Occupant Injury and Response on Oblique-Facing Aircraft Seats: A Computational Study

Aviation safety standards are regulated by the Federal Aviation Administration (FAA) and are aimed toward ensuring aircraft passenger protection during survivable crash scenarios. These standards stipulate the static strength and dynamic evaluation requirements for seats, restraint systems, and other components with mass influencing occupant protection [1,2]. The dynamic tests include two basic tests: Test 1 is primarily a vertical test with a minimum impact velocity of 10.67 m/s with peak acceleration of 14 g and an impact angle of 30 deg of pitch with respect to the vertical axis. Test 2 is a horizontal test with a minimum impact velocity of 13.41 m/s with peak acceleration of 16 g and an impact angle of 10 deg of yaw. These standards were primarily formulated for forward-facing seats (up to 18 deg of yaw to aircraft centerline). In recent decades, airlines have been transitioning their wide-body first and business class seats from the traditional side-by-side parallel row seating to an oblique seating configuration to increase comfort and space efficiency [3–5]. In the oblique configuration, the seats are mounted at an angle relative to the aircraft centerline (18–45 deg). It is anticipated that the oblique seats already pass the vertical test since the loading vector remains the same irrespective of varied yaw angles. However, it is unknown as to how the obliquely installed seats will perform under horizontal loading since the occupant is loaded in a different direction.

In automotive environments, oblique injuries are caused by the intruding door or striking vehicle. Injury biomechanics research efforts in oblique side impacts have typically mimicked this boundary condition by impacting the postmortem human surrogate (PMHS) with a padded or rigid load wall using sled equipment or a pendulum device. The primary focus for those studies was defining injuries to the thorax, abdomen, and pelvic body regions [6,7]. This type of loading mode is uncommon in aviation environments. However, occupant loading in far-side automotive crashes is generally longer in duration than near-side and, hence, more similar to an aircraft emergency landing pulse, but the delta V is lower. Furthermore, with regards to the postcrash event, in an aircraft crash, occupants must be able to escape the cabin within 90 s with little to no assistance from first responders due to the potential for and propagation of postcrash fires [8]. This might put the uninjured occupant in greater risk of injury if the occupants' impairment delays evacuation. Conversely, self-extrication from the vehicle is not generally required in a motor vehicle crash event. Thus, a higher crash pulse and postevent requirements, make the aircraft crash event a unique crash scenario. Occupants on an oblique aircraft seat in such scenario are anticipated to sustain a unique injury pattern.

Postmortem human surrogate sled tests were conducted to assess gross occupant kinematics and evaluate potential injuries for oblique aircraft seat occupants subjected to horizontal loading [9]. The tests reported severe injuries throughout the body. The reported transection injury at the lumbosacral junction and dislocation injuries at the lower cervical levels were of particular concern because they were considered to be incapacitating injuries, limiting the occupant capacity to self-evacuate from the aircraft in an emergency situation. Considering the influence of such spinal injuries on the occupant's postcrash mobility makes these failures a priority which needs to be examined in order to develop and/or improve the existing safety features in the aircraft. Although, the previous sled tests shed light on the underlying mechanisms for spinal distraction injury, only one specimen per condition was tested. Furthermore, the segmental level loads and moments were not available from this test series.

Typically, to understand the orthopedic response and injury mechanisms for a given external insult, it is recommended to test multiple specimens for a given boundary condition in order to provide a corridor with a mean and standard deviation instead of just one data point from a single test. Additional tests were suggested to investigate the influence of different input parameters, such as pulse intensity and seat orientation relative to the aircraft. These parameters are anticipated to influence loads and thus spine injury. Component-level spinal tests are an alternative approach; however, such tests do not include coupling and injuries to adjacent regions or realistic torso mass. A whole-body test provides a response to all orthopedic regions in a single test. However, PMHS cadaveric tests are expensive considering time and labor. As a human surrogate, human body models (HBM) finite element (FE) can overcome the limitations of physical tests. A validated model should have good biofidelity and can be used for parametric studies, particularly the effect of certain input variables can be estimated which would assist in defining future loadings for physical testing. In addition, countermeasures can be evaluated with the validated models. Therefore, to better understand occupant kinematics and spinal loads associated in oblique loading scenario, a computational modeling approach was considered in the current study. The primary focus was directed toward validating the injuries and responses occurring to the cervical and lumbar spinal segments.

The input boundary conditions for the present computational study were obtained from the whole-body PMHS oblique sled test [9]. The HBM was simulated for two PMHS sled tests: condition A: without armrest and condition B: with armrest. Both conditions were completed with the occupant seat orientated 45 deg to the loading vector. The measured sled acceleration pulse served as an input for the simulation. The achieved acceleration had a peak of 16 g and a duration of 145 ms, which meet the requirements in the 14 CFR 25.562 [10] (Fig. 1).

In the cited physical test, specimen ages averaged 60 yr. In order to have an age-specific response, a previously developed elderly male Global Human Body Models Consortium (GHBMC) 50th percentile model (v4.5) was used [11]. This FE model was originally developed for automotive applications. The 50th percentile 26-year-old base model was morphed, and the material properties were updated to account for aging. For the selected HBM, the lumbar spine geometry and material properties of the v4.5 HBM model are detailed elsewhere [12]. The lumbar segments are connected using one-dimensional beam element at the center and two-dimensional shell element at the circumference. The material properties for all the spinal segments were scaled to account for the selected age [11], which were referenced from strength of biological tissue [13]. The disks are typically meant to play a role in load transmission between the vertebrae and act as a shock absorber in an axial loading event. The cited PMHS oblique sled tests suggest that the multi-axial bending and axial tension along the spinal column would be the potential injury mechanisms for the observed spinal failure. The selected HBM's spinal column is capable of measuring the bending and tension loads. Therefore, the selected elderly model was appropriate for the present study. The HBM predicted responses should be considered as conservative responses, which is anticipated to provide better accuracy and will serve in designing future PMHS tests. The HBM was validated for a frontal impact with an average overall correlation rating of 0.62. For the present study, the selected HBM was validated against the PMHS test measured orthopedic responses. Additionally, component simulations on the thorax, abdomen, and pelvis were performed and resulted in a good correlation against the physical test [11]. Furthermore, in the present study, the selected HBM's performance to oblique loading was verified against the oblique PMHS sled tests based on the correlation rating technique.

The selected HBM had an occupant posture simulating the position of a driver with a knee angle at 120 deg. A morphing technique was then used to achieve the test reported knee angle of 90 deg using ANSA morphing technique. The morph boxes were defined for each leg and a rotation operation backward around the knee joint was performed to achieve the desired knee angle. The repositioned HBM simulation setup for oblique impact was referenced from the PMHS sled test [9]. Of note, the physical test described boundary conditions was considered a worst-case loading scenario.

The HBM simulation setup for conditions A and B are shown in Figs. 2(a) and 2(b), respectively. The geometry of the seat fixture model was obtained from a computer aided design (CAD) drawing. The seat fixture FE model included a seat bottom, footrest, seat back bar, and anchors. The rigid fixture was constructed using shell elements and assigned steel material properties. The cushion for the seat was constructed using hexahedral solid elements with material properties defined using MAT 57 LOW DENSITY FOAM with Young's modulus of 4 MPa and a density of 73 kg/m³ to represent the cushion used in the physical test (DAX 47). The cushion FE model was attached to the seat bottom with tied constraints. The stress–strain curve for the cushion model was obtained from an FAA-Madymo computational study [14]. To replicate the effects of gravity, a simulation was performed to settle the model into the cushion. The deformed HBM was then restrained using a generic two-point belt system with anchor points selected to match the experimental setup. The GHBMC pelvis was restrained using a dual belt system, comprised of a traditional lap-belt and a body-center belt. Additionally, the lower legs were restrained at the knee and ankle using a two-point belt system. Shell elements were used to represent the belt with an element size of 10 mm. The pre-impact model posture for condition A is shown in Fig. 2(a). In condition B, the pelvis was restrained using a standard lap belt and an armrest, which was mounted to the sled on the right side of the HBM (Fig. 2(b)). The dimension required for armrest model development was acquired from the CAD drawing. The armrest FE model was constructed in two parts with a rigid plate comprised of shell elements and paper honeycomb padding constructed with solid elements. The honeycomb elements were assigned material properties corresponding to 206 kPa (30 psi) paper honeycomb with a thickness of 25 mm as recorded in the experiments. The stiffness properties for the armrest was obtained from previous pad performance studies conducted for side impact sled tests [15]. Contact interaction was assigned between the model and the environment. The crash pulse (Fig. 1) from the experiment was directly applied to the seat for both simulations using the *BOUNDARY _PRESCRIBED_MOTION ls-dyna keyword. All the simulations were performed using ls-dynamppversion 8.0 single precision solver on a linuxrhel 5.4 computational cluster using 64 cores. Each simulation was run for 200 ms of simulation time and took approximately 40 h of execution time. The principal directions and polarities followed in the current study are the same as those of the SAE J211 sign conventions [16].

HBM biofidelity was based on the resultant accelerations and angular velocities measured at the head, T1, T12, and sacrum regions, obtained from the whole-body PMHS oblique sled tests. Correlation and analysis (CORA) were used to quantify the goodness-of-fit between the simulation and experimental responses. The rating in CORA ranges from 0 (no correlation) to 1 (perfect correlation) and is the average of two techniques, the corridor and cross-correlation methods. A detailed description of each technique can be found in the literature [17]. The corridor method calculates the correlation between the responses using user-defined corridors, while the cross-correlation method quantifies the progression, time-shift, and size of the responses, using three respective subtechniques. The weighted sum of the three subtechniques constitutes the rating for the correlation method, and the equally weighted sum of the corridor and correlation methods provides the overall score. The calculated score is then compared to the original ISO TR 9790 biofidelity rating system (ISO, 1999), shown in Table 3. The user parameters for the CORA score evaluations are listed in Table 4.

The whole-body PMHS sled test reported multiple injuries throughout the body for both test conditions. The injuries to the cervical and lumbar spinal column were of particular interest for the present computational study. Of those, replicating the test-reported lumbosacral transection failure was of key interest because during an emergency evacuation in an actual aircraft crash, such injuries would limit occupant egress from the aircraft. Replicating the test-reported L5–S1 transection injury (condition A) required a failing joint between the L5 and S1 segment. However, the elderly GHBMC model lacks intervertebral disks between lumbar segments as well as those between the T12–L1 and L5–S1 junctions. Instead, vertebral segments were connected using a one-dimensional beam and two-dimensional shell elements. To validate the test-reported transection spinal injuries, the injury threshold obtained from a pure tensile loading study by Pintar et al. [18] on isolated spinal segments was used in the current computational study. Those isolated spine tests determined that injury threshold ranged from 7.6 mm to 8.9 mm for a lumbar spine transection injury [18]. In the model, the distance between the L5 and S1 vertebral body was measured at its anterior, posterior, and lateral aspects. The maximum distance measured was then compared against the literature-reported lower range (7.6 mm) of spinal transection injury threshold.

The dislocation injuries were reported in the lower cervical spine regions for both test conditions (A and B). In order to simulate intervertebral disk avulsion in the cervical spine, the HBM default setting had a failure tensile stress of 30 MPa assigned to the interface between the endplate cartilage and annulus fibrosis [19]. However, the failure control to simulate avulsion injury was not included in the original HBM, in order to ensure numerical stability. Regarding the current modeling study, this was considered appropriate because the primary focus was to replicate the L5–S1 transection injury. Thus, the HBM simulation was completed with the stated settings. With the available modeling outcome, the maximum principal stress was measured on both the upper and lower endplates of the cervical segments (C2–C7). The measured stresses were then compared against the above stated cervical spine avulsion threshold. These approaches resulted in replication of L5–S1 injury and the prediction of lower cervical spine injuries.

Additionally, in the physical test the surrogate also sustained failure at the sacral ala in condition A and multiple rib fractures (conditions A and B). In the selected HBM, the bony materials were described using an elastic–plastic constitutive law, i.e., MAT 81 (MAT_PLASTICITY_WITH DAMAGE) in ls-dyna. The failure criterion for bones was assumed to be based on strain. When maximum plastic strain exceeds the failure threshold, the elements involved fail and are deleted from the model to simulate bone fracture. The failure strain for orthopedic regions were scaled for elderly HBM as per the literature [13].

The validated HBM was then subjected to parametric simulations to examine the influence of peak acceleration on spinal injury occurrence and to predict the spinal loads relative to seat orientation. All the parametric simulations were performed with the condition A boundary setup. For the pulse intensity simulation series, the HBM withstood six input pulses which varied only in peak magnitude over same time duration. The pulse levels considered were: 50%, 60%, 70%, 75%, 80%, and 90% of a standard 16 g pulse (Fig. 3(a)). This corresponds to 8 g, 10 g, 11.2 g, 12 g, 12.8 g, and 14.2 g, respectively. For the seat orientation simulation series, the seat angle relative to the impact vector was varied from 0 deg to 90 deg, with 15 deg increments, corresponding to 0 deg, 15 deg, 30 deg, 60 deg, 75 deg, and 90 deg (Fig. 3(b)). The second simulation series were completed with 16 g pulse. Combinations of varied parameters (for illustration: 50% at 60 deg) were not considered in the present study. Because the physical data considered for validation were obtained from a single specimen per test condition. Furthermore, in each of the stated parametric study the simulation with 16 g and 45 deg served as a control (base simulation); since it is validated against the PMHS study. Whereas other combinations do not have a base simulation to compare the results.

Initially, simulations were completed with the elderly HBM for two oblique loading conditions corresponding to the tests completed with PMHS. The predicted linear acceleration and angular rate responses were extracted from the nodes on the left aspect of the skull, closer to the Frankfurt plane. For the spinal segments T1, T12, and S1, the HBM predicted responses were extracted from the nodes on the posterior element of the spinous process. These nodal regions on HBM were chosen because they were the regions where the responses were acquired in the actual PMHS sled test. The selected HBM biofidelity for oblique loading was quantified using a CORA v3.6 rating scale. The correlation rating for resultant acceleration and angular rate data for conditions A and B are listed in Table 1. Additionally, the seatbelt and armrest loads were extracted from the simulations and were correlated against the corresponding loads measured from the sled tests. The overall CORA ratings for condition A ranged from 0.55 to 0.70 and 0.53 to 0.72 for condition B. According to biofidelity rating the predicted correlation score falls in the fair to good category. Trajectories of the PMHS and HBM are also provided in Figs. 4(a) and 4(b).

In the actual PMHS test, the specimens sustained multiple injuries throughout the body. Table 2 summarizes the PMHS reported injuries and corresponding failures predicted by the HBM for oblique loading. In both conditions, the injuries were mainly concentrated on the spine and rib cage.

The surrogates in both test conditions sustained dislocation injuries at lower cervical levels. The HBM-predicted maximum stress for levels C3–C4 to C6–C7 is shown in Fig. 5(a). For each cervical spine level, the stress was measured at both the superior and inferior endplates. The highest stress magnitude was considered for the maximum stress for that level. For condition A, the test subject reported injury at the C5–C6 level. The corresponding HBM simulation predicted maximum principal stresses were above the injury threshold (30 MPa) at the C5–C6 and C6–C7 levels. However, the highest stress of 37 MPa was estimated at the C5–C6 level. For condition B, the cervical spine injury shifted to the C6–C7 level. For this condition also, the HBM-predicted stresses were above the injury threshold at the lower cervical levels (C5–C6 and C6–C7). However, the highest stress of 40 MPa was estimated at the C6–C7 level. Additionally, in both the simulations, the anterior longitudinal ligaments failed at the lower cervical levels (C5–C6), which further reinforce the failure occurrence of the spine at these levels. The spinal ligament failure was based on strain-based criterion [19]. Therefore, the stress-based failure of the endplate and strain-based failure of the ligaments suggests the occurrence of a dislocation injury at the lower cervical levels. The predicted failure location coincided with the PMHS reported cervical spine injury region. As observed from Fig. 5(a), the maximum principal stresses measured at the upper cervical levels (C3–C5) were below the injury threshold. Furthermore, the spinal ligaments at the upper cervical levels were observed to be intact. Overall, the FE model demonstrated the vulnerability of the lower cervical spine to oblique loading.

In addition to cervical spine injury, the PMHS test for condition A also sustained transection injury at the L5–S1 level. The corresponding HBM simulation predicted maximum axial displacement of 8.60 mm at the L5–S1 level, which is greater than the literature reported distraction injury threshold of 7.60 mm [18]. The HBM predicted maximum displacement for levels T12–L1 to L5–S1 are shown in Fig. 5(b). As noted, the maximum displacement for all levels except L5–S1 were below the injury threshold range in the condition A simulation. This prediction by the HBM was consistent with the actual test for condition A. In condition B, the HBM predicted maximum displacements across all the lumbar spine levels were below the injury threshold. This HBM prediction demonstrates that the addition of an armrest could reduce the possibility of the lower lumbar spine injury and was consistent with the PMHS reported injury summary.

In addition to spinal injuries, the specimen tested under condition A also sustained a fracture at the left sacral ala and illum. Similarly, the corresponding HBM simulation predicted element failure at the left sacral ala region. The HBM-estimated failure location was consistent with the PMHS reported pelvis injury location (Fig. 5(c)). The HBM predicted no failure in the sacrum and pelvis region for the condition B simulation, which was consistent with the actual test.

Other than spinal and pelvis injuries, the specimens in both conditions sustained multiple rib fractures, chiefly on the right side. The PMHS reported rib fracture locations and the HBM estimated rib failures are shown in Figs. 5(d) and 5(e) for conditions A and B, respectively. The HBM predicted rib fracture locations were slightly anterior to the PMHS location and were primarily focused on the lower rib levels. However, the HBM estimated number of rib fracture were in the range with the actual test reported rib failure numbers. The PMHS reported failures were based on one specimen per test condition. Thus, the present study demonstrates the ability of the HBM to reasonably replicate the failure for a given loading condition, additional PMHS tests are warranted to estimate an actual rib failure range. The HBM simulation for condition A demonstrated that rib failure potentially occurred due to the impact of the torso over the upper leg during the event. In contrast, in condition B, the predicted rib failure occurred due to the impact between the torso and the armrest. These observations were very similar to the actual tests.

The HBM validated for condition A was further simulated under six different pulse levels. The maximum stresses between the cervical segments for all seven (six parametric runs plus condition A simulation) pulse levels were predicted and are summarized in Fig. 6(a). The HBM predicted maximum principal stresses were below the injury threshold for peak magnitude ≤9.6 g (60% of 16 g pulse), while the HBM simulations with a peak magnitude ≥11.2 g (70% of 16 g pulse) estimated stresses greater than the injury threshold for lower cervical levels (C5–C7).

Likewise, the HBM predicted no sacrum failure for pulse severities ≤11.2 g (70% of 16 g pulse), while the HBM simulations with peak magnitude ≥12 g (75% of 16 g pulse) predicted sacrum failure at the S1 level. A similar trend was also observed with L5–S1 transection injury. The bar chart in Fig. 6(b) shows the maximum intervertebral distance between the segments for all seven pulse levels. As observed, the model simulations with a peak magnitude ≤11.2 g (70% of 16 g pulse) estimated maximum transection was below the injury threshold range, while the model simulations with a peak magnitude ≥12 g (75% of 16 g pulse) estimated transection was above the injury threshold.

The second parametric series was completed with the HBM validated for the condition A framework. The analysis aimed to explore the relationship between the seat orientation and the load measured at the L5–S1 and C7–T1 spine levels. The HBM-measured peak moments and forces at the C7–T1 level is shown in Fig. 7(a). The moments predicted at the C7–T1 level varied in relation to seat orientation. The HBM measured fore-aft moments at C7–T1 were observed to follow a cubic trend with a gradual increase in magnitude from 0 deg to 30 deg and a steep decline from 30 deg to 90 deg, with 90 deg having the lowest magnitude. The lateral moments at C7–T1 displayed a quadratic trend with a steady increase in magnitude from 0 deg to 60 deg and approximately constant peak from 60 deg to 90 deg. Likewise, the axial load at C7–T1 exhibited a quadratic trend with a rise in magnitude from 0 deg to 45 deg and a decline from 45 deg to 90 deg.

The HBM-measured peak L5–S1 moments and forces are shown in Fig. 7(b). The HBM measured fore-aft moments were observed to follow a cubic trend with a gradual increase in magnitude from 0 deg to 45 deg and a steep decline from 45 deg to 90 deg, with 90 deg having the lowest magnitude. However, lateral moments and tensile loads exhibited a linear trend (R² > 0.9) with a steady increase and decrease in the magnitude as seat angle changed from 0 deg to 90 deg. All simulations demonstrated spinal failure at the cervical and lumbar levels. However, the contribution of the axial loads and bending moments varied relative to seat angle.

As stated in the introduction, age specific HBM was selected in the present computational study to provide a conservative response outcome. The test subject used in the cited sled test study had an average age of 60 yr. The selected elderly male GHBMC model had a BMI of 28.7. According to the anthropometric reference database released by Centers for Disease Control and Prevention (CDC), age groups 60–69 and 70–79 have a BMI value of 28.9 and 28.2, respectively. Therefore, the selected elderly anthropomorphic model falls within this range. Furthermore, the material properties for the hard and soft tissues of the HBM were assigned based on the Yamada's strength of biological materials study [11]. It was observed that tissue stiffness decreases with age; however, the biomechanical response for tissue under a given loading scenario seems to reach a constant magnitude after the age of 60. For illustration, the tensile strength of the cortical bone and thorax soft tissues seem to have constant ultimate strength after the age of 60 yr [13,20–22]. Therefore, it is appropriate to state that the selected elderly model represents an older population ranging from 60 yr to 80 yr of age.

The physical test sustained multiple injuries throughout the body, particularly in the spine and rib cage. To predict injury in crash environments via virtual testing, it is important that accurate injury criteria be defined in the model to estimate the injury of interest. Under oblique forward loading, tension and bending in the spinal column are expected to generate high local strains in the bony structures such as the vertebrae and sacrum, possibly leading to fracture. In the selected HBM, orthopedic failure was defined by a strain-based criterion. Element erosion technique which is widely used in the modeling of traffic and military injuries [12,23,24] was considered in the present study for simulating bone fracture. The desired orthopedic region was considered failed, if the model predicted element deletion site coincided with the test-reported fracture location.

As stated in the method section, the stress and change in displacement failure threshold derived from the isolated component testing in pure loading modes were selected as criteria to replicate test reported spinal failure in the HBM. This selection was considered appropriate because the PMHS and HBM kinematics display occupant's torso going in tension followed by anterolateral bending in the oblique loading scenario. This observation was also noted at the component level of the spine. The reported spinal injury pattern exhibited a distraction type injury, further supporting the failure criteria selected. Basic structural mechanics states that the failure of a material depends on internal responses such as stresses and strains. Therefore, in the present study change in displacement was used as a potential failure criterion to estimate L5–S1 transection injury. While the stress-based criterion included in the original HBM v4.5 was used as a potential failure threshold for estimating lower cervical spine dislocation injury.

For simulating test reported lumbar spine transection injury, maximum intervertebral displacement was estimated and compared against the literature reported injury threshold of 7.6 mm. The spinal kinematics in the present study showed that the maximum distraction was predicted to occur at the posterior and left lateral aspect of the spine. In order to support our selected injury threshold value, it was beneficial to compare against the spinal ligament (particularly the interspinous ligament) failure criteria. Previous study examining the tensile strength of spinal ligaments reported deflection and load associated with failure for varied spinal ligaments [25]. The failure values associated with interspinous ligaments at the L5–S1 level was reported to be a deflection of 7.4±3. This further reinforce the threshold criteria used in the present study to predict lumbosacral failure.

Intervertebral displacement-based failure criteria have been used previously in defining injury in spine modeling studies, particularly in whiplash or rear-impact cervical spine FE models [26]. A recent literature review on the importance of the intervertebral displacement as potential criteria for spinal injuries [27] listed multiple scenarios where the model was validated for gross kinematics but failed to replicate segmental level motions. Such models are only as representative as the experimental data used for their validation and as strong as the agreement of their simulations with such experimental data. This further supports our implementation of intervertebral displacement as a potential failure criterion for replicating test reported lumbar spine transection injury.

Though the delta V is measured to be lower in motor vehicle crashes, the side console in the automobile and lack of structures in the passenger space allow occupant kinematics to resemble oblique-facing occupants in an aircraft. Previously, far-side impact tests at 90 deg and 60 deg have been reported [28]. All were completed with a center console and a leg plate to restrain the hip and the lower extremities of the test subjects. The head excursions predicted by the HBM with armrest condition was similar to that reported in the 60 deg far-side test. Furthermore, the HBM simulation demonstrated that head excursion was greatest without the armrest condition, suggesting that the presence of an armrest limits head motion. The HBM-predicted rib failures with armrest condition correspond to the fracture pattern observed in the oblique far-side study [28]. Similar far-side lateral and oblique impact tests were completed with a delta V of 9.4 m/s [29]. The head excursion and torso rotation reported with 60 deg oblique tests were similar to the HBM-predicted kinematics with armrest, condition B. Additionally, the oblique test in the cited study reported rib fractures as well as C6–C7 joint dislocations; likewise, present HBM simulations showed similar failure patterns.

The PMHS in the cited automotive oblique/lateral impact studies were restrained with a three-point belt system. The head excursion was reported to be higher in a far-side event than a pure lateral impact but was less than the present study estimated head displacement. The potential reason would attribute to higher aircraft crash pulse and lack of shoulder belt in the aircraft seats. The shoulder belt was reported to restrain the occupant to the seat, while in the present study, the HBM kinematics exhibited unconstrainted motion of the torso, resulting in an impact with the armrest or lower extremities during the event. Both the cited far side and present studies reported multiple rib fractures. However, the reason for the fracture varied between the events due to varied loading mechanisms as stated above.

The first parametric analysis was aimed in defining the relationship between the input pulse intensity and the injury occurrence to the spine. The injury summary from PMHS oblique impact tests at 16 g and 9.6 g have been reported [8]. To note the test with 9.6 g in the cited study didn't report the kinematic and orthopedic response, hence the data was not available for validation. The specimen tested with 9.6 g reported no spinal injuries, consistent with the present modeling study. Additionally, the HBM in the present study demonstrated that the HBM simulation with a peak magnitude of 11.2 g produced failure at the lower cervical spine but failure occurrence was mitigated at the lower lumbar spine levels. A previous PMHS–ATD matched paired study [30] recommended 9.6 g as the threshold for injury occurrence for the lower lumbar spine. However, the present modeling effort showed that even with a peak magnitude of 11.2 g, the impact might not produce lower lumbar spine failure.

The second parametric analysis was aimed in investigating the influence of seat orientation on the spine loads. Regardless of the seat angle, the HBM estimated failure at the lower cervical and lumbar spine. This observation leads to the conclusion that pulse intensity would be a potential predictor for spinal failure, since all the simulations for seat orientation parametric analysis were performed with a 16 g pulse. Using head-neck isolated tests, tension-based lower cervical spine injuries were estimated to occur at the force ranging from 800 N to 1900 N [31]. A similar study using ten isolated lower cervical (C4–T1) segments estimated an axial force ranging from 427 N to 587 N would disrupt the disks and produce distraction injuries [18]. The parametric simulations predicted tensile force for orientations (0–45 deg) were >800 N, while they were 776 N, 635 N, and 476 N for orientations 60 deg, 75 deg, and 90 deg, respectively. The predicted loads for orientations 0 deg to 45 deg were in the threshold range of the Yoganandan et al. study. However, the HBM-predicted loads were above the Pintar et al. study for all orientations. Thus, the data supports the occurrence of dislocation-related cervical spine failure for all the seat orientations. The unconstrained nature of the head-neck component in the present study resulted in varied peak flexion and lateral bending moments. Results from the isolated head-neck complex for lateral impact have been reported [32]. A bending moment of 75 N·m was determined to be the lateral moment injury threshold under side impact. The predicted lateral moments for orientations (45–90 deg) were >87 N·m, above the cited literature-estimated failure threshold, while the predicted lateral moments for orientations (0–30 deg) were <60 N·m. Nightingale et al. evaluated the structural properties of the lower cervical segments under flexion and extension and found that in flexion, the mean moment at failure was estimated to be 20±3 N·m [33]. The HBM predicted peak flexion moments were significantly higher than that estimated value. It should be noted that the cited isolated studies were completed with either pure tensile force or pure bending moments, while the HBM in the present study sustained multi-axial loads. Combined metrics with compressive force and flexion have been reported for evaluating lower cervical injury [34–36]. The cited studies estimated flexion-based critical values ranging from 141 N·m to 316 N·m. The HBM predicted peak flexion for orientations (0–60 deg) was >140 N·m and was 50 N·m and 30 N·m for orientations 75 deg and 90 deg, respectively. The predicted flexion moments for orientation 0–60 deg fall between the cited range. It should be noted the cited studies were completed under compressive force and flexion moments, while the HBM was subjected to tensile force, flexion, and lateral bending moments.

An isolated lumbar spine segment study reported failure load ranging from 930 N to 1684 N for producing tension associated lumbar spine injuries [18]. In a similar component study, an average load of 3180 N was estimated to produce lumbar disk rupture under tension [37]. The parametric simulations predicted axial force at L5–S1 level was >3400 N for orientations (0–75 deg), while the axial force was 2549 N for 90 deg orientations. The predicted threshold for all orientations except 90 deg was above Sonoda's estimated mean value. Likewise, the HBM predicted axial force for all the orientations was above the Pintar et al. study. Thus, the predicted axial load supports the occurrence of lower lumbar spine injury irrespective of seat orientations. A study conducted pure flexion testing with 18 isolated lumbar spinal segments and reported moment ranging from 54 N·m to 90 N·m for rupturing the outer annulus of the disks [38]. The HBM predicted peak moments were within 20% of the lower range of the cited study for orientations (0–45 deg), while the flexion moments were significantly lower for seat angles 60–90 deg. An isolated study with 18 T3–L5 spinal segments estimated the average failure flexion moment of 153±82 N·m [39]. A similar study with 6 T12–S1 segments reported an average failure moment of 140±18 N·m [40]. The parametric modeling in the current study predicted peak flexion at the L5–S1 level were significantly lower than the cited studies for all the orientations. Thus, it can be assumed that while flexion may play a role, it may not be the main failure mechanism when spine segments are subject to combined tensile force with flexion and lateral bending loading. Though the predicted L5–S1 tensile force seems to be the dominating contributor in producing the lower lumbar spine and sacrum injuries, orientation-specific flexion and lateral bending moments influence the failure patterns. The HBM simulated with seat angles 0–15 deg predicted bilateral sacral ala failure, while the simulation with the seat angles 30–90 deg predicted failure chiefly on the left aspect of the sacrum. The HBM predicted spinal loads and the cited literature reported measurement are summarized in Table 5.

The selected elderly HBM v4.5 model has simplified lumbar and thoracic spinal segments which lack intervertebral disks. Though the selected model did not mimic the actual spinal anatomy, the HBM served the purpose of predicting the desired responses and spinal failures. However, future modeling efforts can be performed using an elderly HBM with detailed spinal segments, but we anticipate similar responses.

The present study used stress and change in displacement criteria in the HBM, which were obtained from pure tensile testing. Though the spinal kinematics likely support local tension as the primary predictor for the observed spinal injury, the contribution of the bending moments plays a role in defining the failure pattern (as demonstrated by seat orientation parametric analysis). Hence, it is anticipated that a combined criteria including tension and bending moments would be a better injury predictor for oblique loading induced spinal injury. To author's knowledge, such a criterion is not currently available.

The present modeling study was completed with two sets of boundary conditions and the HBM was validated against a single PMHS per test condition. It is recommended to validate the FE model against the response corridors generated using a minimum of three PMHS tests per condition. However, the selected HBM predicted responses had an overall CORA rating ranging from 0.5 to 0.7, which falls between the fair to good biofidelity categories.

The HBM predicted responses were in decent correlation with the cited oblique PMHS sled tests responses. The selected model was able to replicate the PMHS reported spinal and rib cage injuries. The validated HBM was subjected to parametric simulations. Results from the seven pulse level simulations indicated that the HBM could tolerate 11.2 g of the input pulse without the lower lumbar spine failure whereas the cervical spine could sustain a 9.6 g pulse without failure. This observation is anticipated to assist in identifying specific loading conditions for conducting additional tests to obtain statistically relevant data for development of injury criteria in the form of risk curves. The second parametric analysis with seven varied seat angles provided the HBM predicted cervical and lumbar spinal loads. Irrespective of the seat orientation, the HBM predicted spinal failures; however, the contribution of the tensile load and bending moments varied relative to the orientation. This preliminary understanding of the occupant response and injury occurrence with respect to input pulse and seat orientation is anticipated to assist government agencies and aircraft manufactures in formulating safety standards for oblique seats and designing safety countermeasures.

The computational work was conducted under the support of Grant No. 17-G-002, sponsored by the Federal Aviation Administration. The simulation was completed in part with computational resources and technical support provided by the Research Computing Center at the Medical College of Wisconsin. The endowment support for the primary author was from the Dr. D. Robert and Dr. Patricia E. Kern Professorship of Biomedical Engineering (FP). Any views expressed in this presentation are those of the authors and are not necessarily representative of the funding organizations.

• Federal Aviation Administration (Grant No. 17-G-002; Funder ID: 10.13039/100006282).