Includes

Fig 11/12

EVALUATION OF THE 95TH PERCENTILE HIII LARGE MALE DUMMY

Joshua Shaw applied to the HIII 50th percentile male specifications

Transportation Research Center Inc. to develop dimensional, weight and impact
Edward Probst performance requirements for 5th and 95th percentile
Bruce Donnelly HIII dummies.
NHTSA – VRTC
United States Although the HIII 5th percentile small female is
Paper Number 07-0228 currently regulated by the Code of Federal
Regulations, Title 49, Part 572, the 95th Percentile
ABSTRACT Large Male is not; albeit a 95th surrogate is
referenced in FMVSS No. 202 governing head
This study summarizes results from a preliminary restraints.
evaluation of the Hybrid III Large Male Dummy.
The paper reviews the dimensions, certification test The purpose of this evaluation is to support a future
responses, durability, and repeatability and Federalization process by conducting a thorough
reproducibility of two Hybrid III 95th Percentile inspection of component and full-dummy external
Large Male dummies, each produced by a different dimensions of two HIII 95th large male dummies
manufacturer. Response data from multiple repeats from different ATD suppliers. In addition,
of the proposed Calibration and Inspection Test preliminary evaluations of the repeatability and
Procedure sections from the SAE Draft of the “User’s reproducibility (R & R) and laboratory durability of
Manual for the HIII 95th Large Male Test Dummy” each dummy are also performed.
(SAE, 2003) are used as the basis for the analysis.
An overview of the test methodologies employed is Two draft documents currently being developed by
presented as well as an assessment of the compliance the SAE Dummy Test and Evaluation Subcommittee
with the proposed certification specifications, are used as templates for the experimental aspects of
durability at high energy test levels and repeatability this investigation. A draft version of the SAE
and reproducibility. Besides employing the typical Engineering Drawing Package (SAE EA 32) is used
repeatability measures of traditionally-calculated as the basis for a common dimensional and weight
coefficients of variation, a new technique is proposed inspection of both dummies. The R & R test plan is
and discussed involving a time-variant method of based upon multiple repeats of the major certification
calculating coefficients of variation. tests enumerated in the SAE draft of the “Users
Manual for the HIII 95th Large Male Test Dummy”
INTRODUCTION (SAE, 2003). The durability assessment is developed
by scaling the certification tests to energy levels
The first anthropomorphic representation of a large associated with proposed injury levels for the large
male was developed by the Sierra Engineering male.
Company in the late 1940’s and early 1950’s under
contract to the U.S. Air Force. This 95th percentile For the purposes of this project two dummies were
anthropomorphic test device (ATD) known as “Sierra purchased from two manufacturers, one from Denton
Sam,” was primarily used for early aerospace and ATD (DATD) and one from First Technology Safety
high altitude parachute experimentation. With the Systems (FTSS). Dummy #077 is from DATD and
development of more biofidelic ATDs in the 1970’s dummy #226 is from FTSS.
and early 1980’s, such as GM’s Hybrid generation of
50th percentile dummies, a foundation was METHODS
established for the creation of a “family” of these
advanced dummies. Inspection

In 1987 the Center for Disease Control (CDC) The drawing review process involves examining all
awarded a grant to the Ohio State University for the of the engineering drawings for clarity and accuracy.
development of this Hybrid III (HIII) family of These drawings are checked both by comparing the
dummies. The SAE formed a task group of industry, physical part dimensions to the drawing dimensions
government, and academic biomechanics experts to as well as by comparison to the contents of other
support this work through the generation of federally regulated dummies, such as the HIII 50th
geometric and mass scale factors. These factors were percentile male dummy. The inspection process is a

Shaw 1
physical check of the dummy components in both component. Mertz et. al. (1989) also specifies a
segment weights and external dimensions. This can corridor for the neck flexion response. The corridor
also include the inspection of individual part coordinates are shown in Table 1.
dimensions to the corresponding drawing(s). The
segment weights and external dimensions of each Table 1.
dummy are listed in the Results section along with Mertz et. al. OC moment versus D-plane angle
the specification. neck flexion corridor
Upper Boundary Lower Boundary
Certification Testing Coordinates Coordinates
Moment Rotation Moment Rotation
The certification tests utilized to determine the [Nm] [deg] [Nm] [deg]
repeatability and reproducibility of the HIII 95th 0 0 0 34
Large Male dummy are those as specified in the SAE 83 14 37 53
User’s Manual. These types of tests are the same as 83 43 120 73
those specified for the 50th percentile Hybrid III 120 64 258 77
dummy, but the dynamic response requirements have 258 68
been scaled (Mertz, 1989) for a large male dummy
based on geometric similitude. The tests specified by
the User’s Manual include frontal thorax impact, Neck Extension Certification Test - The SAE
frontal neck flexion and extension tests, knee impact, User’s Manual specifies an impact velocity of 5.90 –
and knee slider impact. 6.19 m/s at the pendulum CG and a pulse profile as
measured by the integrated pendulum acceleration.
Head Drop Certification Test - Head drop Mertz et. al. (1989) also specifies a corridor for the
certification tests are conducted by dropping an neck extension response. The corridor coordinates
isolated head from a height of 376 mm onto a 50.8 are shown in Table 2.
mm thick steel plate and measuring the triaxial
acceleration at the center of gravity (CG) of the head. Table 2.
The resultant acceleration is calculated from the Mertz et. al. OC moment versus D-plane angle
filtered acceleration time histories and according to neck extension corridor
the SAE User’s Manual should be between 220 – 265 Upper Boundary Lower Boundary
g at peak and the lateral acceleration (Y axis) should Coordinates Coordinates
be less than 15 g in either direction (positive or Moment Rotation Moment Rotation
negative Y axis). Mertz et. al. (1989) specifies the
[Nm] [deg] [Nm] [deg]
same acceleration range for the head drop.
0 0 0 -48
-42 -19 -11 -77
Neck Flexion Certification Test - The neck
-42 -58 -92 -92
flexion and extension tests are conducted using a
pendulum to which the neck and head are attached. -65 -77
The pendulum is released from a given height to -92 -82
produce a desired velocity at the bottom of the swing
arc. As it reaches the bottom vertical position it is
arrested by an energy absorbing component. Thorax Certification Test - The thorax
Although the pendulum arm motion stops, the inertia certification test is a frontal impact to the dummy
of the head causes the neck to flex in relation to the chest centered midsternally with a 23.36 kg
pendulum. A six axis upper neck load cell was used pendulum traveling at a velocity of 6.71 +/- 0.12 m/s
to record the forces and moments generated about the at impact. In addition to the peak response
occipital condyle (OC) joint while two rotary requirements stated in the SAE User’s Manual, Mertz
potentiometers and a connecting rod are used to et. al. specify a force versus deflection corridor
determine the angle of the head in relation to the (Table 3) for the entire impact.
pendulum, which represents the torso. This angle is
known as D-plane rotation. Knee Impact Certification Test - Knee impacts
are administered at velocities between 2.07 – 2.13
The SAE User’s Manual specifies an impact velocity m/s by a 5-kg pendulum to the knee of a flexed lower
of 6.89 – 7.13 m/s at the pendulum CG and a pulse extremity. It was assumed that both knees of the
profile measured by the integrated pendulum same dummy are the same and therefore only one
acceleration, to define the energy absorbing knee from each dummy is compared.

Shaw 2
 Table 3. of the response, is important. When comparing
Mertz et. al. force versus deflection thorax maximum response values from repeat tests the
corridor values may not be occurring at the same point in time
Upper Boundary Lower Boundary and a small or large CV may not actually be a valid
Coordinates Coordinates assessment of repeatability. Further, it may be that
Force Deflection Force Deflection the maximum values are the only point in time at
[kN] [mm] [kN] [mm] which the dummy responses are similar and the
4.85 6.6 3.65 6.6 repeatability is actually poor at all other points in
5.07 26.4 3.78 2.64 time. In this study of the 95th percentile male dummy
5.34 39.6 3.96 3.96 an assessment of repeatability will be made over a
6.05 64.8 4.45 6.07 portion of the response time history.
4.94 80.8 3.47 6.35
3.47 85.9 0.98 5.41 A set of six force versus time response curves from
0.98 76.7 the thorax certification tests conducted on dummy
077 are shown in Figure 1.

Knee Slider Certification Test - The knee slider

certification test was conducted in nearly the same
manner as the knee impact test. However, the lower
portion of the leg is removed and an impact fixture is
attached to the knee slider. A 12-kg pendulum
strikes the fixture at velocities between 2.70 – 2.80
m/s. Rather than the SAE specified deflection
response, Mertz et. al. specified a stiffness corridor,
137 – 187 N/mm.

Repeatability and Reproducibility

In the traditional method for assessing the

repeatability and reproducibility of a dummy design,
a test is repeated a number of times and the average, Figure 1. Central portion of force versus time
standard deviation, and coefficient of variation (CV) response for certification tests conducted on
of the peak of the response parameters are quantified. dummy 077.
Accepted practice is that a CV between 0 – 5% is
Excellent, 5 – 8% is Good, 8 – 10% is Acceptable, It can be seen that the curves are all quite similar but
and > 10% is Unacceptable (Rhule 2005). Once a that there are slight differences in magnitude and
dummy response is shown to be repeatable (by timing of peaks, e.g., Test 1 curve versus Test 2
displaying an acceptable or better CV), the curve. A mean force versus time curve can be
reproducibility is assessed by combining the results generated from these six digitized curves by
from all tests of all dummies and applying the same averaging the force values at each time increment.
standard for CV. This approach has several Similarly, a standard deviation and a coefficient of
shortcomings and a modified approach will be variation for the six responses can also be generated
followed in this study. at each time increment. At the maximum mean force
(0.0221 seconds) the CV is 1.1%; however, at
The traditional approach for assessing dummy another point in time near the maximum mean force
repeatability, as described above, has been to test a (0.0157 seconds) the CV is 0.5%; less than half. For
dummy several times, in identical configurations, and the portion of the curve that is above half of the
to calculate the CV at the maximum points of a maximum value (i.e., all data above 50% of the peak
relevant response parameter. This approach assesses value of the mean curve), the central portion of the
the dummy response at the point in time at which response data, the CV varies from 0.5% to 8.5% with
most certification criteria are established, e.g., an average CV of 2.2%. Although the majority of
maximum chest force or maximum neck rotation. these CV values are below the accepted standard of
The drawback to this approach is that the dummy 5%, indicating very good repeatability, the CV
response is a time history and the time at which the calculated in the traditional manner (1.1%, see Table
maximum value occurs, as well as the overall shape

Shaw 3
13) is not representative of the CVs over the central lies within a ±5% corridor (actually 4.5%) of the
portion of the curve. mean. Practically speaking, it is 90% certain that the
calculated mean is a good estimate of the true
If it is assumed that the important part of a response response. Similarly, if the sample response data has a
curve is the central portion of the curve, a confidence CV of 10%, we can be 90% confident that the true
interval band can be drawn around the mean based on mean lies within a ±10% corridor (actually 9%) of
the sample response curves obtained from the six the mean. This analysis makes the usual assumptions
tests. It can be stated that we are 90% confident that of statistical independence and a normal distribution
the true mean lies within a confidence interval of response data. It must be noted that the number of
estimated from the sample responses we have. The repeat tests in this example was limited by practical
equation for calculating a confidence interval (CI) is considerations to N=6.
(Bendat, 2000):
In this study repeatability will be assessed by
St
CI = µ ± α / 2
( ) (1)
calculating the CV at all points in the time history of
the central portion and averaging all of these values
N to obtain a representative CV for that portion. If the
average CV is below 5% the repeatability will be
where S = the unbiased sample standard deviation as considered to be excellent.
a function of time
μ = the sample mean value as a function of time Dummy to dummy reproducibility, where each
N = the number of tests dummy has already been shown to be repeatable, is
tα/2 = the two-tailed t-statistic for 90% = 2.015. assessed with a comparison of means by the Student
t-test. A null hypothesis is selected stating that the
Thus, the confidence interval band (CIB) (the average of the absolute value of the paired
projected range of mean variation about the sample differences between the two dummy means is greater
mean) can be defined as: than 10% of the mean of the means. We will test this

( )
hypothesis using the one-tailed T statistic tested
S tα / 2 against the Student t-value at the 10% probability
CI B = 2 ∗ (2) level (α = 0.05). The degrees of freedom will be the
N
number of pairs, which is large and can be considered
as infinity when using the table of t-values. The test
It can be stated that one half of the confidence
statistic uses the standard error for the paired
interval band divided by the sample mean is
differences.
analogous to the CV

(
CI B S tα / 2
= =
)
CV tα / 2 ( ) (3)
d
∑ (0.1)(µ ) − µ − µ 
2µ µ N  1i 2i 
T = i =1
N

Note that the unbiased standard deviation has a

S
d
( )( )
d
(5)

denominator of N-1 whereas the biased standard

deviation, with a denominator of N, is normally used Where μ1 and μ2 are the sample means
in the traditional CV calculation ([N/(N-1)]=1.09 μ = the mean of the means
times the biased CV yields the unbiased CV for d = the number of pairs of data points
N=6). If the “excellent” value of CV=5% is used in Sd = the standard deviation of the differences.
the calculation, the 90% confidence interval is ±4.5%
of the sample mean, as shown below. If the test statistic is larger than the t-value of 1.282
the null hypothesis is rejected and we can assume the
CI B 1.09CV (tα / 2 ) responses from the two dummies are from the same
= population and can be combined. If the two sets of
2 N
(4) data are likely from the same population, it is
1.09(0.05)(2.015)µ
= = 0.045µ reasonable to calculate a mean from all of the data for
6 the dummy and apply the same procedure as was
discussed previously for repeatability. In the case of
In other words, if the sample response data has a CV reproducibility we will apply the more liberal 10%
of 5%, we can be 90% confident that the true mean standard of acceptable reproducibility. If the two

Shaw 4
dummies are significantly different, i.e., if the null 10 kN 12.7 kN 571.208
hypothesis is not rejected, then the dummies are not The data from Table 4 was then used to determine the
reproducible. inputs for durability testing. Rhule et. al. (2005)
simply increased the input energy by a nominal 30%.
Durability Doing this however did not guarantee that a test type
would result in a response at or exceeding the injury
To determine the durability of a dummy component, values. For the current study the outputs were used
the input is incrementally increased to levels above in a variety of engineering relations to determine the
the standard certification tests and the component inputs required to produce a desired response. For
inspected between test intervals. All dummies are example, to achieve the thorax injury level of a force
designed and created with the intention of being used of 6.05 kN and a deflection of 79 mm, the 23.4 kg
in a situation that could potentially cause injury to a pendulum mass must impact the dummy chest at
human. Dummies are routinely subjected to events at some as-yet-unknown velocity. Conducting an
or near the injury level in actual crash testing. energy balance calculation that relates the work done
Therefore it is necessary to have a dummy that is to the input energy, the input velocity can be
durable up to and beyond the injury levels that have determined. The specific approach for each test type
been identified. is described below.

Injury levels for the 95th Male dummy were either Head Drop - The Head Injury Criteria (HIC36) is
determined from the corridors in Mertz (1989) or by specified at 1000 for both the 5th percentile small
scaling the injury values for the Hybrid III 50th Male female and the 50th percentile male dummies (49
listed in 49 CFR 571.208 up to the 95th Male using CFR 571.208), therefore there was no reason to
the scale factors listed in Mertz. It should be noted change the injury level for the larger 95th percentile
that the values employed are not the Injury dummy. HIC36 was calculated for several of the
Assessment Reference Values that would accompany certification tests completed prior and it was noted
a Final Rule for the 95th percentile dummy. Rather that the values obtained were nearly at 1000. As a
they are simply values used to determine an upper result the maximum head drop distance was increased
level that the durability testing should attempt to to the greatest height that the in-house drop fixture
reach. The injury values that were determined to be would allow to achieve HIC36 results above 1000.
sufficient for this testing are listed in Table 4. The head was then dropped from the standard
certification test height (376 mm), an intermediate
Table 4. height (415 mm), and the highest height (450 mm).
Assumed injury levels for 95th Male dummy
Body 50th Male 95th Male 95th Value Neck Flexion - Injury values for neck flexion
Segment Injury Injury Source were derived from the corridors specified in Mertz
Level Level (1989). The angle and moment were then used in an
Head HIC36 HIC36 49 CFR energy calculation to determine the input kinetic
1000 1000 571.208 energy and thus the input pendulum velocity. It was
Neck OC OC Assumed assumed that the input translational kinetic energy
Flexion Moment Moment Mertz would be converted to the output rotational work in
190 Nm 258 Nm (1989) this test type. Before the injury level velocity was
Θ = 80° Θ = 77° Assumed determined a nominal loss term was calculated from
Mertz the certification test averages so that losses could be
(1989) approximated in the higher energy tests. For
Neck OC OC Assumed example, the average input velocity was 6.99 m/s and
Extension Moment Moment Mertz the average moment and rotation were 125 Nm and
57 Nm 92 Nm (1989) 61°, respectively. Solving the energy equation for
Θ = 95° Θ = 92° Assumed the loss portion gives
Mertz
(1989) 1
Loss = mv 2 − Mθ (6)
Thorax Deflection Deflection 49 CFR 2
= 76 mm = 79 mm 571.208
Force = Force = 49 CFR where m is the mass of the head and neck (6.65 kg),
5.43 kN 6.05 kN 571.208 M is the moment, and θ the rotation,
Knee Force = Force = 49 CFR

Shaw 5
 2 certification velocity (5.99 m/s), an intermediate level
1  m
Loss = 6.65kg *  6.99  − 125 Nm * 1.06radians (7) (6.60 m/s), and the injury level (7.24 m/s).
2  s
Thorax - The determination of the input energy
which equals a loss of 30.1 joules. Now if Equation for the thorax impact test to achieve injury level was
6 is algebraically manipulated to solve for velocity similar to that of the neck flexion and extension. For
and the injury level values and the loss term are this test however, the input kinetic energy was related
input, it becomes to the linear work output. Also similar was the
calculation of the nominal energy loss in the
certification tests. Using the average impact velocity,
v=
2
(Mθ + Loss ) (8) peak force, and peak deflection, the loss was
m
calculated as follows

v=
2
(258 Nm * 1.34rad + 30.1J ) (9) Loss =
1
mv 2 − Fd (12)
6.65kg 2

The velocity required to produce the injury level is where m = 23.36 kg and v, F, and d are the averages
10.6 m/s. Therefore the neck flexion tests would be from Table 10 to give
performed at the certification test velocity (6.99 m/s),
at an intermediate level (8.80 m/s), and at the injury 2
1  m
producing level (10.6 m/s). Unfortunately, the Loss = 23.36kg *  6.75  − 5.79kN * 67.7 mm (13)
equipment available could only reach heights capable
2  s
of producing a 7.50 m/s impact velocity. It was
decided to test at the certification velocity (6.99 m/s), which results in a nominal loss of 140 joules.
an intermediate velocity (7.25 m/s), and the upper Equation 12 is algebraically manipulated to solve for
limit of the equipment (7.50 m/s). velocity

Neck Extension - The same activity was

completed for the neck extension test to determine v=
2
(Fd + Loss ) (14)
m
the input required to produce an injury event in the
dummy. The injury values for extension were also and the injury level parameters are used to determine
assumed from the corridors specified in Mertz the velocity required
(1989). Equation 6 was used again to determine the
nominal loss in a standard neck extension
certification test. The average velocity for these tests v=
2
(6.05kN * 79mm + 140 J ) (15)
was 5.99 m/s and the average moment and rotation 23.36kg
were 67.4 Nm and 79.0°, respectively. Solving
Equation 6 with these inputs gives, which is 7.27 m/s. The in-house equipment had the
ability to conduct impacts at greater speeds and
2
1  m therefore it was decided to test at the maximum
Loss = 6.65kg *  5.99  − 67.4 Nm *1.38radians velocity the equipment could produce. The thoraces
2  s
of the dummies were tested at the certification
(10) velocity (6.75 m/s), an intermediate velocity (7.50
m/s) which is near the injury level, and at the highest
where the loss equals 26.3 joules. Similar to the neck velocity attainable (8.33 m/s).
flexion test, the injury level inputs and the loss term
are used in Equation 8 to determine the velocity that Knee Impact - The knee impact test does not
would cause an injury reading measure a deflection, but rather just a force. Instead
of balancing the energy to determine the injury level
v=
2
(92 Nm * 1.61rad + 26.3J ) (11)
input, the velocity requirement was found using
conservation of linear momentum
6.65kg

the required velocity was 7.24 m/s. This velocity was

attainable and the dummy necks were tested at the
∫
mv f − mvi = Fdt + Loss (16)

Shaw 6
where m is the mass of the pendulum and vf is solved using a different mass, indicating a different
approximately 0.73 m/s in the opposite direction after pendulum to be used. The decision was made to use
the impact. The loss was accounted for in the the 12 kg pendulum that was specified for the knee
standard certification test as follows slider tests, rather than the standard 5 kg impactor, to
reduce the required velocity. Solving Equation 23
(
Loss = m v f − vi − Fdt ) ∫ (17)
again with a higher mass results in a velocity of 3.70
m/s, a value much more reasonably attained. Since
the test velocity was to be incrementally increased
 m from the certification velocity, it was decided to use
Loss = 5kg  − 0.73 − 2.10  − .5(.00475s )(− 6451N )
m
the larger mass for all of the durability impacts rather
 s s
than switching part of the way through the test plan.
(18)
Therefore, the certification test velocity would have
and found to be 1.98 kgm/s. The impulse was to be reduced to account for the larger mass. The
approximated by integrating a triangular waveform velocity was scaled based on kinetic energy from the
with a height equivalent to the average peak force 5 kg, 2.10 m/s impact to the 12 kg impact. This
and a base equivalent to the duration of the force resulted in a 1.36 m/s impact required to maintain the
pulse from the certification tests. Now solving input kinetic energy requirement. The knee impact
Equation 16 for vi to obtain the injury level output test velocity was incremented up to the injury level
gives similar to the other tests. The adjusted certification
test velocity was completed first (1.36 m/s), then an
∫
mvi = mv f − Fdt − Loss (19) intermediate velocity (2.53 m/s), and then the
velocity required to produce injury (3.70 m/s).
with the final, or rebound, velocity being a product of
the input velocity. Therefore, the coefficient of RESULTS
restitution, ‘e’, was calculated from the certification
parameters such that Inspection
The inspection of each dummy yielded a positive
mv f 0.73 m result. The weights for all segments of both dummies
e= = s = 0.35 (20) were within the specifications. The majority of the
mvi 2.10 m external dimensions were within the specification for
s
both dummies, however the ‘hip pivot from backline’
which makes Equation 19 become and the ‘head circumference’ were outside of the
specification on dummy 226. The ‘reference location

∫
for chest circumference’ was also slightly high for
mvi − mevi = − Fdt − Loss (21)
both dummies. The measurements for both dummies
can be seen in Tables 5 and 6 and Figure 2.

vi =
∫
− Fdt − Loss
(22)
Certification Testing
m(1 − e )
Head Drop Test – The results for the head drop
Substituting the calculated loss, the desired force certification test are listed in Table 7 and the response
value and the mass into this equation allows for the time histories are shown in Figures 3 and 4 along
determination of the velocity required to produce the with the SAE and Mertz corridors. The head
injury level force response. The duration of the force resultants for dummy 226 were at the top bound of
response was assumed to be equivalent to that of the the corridor while the resultants for dummy 077 were
certification tests. Thus just slightly higher than the top bound of the corridor.
Dummy 226 had essentially no differences in phase
−.5 * .00475s * (−12700 N ) − 1.17 kgm / s while dummy 077 had slight differences. Both
vi = (23)
5kg (1 − 0.35)
dummies had nearly the same duration for the
impacts.

results in a velocity of 8.89 m/s. This velocity is Neck Flexion – The results for the neck flexion
greater than the highest velocities attainable on the certification test are listed in Table 8 and the response
knee impact fixture. Therefore the equation was time histories are shown in Figures 5 and 6 along

Shaw 7
with the SAE and Mertz corridors. The neck flexion the Mertz corridor. All of the test results on dummy
responses were similar for each dummy but were just 226 were within the SAE corridor but all were higher
short of the corridor. Dummy 077 had three of the than the Mertz corridor. Five of the six tests on
six tests enter the corridor but dummy 226 only had dummy 077 were above the Mertz corridor but all
one test enter the corridor. Dummy 226 only were within the SAE corridor. There were some
obtained one value for the OC Moment within the slight variations in phase for each of the dummies.
rotation interval and, as a result, an average, standard
deviation, and CV could not be calculated. Knee Slider – The results for the knee slider
Subsequently a CV could not be calculated for certification test are listed in Table 12 and the
reproducibility for this parameter. All of the tests on response time histories are shown in Figures 13
both dummies did however match the Mertz corridor through 16, along with the SAE and Mertz corridors.
fairly well, with the peak portion being the exception. Nearly all of the knee slider deflections were within
the SAE corridor and all very similar. Six of six tests
Neck Extension – The results for the neck on dummy 077 and five of six tests on dummy 226
extension certification test are listed in Table 9 and were within the deflection corridor. All of the tests
the response time histories are shown in Figures 7 for each dummy were similar in shape and phase for
and 8 along with the SAE and Mertz corridors. The the deflection responses. The knee stiffness for
neck extension responses were quite similar but again dummy 077 matched the Mertz corridor slightly
were just short of being within the corridor. None of better than dummy 226. The majority of the stiffness
the tests for either dummy entered the SAE corridor. response for dummy 077 matched the slope of the
Five of the six tests on dummy 077 produced a Mertz corridors while the loading portion of the
rotation that was within the specification, but all were stiffness response of dummy 226 did not match the
below the minimum OC Moment. This resulted in an Mertz corridor.
average, standard deviation, and CV to be calculated,
but the average was below the lower bound of the OC Repeatability and Reproducibility
Moment specification. None of the tests on dummy
226 produced enough rotation to meet the The traditional CV based on peak values and the
specification, thus the OC Moment parameter does time-based average CV were calculated in the
not have an associated average, standard deviation, or manner described previously for each of the response
CV. Likewise, a CV for reproducibility could not be parameters from the certification tests. If both of the
calculated for the OC Moment parameter. The dummies were repeatable, the time-based means were
majority of the tests on dummy 077 matched the then compared using the t-test. If the two means
Mertz corridor while most of the tests on dummy 226 were from the same population the twelve responses
did not match the corridor, specifically in the peak from both dummies were used to calculate a single
region. time-based mean and the average CV for both
dummies was calculated to assess reproducibility. If
Thorax – The results for the thorax impact the two dummies were found to be from different
certification test are listed in Table 10 and the populations they are considered not reproducible.
response time histories are shown in Figures 9 and 10
along with the SAE and Mertz corridors. The thorax All of the repeatability and reproducibility CVs are
impact responses were excellent and nearly all were presented in Table 13. Under the heading of
within the corridor. Five of the six tests on dummy Repeatability the time-based average CV and the
077 and six of six tests on dummy 226 were within traditional peak CV are shown. Under the heading of
the SAE corridor. In addition, all of the tests Reproducibility the t-statistic, the time-based average
matched the Mertz corridor very well. CV, and the traditional peak CV are shown. If the t-
test indicated the two dummies were not from the
Knee Impact – The results for the knee impact same population the average CV was not calculated
certification test are listed in Table 11 and the because the dummies were not reproducible. The
response time histories are shown in Figures 11 and traditional CV for peak values is presented in
12 along with the SAE and Mertz corridors. The parenthesis although the merit of these CV values is
knee impact responses for both dummies were all questionable since the two dummies were found to be
within the SAE corridor, but most were higher than from different populations.

Shaw 8
 Table 5.
Dummy segment specification and actual weights
Segment Weight Part # Specification [kg] Dummy 226 Dummy 077
Head Assembly 880995-1100 4.94 +/- 0.05 4.92 4.94
Neck Assembly 880995-1250 1.68 +/- 0.05 1.69 1.72
Upper Torso Assembly 880995-1300 22.3 +/- 0.36 22.0 22.5
Lower Torso Assembly 880995-1450 30.3 +/- 0.36 30.4 30.0
Upper Arm, Left 880995-700 2.81 +/- 0.09 2.79 2.83
Upper Arm, Right 880995-700 2.81 +/- 0.09 2.82 2.81
Lower Arm, Left 880995-732 2.06 +/- 0.05 2.05 2.05
Lower Arm, Right 880995-732 2.06 +/- 0.05 2.06 2.06
Hand, Left 78051-208 0.57 +/- 0.05 0.55 0.56
Hand, Right 78051-209 0.57 +/- 0.05 0.53 0.54
Upper Leg, Left 880995-1513 8.21 +/- 0.09 8.20 8.17
Upper Leg, Right 880995-1514 8.21 +/- 0.09 8.16 8.16
Lower Leg, Left 880995-1513 5.75 +/- 0.09 5.81 5.83
Lower Leg, Right 880995-1514 5.75 +/- 0.09 5.80 5.82
Foot, Left 880995-1600 1.59 +/- 0.07 1.58 1.61
Foot, Right 880995-1601 1.59 +/- 0.07 1.53 1.61
TOTAL WEIGHT 101 +/- 1.63 101 101

Figure 2. Reference diagram for dummy external dimensions listed in Table 6.

Shaw 9
 Table 6.
Dummy external dimension specification and actual measurements
External Dimension Key Specification [mm] Dummy 226 Dummy 077
Total Sitting Height A 919 +/- 15 921 927
Shoulder Pivot Height B 536 +/- 15 532 541
Hip Pivot Height C 102 +/- 5 94 97
Hip Pivot from Backline D 155 +/- 5 142 152
Shoulder Pivot from Backline E 91 +/- 5 97 93
Thigh Clearance F 168 +/- 8 168 173
Back of Elbow to Wrist Pivot G 310 +/- 8 311 312
Head Back from Backline H 89 +/- 3 89 89
Shoulder to Elbow Length I 363 +/- 10 356 353
Elbow Rest Height J 213 +/- 10 210 216
Buttock to Knee Length K 648 +/- 13 643 648
Popliteal Height L 470 +/- 13 470 470
Knee Pivot to Floor Height M 533 +/- 13 523 533
Buttock Popliteal Length N 503 +/- 13 503 503
Chest Depth O 246 +/- 8 246 246
Foot Length P 264 +/- 8 269 264
Buttock to Knee Length R 579 +/- 13 577 579
Head Breadth S 155 +/- 5 160 156
Head Depth T 196 +/- 5 201 199
Hip Breath U 404 +/- 10 409 403
Shoulder Breath V 475 +/- 10 472 475
Foot Breath W 99 +/- 8 101 99
Head Circumference X 572 +/- 5 584 574
Chest Circumference with Jacket Y 1135 +/- 20 1130 1146
Waist Circumference Z 1008 +/- 20 988 988
Reference Location for Chest Circumference AA 483 +/- 5 508 508
Reference Location for waist Circumference BB 203 +/- 5 203 203

Table 7.
Head drop certification test results for R&R
Dummy No. Maximum Resultant Acceleration
220 – 265 g
Average 278
077 Std Dev 1.99
%CV 0.72
Average 265
226 Std Dev 1.03
%CV 0.39
Average 271
Both Std Dev 6.95
%CV 2.56

Shaw 10
Figure 3. Dummy 077 head drop response. Figure 4. Dummy 226 head drop response.

Figure 5. Dummy 077 neck flexion response. Figure 6. Dummy 226 neck flexion response.

Figure 7. Dummy 077 neck extension response. Figure 8. Dummy 226 neck extension response.

Shaw 11
Figure 9. Dummy 077 thorax impact response. Figure 10. Dummy 226 thorax impact response.

Table 8. Neck flexion certification test results for R&R

OC
OC
Peak Moment
Impact Moment Peak OC
Pendulum Velocity Profile D-Plane Peak in
Velocity 10 Nm Moment
Rotation Rotation
Decay
Interval
m/s @ m/s @ m/s @
m/s deg Nm ms Nm
10ms 20ms 30ms
6.89-7.13 2.2-2.7 4.0-5.0 5.7-6.9 61-75 110-130 77-97
Dummy 077
Average 6.99 2.44 4.60 6.62 62.1 123 80.3 123
Std Dev 0.00 0.01 0.03 0.04 4.60 6.31 0.35 5.72
CV % 0.00 0.56 0.75 0.60 7.40 5.13 0.44 4.63
Dummy 226
Average 6.99 2.50 4.62 6.53 59.9 N/A 79.7 126
Std Dev 0.00 0.03 0.03 0.03 1.37 N/A 0.44 0.95
CV % 0.00 1.31 0.74 0.48 2.29 N/A 0.55 0.75
Both
Average 6.99 2.47 4.61 6.58 61.0 N/A 80.0 125
Std Dev 0.00 0.04 0.03 0.06 3.43 N/A 0.47 4.22
CV % 0.00 1.54 0.76 0.85 5.63 N/A 0.59 3.38

Shaw 12
 Table 9.
Neck extension certification test results for R&R
OC
OC
Peak Moment
Impact Moment Peak OC
Pendulum Velocity Profile D-Plane Peak in
Velocity 10 Nm Moment
Rotation Rotation
Decay
Interval
m/s @ m/s @ m/s @
m/s deg Nm ms Nm
10ms 20ms 30ms
5.95-6.91 1.8-2.2 3.2-4.2 4.8-5.8 81-98 66-84 100-120
Dummy 077
Average 5.96 1.93 3.69 5.32 81.2 62.5 106 64.1
Std Dev 0.02 0.02 0.03 0.05 2.02 0.90 1.48 1.10
CV % 0.41 1.22 0.74 1.01 2.49 1.45 1.39 1.72
Dummy 226
Average 6.01 1.94 3.71 5.35 76.8 N/A 104 70.8
Std Dev 0.00 0.06 0.04 0.04 1.75 N/A 1.11 2.13
CV % 0.00 2.88 1.11 0.77 2.28 N/A 1.08 3.01
Both
Average 5.99 1.93 3.70 5.33 79.0 N/A 105 67.4
Std Dev 0.03 0.04 0.03 0.05 2.91 N/A 1.80 3.89
CV % 0.52 2.13 0.92 0.90 3.69 N/A 1.72 5.77

Table 10.
Thorax certification test results for R&R
Impact Velocity Chest Dmax Fmax in Defl Corridor Hysteresis Fmax
m/s mm kN % kN
6.59 - 6.83 66.0 - 76.0 5.10 - 5.91 69 - 85
Dummy 077
Average 6.75 68.6 5.89 71.7 5.90
Std Dev 0.04 0.47 0.06 0.82 0.05
CV % 0.53 0.68 1.06 1.14 0.84
Dummy 226
Average 6.75 66.8 5.56 73.7 5.68
Std Dev 0.01 0.74 0.10 0.52 0.05
CV % 0.15 1.11 1.86 0.70 0.90
Both
Average 6.75 67.7 5.74 72.7 5.79
Std Dev 0.02 1.10 0.19 1.23 0.12
CV % 0.35 1.62 3.33 1.69 2.15

Shaw 13
 Table 12.
Knee slider certification test results for R&R
Knee Pendulum
Velocity
Slider Force
m/S mm N
2.7 - 2.8 15.0-18.3
Dummy 077
Average 2.75 17.6 3710
Std Dev 0.02 0.18 86.7
%CV 0.66 1.04 2.34
Dummy 226
Average 2.73 15.3 3920
Figure 11. Dummy 077 knee impact force Std Dev 0.00 0.42 36.3
response. %CV 0.00 2.74 0.93
Both
Average 2.74 16.4 3810
Std Dev 0.02 1.22 129
%CV 0.62 7.39 3.38

Figure 12. Dummy 226 knee impact force

response.

Table 11.
Knee impact certification test results for R&R Figure 13. Dummy 077 knee slider deflection
Velocity Pendulum Force response.
m/S N
2.07 - 2.13 4900 - 7300
Dummy 077
Average 2.10 6190
Std Dev 0.02 166
CV % 0.78 2.68
Dummy 226
Average 2.10 6710
Std Dev 0.00 148
CV % 0.17 2.20
Both
Average 2.10 6450
Std Dev 0.01 309
CV % 0.54 4.80 Figure 14. Dummy 226 knee slider deflection
response.

Shaw 14
Figure 15. Dummy 077 knee slider stiffness Figure 18. Neck flexion response for both
response. dummies at increasing pendulum velocities.

Figure 16. Dummy 226 knee slider stiffness Figure 19. Neck extension response for both
response. dummies at increasing pendulum velocities.

Figure 17. Head drop response for both dummies Figure 20. Thorax response for both dummies at
at increasing drop heights. increasing pendulum velocities.

Shaw 15
 226, respectively, nearly four times higher. This
result can be explained by examining Figure 22. The
six force versus time curves are relatively similar in
shape, duration, and especially peak values which
translates into an excellent CV value from the
traditional approach. However a distinct difference
in phase can be observed among the test results
which is attributed to the knee itself and not the test
setup (time zero defined by electrical contact switch).

Figure 21. Knee response for both dummies at

increasing 12 kg pendulum velocities.

DISCUSSION

Certification Testing

For the majority of the certification test types

conducted the responses fit within the proposed SAE Figure 22. Knee force versus time histories near
corridor. The tests that did not result in a response peak for dummy 077.
that was inside the corridor could easily be made to
fit the corridor with a slight adjustment of the Figure 23 shows the time-based average curve and
corridor bounds. For example, the head drop test plus and minus one standard deviation curves for the
could have the response corridor increased by 20g’s same six knee force responses. It can be seen that the
and then all of the tests that were conducted would be standard deviation at the peak of the curves is smaller
contained within the allowable region. Likewise, the than at other times. Further, it can be seen that the
neck flexion response corridor could be moved so standard deviation on the sloped sections of the curve
that the moment requirement is held the same, but the is essentially constant. This is due to the phase shift
rotation requirement reduced by 7 – 10 degrees and among the curves seen in Figure 22. As a result of
all responses would fall within the corridor. The this situation the CV (standard deviation divided by
neck extension corridor would have to be reduced the mean) increases as the mean value decreases and
both in terms of moment, 6 Nm, and rotation, 6 – 7 the time-based average CV is larger than the CV at
deg, so that all of the responses would fall within the the peak.
corridor. The thorax and knee impact tests already
fall within the corridor and would not necessitate a
change. However, the knee slider deflection corridor
could be reduced by 2 mm and all of the responses
would be within the specification.

Repeatability and Reproducibility

In most instances the traditional approach to CV

provides a lower value indicating a better
repeatability or reproducibility than does the time-
based average CV. This is illustrated by the knee
impact force response CVs. The traditional approach
results in a CV of 2.7% and 2.2% for dummy 077 and
226, respectively, while the time-based average CV
Figure 23. Knee force versus time mean and
results in a CV of 9.2% and 7.8% for dummy 077 and
standard deviation.

Shaw 16
Although the knee force responses for the two CV is 13% (not shown) which is larger than the
dummies were both repeatable, the t-test indicates acceptable level of 10% for reproducibility. The
that they are not reproducible. A look at the knee time-based average CV methodology is better able to
force response curves of Figures 11 and 12 indicate identify the fact that the knee force responses are not
that the two dummy knees are quite different in both reproducible.
magnitude and timing. The traditional CV, as shown
in Table 13, is 4.8% whereas the time based average

Table 13.
Time series CV and peak CV comparison along with t-statistic for repeatability and reproducibility
Repeatability Reproducibility
CV Avg CV peaks T statistic CV Avg CV peaks
Head 077 4.93 0.72
2.45 6.28 2.56
Head 226 1.31 0.39
Neck Flx Angle 077 5.77 7.40
151 4.65 5.63
Neck Flx Angle 226 2.47 2.29
Neck Flx Mom 077 2.58 4.63
11.3 5.08 3.38
Neck Flx Mom 226 1.96 0.75
Neck Ext Angle 077 3.53 2.49
2.91 5.94 3.69
Neck Ext Angle 226 2.76 2.28
Neck Ext Mom 077 2.32 1.72
-17.0 (5.77)
Neck Ext Mom 226 4.12 3.01
Thorax Force 077 2.19 0.84
56.2 3.10 2.15
Thorax Force 226 1.81 0.90
Thorax Defl 077 1.63 0.68
17.9 1.97 1.62
Thorax Defl 226 1.58 1.11
Knee Impact 077 9.18 2.68
-7.37 (4.80)
Knee Impact 226 7.80 2.20
Knee Slider Defl 077 2.20 1.04
-19.0 (7.39)
Knee Slider Defl 226 3.82 2.74
Knee Slider Force 077 3.52 2.30
-16.4 (3.38)
Knee Slider Force 226 3.50 0.93

Coefficient of Variation <= 5%

Coefficient of Variation > 5% <= 10%
CV for dummies from different populations ( # )

Considering the other body components of these CONCLUSIONS

dummies that were found to be not reproducible
(knee slider force and deflection and neck extension When evaluating the averages of six repeats of the
moment) the time-based average CVs (9%, 9%, 8%, various certification tests for each dummy, it can be
respectively) are larger than the traditional CVs but seen that neither dummy fully complies with the
the values are within the acceptable limit of 10%. proposed SAE certification requirements. Dummy
This indicates that merely averaging all of the data 077 does not certify with respect to the head drop
and calculating a CV without performing the t-test requirements while dummy 226 does not certify with
does not identify non-reproducibility; the t-test is respect to the peak D-plane angular requirements for
required. either neck flexion or extension. However, if the
certification specifications were to be adjusted both
dummies could meet the revised specifications.
Because the certification specifications were based
on scaled 50th percentile male specifications it is

Shaw 17
reasonable to adjust the specifications to fit the REFERENCES
dummies so long as the dummies being evaluated are
reproducible. It is to be expected that design and Bendat, J., Piersol, A., Random Data: Analysis and
fabrication variances would result in a 95th sized Measurement Procedures, 3rd Edition, John Wiley &
dummy that is not a perfectly scaled replica of the Sons, New York, NY. 2000.
50th sized dummy.
Code of Federal Regulations, Title 49. United States
Traditional methods of calculating repeatability CV Government. Washington, DC. 2004.
values (CV peak method) show both dummies,
individually, provide good or excellent response CVs Mertz, H., Irwin, A., Melvin, J., Stalnaker, R., Beebe,
for all of the certification tests. A proposed method M., “Size, Weight and Biomechanical Impact
of calculating a time-based CV (CV average method) Response Requirements for Adult Size Small Female
also shows the CVs for both dummies to range from and Large Male Dummies.” SAE International
good to excellent although the time-based method Congress and Exposition, Detroit, MI, 1989.
appears to be more demanding.
Rhule, D., Rhule, H., Donnelly, B., “The Process of
The reproducibility CV is good to excellent for all Evaluation and Documentation of Crash Test
certification test responses when looking at the Dummies for Part 572 of the Code of Federal
combined data from both dummies using the Regulations.” 19th International Technical
traditional method of calculating CV. However, Conference on the Enhanced Safety of Vehicles,
when using hypothesis testing of the time-based Washington, D.C., USA, 2005.
responses to study the reproducibility of the two
dummies, the Neck Extension Moment, Knee Impact, SAE Dummy Testing Equipment Subcommittee.
Knee Slider Deflection, and Knee Slider Force are “User’s Manual for the Hybrid III Large Male Test
shown to be statistically different. If the two Dummy.” Engineering Aid 26 Draft. Warrendale,
dummies are not from the same population the PA. 2003.
pooling of all responses from both dummies in order
to calculate a CV is inappropriate.

Applying energy balance and conservation of

momentum techniques to previously proposed injury
criteria for the 95th percentile large male, certification
test input energy levels were increased and used to
evaluate dummy durability. The head, neck, thorax,
and knee from both dummies were repeatedly
subjected to testing at these higher energy levels with
no adverse effects observed by physical inspection or
by examination of the dummy responses.

Only two dummies have been examined: one each

from two manufacturers. It appears that the dummies
could meet certification specifications if the current
specifications were adjusted. Durability of both
dummies was satisfactory even when tested at very
high input energy levels. Repeatability of each of the
dummies was acceptable to excellent but
reproducibility of several components was less than
desirable when tested using the Student t-test with a
time-based tolerance of 10%.

ACKNOWLEDGEMENTS

The authors would like to thank Rick Waldeck and

Ron Puckett of Transportation Research Center for
their efforts in the completion of this project.

Shaw 18