Factors Influencing Occupant-To-Seat Belt Interaction in Far-Side Crashes

Abstract

Seat belt interaction with a far-side occupant’s shoulder and thorax is critical to governing excursion towards the struck-side of the vehicle in side impact. In this study, occupant-to-belt interaction was simulated using a modified MADYMO human model and finite element belts. Quasi-static tests with volunteers and dynamic sled tests with PMHS and WorldSID were used for model validation and comparison. Parameter studies were then undertaken to quantify the effect of impact direction, seat belt geometry and pretension on occupant-to-seat belt interaction.

Results suggest that lowering the D-ring and increasing pretension reduces the likelihood of the belt slipping off the shoulder. Anthropometry was also shown to influence restraint provided by the shoulder belt. Furthermore, the belt may slip off the occupant’s shoulder at impact angles greater than 40 degrees from frontal when no pretension is used. However, the addition of pretension allowed the shoulder to engage the belt in all impacts from 30 to 90 degrees.

Research attention and government regulations have traditionally focused on protecting nearside (or struck side) occupants of the vehicle, with little attention being paid to protecting far-side (or non-struck side) occupants. Research by Gabler et al. (2005a) using NASS/CDS and FARS data from 1997–2002 indicated that far-side occupants account for 43% of the AIS 3+ injured persons and 30% of the Harm in U.S. side impact crashes. Using MUARC in-depth data (MIDS) from 1993–2002, Gabler et al. (2005b) observed that far-side occupants accounted for 20% of the AIS 3+ injured persons and 24% of the Harm in Australian side impact crashes. Using English data from 1992–1996, Frampton et al. (2000) showed that far-side occupants accounted for 42% of MAIS 2 and 31% of MAIS 3+ injured persons in European side impact crashes.

The primary form of restraint for a far-side occupant is the outboard mounted three-point seat belt. However, it has been recognized that this design does not provide adequate thoracic restraint in far-side impacts. In the early 1990’s, Mackay et al. (1993) observed that of those far-side occupants sustaining AIS ≤ 2 head injuries, 35% came out of the shoulder section of the seat belt. More recently Gabler et al. (2005a) observed that head and thorax injuries accounted for over half the serious injuries sustained in far-side crashes. Furthermore, the seat belt was recorded as the injury source in 86% of AIS2+ abdominal injuries sustained in far-side crashes [Gabler et al., 2005a].

There have been few laboratory tests investigating this problem of seat belt performance in far-side impacts [Adomeit et al., 1977; Horsch, 1980; Kent et al., 2003; Törnvall et al., 2005]. The primary reason being that no regulation exists for protecting occupants in far-side crashes. As a result, no ATD has been designed specifically for far-side impacts. WorldSID has been suggested to be the best of the available ATDs [Fildes et al., 2002], however, thorough validation is yet to be seen. THOR Alpha has also been shown to mimic head trajectories from PMHS tests quite well in far-side impacts up to 30 degrees from frontal, however this too is based on a limited amount of data [Törnvall et al., 2005].

A major limitation of most ATDs is the ability to mimic the seat belt to shoulder-complex interaction. This has come primarily from the fact that ATDs are designed to work within a narrow crash configuration band. In side impacts, ATDs are largely not validated using shoulder belts. In frontal impacts, Hybrid III ATDs only have a single device in the chest to measure the effect of shoulder belt load. However, up to half the belt load gets distributed through the shoulder where no measurement device exists [Kent et al., 2003]. Törnvall et al. (2005) indicated that weaknesses in the kinematic shoulder response of the Hybrid III and THOR ATDs are possibly related to limitations in shoulder range-of-motion and the lack of human-like shoulder-complex design.

Due to the importance of this belt to shoulder interaction, this study aimed to investigate the influence of occupant, impact and restraint parameters on the lateral restraint provided to the occupant by the three-point seat belt in a far-side impact.

METHODS

This paper is separated into three studies: Study 1: Quasi-static far-side tests; Study 2: Far-side sled tests; and Study 3: Parameter study investigating impact and restraint factors. Each study is reported separately and includes details of experimental design and results.

All modeling was completed in MADYMO 6.3 using the 50^th percentile male TNO Human Facet Model (TNO, 2005). The model used was not standard as it included modifications to the shoulder region (described in Douglas et al., 2007). These modifications were made as the standard TNO Human Facet Model was not capable of mimicking the contour variation of the shoulder-complex’s boney structures. To address this, rigid ellipsoids were inserted into the shoulder region.

STUDY 1: QUASI-STATIC FAR-SIDE TESTS

TEST DESCRIPTION

These tests aimed to characterize the influence of belt geometry and pretension on the belt to shoulder-complex interaction in a lateral far-side impact (Douglas et al., 2007). This was achieved using a test rig that rotated the subject in the frontal plane, about an axis running horizontal to the ground through their thorax. When rotated 90°, the subject experienced a 1g lateral force. The subject was seated normally in a Volvo V70 seat with the belt in the drivers position. The seat X-position (fore/aft) was instrumented such that 5 positions: 0, 60, 120, 180, and 240 could be determined. These positions (measured in millimeters from most-rearward) represented 0%, 25%, 50%, 75%, and 100% forward.

Three subjects were put through the entire matrix of tests: a Hybrid III 50^th Percentile Male; a Hybrid III 50^th Percentile Male with a Spring-Spine (as seen in Boström et al., 2005); and a mid-sized male volunteer (V1) (180cm, 80kg). The Spring-spine was added to the Hybrid III ATD to allow the spine to shear, bend and elongate. A second volunteer (V2) with broader shoulders and greater chest depth was exposed only to the X=120mm, 0N pretension configuration to highlight the difference body size had on the resulting restraint. V1’s shoulder breadth was approximately 480mm, whereas V2’s was 560mm. The shoulder breadth of the human model was approximately 460mm.

The only measurement from the physical quasi-static tests was whether the seat belt slipped off the shoulder or not. Slip was deemed to have occurred if the belt slipped off the shoulder and loaded the upper arm. Five physical tests were conducted with each subject at the same configuration. As such, the percentage of times belt slip occurred for each configuration was determined.

As the model’s shoulder was not standard, shoulder anteroposterior (AP) thickness was varied to quantify the model’s sensitivity to this dimension. The standard AP thickness in Douglas et al. (2007) was 106mm, which represents a 50^th percentile male [Tilley et al., 2002]. Here, shoulder AP thickness was increased to 116mm and also decreased to 82mm to replicate dimensions for 99^th and 1^st percentile males respectively [Tilley et al., 2002].

RESULTS

Results from the 4 rearmost D-ring positions are shown (Table 1). For V1, the most-forward D-ring (X=0mm) gave the same results as X=60mm. Hence those results are not shown.

Table 1.

1g quasi-static test results. Numbers represent the proportion of time slip occurred at that configuration. Shading represents cases that match V1’s response

VOLUNTEER 1 (V1)				HYBRID III SPRING-SPINE
D-ring Pos.	0N	100–150N	200–250N	D-ring Pos.	0N	100–150N	200–250N
60	100	100	100	60	100	60	0
120	100	100	20	120	100	0	0
180	100	0	0	180	0	0	0
240	100	0	0	240	0	0	0
HUMAN MODEL				HYBRID III
D-ring Pos.	0N	125N	225N	D-ring Pos.	0N	100–150N	200–250N
60	100	100	100	60	60	0	0
120	100	100	0	120	0	0	0
180	100	0	0	180	0	0	0
240	100	0	0	240	0	0	0

Results from the volunteer tests indicated that a trend exists between moving the D-ring rearward, increasing pretension, and thus, an increased likelihood of the belt engaging the shoulder. A visual example of subjects in cases of belt slip and shoulder or thorax engagement can be seen in Figure 2.

Figure 2.

Volunteer 1 (Left), Hybrid III Spring-Spine (Middle) and Human model (Right). Cases indicative of belt slip are seen on Top, with cases engaging the shoulder or thorax on the Bottom

Results also highlighted that the model correctly predicted the binary outcomes from the mid-sized volunteer tests, in addition to the trend observed between D-ring position, pretension and belt slip. The only difference being the case of X=120mm, 200–250N, where the model was not able to predict a 20% likelihood of slip. This model was only capable of predicting 0% or 100% likelihood, as the input parameters are fixed for a given configuration. In the physical testing, some minor differences in test setup may have been present. Hence the reason for completing five tests at any one configuration.

In addition to the slip or no-slip condition, model T1 lateral (Y) displacements were plotted to quantify the effect of D-ring position and pretension on the thorax lateral displacement (Figure 3).

Figure 3.

T1 lateral displacement vs. time for cases with belt slip (Left) and those engaging the shoulder (Right)

Results indicated that the belt slipping (or not slipping) off the shoulder is a crucial factor influencing the magnitude of thorax lateral displacement. For cases where the belt slips off the shoulder, T1 displacements average 138mm, whereas when the belt engages the shoulder the average displacement is 126mm. This only equates to an average 9% reduction in T1 lateral displacement. It should be noted however that the maximum displacements for cases with slip occurred approximately 200ms earlier that those with engagement.

As previously mentioned, V2 was only tested in the X=120mm, 0N pretension case. At that configuration, no belt slip was observed for the larger occupant. Despite this only being a single configuration, it suggests that anthropometry plays a major role in governing restraint. It also implies that outboard mounted three-point belts may better restrain larger occupants in far-side impacts.

Due to this effect, it was expected that varying the model’s shoulder AP thickness would affect belt to shoulder interaction. Results suggest that this dimension is an important factor in governing restraint (Table 2). Specifically, that an increased shoulder thickness increased the likelihood of the shoulder engaging the belt and vice versa. Other factors such as shoulder breadth and chest depth may also be contributors, however this was not possible to vary with the model at hand. Hence, thus far, all these results suggest is that the model is sensitive to shoulder AP thickness. Further investigations should be carried out with more volunteers in an attempt to better quantify this relationship.

Table 2.

Sensitivity of model to shoulder AP thickness

HUMAN MODEL (1% SHOULDER)				HUMAN MODEL (99% SHOULDER))				HUMAN MODEL (50% SHOULDER)
D-ring	0N	125N	225N	D-ring	0N	125N	225N	D-ring	0N	125N	225N
60	Y	Y	Y	60	Y	Y	Y	60	Y	Y	Y
120	Y	Y	Y	120	Y	N	N	120	Y	Y	N
180	Y	N	N	180	N	N	N	180	Y	N	N
240	Y	N	N	240	N	N	N	240	Y	N	N

STUDY 2: FAR-SIDE SLED TESTS

TEST DESCRIPTION

Shoulder belt forces and head displacements from far-side sled tests conducted with unembalmed PMHS and WorldSID ATD were used as means of model validation and comparison in this first phase of impact (Tables 3 and 4). Impacts at 60° (30° from lateral) and 90° (pure lateral) were conducted at 30km/h (19mph) using a unique far-side impact buck which included, as a standard configuration, a center console and outboard three-point belt system (Figure 4) [Pintar et al. 2006]. Impact speed was chosen based on median delta-v estimations by Gabler et al. (2005b) for occupants sustaining AIS3+ injuries in far-side crashes. The effects of adjacent occupants, airbags and intrusion were not considered in these particular sled tests.

Table 3.

Sled Test Matrix

Test	Impact Direction	D-Ring Position	Pretension	Test Subjects
1	90	Forward	0N	PMHS 1, WorldSID
2	90	Middle	100N	PMHS 2, WorldSID
3	60	Middle	0N	PMHS 3, WorldSID

Table 4.

PMHS Sex, Age and Anthropometry

PMHS	Sex (M/F)	Age (Years)	Height (m)	Weight (kg)
1	F	74	1.60	70
2	M	80	1.73	67
3	M	81	1.75	70

Figure 4.

Human model in simulated far-side buck

For this study, three configurations of impact angle, seat belt geometry and pretension were investigated (Table 3). The D-ring positions were not arbitrary but meant to replicate realistic real-world belt positions at average and extreme conditions. As a realistic worst-case scenario for the shoulder and thorax escaping the belt, a 90° impact with a forward mounted D-ring (located 120mm above and 30mm rear of the shoulder) was performed. A second 90° test was conducted with a mid mounted D-ring (located 120mm above and 90mm rear of the shoulder) and 100N of pretension applied. This middle position was deemed to be an average D-ring location for a B-pillar mounted belt. A third test was conducted at 60° (middle D-ring) to investigate model behavior in angled far-side impacts.

RESULTS

For Test 1 (Forward D-ring, 90°, No pretension), all test subjects (PMHS, WorldSID and Model) slipped out of the shoulder portion of the seat belt. In all cases, the belt provided restraint via loading the thorax in the early phases of impact, however this was more prominent in the model. The belt subsequently slipped past the shoulder and got caught on the upper arm near the elbow (Belt force-time traces found in Appendix).

The resulting lateral (Y) head displacements for all test subjects were similar (within 5%) (Figure 5). In contrast to the physical test results, the Model spent in excess of 100ms at 95% of maximum displacement, whereas the PMHS and WorldSID only spent 60ms and 65 ms respectively. This was related to the human model continuing to slip and not rebound as quickly as the PMHS and WorldSID. This being partly related to the model’s slightly lower lateral velocity compared to both PMHS and WorldSID.

Figure 5.

Head Trajectories in Test 1: 90° far-side test with Forward D-Ring and No pretension. Transverse plane - Left and Coronal Plane – Right. (0 - 240ms shown)

For Test 2 (Middle D-ring, 90°, 100N pretension), the shoulder-complex of all test subjects engaged the seat belt. Both WorldSID and the Model predicted the peak magnitude of shoulder belt force very well (within 4%), however this peak was delayed 10–12ms compared to the PMHS test (Table 5).

Table 5.

Peak shoulder belt force and timing for test subjects. Test 1 – 90° Fwd D-ring 0N Pretension. Test 2 - 90° Mid D-ring 100N Pretension. Test 3 - 60° Mid D-ring 0N Pretension

	TEST 1		TEST 2		TEST 3
	Peak Force (N)	Timing (ms)	Peak Force (N)	Timing (ms)	Peak Force (N)	Timing (ms)
PMHS	2410	89	2982	96	3808	101
WorldSID	1700	99	2806	106	3030	98
Model	2629	99	2927	108	2867	96

The magnitude of head lateral displacement for all three subjects in Test 2 was within 6%, despite the Model and WorldSID reaching these maxima up to 10ms later than the PMHS (Figure 6). The Model did however predict less inferior (Z) motion than both WorldSID and the PMHS. In the AP (X) direction, WorldSID moved very little, whereas the Model and PMHS showed posterior rebound.

Figure 6.

Head Trajectories in Test 2: 90° far-side test with Middle D-Ring and 100N pretension. Transverse plane - Left and Coronal Plane – Right. (0–190ms shown)

For Test 3 (Middle D-ring, 60°, No pretension), the PMHS and WorldSID engaged the belt at the shoulder and then subsequently slipped out. The Model on the other hand did not engage the belt at the shoulder, instead hooking on the upper arm. Despite the difference in loading locations, the shoulder belt force magnitude was similar for both WorldSID and the Model, however the PMHS loading was much higher (Table 5).

As can be seen in Figure 7, despite the difference in shoulder belt loading locations, the PMHS, WorldSID and Model head coronal trajectories are similar. In the transverse plane, variations are observed between human surrogate head displacements. This observed difference is explained by the way in which all three human surrogates interacted with the shoulder belt. For the Model, the belt slipping off the shoulder allowed more anterior motion to be achieved.

Figure 7.

Head Trajectories in Test 3: 60° far-side test with Middle D-Ring and No pretension. Transverse plane - Left and Coronal Plane – Right. (0–190ms shown)

In light of these results, it appears that the human model is generally capable of mimicking events seen in the sled tests. It should be noted that these impacts only represent one PMHS for each configuration. As already mentioned in the quasi-static tests, the fact that anthropometry of the model and PMHS are not the same makes it difficult to draw definite conclusions pertaining to model biofidelity. Furthermore, since factors such as airbags, adjacent occupants and intrusion were not simulated, the head excursions reported only reflect what is seen in cases where these factors are not influential. When those factors are present, issues of model rebound after maximum excursion are of less significance.

STUDY 3: PARAMETER STUDY

TEST DESCRIPTION

Two separate parameter studies were conducted to investigate the effect of D-ring position, pretension levels and impact direction on occupant-to-seat belt interaction (Table 6). Vehicle interior geometry was identical that used in the far-side sled tests (Study 2). D-rings located in the forward and middle positions were as previously described. High and low D-rings were located 90mm rear of shoulder and 0mm and 150mm above the shoulder respectively [Pintar et al., 2006]. The low D-ring position was aimed to replicate a vehicle with a seat-mounted retractor. The high belt position was the realistic worst-case position for neck loading (if positioned inboard).

Table 6.

Parameter study matrix

	D-ring position(s)	Pretension Levels	Impact Directions
Study A	Fwd, High, Mid, Low	0N, 100N, 200N	60, 90
Study B	Middle	0N, 100N	30, 40, 50, 60, 70, 80, 90

RESULTS

Study A investigated the effect of 4 D-ring positions and 3 pretension levels on whether the belt slipped off the shoulder and the resulting head trajectory (assuming no adjacent occupants, airbags or intrusion). This was performed for both 60° and 90° impacts (Figures 8 and 9). Head displacements in the coronal (YZ) plane were focused on.

Figure 8.

Head coronal plane trajectories in 60° far-side impacts with varying D-ring positions and pretension levels (0–190ms shown)

Figure 9.

Head coronal plane trajectories in 90° far-side impacts with varying D-ring positions and pretension levels (0–190ms shown)

It was firstly observed that the belt slipped off the shoulder in all 60 and 90 degree impacts without pretension, regardless of D-ring location. When pretension was added, regardless of severity, the belt did not slip off the shoulder. Additionally, these results suggest lower or more rearward D-rings and the addition of pretension reduce lateral head excursion in both 60 and 90-degree impacts.

Study B investigated the effect of impact direction and pretension on occupant-to-belt interaction and the resulting head trajectory. Head trajectories in the transverse and coronal planes can be seen in Figures 10 and 11. Results indicated that the model’s shoulder escaped the belt at impact angles greater than 40° when no pretension was used. When pretension was used, the belt restrained the shoulder in all cases and generally reduced lateral displacement. Furthermore, these results indicate that if no pretension is applied, head lateral excursion is similar for impact angles 70° to 90°. The addition of pretension was shown to generally reduce head inferior (Z) motion as impact angle increases from 30° to 90°.

Figure 10.

Head transverse plane trajectories in 30° – 90° impacts without pretension (Left) and with 100N of pretension (Right)

Figure 11.

Head coronal plane trajectories in 30° – 90° impacts without pretension (Left) and with 100N of pretension (Right)

DISCUSSION

The primary aim of this study was to quantify some of the factors influencing occupant-to-seat belt interaction in a far-side impact. This was firstly achieved via simulating quasi-static volunteer tests, serving as preliminary model validation. The model’s dynamic performance was then compared against shoulder belt forces and head trajectories from a limited number of PMHS and WorldSID ATD tests. Subsequent to this, estimations were made regarding the influence of certain factors on the occupant’s interaction with the belt during a 30km/h far-side impact.

In the 1g quasi-static tests, the human model was able to demonstrate both of the critical findings from the volunteer tests. Firstly, thorax lateral restraint is dependent on seat belt geometry and level of pretension applied to the belt. Specifically, the seat belt is less likely to slip off the shoulder with a more rearward D-ring and increasing levels of pretension. However, modeling suggested that once the belt engages (or slips off) the shoulder, the effect of D-ring position and pretension is negligible.

Secondly, the relationship between the shoulder engaging the belt (or slipping) and seat belt geometry and pretension is highly dependent on human anthropometry. Only two volunteers were needed to demonstrate the uniqueness of humans in this sense. While the modeling only focused on a single parameter change (shoulder AP thickness), it was able to replicate the same trend as seen in the volunteer tests. Specifically, a person with greater shoulder depth is less likely to slip out of the shoulder belt.

In Parameter Study A, it was shown that without pretension in 60 and 90-degree impacts, the belt is likely to slip off the shoulder regardless of D-ring position. The addition of pretension facilitated the shoulder engaging the belt in all cases, with the effect of pretension level being minor. Furthermore, as seen in the quasi-static tests, D-ring position generally has little effect on lateral displacement once the belt has either slipped off or engaged the shoulder. Results did however suggest that a low D-ring yields lower lateral displacements.

The concept of a low D-ring decreasing lateral restraint contradicts previous work by Rains et al. (1998) who claimed that raising the D-ring reduced lateral excursion in far-side rollover tests. However Rains et al. based their conclusions on results from a Hybrid III ATD, which this paper and Douglas et al. (2007) have already shown is not capable of accurately mimicking seat belt to shoulder-complex interaction.

Whilst lowering a D-ring and applying pretension seem relatively simple methods of increasing lateral restraint, due care must be taken regarding the use of a low positioned D-ring. In severe frontal impacts, such a design may potentially increase restraint to the upper torso, causing the kyphotic thoracic spine to straighten and press on the thoracolumbar spine, resulting in anterior wedge fractures [Begeman et al., 1973]. Despite this model not being validated for severe vertical loading, some spinal compression was observed in these far-side impacts. Further research should be undertaken to ensure that such a design is not going to have adverse effects on other body regions.

In Parameter Study B, the influence of impact direction and pretension on occupant-to-seat belt interaction was estimated. When no pretension was used, the model’s shoulder did not escape the belt until impact angles exceeded 40 degrees (from frontal). This result agrees with previous findings by Adomeit et al. (1977) and Horsch (1980). Interestingly though, despite the difference in shoulder belt loading locations for the PMHS and Model in the 60 degree impacts, the resulting lateral head excursion was similar. Further research needs to be conducted to better understand this shoulder to belt interaction in far-side oblique impacts. The reason for this is that evidence already suggests that 60-degree impacts are the most important far-side crashes to understand (Gabler et al., 2005b). Results also indicated that when the impact angle is between 70 and 90-degrees, lateral head excursion is similar. This suggests that the shoulder portion of the seat belt provides very little thoracic restraint at these angles compared to more frontal impacts.

In light of the limitations of this study, it is hypothesized that introducing methods that encourage the belt to engage the shoulder will reduce both the likelihood and severity of head and thorax contacts with intruding structures or other occupants. However, further research needs to be conducted taking these additional factors into consideration. Additionally, the influence of both driver and passenger airbags in angled far-side impacts needs to be explored.

CONCLUSION

This study has quantified some of the factors influencing occupant-to-seat belt interaction in far-side impacts. Results from both quasi-static and dynamic tests indicate that lower positioned D-rings with the addition of pretension offer potential benefit in far-side impacts. Specifically by increasing the likelihood of the shoulder engaging the seat belt. It was also observed that occupants are likely to escape the shoulder portion of the belt at far-side impact angles greater than 40 degrees from frontal when no pretension is used. The addition of pretension allowed the shoulder to engage the belt in all impacts from 30 to 90 degrees.

Figure 1.

Rotating quasi-static rig (Actual – Left, Simulated – Right)

APPENDIX: SHOULDER BELT FORCES FOR SLED TESTS

Figure A1.

Forward D-ring. 90-degree impact. No pretension

Figure A2.

Middle D-ring. 90-degree impact. No pretension

Figure A3.

Middle D-ring. 60-degree impact. No pretension