A Computational Biomechanical Analysis to Assess the Trade-off Between Chest Deflection and Spine Translation in Side Impact

Abstract

Objectives: The objective of this study is to evaluate how the impact energy is apportioned between chest deflection and translation of the vehicle occupant for various side impact conditions.

Methods: The Autoliv Total Human Model for Safety (modified THUMS v1.4) was subjected to localized lateral constant velocity impacts to the upper body. First, the impact tests performed on postmortem human subjects (PMHS) were replicated to evaluate THUMS biofidelity. In these tests, a 75-mm-tall flat probe impacted the thorax at 3 m/s at 3 levels (shoulder, upper chest, and mid-chest) and 3 angles (lateral, +15° posterolateral, and −15° anterolateral), for a stroke of 72 mm. Second, a parametric analysis was performed: the Autoliv THUMS response to a 250-mm impact was evaluated for varying impact levels (shoulder to mid-thorax by 50-mm increments), obliquity (0° [pure lateral] to +20° [posterior impacts] and to −20° [anterior impacts], by 5° steps), and impactor pitch (from 0 to 25° by 5° steps). A total of 139 simulations were run. The impactor force, chest deflection, spine displacement, and spine velocity were calculated for each simulation.

Results: The Autoliv THUMS biofidelity was found acceptable. Overall, the predictions from the model were in good agreement with the PMHS results. The worst ratings were observed for the anterolateral impacts. For the parametric analysis, maximum chest deflection (MCD) and maximum spine displacement (MSD) were found to consistently follow opposite trends with increasing obliquity. This trend was level dependent, with greater MCD (lower MSD) for the higher impact levels. However, the spine velocity for the 250-mm impactor stroke followed an independent trend that could not be linked to MCD or MSD. This suggests that the spine velocity, which can be used as a proxy for the thorax kinetic energy, needs to be included in the design parameters of countermeasures for side impact protection.

Conclusion: The parametric analysis reveals a trade-off between the deformation of the chest (and therefore the risk of rib fracture) and the lateral translation of the spine: reducing the maximum chest deflection comes at the cost of increasing the occupant lateral displacement. The trade-off between MCD and MSD is location dependent, which suggests that an optimum point of loading on the chest for the action of a safety system can be found.

Keywords:

• side impact
• PMHS
• human body model
• spine translation
• ribcage deformation

Introduction

Side impacts represent approximately 30% of all crashes involving severe or fatal injuries (Samaha and Elliott 2003). Compared to frontal crashes, the proportion of side impact fatalities is steadily increasing (Kahane 2007). In general, about 70% of side impacts resulting in fatal and serious injuries are near-side impacts (occupant adjacent to the intruding structure), and about 30% are far-side impacts (occupant opposite the intruding structure; Gabler et al. 2005). Injuries to near-side occupants are associated with door contact: the change in velocity and the maximum door crush affect injury severity, with a majority of serious injuries to the thorax, abdomen, and pelvis occurring in vehicles with a maximum door crush between 250 and 350 mm (Tencer et al. 2005). On the other hand, injuries to far-side occupants occur when the occupant slips out of the belt and contacts the vehicle interior structures. Such injuries are often to the head and thorax (Gabler et al. 2005). If 2 front seat passengers are present, the injury risk increases due to occupant-to-occupant interaction (Newland et al. 2008; Stigson and Kullgren 2011). Both improved vehicle structure and side airbags for the head and chest reduce fatality risk (Kahane 2014; Teoh and Lund 2011).

Side airbags protect the occupant by distributing load, dissipating impact energy, and moving the occupant away from the intruding structure. Side airbags can load the pelvis and the shoulder, but if these load paths fail, then the severity of injury to the thorax and abdomen increases (Yoganandan et al. 2013). Furthermore, extensive movement of the occupant increases the risk of injuries from contact with the interior structure of the vehicle (for example, the central console) or with another occupant. An optimal side-impact restraint system must therefore balance both load distribution and occupant displacement. This balance requires a thorough investigation of the relationship between thorax deformation and spine translation. In the current study, the apportionment of impact energy between chest deformation and translation of the spine for a vehicle occupant subjected to various localized lateral loadings was evaluated.

Method and Materials

Because this type of parametric analysis would be prohibitively expensive and time-consuming if performed with physical human surrogates (anthropomorphic test devices and postmortem human subjects [PMHS]), a computational approach was developed, and the Autoliv THUMS was used in the current study. The Autoliv THUMS was derived from the THUMS model (Total Human Body Model for Safety, version 1.4). The THUMS model was updated with a number of in-house modifications to improve its biofidelity in frontal impacts using table top and sled PMHS tests (Pipkorn and Kent 2011; Pipkorn and Mroz 2008). The rhomboids major and rhomboids minor muscles that connect the medial border of the scapula to the spine were missing in the original THUMS and added to the Autoliv THUMS. This connection between the scapula and the upper body is believed to be important in the side impact response. However, the biofidelity of Autoliv THUMS in constant velocity side impacts has not been verified yet. Therefore, the study was designed to (1) evaluate THUMS performance in side impact by replicating the impact tests performed on PMHS by Subit et al. (2010) and (2) perform a parametric analysis to determine how the human body deforms and translates in localized impacts.

Fig. 1. THUMS in its initial position. The subject coordinate system is shown, as well as the vertical direction about which the subject was rotated for the oblique impacts. The vertical direction was the upward vector in the laboratory frame. For visibility purposes, the impactor coordinate system is not shown at its origin (center of the probe impacting face) but at the side of the impactor probe.

Evaluation of Autoliv THUMS for Localized Impacts

For the evaluation of the Autoliv THUMS in side impact, the PMHS impactor tests carried out by Subit et al. (2010) were mimicked. The model was evaluated for pure lateral as well as angled impacts. These PMHS tests were selected because each PMHS was impacted several times, which reduces the effect of intersubject variability. Three PMHS were impacted at 3 levels (shoulder, upper chest, and mid-chest) and at 3 angles (lateral, +15° posterolateral, and −15° anterolateral). The impactor was a 75-mm-high and 400-mm-wide probe driven at a constant velocity of 3 m/s, throughout the interaction with the PMHS, until the target stroke was reached and the impactor was stopped. The impacts to the shoulder were performed with the arm along the side of the chest, and the upper and mid-chest impacts were performed with the shoulder joint in 90° flexion. For all impact angles, the 3 PMHS were impacted at 3 levels, defined based on external anatomical landmarks. The same definitions were used in the simulation work:

• Shoulder impact: The center of the impactor face was aligned with the head of the humerus. This position was referred to as Z = 0 in the simulation.

• Upper chest impact: The superior face of the impactor was just inferior to the posterior aspect of the upper arm. This position was 50 mm below the shoulder impact in the simulations (Z = 50).

• Mid-chest impact: The superior face of the impactor was just inferior to the inferior angle of the scapula. In the simulations, this impact level was 150 mm below the level of shoulder impact.

The time history data for the impactor force and the 3-dimensional kinematics of the head; T1, T8, and L1 vertebrae; sternum; and pelvis were output from the simulations. The Autoliv THUMS was seated on a seat plate that was at a 15° angle relative to the horizontal (Figure 1), similar to the seat used in the PMHS tests. The model was rotated 21° from the vertical plane to mimic the setup in the PMHS tests. The friction coefficient between the seat and the pelvis was 0.3. The chest deflection and spine displacement were calculated using the method outlined in Subit et al. (2010): a virtual marker was defined on the spine at the same Z-level as the impactor center point, and its trajectory was linearly interpolated from the kinematics of the adjacent vertebrae (Figure 2). The chest deflection was the displacement of the impactor relative to the virtual marker along the Y direction in the seat coordinate system, with the deflection initialized to 0 at time 0 (time of first contact between the impactor and THUMS). The spine displacement was taken as the displacement from time 0 of the virtual markers along the Y direction in the seat coordinate system. In addition, the chest stiffness was estimated using a linear model, by performing a linear regression of the impactor force versus the chest deflection curve.

Fig. 2. Impact levels and virtual markers used to calculate the chest deflection. The coordinate system shown corresponds to the seat coordinate system.

Three subjects were included in Subit et al. (2010): S1 was evaluated in pure lateral impacts (0°), S2 in posterolateral (+15°) and anterolateral (−15°) impacts, and S3 in pure lateral (0°) and anterolateral (−15°) impacts. The response of the Autoliv THUMS was compared to that of the PMHS S1 for the lateral impacts (0° angle) and S2 for the antero- and posterolateral impacts (+15° and −15° angles), because the anthropometry of these 2 subjects was closer to that of the Autoliv THUMS than the anthropometry of S3. The maximum stroke was 72 mm for both S1 and S2. For S3, the maximum stroke was only 47 mm because this subject was underweight and possessed a lower body mass index compared to S1 and S2.

The performance of the Autoliv THUMS was quantified by calculating the quantity δ defined as the normalized difference between the model prediction and the experimental results: (1) where M was the maximum chest deflection (MCD), or the maximum spine displacement (MSD; calculated between 0 and 72 mm of impactor displacement), or the chest stiffness for a stroke of 72 mm, and reference PMHS represents the value obtained for the PMHS taken as reference (S1 for 0° impacts and S2 for +15° and −15° impacts). There are, however, no threshold values for the normalized difference to assess whether the THUMS response is biofidelic. To address this issue, the normalized difference was calculated for subject S3 (for a stroke of 47 mm) for the −15° and 0° impacts (with S2 and S1 used as reference) to quantify the inter-PMHS variability and determine what threshold should be used to assess THUMS performance. Nine simulations were run with the default (or intact) model (3 levels × 3 directions; Table 1.).

Table 1. Simulation matrix for the evaluation of Autoliv THUMS

THUMS model	Level	Direction	Simulations
Intact	Shoulder (Z0) Upper chest (Z50) Mid-chest (Z150)	−15°, 0°, +15°	9
Fractured	Z0, Z50, Z100, Z150, Z200	−20°, 20°	10

Table 2. Simulation matrix for the impactor obliquity and pitch (24 cases). This matrix was applied for each of the 5 impact levels (5 × 24 = 120)

	Obliquity (°)
	−20	−15	−10	−5	0	5	10	15	20
Pitch (°)
0	y	y	y	y	y	y	y	y	y
5	y				y				y
10	y				y				y
15	y				y				y
20	y				y				y
25	y				y				y

Table 3. Evaluation results for the Autoliv THUMS vs PMHS

	Maximum chest deflection (mm or%)			Maximum spine displacement (mm or%)			Stiffness (N/mm or%)
	−15° (ant)	0°	15° (post)	−15° (ant)	0°	15° (post)	−15° (ant)	0°	15° (post)
PMHS
Shoulder	66.9^a	60.4^b	64.4^a	4.0^a	13.4^b	7.2^a	23.4^a	31.3^b	31.4^a
Upper chest	76.8	60.9	53.0	3.3	9.9	17.7	17.1	24.5	41.3
Mid-chest	67.5	62.0	61.7	0.4	6.2	7.3	17.9	19.2	28.4
THUMS
Shoulder	55.2	59.4	62.6	17.2	15.6	9.7	36.1	33.4	25.6
Upper chest	60.0	63.6	62.8	12.5	11.4	9.5	37.0	30.4	22.8
Mid-chest	62.9	66.4	63.6	9.5	8.6	8.6	27.2	22.4	19.6
Difference
Shoulder	−17	−2	−3	330	16	35	54	7	−18
Upper chest	−22	4	18	279	15	−46	116	24	−45
Mid-chest	−7	7	3	2,275	39	18	52	17	−31
Inter-PMHS variability
Shoulder	−22^c	−24^c		153^c	88^c		48^c	3^c
Upper chest	−9	−7		399	147		131	31
Mid-chest	−9	−7		−708	188		72	21

^aResults from subject S1 in Subit et al. (2010).^bResults from subject S2 in Subit et al. (2010).^cResults from subject S3 in Subit et al. (2010).

In addition, a set of simulations was run to evaluate the effect of fractured ribs on THUMS response. Indeed, rib fractures were reported in the experiments but they could not be linked to a specific impact in many instances because of the repeated tests performed on a single PMHS. In addition, the ability of the THUMS model to predict rib fractures has not been evaluated. Therefore, though all the simulations were run without allowing the ribs to fracture, a fractured THUMS model was created where ribs 1 to 10 were cut into 2 sections by deleting a row of elements in the posterior aspect of the ribs cortical shell and trabecular bone. The posterior aspect was chosen to make certain that the impactor was in contact with the fractured part of the rib to greatly alter the distribution of the load within the ribcage. This represents an extreme fracture scenario that was used to evaluate how fractured ribs alter THUMS impact response without attempting to match the fracture profiles reported in the experiments. The fractured model was evaluated for all the impact levels (0, 50, 100, 150, and 200 mm) in the extreme oblique loading (+20° posterolateral and −20° anterolateral; see next section). Ten simulations were run with the fractured Autoliv THUMS model (5 levels × 2 directions; Table 1).

Parametric Analysis to Evaluate the Deformation and Translation of the Thorax

The intact Autoliv THUMS model (without fractures) was used in a parametric analysis. Three impact parameters were varied: the impactor level, its obliquity (angle about the vertical direction; Figure 1), and its pitch (inclination about its direction of travel, y_imp in Figure 1). The simulation matrix shown in Table 2 was applied to the 5 levels of impact (0 to 200 mm by 50-mm increments; Figure 2). In the parametric analysis, the Autoliv THUMS impact response was analyzed up to a maximum impactor displacement of 250 mm. The maximum stroke of 250 mm was chosen because it was in the range of the door intrusion reported in the literature (Arbelaez 2005), while causing less than 100 mm of pelvis y displacement, based on the simulations run for the evaluation of THUMS: because simplified boundary conditions were used compared to a real production seat (rigid seat plate, no back support, no interaction with an actual seat), it was desirable to limit the lateral translation of the model. The chest deformation and the spine displacement were calculated in the same fashion as what was done for the evaluation of the PMHS. In addition, the velocity of the T8 vertebra and the displacement of the struck scapula were calculated along the Y direction in the seat coordinate system. Because the subject starts immobile, the spine velocity increases during the impact. The rate of change of spine velocity was shown to be related to injury (Cavanaugh et al. 1993). The displacement of the struck scapula was defined as the displacement of its acromion angle. In total, 120 simulations were carried out in the parametric analysis (Table 2).

Results

Evaluation of Autoliv THUMS for Localized Impacts

The responses predicted by the Autoliv THUMS were found to be qualitatively in good agreement with the responses measured in the experiments (Appendices A-1 and A-2, see online supplement). The normalized deviation was calculated for the MCD and MSD and stiffness (Table 3) to compare THUMS response to the PMHS response (third row in Table 3, “Difference”), as well as the response of S3 to that of S1 and S2 (fourth row in Table 3, “Inter-PMHS variability”). For the MCD, the difference between THUMS response and the PMHS response was within the same range as the inter-PMHS variability (difference between S3 and the subjects 1 and 2, S1 and S2). This was also the case for the stiffness. For the MSD, the difference between THUMS and PMHS was within the range of the inter-PMHS variability, expect for the anterolateral mid-chest impact. In that loading condition, the MSD was very small (0.45 mm), which explains why the normalized difference was so large. In addition, the inter-PMHS variability was found to be large for the anterolateral impacts compared to the other impact directions, which suggests that the loading mechanisms involved in this impact direction differ from that involved in the pure lateral and posterolateral impacts.

The results from the fractured Autoliv THUMS were processed using a similar approach (Appendix A-3, see online supplement). When the fractured ribs were introduced, the stiffness decreased on average by 15.7 % (standard deviation: 4.8%, range: 4.1 to 20.3%), and the MSD decreased on average by 40.1% (standard deviation: 14.2%, range: 16.6 to 61.6%), while the MCD increased by 5.3% (standard deviation: 2.3%, range: 2.8 to 10.2%). This indicates that the Autoliv THUMS predicts an increase in the deformation of the chest and a decrease in the spine displacement when fractured ribs are introduced in the chest computational model. However, the variation in THUMS response remains less than the inter-PMHS variability, which shows that a model that does not fracture can be used to replicate PMHS tests where rib fractures occurred bcause their effect of the measured impact response is less than the variation observed between PMHS.

Parametric Analysis to Evaluate the Deformation and Translation of the Thorax

The MCD and MSD were calculated for an impactor displacement between 0 and 250 mm. Various trends are visible (Figure 3):

Fig. 3. Variation of the maximum chest deflection and spinal displacement as a function of the loading conditions.

• For all impact levels, MCD and MSD vary in opposite directions.

• The variation in MCD and MSD is level dependent: for Z = 50, 100, and 150 mm, MCD and MSD have monotonous trends, with MCD increasing from the anterior to the posterior impacts, whereas for Z = 0 and 200 mm, the curves exhibit optima (around 5° of obliquity at Z = 0, and 0° of obliquity at Z = 200). There is some variation in both MCD and MSD as a function of the impactor pitch; however, the effect is not as large as that of the obliquity.

The trade-off between the deformation of the chest and the kinetic energy transmitted to the ribcage was evaluated by plotting MCD against the velocity of the spine measured at the T8 location (Figure 4, left) for all simulations. This shows a nonproportional relationship between MCD and the T8 velocity that varies with impact level and obliquity. The lower level impact (Z = 200) generated the lowest spine velocity but the greatest MCD, whereas the impacts in the mid-thorax area (Z = 50, 100) generated the largest spine velocity. For the impacts to the shoulder (Z = 0), the responses are clustered with low MCD and low T8 velocity, showing little sensitivity to the impact condition. For the impacts immediately below (Z = 50, 100, 150, 200) and within each impact level, the anterior impacts generated smaller MCD than the posterior impacts, and for the 2 lowest impact levels (Z = 150, 200), the T8 velocity increases with decreasing MCD. The displacement of the struck scapula (Figure 4, right) was found to be minimal (less than 11 mm) for the mid-thorax area (Z = 100, 150) regardless of the impact direction, and there was more variability for the impacts to the higher (Z = 0, 50) and the lower thorax (Z = 200). For the higher thorax area, the anterior impacts led to a greater scapula displacement compared to the posterior impacts. For the lower thorax (Z = 200), the scapula displacement was roughly constant for the posterior impacts, though there was a large variability for the anterior impacts.

Fig. 4. Maximum chest deflection versus T8 lateral velocity (left) and versus the maximum displacement of the struck scapula (right). Markers with white face color: anterior impacts (−20 to −5°), markers with grey face color: pure lateral (0°), markers with solid face color: posterior impacts (5° to 20°).

Discussion

Evaluation of the Autoliv THUMS Model in Localized Lateral Impacts

The response of the Autoliv THUMS model was evaluated by means of 9 PMHS tests performed with 2 PMHS. The use of repeated tests strengthens the results because the effect of intersubject variability is minimized. Ideally, several PMHS tested in all 9 impact conditions (3 levels × 3 directions) would be used to define the target response, but such a data set is not available. An important characteristic of Subit et al.'s (2010) data set is the use of fixed velocity impact rather than a fixed energy impact (which means that the velocity of the impactor does not depend on the stiffness and mass of the impacted PMHS), which makes the simulation and interpretation of these impacts straightforward. A potential drawback, however, with repeated tests is the creation of damage (e.g., rib fractures) that may accumulate throughout the test series and modify the properties of the subject. The analysis performed in Subit et al. (2010) suggests that the effect of rib fractures on the subject response may be small and hardly measurable. The simulations performed in the current study with the THUMS model with fractured ribs concur with Subit et al.'s (2010) analysis, because the variations in THUMS response due to fractured ribs were small compared to the inter-PMHS variability. Furthermore, the THUMS model with fractured ribs is an extreme case of injuries because all of the nonfloating ribs were made fractured. Interestingly, the addition of the fractured ribs increased the compliance of the chest (Appendix A-3), which contributes to bring THUMS response closer to that of the PMHS (Appendix A-1).

The greatest difference between THUMS response and the PMHS response was observed for the anterolateral impact. In this impact, the PMHS spine displacements were small, which led to very large normalized differences in terms of percentage between the PMHS and THUMS.

The THUMS chest deflection, spine translation, and spine velocity were evaluated for an impactor displacement of 72 mm. It was confirmed that no rotation of the PMHS or THUMS about the pelvis occurred during the impacts and therefore defining the chest deflection as the y displacement of the impactor relative to the spine virtual marker in the seat coordinate system was justified. For the parametric analysis, an impactor displacement of 250 mm was confirmed to create negligible rotation and therefore using the same definition for the chest deflection for the evaluation simulation and the parametric analysis was justified.

The rating scheme developed in the current study proved to provide valuable insight about inter-PMHS variability, because it allows quantification of this variability. Based on this rating scheme, the agreement between the predictions from the model and the results of the PMHS test was considered to be sufficient to use the Autoliv THUMS for the parametric analysis.

Translation and Deformation of the Thorax in Side Impact

The parametric analysis was designed to assess the apportionment between the deformation of the thorax and its translation. Current injury risk functions are available only for the deformation of the thorax and are used to assess the effectiveness of safety systems. However, the current study demonstrates that controlling the chest deformation is not sufficient to capture all of the physical phenomena involved in a side impact, because optimizing a restraint system to minimize chest deflection may increase momentum transfer to the vehicle occupant. There was no quantified information available, because experiments with PMHS do not provide enough data points and enough control of the subject anthropometry, posture, and impact conditions to generate such results. The current study relies on a computational model to identify trends that have been suspected but never shown: minimizing the deformation of the chest leads to reducing the amount of impact energy dissipated by deformation of the ribcage and increasing the kinetic energy transferred to the subject's torso. This trend was shown to be complex as the obliquity of the impactor and its pitch played an important role in the thorax kinematics. The upper thorax (Z = 0) was found to be less sensitive to the impact conditions than the other levels that were evaluated. However, the response of the model suggests that a variation of 50 mm in the impact level (Z = 0 to 50, anterior impact) can lead to an increase of up to 1 m/s of T8 velocity, while keeping MCD constant. This finding needs to be confirmed once (1) the biofidelity of THUMS in terms of its kinematics is further confirmed and (2) a back seat is included in the seat model.

To “push” the shoulder in order to off-load the thorax is a known strategy in restraint design. The need for a balanced MCD and MSD is due to 2 reasons: first, to avoid damage to the shoulder with potential increase of thoracic injuries as observed in previous studies (Yoganandan et al. 2013) and, second, to prevent the occupant from sliding out of the shoulder belt or moving too far inboard into the vehicle. If the occupant were to slip out of the belt, the risk of injuries due to contact to vehicle interior would increase, and the occupant could end up unrestrained in case of a second crash event. If the occupant head or upper body moved too far inboard in the vehicle there would also be a risk for injurious contact with an accompanying passenger. With a system designed to distribute the load according to the characteristics of the impacts at Z = 0 and Z = 50 mm, T8 would move approximately 200 mm inboard in the vehicle (Figure 4). This level of movement would be sufficient to result in the occupant slipping out of the shoulder belt or his or her head impacting an accompanying occupant. Conversely, allowing a large CLD to limit MSD, as for impacts at Z = 150 and Z = 200 mm where apportionment between MCD and MSD follows a similar trend, may be risky because the level of deflections estimated in the current study (greater than 80 mm) is likely to generate rib fractures. Based on the parametric analysis carried out in this study, the optimal load distribution is for an impact at Z = 100 mm where 40% of the impactor displacement is converted into MCD and 60% into MSD (Figure 3) in the posterior direction to minimize the spine velocity (Figure 4).

The parametric analysis results from this study quantify the apportionment between MCD and MSD for different impact levels and directions. The results can be used to guide the design of restraint systems for near-side impact protection; more important, this information can be used as one input for defining corridors for kinematic biofidelity of anthropomorphic test devices.

The analysis developed in this study leads to the following conclusions:

• The Autoliv THUMS model predictions were in good agreement with the PMHS results for the estimation of the chest deflection, because similar variability was observed between the THUMS results and the PMHS results on the one hand and between the PMHS results on the other hand; however, the kinematics response of the Autoliv THUMS model needs to be further evaluated for whole-body and shoulder kinematics in side impact by simulating other experimental tests.

• The variation in the spine velocity as a result of a lateral impact was quantified using a whole-body model and can be used as a proxy to estimate the kinetic energy transferred to the torso.

• For the 0° impactor angle, the chest stiffness increased as the impact level decreased from the shoulder to the mid-torso.

• For 250 mm of impactor stroke, which would be equivalent to the intrusion of a door in a side impact, the impactor angle that resulted in MCD and MSD depends on the impact level.

• For 250 mm of impactor stroke, the MCD increased and the MSD decreased with increasing obliquity.

Supplemental Materials

Supplemental data for this article can be accessed on the publisher's website.