimize protection design. Detailed FE models, which are validated under a wide range of impact conditions, have the potential to enhance the performance of current protection by providing information that would be challenging to retrieve by laboratory methods; namely video analysis,26 in-game accelerometer data,6,32 physical helmet experiments,35,38 and physical component level experiments.2,4 The understanding of the behaviour of energy absorbing components is of interest at the stage of designing helmets, evaluating their performance, and in reducing head injury risk. Johnston et al.15 developed a football helmet FE model (Schutt Ion 4D) to investigate changes of performance due to numerical alteration of energy absorbing padding within the helmet model. Both the physical helmet and the numerical equivalent were donned on a NOCSAE headform23 coupled with a rigid neck, subjected to drop test impacts (heights of 304.8 mm to 1524 mm), and lateral linear impacts (velocity between 3.2 and 3.3 m/s) where axial rotation of the head was allowed. Although this study demonstrated the potential for FE models to provide information on helmet performance, no details were provided for: the headform model and validation, the geometry development of the helmet model, or model validation were provided.

In another study, Johnson et al.14 integrated a simplified generic football helmet FE model with a simplified human head FE model comprising three components (skull, cerebrospinal fluid, and brain) to perform a topological optimization on the helmet facemask. The optimization-targeted reduction of shear strain and maximum tensile pressure predicted in the FE brain model under a linear impact. However, no details on the helmet or head model validation were provided.

Darling et al.5 integrated an FE model of a current American football helmet (Riddell Attack Revolution Youth helmet) with the Global Human Body Models Consortium (GHBMC 2014) average stature male Human Body Model (HBM) to evaluate TBI risk under helmeted and non-helmeted head impacts. The FE helmet model geometry was based on the youth football helmet (Riddell Attack Revolution Youth) and simplified to comprise the shell, a uniform foam energy absorbing layer, and a uniform soft foam comfort liner. A linear elastic constitutive model was used for the helmet shell, whereas a non-linear and non-strain rate-dependent constitutive model was used for energy absorbing and comfort foam layers. The physical helmet was assessed using a low-severity (3 m/s) experimental drop test, where the helmet was donned on a hemispherical cap representing the headform in a non-standard test configuration impacting the crown of the helmet. The integrated FE HBM-helmet assembly was subjected to crown, and frontal (forehead) impacts to investigate brain response. However, the responses were only tracked for 20 ms, which corresponded to the initial portion of the loading phase.5 Due to the considered impact direction, the helmet model did not contain the facemask, chin cup, and a strap system. Other previous studies have focused on energy attenuation of football helmets fitted on headforms7,39 or focused on the energy attenuation of helmet liner components in isolation.13,18 However, no previous studies have quantified the energy distribution of the individual helmet components over a representative range of impacts conditions, and most studies have focused only on the helmet liner materials, but have generally not considered the helmet shell, facemask and chin straps.

Giudice et al.11 developed an open source set of headforms and impactors based on NOCSAE and STAR protocols,23,28 namely the NOCSAE head, and HIII head and neck models. Typical impact boundary conditions were also developed including the drop test, linear impactor, and pendulum impactor. Giudice et al.11 validated the FE models against 20 commonly used load cases for helmet evaluation, subjecting the HIII head and neck, and NOCSAE headform to the linear impactor, pendulum, and drop test loading. The models were compared to the experiments using International Organization for Standardization (ISO)1 and CORrelation and Analysis (CORA)9 cross-correlation methods, where a rating of 0 means no correlation and 1 means strong correlation between the measured experimental data and the simulation response. The models achieved good ISO cross-correlation [0.65–0.81], and good to excellent CORA ratings [0.71–0.89]. Recently, a set of 240 linear impactor, drop tower and pendulum impact experiments using four modern football helmets (60 experiments per helmet) donned on the NOCSAE and HIII headforms was published3 under the NFL engineering road map,37 providing a valuable data set for assessing the performance of FE helmet models.

Within the present study, a detailed FE model of a modern football helmet (Xenith X2E 2016, Fig. 1b) comprising the compression shocks (thermoplastic polyurethane (TPU)), helmet shell (polycarbonate), face mask (steel coated by polyurethane), shock bonnet (TPU), comfort pads (bilayer of high and low density foam encapsulated in vinyl) (Fig. 1a), was developed to assess helmet performance, quantify the strain energy per component, identify potential areas for improvement, and optimize future helmet designs. FE models of the helmet components (helmet shell, facemask, compression shocks, comfort pads, straps) were previously developed and validated.2,4 Material samples were extracted from the helmet and tested at quasi-static and dynamic rates (0.01 to 1400 s−1), in a loading mode corresponding to the component function in the helmet, and the isolated components were tested in the primary mode of loading seen in an impact scenario (e.g., compression shock and comfort pads were tested in through-thickness compression) at quasi-static and dynamic rates (0.0002 to 120 s−1). Constitutive models were identified and implemented based on the material test data for validation of the components with the experimental data, demonstrating good (0.65 to 0.85) to excellent (0.86 to 1) CORA ratings (0.95–0.98, 0.93, 0.94–0.97, and 0.85–0.96 for the helmet shell, facemask, compression shocks, and comfort pads, respectively). Mesh sensitivity studies were previously performed on the individual helmet components4 and were found to be within 2 to 7% of the theoretical zero mesh size.27

Figure 1
figure 1

Modern American football helmet (Xenith X2E) incorporating energy absorbing components.

Full size image

In this study, the helmet components were assembled into a full helmet FE model and donned on HIII and NOCSAE head FE models.11 The assembled helmet model was extensively validated using 60 physical impact experiments3 and the helmet model was used to assess the component energy distribution for varying impact velocity and location.

Materials and Methods

The physical helmet (X2E 2016, Xenith, LLC, size large) (Fig. 1b) was scanned using Computer Tomography (CT) (General Electric Phoenix-V|tome|x-s, 2014) in the fully assembled form to provide a reference for the assembled helmet model. The helmet was then dissembled into individual components (helmet shell, facemask, chin cup and straps, shock bonnet suspension system with compression shocks, and foam comfort pads) (Fig. 1). The individual components were then scanned and processed into surfaces using GeoMagic Wrap (3D Systems, Rock Hill, SC), which were then meshed using HyperMesh 14.0 (Altair Engineering, Inc., Troy, MI).

Helmet Model Assembly

The facemask was fixed to the helmet shell (Fig. 1e) at four locations using rigid beams representing clips in the physical helmet (Fig. 1b). The compression shocks (Fig. 1d) were integrated with the shock bonnet (Fig. 1d) through tying edges of the shocks to the bonnet using a nodes-to-surface contact algorithm to model the thermal joining between the components. Since the reference surfaces for the shock bonnet mesh were CT-scanned for a dissembled component, separate from the helmet shell, a pre-simulation was required to integrate the bonnet within the helmet shell. To locate the bonnet within the shell, the helmet shell was treated as rigid and kept stationary, while the bonnet and compression shocks were assigned physical material properties. The shock bonnet mesh was located inside the helmet shell, and a NOCSAE headform with prescribed motion was used to push the bonnet components to the final locations. The target location was confirmed using the full helmet scan, and physical helmet. The separate jaw compression shocks were positioned inside the helmet shell through graphical pre-processing (LS-PrePost 4.5, Livermore Software Technology Corporation, US), matching their final location in the assembled helmet. A fit was achieved when the bonnet and compression shock location matched the assembled helmet scan locations, ensuring there was no deformation of the shocks due to contact with the helmet shell. Tied contacts were used to represent screws connecting the bonnet to shell in the physical helmet.

The foam comfort pads (Fig. 1c) were scanned, segmented, and meshed as individual components; therefore, they were required to be fitted inside the shock bonnet to match their final positions. For the fitting process, the foam materials and corresponding vinyl covers were assigned elastic material properties with high stiffness (20 times the actual material properties) to avoid deformation of the foam during initial positioning. The NOCSAE headform with prescribed motion was used to move the comfort pad models to the target locations in the assembled physical helmet (Fig. 1b). Tied contacts were then used to connect the comfort pads to the shock bonnet (Fig. 1d), representing the Velcro connections in the physical helmet.

The strap system was modelled using shell elements at locations where the straps could contact the headform, and with 1D tension-only elements otherwise to numerically simplify guiding the straps through eyelets of the shock bonnet suspension system. This approach was adapted from seatbelt modelling practices in the automotive industry.34 The chin cup was positioned in the vicinity of the chin, based on images of a helmet donned on a headform. To ensure that helmet straps remain tightened, seatbelt retractor elements were included in the model at four locations corresponding to clips connecting straps to shell in the physical helmet. Prior to an impact simulation, the retractor elements were activated an applied an initial tension of 50 N per side to represent the strap pretension in a helmet donned on the headform or worn.

Donning the Helmet Model on the Headform

The helmet model was donned on the NOCSAE and HIII headform models developed by Giudice et al.11. Firstly, the headform and helmet models were overlaid in a graphical pre-processor (LS-PrePost 4.5) in accordance with the experimental tests (Biocore LLC). The helmet shell and facemask nodes were then constrained in all directions, and the headform was initially uniformly scaled down to 75% of the original size, to fit inside the helmet assembly without contacting any components, and then expanded uniformly until it reached full size through pre-simulation. All components of the helmet were assigned the average reported material properties4 to ensure a realistic fit between the headform and the deformable internal components of the helmet. A final pre-simulation was conducted to position the chin cup and adjust the straps around the coupled head-helmet model to match the position of the helmet relative to the head as reported in the experiments. The process was conducted for both the HIII and NOCSAE headforms separately to account for their specific geometry. The integrated helmet and headform models were then subjected to impact scenarios and assessed using the corresponding experimental data.

Helmet Model Validation for 60 Impact Cases

The helmet finite element model was analyzed using a commercially available explicit finite element solver (LS-DYNA, version R7.1.2). The pre and post processing was done in LS-PrePost (v4.2, LSTC, Livermore, CA, USA) and custom scripts to process the data were written in VBA (Excel, Microsoft, Santa Monica, CA, USA). Details on the constitutive models, contact algorithms and element types within the helmet model are included in the supplemental material (Table S8) and can be found in the publically available helmet model.37

Sixty experimental tests (Table 1) performed on the physical Xenith X2E helmet in a laboratory environment3,11 were used to validate the numerical helmet model. The experimental tests were performed following the NFL helmet test protocol8 and STAR rating system30 standards, namely a linear impactor and pendulum impacts including the HIII head and neck assembly, and a drop test impact with the use of two headforms (NOCSAE, and HIII head and neck) at different impact velocities and directions. For the drop test configurations, an initial velocity was applied to the parts that were in motion: helmet, headform, and the drop test carriage. The motion of the base of the deformable Modular Elastomer Programmer (MEP) pad and the load cell was constrained in all directions. The initial velocity applied to the parts in motion represented the velocity at the moment before impact for four different drop tests with the NOCSAE headform (2.9, 3.7, 4.9, and 6.0 m/s) and three with the HIII headform (2.9, 4.9, and 6.0 m/s). Headform kinematics, carriage kinematics, and load cell force were extracted from the model simulations. For the pendulum (3.1, 4.6, and 6.1 m/s) and linear impactor (5.5, 7.4, and 9.3 m/s) models, the initial velocity was applied to the impactor only. The deformable model of the helmet, head and neck were free to move following the impact, as in the experimental test set up (Table 1).

Table 1 Summary of helmet impact boundary conditions.
Full size table

Headform kinematics, impactor kinematics, and impactor load cell force predicted by the model were compared to the experimental data and assessed objectively using cross-correlation. The drop tests did not allow for head rotation, and the pendulum test series did not include a force transducer in the pendulum impactor. Therefore, these metrics were not incorporated into the CORA cross-correlation rating assessment. A list of the metrics used for the model validation can be found in the Supplementary Material (Table S1).

The termination time for all simulations was set to 50 ms, which was well beyond that required for analysis of the relevant kinetics and kinematics. The loading and unloading phases of the impacts, and relevant kinematics, were all complete by a time of 30 ms, and so this was the timeframe used for the CORA analysis. The experimental curves and model results for a given impact case were synchronized in time using the initial rise of the impactor acceleration trace from the experiments and impactor acceleration trace from the model (Fig. 2d). The same synchronization or time shift was applied to all model data. The weighting factors for the shape, size, and phase ratings were 0.5, 0.25, and 0.25 respectively, and the maximum allowable phase shift was 4.5 ms (15%) for each signal comparison. The phase rating was equal to 1 if the phase shift was less than 0.3 ms (1%), and equal to 0 if the shift was greater than 3.0 ms (10%). The CORA ratings for the three components of head translational acceleration and rotational velocity were combined into one rating each using the Signal Magnitude Factor31 based on the experimental peak magnitude.

Figure 2
figure 2

Linear impactor rear oblique configuration at 5.5 m/s. (a) simulation, (b) experiment video, (c) experimental (black) and simulation (red) plots of the head kinematics in the impact direction and (d) impactor kinematic used for signal alignment.

Full size image

In addition to the measured head kinematics, for the linear impactor load case, the rotation of the helmet shell in the XZ plane was extracted from the experimental videos and compared to the model rotation response using CORA. A video analysis package (Tracker v4.96), was used to measure the position of two landmarks in the video, and the same two landmarks were used to calculate the rotation of the helmet in the model. The location of the landmarks can be found in the Supplementary Material (Fig. S1).

Strain Energy Distribution in Helmet Component Groups

The linear impactor test evaluated the helmet over a wide range impact speeds and locations; hence, the strain energy or internal part energy of each component was extracted from the model (LS-PrePost, Internal energy in the matsum database) for these load cases. The helmet was assessed in five component groups: helmet shell, compression shocks, facemask, straps system, and comfort pads. The total strain energy of each group was defined as the sum of the strain energy of all components within the group (e.g., the strain energy of the compression shocks was defined as a sum of strain energy of each shock in the helmet at a given time step). The total energy of the comfort pads was the sum of the total strain energy of the foam and the vinyl cover, where the vinyl accounted for no more than 3% of the total strain energy. The total strain energy of the helmet was calculated by summation of the strain energy of the five energy absorption groups, and the contribution of each group to the total helmet strain energy was calculated as a percentage, that is group strain energy divided by the total helmet strain energy. The potential for head injury risk was evaluated by calculating the Head Injury Criterion (HIC15)20 based on resultant linear head acceleration for each impact scenario, and Brain Injury Criterion (BrIC),33 based on maximum head rotational velocities. Correlation of the HIC15 and BrIC values with strain energy in the five groups of components was then evaluated. A linear regression and Spearman’s correlation were used to identify the presence and strength of positive or negative correlation.

Results

Helmet Model Validation for 60 Impact Cases

The CORA ratings obtained for the head and impactor kinematics (Table 2) varied from good (0.65 to 0.85) to excellent (0.86 to 1), where the drop test cases had the highest ratings, 0.88 with the HIII headform and 0.93 with the NOCSAE headform, while the linear impactor had a rating of 0.82 and the pendulum impact 0.83. The model response for helmet rotation was in good agreement with the experimental data measured from video analysis, with an overall CORA rating of 0.79 (Table 2). CORA ratings for all the load cases and impact velocities can be found in the Supplementary Material (Tables S1–S6).

Table 2 CORA analysis results.
Full size table

Within the drop test with the rigid neck, all the impact speeds and locations performed similarly as measured using the CORA rating, and the ratings of the individual metrics within impact cases were also similar. Higher CORA rating variability was observed when using the HIII deformable neck, where lower scores were obtained in the impact cases that involved the facemask. Within impact cases, the rotational acceleration of the head demonstrated the lowest correlation ratings (e.g., 0.48) amongst all metrics. Generally, the CORA rating increased with increasing impact speed for the linear and pendulum impacts.

Strain Energy Distribution in Helmet Component Groups

In an exemplary linear impact condition (linear impactor, rear 5.5 m/s), the energy of the impactor mass and initial velocity (232.9 J) was distributed primarily into kinetic energy of the head (26%) followed by the helmet strain energy (20%). The helmet kinetic energy accounted for 17% while the neck strain and kinetic energy accounted for 11% and 7% of the total energy, respectively. The impactor and head strain energy contributed to 16% to the energy distribution. Hourglass and contact energy accounted for the remaining energy balance (3%). The maximum total strain energy of the helmet varied with impact direction and impact speed. The highest total strain energy occurred in rear impact at 39.7 to 66.7 J for impact speeds from 5.5 to 9.3 m/s. The lowest strain energy occurred in the facemask impact (19.1 to 44.5 J for 5.5 m/s to 9.3 m/s).

The HIC15 and BrIC values calculated from the head kinematics were plotted against the peak strain energy of each component for all linear impact conditions (Fig. 3). Overall, HIC15 was positively correlated with increasing maximum strain energy in each component. The compression shocks exhibited the strongest positive trend (Spearman’s Rho = 0.73) of increasing HIC15 with increasing strain energy, followed by the comfort pads (Spearman’s Rho = 0.70), shell (Spearman’s Rho = 0.60), straps (Spearman’s Rho = 0.18), and facemask (Spearman’s Rho = 0.10). The increase of BrIC with an increase of maximum strain energy per component was modest and positively correlated with the shell and facemask having the strongest positive correlation (Spearman’s Rho = 0.32 and 0.31 respectively).

Figure 3
figure 3

BrIC (left) and HIC15 (right) vs. helmet component group strain energy for all linear impact conditions and Spearman’s coefficients (Sp Rho), Spearman’s R2 (Sp R2) and linear regression Coefficient of Determination (R2).

Full size image

Overall, when the component energies for all impact directions and velocities were averaged, the helmet shell dominated the energy distribution, followed by the compression shocks (Fig. 4). This distribution was particularly pronounced and velocity-independent in configurations where there was no contact between the impactor and facemask. The contribution of the facemask progressively increased with impact velocity for all impact directions.

Figure 4
figure 4

Strain energy distribution (Helmet shell in red, shocks in black, facemask in green, comfort pads in blue, and chinstrap in purple), averaged for all impact directions at three impact velocities.

Full size image

For the same impact velocity, the strain energy for each component group was dependent on the impact location. Facemask impacts (Fig. 5a) were primarily absorbed by the strap system while the helmet shell dissipated much of the impact energy when impacted directly at locations distant from the facemask (lateral, rear, and rear oblique) (Fig. 5b). The compression shocks absorbed much of the impact energy under rear oblique, low side, and front top impact configurations (Fig. 5c) that also engaged edges of the facemask. The magnitude of the strain energy in each component group for all impact directions and velocities is summarized in the Supplementary Material (Table S7).

Figure 5
figure 5

Strain energy distribution (helmet shell in red, shocks in black, facemask in green, comfort pads in blue, and chinstrap in purple) for three different impact directions at 5.5 m/s.

Full size image

For the same impact direction, the strain energy of each component group changed with increasing impact velocity, which was particularly pronounced in impacts involving the facemask (Fig. 6a). In the oblique facemask impact at 5.5 m/s, the strap system shared 29% of the total strain energy (Fig. 6b); however, at 9.3 m/s strain energy shared by the strap system reached only 12%, falling behind the facemask (33%) and compression shocks (23%) (Fig. 6c). In the facemask lateral impact, the facemask largely dominated the total strain energy distribution (43%), followed by the compression shocks (19%). At lower impact speeds (5.5 m/s) the energy contribution was more equally distributed between the shocks, helmet shell and facemask (21%, 22% and 30% respectively).

Figure 6
figure 6

Strain energy distribution in component groups for two impact velocities, compared to the average for all impact velocities (helmet shell in red, shocks in black, facemask in green, comfort pads in blue, and chinstrap in purple).

Full size image

Discussion

This study summarized the development and validation of a detailed numerical representation of a modern American football helmet (X2E, 2016, Xenith, LLC). Components of the helmet were previously validated at a material and component level 1,2 and successfully assembled into a full helmet finite element model using pre-simulation methods to position the deformable internal components (foam, compression shocks) within the helmet shell. The full helmet assembly was donned on the NOCSAE and HIII headforms using a novel headform expansion technique to fit the helmet to the head. Lastly, a strap pre-tensioning method was applied to tighten the strap system and fit the helmet to the headform at the start of the simulations, ensuring fit and initial stresses were achieved in the deformable components prior to the impact simulation. Experimental tests of the Xenith X2E helmet were modelled using two validated headforms and impactor models11 and sixty impact configurations were used to validate the assembled full helmet model. A range of impact locations and speeds provided a good assessment of the model performance objectively compared to experimental data, with cross-correlation ratings from good to excellent among the four impact configurations. The lower ratings obtained in the rotational acceleration in the pendulum and linear impact were attributed to the strap-headform interaction, where initial strap tension was not reported in the experimental data.

The helmet component strain energy distribution resulting from impact simulations demonstrated a complicated relationship, depending on both impact location and speed. The helmet shell was found to have the highest strain energy, while the internal deformable structures demonstrated a positive correlation between strain energy and the head injury criterion. The maximum strain energy of each component was positively correlated with HIC15 (Fig. 3) for varying impact locations and velocities, where the comfort pads were associated with an increase in HIC15 value with increasing strain energy. The increase of HIC15 with the strain energy in the comfort pads relative to other components was attributed to the consolidation of the comfort pads for all impact conditions, despite accounting for only 15% of the shared strain energy overall. In contrast, due to the higher energy required to reach consolidation and engagement directly or indirectly through the helmet shell, the compression shocks (strongest correlation between HIC15 and strain energy) accounted for 22% of the overall strain energy for all impact conditions. The correlation of BrIC with increased maximum strain energy was modest when compared to the correlation to HIC15, suggesting a stronger correlation of component deformation to linear kinematics than to rotational kinematics where helmet fitting and impact location could play a substantial role. The low R2 values suggest a complex relationship between the deformation of each component and the outcome as quantified by HIC15 and BrIC. The strain energy share of each component was demonstrated to be dependent on impact velocity (Fig. 4) and impact direction (Fig. 5). Helmet components that were not included in past literature investigations, such as chin cup and straps5,14 were found to share a major portion of the strain energy (up to 30%) in impacts engaging the facemask. While an essential contribution of the straps was demonstrated, no information on strap system pre-tension, which could vary significantly depending on the player preference, was available. This factor could influence the energy sharing distribution and requires further investigation.

Limitations of the model included a simplified representation of the straps at locations that were not in contact with the headform (1D beam elements rather than having a 2D elements representation in the entire helmet). The strap system simplifications were based on a common methodology of guiding seatbelts through anchoring points for automotive applications. The sensitivity of strap pretension was not investigated and can potentially affect the deformation of the components as well as head kinematics. Data regarding in-game strap tension is needed for future studies. Also, variations in helmet position on the head were not investigated. In each impact case, the helmet was positioned as reported in the experiments, but variations in orientation could affect the helmet performance. Lastly, the helmet material properties were measured at room temperature, therefore, the effect of temperature effect on material properties and helmet response was not investigated in the present study.

It was demonstrated that while an energy dissipating structure, such as helmet shell, may dominate the overall model response (Fig. 4), other structures become equally or more critical under different impact directions (Fig. 5). The understanding of how the different components contribute in different impact conditions can inform further development of position-specific helmets since it has been demonstrated that players at different positions were likely to experience impacts from different sources and to different head regions during the game.36 In addition, strain energy shared between the components was impact velocity dependent, highlighting the importance of assessing helmet performance for different impact velocities, which further advances findings of past studies performed at a limited range of impact velocities. The distribution of energy could inform future developments of position-specific helmet designs considering the impact location dependency, and future research should consider optimizing helmet performance for varying impact velocity and location.