Documentation:FIB book/Child ATD Development

Introduction

Conducting experiments with Anthropometric Testing Devices (ATDs) is a common method for testing car safety for regulatory purposes, accident reconstructions, and research purposes. Many motor vehicle manufacturers go through car safety testing with adult ATDs to measure the impact response and to ensure the design of the car meets the safety requirements listed by regulatory bodies, for example, the Federal Motor Vehicle Safety Standards (FMVSS) in the US.

The first adult crash test dummy was created in 1949 for testing aircraft ejection seats. It wasn't until 1970, when people realized how a child's response to impact cannot be represented by an adult ATD, were the first child ATDs developed. Nowadays, for children's car safety, many Car Restraint Systems (CRSs) and airbags are tested to ensure conformance with dummy Protection Reference Values (PRVs). However, due to the lack of biomechanical data for children, these child safety standards are based on injury criteria and PRVs which are often derived from adult data, scaling from a 50^th percentile adult male according to the differences in the size of the occupants. Similar assumptions are made when testing the biofidelity of the ATDs. This raises ongoing concerns about the effectiveness of child crash test dummy testing for the safety of cars and CRSs. Even though child dummies have many different sets of instrumentation to measure impact response, whether the results of the data are an accurate representation of what actually happens to a child in a car crash will always require careful investigation and analysis.

Hence, the main objectives of this page are to investigate the type and severity of injuries children get after car accidents, the differences in physiology and biomechanical characteristics between adults and children, the design specifications of existing child ATDs, and the future work that is required for improving car safety for children.

Background

Epidemiology on Child Injuries

The child ATDs that are currently used in car safety testing, such as CRABI series, the Hybrid III child series, and the Q-series, were mainly developed in the early 90s. However, the epidemiological data that was available at that time is very different from the data that is available now.

As shown in Table 1, data from 1988 to 1991, which was available when the current child ATDs were being developed, showed that 58.5% of the children left automobile accidents with no injuries and 37.5% of them left with injuries of MAIS score not higher than 2. Only 4.0% of the children who were involved in car accidents had an injury with a MAIS score above 3.

Table 1. MAIS Scores of Children in 1988 - 1991 (Table 3 from “Techniques for Developing Child Dummy Protection”)

MAIS	Younger (0-5) Restrained	Younger (0-5) Unrestrained	Older (6-12) Restrained	Older (6-12) Unrestrained	Total
0	70.8%	44.6%	62.8%	36.7%	58.5%
1	24.8%	45.3%	31.1%	55.2%	35.1%
2	1.2%	3.5%	2.8%	3.6%	2.4%
3	0.5%	1.2%	0.5%	1.5%	0.8%
4	0.2%	0.3%	0.1%	0.2%	0.2%
5	0.0%	0.2%	0.1%	0.2%	0.1%
6	0.1%	0.2%	0.0%	0.1%	0.1%
7	2.4%	4.6%	2.5%	2.6%	2.8%
Total	100%	100%	100%	100%	100%

However, looking at the data for 2005 - 2019, as shown in Table 2, only 28.9% of the children involved in car accidents were not injured, which is equivalent to AIS 0 in the table above. Moreover, a total of 35.0% of the children who were involved in car accidents had either fatal or incapacitating injuries.

Table 2. Person Injury Types for Children at the Age from Newborn to 12 Years Old in 2005 - 2019

Personal Injury Type	0-5 Years	6-12 Years	Total
Fatal	20.2%	19.5%	19.8%
Injured, Incapacitating	13.5%	16.6%	15.2%
Injured, Non-incapacitating	18.5%	21.1%	19.9%
Injured, Other	16.5%	15.3%	15.8%
Not Injured	30.8%	27.3%	28.9%
Unknown	0.4%	0.2%	0.3%
Total	100%	100%	100%

The large discrepancy of the percentage of each injury type between the two time periods could be due to the increase in traffic volume and the increase of children in cars and on the road. Additionally, it could be due to the lack of child biomechanics data to inform a biofidelic method for testing cars and CRSs safety, which will be investigated more in later sections of this paper.

Crash Modes

To design biofidelic ATDs and effectively test child safety in cars, it is important to understand the common crash modes and what kinds of collisions to test for. The most common type of crash that resulted in fatalities or injuries of children are accidents that don't involve collision with another vehicle in transport. For collision-type accidents, collisions at an angle are the most common.

Table 3. The Manner of Collision for Car Crashes Involving Child Fatalities and Injuries (Aged from 0 to 12 years) in 2005 - 2019

Manner of Collision	Person Injury Type
	Fatal	Injured, Incapacitating	Injured, Non-incapacitating	Injured, Other	Total
Not collision with motor vehicle in transport	56.1%	35.9%	35.8%	28.2%	39.8%
Rear-end	9.0%	7.8%	8.8%	12.7%	9.6%
Head-on	10.2%	22.4%	31.4%	35.5%	29.0%
Angle	21.2%	29.3%	31.4%	35.5%	29.0%
Sideswipe	2.6%	4.0%	4.3%	4.9%	3.9%
Other	0.5%	0.4%	0.4%	0.7%	0.5%
Unknown	0.2%	0.1%	0.1%	0.2%	0.2%
Total	100%	100%	100%	100%	100%

As shown in table 3, side impacts in collisions seem to be more common. In terms of side impacts, there are two main types, including near-side impact and far-side impact. Near-side impact means that the occupant is close to the side of impact while far-side impact means that the occupant is far from the side of impact. Looking at the fatalities of children in side impacts and the second seating position, near-side impacts have a higher rate of fatality as compared to far-side impacts, which can be explained by the fact that near-side impacts can cause a higher acceleration and more energy exerted on the occupants.

Table 4. The Secondary Seating Position Relative to Child Fatality Rates (1991 - 2000)

Second seating position	<1 year old	1 - 3 years old	4 - 8 years old	Total
Near-side	55.4%	55.8%	59.9%	57.7%
Middle	25.2%	22.3%	17.9%	20.5%
Far-side	19.4%	22.0%	22.2%	21.7%
Total	100%	100%	100%	100%

From the data above, frontal and side collisions appear to be more common in accidents that have caused child fatalities. This correlates with the fact that most of the child ATDs are designed for both frontal- and side-impact testing.

Biomechanical Basis

As mentioned, the epidemiological data on child automotive injuries from 1988-1991 may now be considered outdated. Whether this is due to incomplete sampling of the population or, more likely, because risks have increased along with the traffic population, this early data formed the groundwork for the development of child ATD development - including for the Q-series, the Hybrid III child, and the CRABI, ATDs that are still in use. The same may be said for the biomechanical and biofidelic properties involved (though anthropometry data has remained fairly consistent), many of which were based on assumptions made necessary by the low availability of child injury data. Problems with some conventional data sources are as follows:

• Volunteer Impact Testing: The presence or lack of injuries is correlated to the force and acceleration measurements taken. This is not an option with children for the purposes relating to health, safety, and consent. Also, as is always the case with volunteer testing, data regarding severe injuries is not available here. Inaccuracies arise from placing instrumentation in ways that do not cause injury (leading to less reliable results), and muscle tension and human reactions may act as confounding variables.
• Cadaver Testing: Cadavers are instrumented and subjected to impact, and autopsies may show the presence and nature of injury. This is a scarce source, given the low availability of child cadavers. Additionally, “Previous cadaver testing of child subjects has also raised ethical objections, making subjects extremely rare. Only eleven cadaver tests on children have ever been reported” prior to the development of common child ATDs. Moreover, all eleven simulations were of frontal impacts without airbags, and “the limited d amount of instrumentation on the cadavers makes injury criteria recommendations speculative”.
• Animal Testing: Animal surrogates are used to estimate human response characteristics. This has provided a significant portion of the foundational data, using animals of similar size to children (commonly, primates and pigs). Differences in anatomy must be accounted for with assumptions, making the translation from experimental data to human injury criteria difficult.

As a result, “the child dummies generally have even less human-like characteristics than adult dummies, which makes accurate reconstructions a problem. Computer models of children have similar drawbacks”. This made it difficult to create adequate PRVs for child ATDs and injury criteria for human children.

To make up for a lack of child data, dimensional analysis was often applied, scaling adult data to suit child impact scenarios for use in PRV and injury criteria development. “Using geometry, mass, and biomechanical material property ratios, PRV for the smaller dummies can be estimated by scaling adult values with dimensional analysis”. However, the geometric and anthropometric assumptions involved in this technique become less appropriate as age decreases, and approximations are introduced.

Differences in Adult and Child Physiology

As shown in Table 5, the head and face are the most commonly injured body regions, seen in 55.1% of injuries in children in automotive accidents between 1988 and 1991. This is particularly dangerous to children due to underdeveloped skulls being more susceptible to fracture. Neck injuries, while making up only 4.4% of cases, often see more extreme results, for example C1/C2 vertebral fractures (associated with brain stem and spinal cord damage), or dislocation at the OC-C1 head-neck junction (associated with transection of the spinal cord). These are more frequent among children, as 30% of one’s total body mass is found in their heads at birth, tapering down to 6% by adulthood. Hence, during an impact, the larger neck load / moment is being counteracted by less developed muscles and ligaments. Frequency and severity draw attention to the head and neck injuries, forming the main focus of Table 6, a summary of physiological differences between children and adults with respect to collisions. Also note the discrepancy in pelvis/abdomen injuries between age groups, at 20.6% of the retrained younger group and 11.0% for the retrained older group. Klinich et. al. explains:

“A child’s abdomen protrudes more than an adult’s, and the liver and spleen are not as protected by the rib cage as they will be later in life. Children may be more likely to suffer a higher incidence of multiple organ injury, because kinetic energy is dissipated into a smaller mass. The iliac crests of the pelvis do not develop until approximately age 10. The absence of the iliac crests to help position lap belts properly over the pelvis poses a particular challenge to restrain designers.”

Table 5. Child Injuries by Body Region in 1988-1991 (Table 4 from “Techniques for Developing Child Dummy Protection”)

Body Region	Younger (0 - 5) Restrained	Younger(0 - 5) Unrestrained	Older (6 - 12) Restrained	Older (6 - 12) Unrestrained	Total
Head	12.7%	13.3%	11.3%	12.5%	12.5%
Face	39.7%	47.0%	40.0%	42.5%	42.6%
Neck	4.0%	2.2%	7.5%	4.6%	4.4%
Thorax	8.3%	6.1%	9.3%	7.8%	7.7%
Pelvis/Abdomen	20.6%	13.5%	11.0%	5.2%	11.5%
Whole Body	2.7%	2.9%	1.2%	3.1%	2.6%
Lower Extremities	6.6%	9.0%	13.7%	15.0%	11.5%
Upper Extremities	5.1%	5.2%	5.9%	8.9%	6.6%
Unknown	0.3%	0.8%	0.1%	0.4%	0.4%

Table 6. Relevant Differences Between Children and Adults

Property	Difference	Relevance During Collisions
Proportion of total mass at head	At birth, the head is 30% of one’s body weight and 1/4 their total height, tapering to 6% and 1/7 by adulthood.	Overall center of gravity is higher, causing bending over a lap belt or around a shoulder belt. Underdeveloped neck muscles and ligaments must counteract a higher moment. Figure 1. Rough proportions at birth, age 6, and age 25 (left to right).
Skull structure	At birth, the skull is flexible, consisting of six fontanelles which don’t fuse until ~18 months of age.	The skull deforms more easily under loads and is more susceptible to fracture. In designing adult ATDs, the skull is assumed to be rigid; this is incorrect when applied to infants. Adult HIC that links the likelihood of brain injury to skull fracture cannot be scaled to younger ages. Also cannot make assumptions of geometrical similarity during scaling.
Neck vertebrae / cervical spine structure	At birth, the neck vertebrae are three bones joined by cartilage, growing together at ~3 years of age. The C1 and C2 do not join until ~4-6 years. Vertebrae do not finish developing until ~25 years.	During this early development, the upper neck’s facet joints are more horizontal. Minimal forces may cause partial dislocations. Here, 60-70% of cervical fractures occur at C1 or C2; the natural neck pivot is at C2 or C3. (as opposed to adults, who have ~16% at C1 or C2, with a natural pivot near C6).
Neck muscles / ligaments	At early ages, muscles are underdeveloped and ligaments are lax. Heads cannot be held up until ~3 months of age.	Combined with head size, this leads to spinal cord stretch injury; underdeveloped soft tissues allow the flexible vertebrae to displace more before fracture, stretching the spinal cord. Children can receive spinal cord injuries without vertebral damage.
Ribs flexibility	Ribs become less flexible with age.	The lack of rigidity offers less thorax protection. Children have less chance of rib fracture, but more chance of thoracic organ damage. Broken ribs indicate high impact energy, making internal damage likely.
Abdomen and pelvis	At early ages, the abdomen protrudes more. The iliac crests don’t develop until ~10 years of age	The liver and spleen are less protected by the ribcage, leading to a higher incidence of multiple organ injury (also because kinetic energy is applied to a smaller mass). In adults, the iliac crest helps position lap belts over the pelvis - belt position is unpredictable in children.

Animal Surrogate Testing

Animal testing was not uncommon prior to 1998, whereby human tolerance levels in impact conditions were approximated by test results from primates or pigs. Animal injury criteria estimates are scaled to match humans and, as these animals are more similar in size to children than to adults, mass and size ratios may be applied here with (arguably) less concern. Objections arise, however, to do with interspecies and age-related differences; an adult primate will not have the same developmental deficiencies to their spine, head, or neck muscles, for example, as those discussed above. This is important, as “a major area of this type of research deals with human tolerance to head angular acceleration”.

In the study of traumatic brain injuries, at least, these concerns are less pressing; the response of the brain matter is what is important here, rather than the resistance to or protection from injury brought about by muscle and bone reactions. Diffuse axonal injury, the shearing of the brain’s axons due to sudden rotational accelerations, may result in concussions or comas - numerous studies have examined primate threshold levels of angular accelerations and velocities, which were then scaled to apply to adult humans, as shown in Figure 2.

Figure 2. Rotational velocity and acceleration for 50% probability of concussions, scaled from primates to humans (Figure 10 from "Techniques for Developing Child Dummy Protection")

Pigs and baboons have also been used to simulate children being struck by passenger side air bags, with varying conditions being applied to give a range of injury severities. Here, tests were conducted alongside existing 3-year-old ATDs to inform PRV development for that dummy, providing the reference values shown in Table 7.

Table 7. IARV for General Motors' 3-Year Old ATD, Scaled from Pig and Baboon Test Data (Table 10 from "Techniques for Developing Child Dummy Protection")

Body Region	Parameters	Risk of Serious Injury
		1%	10%	25%
Head	HIC (15 ms)	1480	1530	1570
Neck	Axial Tension (N)	1060	1125	1160
Thorax	Upper Spine Acceleration (g)	55	59	62
	Upper and Midsternal Delta V (km/h)	9	16	19
Abdomen	Lower Spine Acceleration (g)	34	42	45
	Lower Sternal Delta V (km/h)	19.5	19.9	20.4

This does not tell the full story for these body regions. Note that there is no mention of a neck tension-moment relationship / limit for injury; this did not come about until later. Additionally, this data would lead one to believe that 3-year-olds are more durable than they actually are; values were later adjusted to better suit reality. For example, a 3-year-old’s 15 ms HIC was adjusted in 2003 to just 568 at 5% risk of injury. This is potentially due to the previously discussed interspecies and developmental differences.

Scaling Adult Data

It is rare to find data on biomechanical properties as a function of age; in many cases, scaling must be done under the assumptions of volumetric and geometric similarity using ratios of mass and dimension. The data in Figure 2, for example, were scaled from primate to human contexts according to the scaling laws for angular acceleration and velocity:

${\ddot {\theta }}_{m}={\ddot {\theta }}_{p}(1/\lambda _{M})^{2/3}$

${\ddot {\theta }}_{m}={\ddot {\theta }}_{p}(1/\lambda _{L})$

where λ_M and λ_L are the head mass and head length scaling ratios, respectively. The scaled data, along with injury correlations from the primate tests, resulted in the head angular injury criteria shown in Table 8.

Table 8. Head Angular Injury Criteria, Scaled from Primate Tests (Table 8 from "Techniques for Developing Child Dummy Protection")

Age	Head Mass Scaling Ratio λM	Head Length Scaling Ratio λL	Angular Velocity Limit (rad/sec)	Angular Acceleration Limit (rad/sec2)
Adults	1.000	1.00	>=30	<1700
6-year-old	0.725	0.9	>=33	<2106
3-year-old	0.655	0.87	>=34	<2255
12-month-old	0.553	0.81	>=37	<2524

In theory, dimensional analysis could be applied to any scenario where engineering units can be broken down into the base units of mass, time, temperature, and length. “For example, head acceleration units are $length/time^{2}$. If we know the length and time ratios between two dummies, we can scale the acceleration limits”. Characteristic head lengths ($Lr$) at different ages may be measured or gathered from anthropometric data, but time ratios must be derived from some other relationship such as a modulus of elasticity ($E_{r},mass/length*time^{2}$). An acceleration ratio ($A_{r}$) can thus be derived (as can force, moment, velocity, and HIC, in similar fashion):

$A_{r}={\frac {L_{r}}{(T_{r})^{2}}}={\frac {\frac {L_{r}}{(M_{r}L_{r})}}{E_{r}}}={\frac {E_{r}}{M_{r}}}$

Scaling through dimensional analysis does not in itself offer much consideration to assumptions - the ratio above, for example, assumes that the density of what is being scaled is equal for both participants in the ratio. A potential for error is introduced by the approximations made during analysis. Scaling head acceleration from adult to child populations, for example, encounters the following problems:

• The amount of skull modulus data is insufficient to define a modulus ratio,
• There are geometric differences besides characteristic lengths between skulls,
• Head density changes with age.

In some cases, even anthropometry data is incomplete where needed; neck circumference, for example, is not listed, so a formula was created to derive neck ratio from a perceived linear relationship between head and chest ratios: neck ratio = chest ratio + 0.347 * (head ratio - chest ratio). Bone and tendon moduli are a preferred choice of scaling variables but, again, information at a specific location is often missing. No chest bone modulus or neck tendon modulus was available during the 1990s, and it was assumed that cranial bone data and general tendon data would be suitable representations.

All of this is particularly concerning when it has to do with human injury criteria that was derived from animal tests in the first place. Nevertheless, child ATDs and their PRVs are mostly scaled down versions of their adult counterparts, the argument being that dimensional analysis is appropriate because the same materials are used.

Child ATD Development

This section will give a brief overview of some commonly used child ATDs and their designs. Key features and assumptions have been noted, some of which serve to compound the sources of error seen in the Biomechanical Basis section. In particular, note the frequency of scaling methods being used in their design, and other ways in which the lack of experimental data and material characterization with age has been bypassed. Also note how instrumentation decreases with ATD age, with various justifications. In this section, the justifications for what was considered acceptable are presented as expressed by the publications they are sourced from, and may themselves rely on faulty assumptions and data. For example, where tests and biofidelic corridors are mentioned, it is safe to assume that criteria have been scaled down from adult ATDs.

CRABI

The CRABI (Child Restraint / Airbag Interaction) series of ATD is made up of 6, 12, and 18 month old infant models (the CRABI-6MO, CRABI-12MO, and CRABI-18MO, respectively), was designed in 1990 for use in side and frontal impacts involving child restraint and airbag interaction (though they are considered suitable for use in crash scenarios of any direction, with or without airbag interaction). Much like the Hybrid III Child, their designs resemble the Hybrid III 50th percentile adult male dummy, as seen in Figure 3. Similarities are expected, as the CRABI and Hybrid III child series were developed in parallel by task forces within the Society of Automotive Engineers.

Figure 3. Sketch of the CRABI six-month-old (Figure 4 from "Techniques for Developing Child Dummy Protection")

Note the scaled down Hybrid III skeletal structures used in the neck and spine, notched to reduce stiffness. Rubber elements are included at the limb joints for improved biofidelity and posability. Instrumentation in the CRABI-6MO measures head, pelvis, and chest accelerations, and neck and lumbar spine forces and moments. The CRABI-12MO and -18MO also measure linear loads at the pubis and shoulder. The instrumentation, while less than that of the Q-series, is enough at the head and neck to fill an intended niche: evaluation of infant responses during impact when seated in rear- or forward-facing child seats.

In a frontal crash, with a forward-facing child seat, the underdeveloped skeleton, muscles, and ligaments of an infant cannot withstand the forces delivered via the belt restraints, or the moments due to large head mass and dimension ratios. Additionally, a forward-facing child seat in the front passenger space exposes the infant to full contact with airbags designed for adult passengers. Rear-facing seats help distribute crash deceleration forces over the area of the chair. When developing PRVs for these scenarios, the CRABI series is made adequate through its instrumentation; “Although the CRABI 6-Month dummy does not necessarily possess realistic biomechanical responses, it does have up-to-date instrumentation which allows head translation and angular accelerations, neck forces and moments, and chest translational accelerations to be measured”. Most masses and characteristic dimensions used in design were available as anthropometric data but, where scaling was necessary, the elastic bending moduli of bone for infants and adults were estimated using parietal bone fragments.

As with the previous dummies, the biofidelity of the CRABI’s head impact response was measured by a drop test, here from a height of 376 mm. During tests with adult cadaveric heads, adult male forehead impacts showed peak resultant head accelerations between 225 and 275 G. Dimensional scaling of this requirement was applied to suit the three infant head sizes, with the skull flexibility added by an infant’s un-fused fontanelles being overestimated as a safety factor. The cartilage between the fontanelles were expected to dictate head stiffness, but its viscoelastic properties during impact were ignored; “No attempt was made to model this interaction since time-dependent material properties for the interconnecting cartilage are unknown”. As a result, head response requirements may have been too low, and the CRABI heads may have been too soft. “However, if the child restraint system is designed so that the child does not experience a hard surface head impact under the design test condition, then the CRABI heads will give a childlike response”. Neck and chest response requirements were scaled from adult counterparts in similar fashion, to try and account for weak neck muscles / ligaments and more flexible ribs. Some criteria seen above, such as knee impact responses, were not addressed for these infant dummies.

Q-Series

As successors to the P-series of child dummies, Q-series child dummies were developed in 1993 and were designed for testing frontal or side impacts, which is reasonable given that there are more child fatalities in frontal and side collisions as mentioned in the section in which we talked about crash modes. Just like the adult ATDs, the instrumentation of the Q-series is interchangeable within the series. The available dummies in the Q-series are the Q0, Q1, Q1.5, Q3, Q6, and Q10, with numbers denoting the child’s age in years. All of these models have similar design principles to improve anthropometry, biofidelity, and kinematics as compared to the P-series.

The Q-series is designed to look at child safety during car crashes and to validate CRSs and to inform engineers for designing better protection for children in cars. As compared to P-series, the Q-series has improved in different perspectives, including anthropometry, biomechanics, and kinematics. The anthropometry designs of the Q-series are informed by the Child Anthropometry Database (CANDAT) which contains child data from 0 to 18 years old collected from different regions in the world including Europe, US, and Japan. The biomechanical response requirements for the Q-series dummies, however, are derived from the set of existing data for adult ATDs for frontal and side impacts by scaling according to some differences in body tissue characteristics between adults and children. For frontal impacts, the biofidelity requirements are set up for the head, neck, thorax, abdomen, and lower extremities; while the biofidelity requirements for side impacts have additional sets for the shoulder and the pelvis. The following subsections describe the design modifications made to these Q-series and the corresponding biomechanical response testing that is done for each body part. As the Q-series is very well documented, we took the opportunity to include some detail addressing test methods and referencing biofidelic corridors.

Q0

Q0 is an ATD created to represent the kinematic and biomechanical characteristics of a 6-week old newborn. However, due to the restriction in size, there were limitations in the design of the body parts as well as the instrumentation. The Q0 has eleven body parts: head, neck, shoulder, two arms, thoracic spine, lumbar spine, thoracic flesh, pelvis block, and two legs. However, looking at the arms and the legs, the angles between the upper and lower arm and the upper and lower leg are fixed since there are no elbow joints or knee joints, respectively. Unlike the other models, Q0 only has one part with torso flesh foam representing the ribcage and the abdomen. The designs for the neck and the lumbar spine have a relatively similar design as the other models in the Q-series. In terms of instrumentation, the Q0 is only equipped with a head, T1, and pelvis accelerometers, and an upper neck load cell.

Q3

The rest of the models in the Q-series, except Q10, have a very similar layout and design for each body part to Q3. The head and the clavicle are made completely with plastics. On the head, there is a semi-rigid component that allows multiple linear accelerometer arrays to be mounted on to measure linear acceleration and rotation. The biomechanical response test was done by simulating the cadaver 130-mm head-only drop tests conducted by Hodgson and Thomas. The test was not performed back in the 90s but was later on performed by da Jager et al. and was found that the peak resultant head accelerations were close to the targeted values for all Q-dummies.

The neck consists of a flexible column that is made out of a combination of metal and natural rubber. For the neck, there are separate joints at the occipital condyles and atlas axis (C1 - C2) to improve the kinematic response. These joints have low resistance and a large range of motion allowing more flexibility. C2 - C7 is simulated by a flexible column that allows bending, twisting, and limited shear, compressing, and elongation. There are also an upper neck and a lower neck load cells measuring forces and moments in 3 axes for injury measurements. The neck is tested for flexion-extension and lateral flexion bending responses. As with the head, the neck response was not tested in the 90s, but in the study by de Jager et al., the test results for the neck flexion fall within the biofidelity corridor.

The shoulders are made out of natural rubber with metal end plates connecting to the upper arm and the thoracic spine. It also incorporates a ball-and-socket joint to simulate the humerus-scapula joint. Unfortunately, there is no instrumentation mounted on the shoulder for injury assessment. The shoulders are tested for lateral pendulum impactor tests with a 3.8-kg pendulum at 4.3 m/s. Unfortunately, there was no test data on the biofidelity of the shoulder structures from both papers reviewed.

For the thorax, there is a deformable rib-cage type structure mounted onto a rigid thoracic spine box. The shape is derived from anthropometry data and analyses of X-rays of real children. In terms of instrumentation for injury measurements, there are accelerometers on both the thoracic spine and the thoracic rib-cage and transducers for chest deflection measurements. The thorax is tested for a pendulum impactor test and is tested for both frontal and side impacts. From the test results on both studies, the one by van Ratingen et al. in 1997 and the one by de Jager et al. in 2005, it seems that the test results don’t fall within the corridors exactly and could be caused by the stiffness in the thoracic spine.

The abdomen is a deformable structure covered by skin and is enclosed by the rib-cage and the pelvis. There is a load cell measuring force in the abdomen structure for injury assessment. The abdomen structure is tested for dynamic belt test and pendulum impactor test and the test results show that they are mostly within the biofidelic corridor.

The pelvis is made out of a semi-rigid plastic skeletal structure covered with flesh and skin and consists of a metal pelvic bone. There is an accelerometer in the pelvis for injury measurements. Since submarining starts at the legs and pelvis, the pelvis is only tested for side impacts. Its performance is tested for a pendulum impact test with a cylinder of 3.96 kg at 5.2 m/s impact velocity. The test results of the pelvis fall within the biofidelity corridor.

The thighs are tuned according to the compression characteristics according to biomechanical data. There are no instruments in the body part but the performance of the thighs is tested for the strap force-lap belt penetration test and the displacement is measured by the string potentiometer which measures the deformation of the left and right legs. The test results show that the performance falls under the biofidelic response corridor.

For the lumbar spine, it is similar to the neck structure with a flexible column. There is a load cell at the lower lumbar spine for injury assessment. However, there are no tests or sets of biofidelic response corridors used to test the performance of the lumbar spine. This may have led to some inaccuracy in representing the response of the lumbar spine of an actual child in a car accident as discussed later in the section of “Comparison with Other Models”.

There are a lot of improvements in anthropometry, biomechanics, and kinematics in Q3, similarly for other Q-series models except Q10, as compared to the P-series, especially with the access to a large anthropometric database. However, there are still some concerns with the biofidelity of these dummies, especially for the biomechanical characteristics. As discussed above, the test data for the thorax shows that the thorax is slightly stiffer than the biofidelic model, and there are no available data and tests for the lumbar spine.

Q10

The Q10 model is the latest ATD model of the Q-series and is designed to test the biomechanical response of children for frontal impacts mainly, with some side impact applications. The size of the ATD was scaled according to the anthropometry of the 50th percentile of 10.5-year-old children. Some of the features that differentiate the Q10 from other Q-series ATDs are the pelvis design which focuses on preventing the seat belt from being trapped by reducing the gap between the pelvis and the legs as well as introducing stiffer hip shields. This is particularly important since the trapping of the belt can have some negative repercussions on the ATD’s biomechanical response during impact. Another Q10 notable feature are tilt sensors integrated for positioning of the dummies.

When it comes to how biofidelic the Q10 is, Lemmen et al. designed prototypes with anthropometric specification mentioned above and different instrumentation for each body part. Biofidelity of each Q10 part was tested separately and parts were further improved if needed depending on the results obtained. Tilt sensors were introduced to different spots on the ATD which helped with the positioning aspect of the tests.  Lemmen et al. decided to find Q10 biofidelity targets for the following test by scaling Q10 targets to be in between those of smaller child ATDs and adult ATDs. As for the head, the Q10 adhered to the same design as the Q3 head and was instrumented with 3D angular and linear accelerometers. Three criteria biofidelity drop tests on a rigid plate were performed from 130 mm and 200 mm high and the test results fall within the biofidelic corridor considering only the tests for EEVC standards.

The neck design was also based on the Q3 model and was instrumented with 6-axis load cells on both upper and lower neck. For biofidelity, multiple part 572 pendulum tests were performed; neck flexion bending, neck extension bending and neck lateral flexion, and the performance of the ATD was in the lower end of the corridor, center of the corridor, semi-fitted in the corridor (started off in the corridor then started outlying) respectively. Therefore, the neck is not fully biofidelic and improvements are required since head and neck are two of the most crucial parts for frontal impact analysis.

The shoulder biofidelity test was performed using an 8.76-kg pendulum at a lateral impact speed of 4.5 m/s and it was noted that the initial response of the shoulder is too stiff and results mostly lie outside the corridor limits. This proves that shoulder biofidelity is very low, and it was mentioned that since the Q10 is an omni-directional dummy, performance tuning in any direction would affect performance in another direction; hence, it would be tricky to address this issue.

As for the thorax, instrumentation included 3D linear accelerometers, at spine T4 and T12, 3D angular accelerometers and 3D angular rate sensors along with 2D displacement sensors on the upper and lower rib cage and 3D linear accelerometers on the sternum. Two pendulum test impact speeds were performed during frontal and lateral thorax biofidelity tests, 4.31 and 6.71 m/s for each. Biofidelity for the frontal seems to be very biofidelic, with higher speed being more biofidelic, which is a great improvement from previous Q-series models. However, lateral biofidelity seems to be quite low due to the initial response overestimating shoulder stiffness and any modifications to address this issue would cause an imbalance when it comes to the performance of the shoulder and pelvis during lateral loading.

The lumbar spine is composed of a cylindrical rubber column, with 6-axis load cells included and was tested by pendulum impact tests. The test resulted in dynamic stiffness of 81.9 Nm/rad which was slightly greater than set targets of 68.6 Nm/rad for normal flexion and 71.4 Nm/rad for lateral flexion. Note that Lemmen et al. assumed that performances in frontal and lateral flexion are similar due to the structural shape of the spine, and the results, although are a bit higher than targets, are deemed acceptable and biofidelic. Finally, the pelvis incorporated 3D linear accelerometers, 3D angular accelerometers and 6-axis load cells on the sacroiliac joint. Pelvis biofidelity was tested using lateral impact and pelvis showed quite high stiffness. There was an incidence of bottoming-out contact between the iliac wing and the sacrum block at an impact speed of 4.0 m/s which shouldn’t have occurred until 5.2 m/s. Hence, the pelvis is considered not biofidelic.

As portrayed above, some parts were biofidelic and performed splendidly; however, others seemed to have failed to reach required standards, along with difficulty to adjust those issues without affecting the dynamics of other parts. Additionally, there are a couple of other issues with the Q10 such as ballooning of the abdomen due to air not escaping fast enough, fracturing of wrist flesh upon contact with front seat and pelvis flesh damage due to sliding of upper leg flesh towards the knee. To summarize, the Q10 model seems to be more sophisticated and an advancement compared to previous models such as Q3; however, there are multiple points of improvements and optimization needed to further improve the biofidelity of this ATD as a whole and present more reliable child-like responses during crash tests.

Hybrid III 10-Year-Old

In the Hybrid series, the Hybrid III 10 year old child ATD was designed to fill the gap between small female dummies and 6 years old dummies and are utilized for booster seats evaluation and the inflation induced injury potential of passenger air bags. The masses for each body component were found using multiple scaling factors and using respective proportions of small female ATD masses. Some notable features that differentiate this ATD from other Hybrid III ATDs include a more sloped upper surface of the shoulder contour and an adjustable bracket between the lumbar spine and the pelvis allowing the ATD to be oriented in both “normal” and “slouched” postures so it could be used with and without a booster seat.  As for the biofidelity corridor limits for Hybrid III child ATD, they were obtained by scaling midsize male limits using the scaling relationships developed by Irwin and Mertz (1997) and dummy parts are deemed biofidelic if they lie between those limits.

Starting from the head, a Hybrid III small female head was used, due to similar anthropometry and was instrumented by triaxial accelerometers at centre of gravity and a tilt sensor.  To test the head’s biofidelity, it was dropped from a 376-mm height and impacted a flat rigid surface with its forehead. Since a commercially available Hybrid III small female head was used, it was not surprising that the head acceleration was in between the limits (295 and 240G) and was deemed biofidelic. The neck of the dummy is a scaled version of the segmented Hybrid III neck design and incorporated a six-channel load cells at the occipital condyle and six-channel C7/T1 interfaces. A neck pendulum calibration fixture was used to evaluate the neck’s flexion and extension biofidelity along the sagittal plane. The resulting moment values seem to have followed a similar shape to the corridor limits so the SAE Task Group concluded that the neck’s biofidelity was acceptable.

Another notable feature of the Hybrid III child ATD is the thorax, which includes six steel ribs with bonded-on damping material and a 50 mm minimum distance between the sternum and the rib stops. The thorax had a triaxial accelerometer attached to T4, upper and lower sternum accelerometers, upper and lower spine accelerometers, and a rotary potentiometer for sternal deflection. To investigate sternum response biofidelity, the sternum was impacted by a 6.89-kg cylindrical pendulum, with a 121-mm diameter flat surface, at a speed of 6.7 m/s. Similar to the neck, the results lied within the corridor limits and the sternum/thorax was deemed biofidelic.

Finally, the dummy’s knee impact response was analyzed using a 1.9 kg cylindrical pendulum impactor, with a 76-mm diameter, impacting the knee at a speed of 2.1 m/s. In spite of the fact that the results obtained from this test were close to the upper biofidelic force limit of 3.14 kN, the SAE Task Group accepted the knee design’s level of biofidelity since the only way those values could be lowered was by inserting a pad between the knee structure and the flesh, which was deemed unacceptable by the SAE Task Group.

As portrayed, the Hybrid III child ATD seems to be more biofidelic than the Q10 model. Nevertheless, the head and neck, which are two of the most important points of interest in frontal impact analysis, were not quite representative of child anthropometry (i.e. head from small female Hybrid III and neck from the Hybrid III model). In addition, Hybrid III had less biofidelity tests when compared to Q10, but Hybrid III series had more robust and extensive biomechanics analysis and improvements throughout the years, which explains why Hybrid III seems more representative than Q-series. With that being said, Hybrid III child has its own issues too such as the lumbar angle being difficult to adjust along with some extremity joints not maintaining their post-test torque settings. It should be noted that the criteria used to validate biofidelity may be based on faulty assumptions; the validations performed here should be carefully analyzed. Thankfully, at 10 years of age, this is a relatively older dummy than other representations discussed here, and may be subject to less compounding variables such as the softer skulls and weaker muscles of infants.

Comparison Between Q-series and Hybrid III

De Jager et al. emphasized that Q-dummies are used to look at child safety during car crashes and to validate CRSs while the child dummies in the US (Hybrid III and CRABI) are mainly developed to assess airbag interaction with rear-facing child restraints. The biofidelity of pediatric ATD neck response and loads has been a much debated topic among the automotive safety community, which drove Seacrist et al. to compare Hybrid III and Q-series ATDs to pediatric volunteers in low-speed frontal crashes.

Previous researchers (Sherwood et al.) have looked into 10-year-old Post-Mortem Human Subject (PMHS) data from Kallieris et al. (1976) and compared them to Hybrid III 6 year old ATD. The ATDs seemed to yield higher neck loads due to a rigid thoracic spine which was not indicative of true injury potential. Another study investigated the kinematics of pediatric volunteers and compared those data to those of Hybrid III and Q-series 6 and 10 year olds. That study deduced that the ATDs response was primarily through rotation of the head about the upper neck; on the other hand, pediatric volunteers’ response involved a combination of head, spine and pelvis rotation. These differences could result in overestimations of forces/loads on the ATD upper necks, which is not representative of what happens in real life; hence, why this study is quite important for the advancement of ATDs.

For this study, researchers screened pediatric volunteers by having a size criteria which selected subjects whose masses and erect seated heights lied within ±15% of average ATD values. Three volunteers met the requirements for the 6 year old ATDs and seven for the 10 year old ATDs, making a total sample size of ten volunteers which is not a bad number considering they are volunteers and are of young age. For the tests, subjects (volunteers and ATDs) were “seated in a pneumatically-actuated, hydraulically-controlled low acceleration sled” and “a safe, non-injurious crash pulse applicable to the pediatric population was derived from an amusement park bumper car impact”. For restraining the subjects, researchers used “an automotive three-point belt system including an emergency locking retractor with automatic locking retractor function, webbing with 10–12% elongation, and cinching latch”. Load cells were added along with angular adjustments and tests were performed.

Results for peak value comparison are as follows:

• Hybrid III 6, Q6 and Q10 exhibited much higher peak for upper neck shear in comparison to volunteers, whereas Hybrid III 10 peak shear matched the volunteers.
• Hybrid III 10, Q6 and Q10 portrayed much lower peak axial force in comparison to volunteers.
• Hybrid III 6, Q6 and Q10 showed much greater Nij and bending moment in comparison to volunteers, while Hybrid III 10 showed lower values than its volunteers.
• Hybrid III 10, Q6 and Q10 exhibited much smaller peak head flexion angular acceleration than volunteers, whereas Hybrid III 6 had greater acceleration than volunteers.

Results for time-to-peak value comparison are as follows:

• Q-series ATDs demonstrated faster times to peak response for all metrics except for Q10 during shear force testing.
• Hybrid III 10 ATDs reached peaks for upper neck shear and bending moment at similar times to those of volunteers, whereas ATDs reached peaks for angular acceleration and axial force faster than volunteers.
• Hybrid III 6 ATDs had a uniquely late time to reach peak shear force compared to volunteers.

Overall, the data portrayed the extreme peak axial load underestimations by Hybrid III 10, Q6 and Q10 which is a very crucial miscalculation due to pediatric personnel having low axial loading tolerance and hence crash tests using these ATDs might under-predict pediatric cervical spine injury potential. These underestimations of axial force and bending moment would also result in a lower Nij value and thus resulting in mistakes during neck injury predictions. We have noted earlier, while discussing different children ATDs, how both Q10 and Hybrid III passed neck biofidelity tests; a possible explanation for this sudden decrease in biofidelity during low speed impacts could be muscle activations in the neck while volunteers react to impact, which the ATDs obviously lack.

This study proves that child ATDs in the market are not good enough to represent children during impacts and not enough research and improvements have been done on them. A limitation mentioned in the study is the fact that they used a standard three-point seat belt system which does not mimic any current automobile. Hence, in the future, a normal seat belt can potentially be used instead to have a more realistic environment and hopefully, more biofidelic results.

Comparison Between Q-Series and Other Models

With the advancement of technology, computational models have been more and more common in crash analysis. Li et al. looked into the differences in kinematic and biomechanical measurements with the computational Q3 side-impact child crash test dummy model, a FE child human model that was developed by Oshita et al., and a modified FE child human model that was developed by Zhang et al. The original FE child human model was developed just by scaling Total Human Model for Safety (THUMS) data due to the lack of anthropometric and biomechanical data for children. Later on, Mizuno et al. and Zhang et al. performed testing with child cadavers and realized that there are huge differences in the density, geometry, flexibility, and mechanical properties in a child human model as compared to just a scaled down adult human model.  Hence, Mizuno et al. modified the model in 2006 and Zhang et al. performed more modifications in 2008. The study done by Li et al. looked into far-side impact and near-side impact for both proper use and misuse of CRSs with the three different models. The test results have informed some biomechanical characteristics that are missing from the Q3 model.

In terms of kinematics, the neck and the spine of the modified human model showed more flexibility than the Q3s dummy model and the original FE model. “After contact between the head and CRS, the Q3s dummy maintained a somewhat upright position except for the far-side misuse condition.” The qualitative analysis shows that the human models had a forward lean and significant head sag in all impact conditions while the Q3s dummy model was a bit more rigid. This could be due to a stiffer neck and thoracic spine.

In terms of biomechanical data, the Q3s model over-predicted all the quantitative values, including the chest accelerations, HIC15, head contact forces, and upper and lower neck forces. Some of these over-predictions reach up to 152%, 198%, 213%, 484%, and 1175% respectively. This is mainly due to the more stiff modelling of the cervical and thoracic spines of the Q3s model.

As the modified FE model has taken actual child cadaver data (which was available after the the release of these ATDs) into account in terms of anthropometry, biomechanics, and kinematics, it should be a more accurate representation of a biofidelic response of a child in car accidents as compared to data that was scaled from adult data. This is because of the huge difference in the physiology, density of bones, geometry, and other factors as mentioned in the “Biomechanics Basis” section on this page. Hence, it is important to note that, from the data collected from a computational model created by Li et al., there are some design specifications that need to be modified on the Q-series, especially the cervical and thoracic spines, in order to achieve a more accurate and biofidelic representation of a child in response to different impacts.

Challenges and Future Work

Throughout this wiki page, challenges of designing a biofidelic child ATDs have been mentioned. A critical issue is the lack of biomechanics data for children; because volunteer and cadaver tests are not options for this demographic, data has been gathered using animal surrogate testing and through the scaling of adult data. Both these methods make broad assumptions that often fail to recognize the physiological and developmental differences between a human child and a human or animal adult. This is especially important when examining the structures above the shoulders, which see the most common and most severe injury modes - the accuracy of child ATD biofidelity and PRVs has suffered due to assumptions made regarding deficiencies in skull and neck structures. For example in 1996, General Motors set a 3-year-old's 15 ms HIC value at 1480 for a 1% risk, based on data scaled from animal tests. By contrast, researchers in 2003 considered a 3-year-old's 15 ms HIC to be just 568 for a 5% risk. These concerns were acknowledged by early researchers as possible sources of error, but likely failed to recognize the magnitude of that error. The data that forms the foundation of child ATDs that are still in use today are questionable at best.

More assumptions have been made in the design process of child ATDs that serve to exacerbate the challenges stemming from the poor biomechanical basis. The CRABI, for example, accounted for decreased head stiffnesses by applying a general safety factor. This was done even with the knowledge that the cartilage between fontanelles would be a deciding factor in skull flexibility - due to the complex nature of the system, however, this was not modelled. As a result, CRABI heads likely don't have a representative stiffness and don't account for any viscoelastic properties. Response requirements for the examined dummies were often scaled from adult data in ways that failed to capture the relevant behaviors discussed in Table 6.

In papers that do acknowledge these sources of error, it is noted that “many of the methods of applying these data as described [...] have not been substantiated through biomechanical testing. These methods, as well as the basic criteria, are subject to change with future analyses and testing”. In other words, researchers did the best they could with the data they had. While dummy construction remains largely the same over two decades later (due to their tendency to become more entrenched in common use as they produce more values, and to the time and expense required to make replacements), there have been updates to child dummy PRVs and to injury criteria, and efforts continue to make ATD use more representative of the real world scenario. As discussed in summaries by Mertz et. al., organizations and researchers have worked to re-evaluate the assumptions carried out during scaling and animal testing exercises, updating PRVs to more reasonable values. As more anthropometric, biomechanical, and biomaterial data becomes available, these evaluations may continue. Additionally, advancements in computer analysis offer opportunities to confirm or reject models and criteria, as described briefly in the study done by Li et al. in 2011 - entire papers could be written on how finite element analysis can fill the data gaps in this unique scenario, where volunteer, cadaver, and animal testing is insufficient.

Another challenge is that, as seen in the development of Q-series and Hybrid III, even with more representative data, to address a particular issue would result in negative repercussions related to other biomechanical features. A potential idea to address this is using sophisticated computational models with highly detailed material characteristics and dimensions to try and find the close-to-perfect material combination to reach our biofidelic needs. Although this would be very resource intensive, it would help solve this unreliability with child ATD performance and establish a dependable dataset for future injury analysis.

Conclusion

From 1988 to 2019, there has been a great increase in child fatality rate in car crashes which could be due to the increase of the number of children in traffic; at the same time, this could be an indication that car safety needs to be assessed more effectively for children. Since child ATDs were created in 1970, there have been many advancements to improve their biofidelity. However, studies have shown that the current ATDs, including CRABI, Q-series, and Hyrid III 10-year-old, are still not exactly biofidelic according to the biofidelic corridors, even those scaled from adult data. This is due to a few challenges including the lack of biomechanics data for children and addressing a particular issue can result in negative repercussions to other biomechanical features. Some work that could be done in the future includes using computational models to help find improvements required on ATDs and to optimize material characteristics and dimensions for the biofidelic needs.