Airbag 2010 – 10th International Symposium and Exhibition
on Sophisticated Car Occupant Safety Systems,
December 6-8, 2010
CHARACTERIZATION AND EVALUATION
OF FRONTAL CRASH PULSES FOR USNCAP 2011
Jürgen Metzger, Lars Kübler and Simon Gargallo
TRW Automotive GmbH
Industriestr. 20, 73553 Alfdorf, Germany
[juergen.metzger;lars.kuebler;simon.gargallo]@trw.com
Keywords: Frontal Crash, Crash Severity, Crash Pulse Criterion, USNCAP 2011.
Abstract. Crash pulse characterization is of great importance in many fields of occupant
restraint system development. It allows assessment of severity of a specific crash pulse with
respect to the dummy. A measure for the quality of pulse criteria is given by the correlation
between dummy values and pulse criteria values. In this paper focus is given on the modified
USNCAP rating starting 2011. At first it is analyzed whether it is generally possible to find
one single pulse criterion that gives a sound assessment regarding all relevant dummy values
of the USNCAP 2011 rating. Then the correlation of existing pulse criteria to those dummy
values is evaluated. Further, an approach is proposed how to derive specific criteria for the
new rating and it is discussed how such criteria could be used in future to support restraint
system development.
1
INTRODUCTION
Occupant restraint systems are essential parts of today’s vehicles to reduce occupant
injuries during collisions. In order to evaluate the restraint performance, computer
simulations, sled tests and vehicle crash tests are conducted for several frontal collision types.
A substantial parameter in this context is the acceleration field, effective on occupants during
a crash test, the so-called crash pulse. Crash pulses are input for sled tests and simulations and
strongly influence the development of restraint systems, since their variations have significant
influence on the overall system responses.
Pulse characterization is of great importance in many fields of occupant restraint system
development. In order to allow a sound comparison of pulse “severity” TRW proposed an
enhanced criterion: OLC++ [1,2]. In addition, the possibility to estimate OLC++ threshold
values, which give an indication for restraint component selection, has been discussed in
[1,2].
The USNCAP rating will change in 2011 with modifications in utilized dummies and
considered dummy values. In particular the 5th percentile female is used on passenger side,
chest deflection is evaluated instead of chest acceleration and the new rating comprises neck
1
J. Metzger, L. Kübler, S. Gargallo: Characterization and Evaluation of Frontal Crash Pulses for USNCAP 2011
dummy values. An open question is, whether existing criteria and OLC++ are suitable for the
new rating or if modifications are necessary to cover the mechanisms driving the modified
dummy values for both dummies.
At first the correlation between dummy values themselves is analyzed. Purpose of this
investigation is to answer the question, if it is possible to establish a general crash pulse
criterion which shows good correlation to all relevant dummy values. This provides more
insight in the driving mechanisms and supports further criteria development.
An important question is also, to which extent existing pulse criteria derived for the current
USNCAP rating and other load-cases are capable to assess pulse strength with respect to the
relevant dummy values of the USNCAP 2011 rating. A measure for the quality of pulse
criteria is given by the correlation between dummy values and pulse criteria values. Therefore,
OLC++ and other existing criteria for pulse characterization are compared regarding their
correlation to dummy values in the new USNCAP load-case for a large number of crash
pulses over a range of vehicle types. The investigation is carried out in different vehicle
environments for driver and passenger.
Thereafter, an approach is proposed how to derive specific criteria for the new rating, and it
is discussed how such criteria could be used to support future restraint system development.
2 CORRELATIONS OF DUMMY VALUES AMONG THEMSELVES
In [1, 2] one single criterion has been established in order to evaluate pulse severity for the
current USNCAP rating, i.e. one criterion was feasible to give an indication for the relevant
dummy values: chest acceleration and HIC36 for the 50th perc. male dummy.
The situation gets more complicated for the new USNCAP rating with 5th perc. and 50th
perc. dummies and an increasing number of dummy values: HIC15, chest deflection, Nij, neck
tension/compression, and compressive femur forces. The question arises, whether it is still
possible to find one common pulse criterion for all dummy values or not. To answer that
question, in this section correlation of the relevant dummy values among themselves is
investigated for both dummies. A necessary condition for the possibility to derive one single
crash pulse criterion is that each injury parameter shows significant correlation to all other
dummy values under variation of the crash pulse.
For the correlation analysis approximately 400 USNCAP crash pulses from the TRW
database are imposed on MADYMO [9] occupant restraint system models. The crash pulse
database includes pulses of all vehicle classes and manufacturers. Potential influence of
vehicle structure, interior, restraint system, occupant and its position is taken into account by
using different types of cars and both 5th perc. and 50th perc. dummy on passenger and driver
side.
In the following, exemplary passenger results for a middle class sedan vehicle with a
standard restraint system, consisting of passenger airbag, constant load limiter and standard
retractor pretensioner, are discussed. Focusing on US market, the airbag is configured in order
to fulfill US legal requirements in the unbelted load cases. To ease the simulation effort, the
airbag vent definition was done once for one of the most severe unbelted crash pulses in TRW
database. Load limiter configuration was set in order to prevent bottoming out or head contact
to the instrument panel in USNCAP load case for all crash pulses in the TRW database.
For the evaluation relevant correlation was assumed for quadratic correlation coefficients
of 0.7 or larger. Strong correlation is indicated by values larger than 0.9. In Figure 2.1
correlation coefficients are given for the relevant dummy values for the 50th perc. dummy.
2
J. Metzger, L. Kübler, S. Gargallo: Characterization and Evaluation of Frontal Crash Pulses for USNCAP 2011
Chest acceleration
Neck tension
0,81
0,80
0,92
0,74
0,82
1,00
0,72
0,57
0,58
0,90
0,59
1,00
0,82
HIC 15
0,91
0,92
0,81
0,80
0,92
0,82
1,00
Neck tension
0,62
0,47
0,69
0,17
0,63
0,85
0,79
0,64
1,00
0,60
0,81
Neck compression
0,88
0,95
0,57
0,58
0,90
1,00
0,82
Chest acceleration
0,93
0,93
0,68
0,69
1,00
0,88
0,92
Femur right
0,77
0,69
0,97
1,00
0,69
0,59
0,80
HIC 15
Chest acceleration
0,93
0,66
0,93
0,68
0,72
0,68
0,69
1,00
0,57
0,88
0,92
Femur right
0,77
0,48
0,69
0,37
0,56
0,97
1,00
0,69
0,69
0,59
0,80
HIC 15
Femur right
0,72
0,71
Neck compression
Femur left
0,56
0,95
Femur right
NCF
0,92
0,55
Femur left
NCE
0,57
0,88
NTF
NTF
0,91
Neck compression
Chest deflection
NTE
HIC 15
Neck compression
Chest deflection
Chest acceleration
Because of its relevance to the FMVSS 208, chest acceleration is also observed. Green color
indicates a correlation higher than 0.9, yellow color represents values between 0.7 and 0.9.
Values without significant correlation are not highlighted.
Femur left
0,77
0,51
0,69
0,33
0,54
1,00
0,97
0,69
0,68
0,57
0,82
NCF
0,68
0,55
0,76
0,32
1,00
0,55
0,56
0,76
0,64
0,79
0,76
Femur left
0,77
0,69
1,00
0,97
0,69
0,57
0,82
NCE
0,68
0,35
0,69
1,00
0,32
0,34
0,37
0,67
0,20
0,71
0,58
NTF
0,93
1,00
0,69
0,69
0,91
0,91
0,93
Chest deflection
1,00
0,93
0,76
0,77
0,92
0,87
0,92
NTF
0,93
0,52
1,00
0,72
0,67
0,69
0,69
0,91
0,58
0,91
0,93
NTE
0,64
1,00
0,58
0,34
0,56
0,50
0,49
0,68
0,46
0,59
0,57
Chest deflection
1,00
0,61
0,93
0,71
0,63
0,76
0,77
0,92
0,54
0,87
0,92
Figure 2.1: Quadratic correlation matrix between dummy values for 50th perc. dummy (yellow: relevant
correlation, green: strong correlation, white correlation not significant)
Left: all criteria, Right: reduced to criteria with overall significant correlation
Figure 2.1 (left) shows all dummy values. Obviously not for all values significant
correlation is given to all other values. In Figure 2.1 (right) values without sufficient
correlation to most of the other criteria are removed: NTE, NCE, NCF and neck tension. For
femur forces sufficient correlation to HIC15 and chest deflection are given, however not for
neck values. In principle this result indicates for the 50th perc. male dummy:
1) Correlation of one criterion to all values in same quality is not possible.
2) For HIC15, chest deflection, neck compression, chest acceleration, and NTF it might
be possible to find a single criterion1 for crash pulse assessment.
3) For femur forces it might also be possible to be covered by such a criterion, if
reduced correlation to femur is accepted.
4) The remaining neck values: NCE, NCF, NTE and neck tension even do not fully
correlate to each other. That could indicate that all neck values could not even be
covered by a separate criterion. This will later be observed in more detail.
In Figure 2.2 correlation coefficients are given analogously for 5th perc. female dummy.
Figure 2.2 (left) shows all dummy values. Again not for all values significant correlation is
given to all other values. In Figure 2.1 (right) values without sufficient correlation to most of
the other values are removed: NTE, NTF, Ncf, neck tension and femur right. For femur forces
left and NCE correlation to some other values are given, not to all though.
Regarding femur forces, in the USNCAP rating only compressive forces are considered.
This explains low correlation for femur right, where compressive forces are almost not excited
for all pulses for this vehicle. For femur left compressive forces increase with pulse severity,
but over the full pulse range the values are still relatively low.
1
Significant correlation is only a necessary condition, i.e. that does not mean that such a criterion necessarily
exists.
3
HIC 15
Neck compression
Chest acceleration
Femur left
NCE
Chest deflection
HIC 15
Neck compression
Neck tension
Chest acceleration
Femur right
Femur left
NCF
NCE
NTF
NTE
Chest deflection
J. Metzger, L. Kübler, S. Gargallo: Characterization and Evaluation of Frontal Crash Pulses for USNCAP 2011
HIC 15
0,84
0,35
0,51
0,75
0,60
0,75
0,28
0,88
0,54
0,85
1,00
Neck compression
0,81
0,32
0,67
0,72
0,87
0,73
0,19
0,91
0,39
1,00
0,89
HIC 15
0,84
0,75
0,75
0,88
0,85
1,00
Neck tension
0,55
0,77
0,56
0,74
0,05
0,70
0,31
0,75
1,00
0,53
0,77
Neck compression
0,81
0,72
0,73
0,91
1,00
0,89
Chest acceleration
0,72
0,44
0,55
0,89
0,55
0,84
0,26
1,00
0,69
0,84
0,89
Chest acceleration
0,72
0,89
0,84
1,00
0,84
0,89
Femur right
0,21
0,43
0,15
0,05
0,24
0,07
1,00
0,05
0,29
0,06
0,26
Femur left
0,67
0,80
1,00
0,84
0,69
0,77
Femur left
0,67
0,46
0,58
0,80
0,47
1,00
0,21
0,84
0,67
0,69
0,77
NCF
0,67
0,25
0,59
0,58
1,00
0,62
0,20
0,84
0,26
0,86
0,81
NCE
0,63
1,00
0,79
0,88
0,80
0,77
NCE
0,63
0,56
0,54
1,00
0,48
0,79
0,31
0,88
0,73
0,80
0,77
Chest deflection
1,00
0,58
0,64
0,75
0,81
0,84
NTF
0,56
0,51
1,00
0,54
0,50
0,61
0,14
0,61
0,48
0,64
0,51
NTE
0,44
1,00
0,55
0,55
0,02
0,45
0,31
0,42
0,80
0,49
0,56
Chest deflection
1,00
0,28
0,57
0,58
0,67
0,64
0,22
0,75
0,35
0,81
0,84
Figure 2.2: Quadratic correlation matrix between dummy values for 5th perc. dummy (yellow: relevant
correlation, green: strong correlation, white correlation not significant)
Left: all criteria, Right: reduced to criteria with overall significant correlation
To support that conclusion, in Figure 2.3 femur forces over time are given for three pulses
(compressive forces have negative sign). A very soft (green), a moderate (blue) and a very
strong pulse (red) are chosen exemplarily.
Figure 2.3: Femur forces over time:
green: very soft pulse, blue: moderate pulse, red: very strong pulse
In summary it follows for 5th perc. female dummy:
1) Correlation of one criterion to all values in same quality is most likely not possible.
2) For HIC15, chest deflection, chest acceleration and neck compression it might be
possible to find a single criterion1 for crash pulse assessment.
3) For femur forces and NCE it might also be possible to be covered by such a
criterion, if reduced correlation is accepted.
4) Again between all other neck criteria relevant correlation is not fully given. That
could indicate that also for 5th perc. dummy the neck could not be fully covered
even by a separate neck crash pulse assessment criterion.
In order to get a better insight into the root cause for neck results for both dummies, in
Figure 2.4 NTF and NCF are shown over time for the 50th perc. dummy for three pulses
analogous to Figure 2.3.
4
J. Metzger, L. Kübler, S. Gargallo: Characterization and Evaluation of Frontal Crash Pulses for USNCAP 2011
Figure 2.4: Nij values over time for 50th perc. male dummy.
The observed good correlation of NTF to other dummy values can be explained by relatively
high values in comparison to the other Nij and more important continuity, i.e. maxima do
appear in the same time ranges and do not jump between peaks. For other Nij to some extent
discontinuities appear, i.e. discrete changes between dominating peaks are found. For example
in Figure 2.4 (right) with increasing pulse severity NCF are more and more excited at approx.
80ms (no effect for weak, low effect for moderate and strong effect for strong pulse) and the
effect reduces respectively at approx. 140ms. The maximum for NCF changes from peak at
140 ms to peak at 80 ms for the strong pulse.
Analogous effects can be found for the 5th perc. dummy, compare Figure 2.5. For NCE
excitation maxima appear only at one time frame while for NTF the maximum changes from a
first excitation time range (50-70ms) to a later one (120ms) for the weak pulse.
Figure 2.5: Nij values over time for 5th perc. female dummy.
It is expected that these discontinuities are the main reason for limited correlation. As
potential cause for the discrete effects, a strong sensitivity of Nij to changes in dummy
kinematics and, hence, different interaction with seat belt, seat, etc. and in particular the
airbag is assumed. In other words: small changes in dummy kinematics due to strong pulse
variations lead to bifurcations between excitation of different mechanisms. In order to analyze
these effects in more detail, correlation of the Nij components neck force and neck moment
will be observed in more detail. This will be subject of further investigation.
5
J. Metzger, L. Kübler, S. Gargallo: Characterization and Evaluation of Frontal Crash Pulses for USNCAP 2011
The analysis of correlation between dummy values in this section leads to the following
conclusions for 5th perc. and 50th perc. dummies:
1) Correlation of one criterion to all values in same quality is most likely not possible.
2) For HIC15, chest deflection and chest acceleration it might be possible to find a
single criterion for crash pulse assessment.
3) For femur forces it might also be possible to be covered by criterion in (2), if
reduced correlation of femur values is accepted.
4) To find one criterion that covers all neck values over a large pulse range is
unlikely, due to strong influence of small changes in dummy kinematics under
pulse variation. Still a generic indication and pulse comparison seems possible also
for neck under the assumptions:
- Bifurcations are not considered in the criterion. It just gives a relative
change assuming continuous effects.
- Highly non-linear effects like bifurcations are expected to be reduced in a
typical pulse range during development of a specific vehicle, i.e. within that
range good correlation to dummy values seems possible for robust airbag
definition over its working range.
5) Separate criteria with high correlation are recommended at least for neck and
femur forces. The other values could potentially be assessed by a combined
criterion as in (2) or also separately. All criteria can then be combined by the
weighted criteria method or utilizing a Pareto approach.
3
CORRELATION OF EXISTING PULSE CRITERIA
After the analysis of necessary conditions for a crash pulse criterion for the new USNCAP
rating, in the following chapter existing crash pulse criteria are evaluated regarding their
correlation to the relevant dummy values over approx. 400 USNCAP crash pulses from the
TRW database. Purpose is to gain more insight in the possibility to utilize existing pulse
criteria with respect to the USNCAP 2011 rating, i.e. to find out what indications are possible
and where are the limitations.
3.1 Existing pulse criteria
In literature many crash pulse criteria are known and thereof several are used in the
automotive industry. A detailed description is given in [1] and [4]. In this paper it will be
focused on the criteria that were already identified in [1]: occupant load criterion (OLC), point
in time when the vehicle velocity is zero Tv=0, sliding mean SM25 and SM35 (window size for
averaging 25ms and 35ms) and the OLC++ criterion proposed by TRW for current USNCAP
rating. Further, the maximum deformation motion of the vehicle smax is investigated.
3.2 Approach
In this study three different simulation models are utilized for the correlation analysis
Vehicle 1: System model of a middle class sedan vehicle, driver side
Vehicle 2: Same vehicle as 1, passenger side
comparison to vehicle 1 for analysis of influence driver vs. passenger
side
Vehicle 3: System model of a sports car, passenger side
comparison to vehicle 2 for analysis of influence of vehicle type
6
J. Metzger, L. Kübler, S. Gargallo: Characterization and Evaluation of Frontal Crash Pulses for USNCAP 2011
Before using the three system models for the correlation study, the restraint system
configuration was accomplished as discussed in Chapter 2.
The approach for correlation analysis follows [1]. It is illustrated in Figure 3.1. A number
of N USNCAP crash pulses from the TRW-database are imposed on each of the three system
models. The system models are set up in MADYMO [9] including 5th perc. and 50th perc.
Hybrid III dummy. In parallel all crash pulse criteria are calculated for each crash pulse.
Finally, dummy values generated by a system model are displayed versus corresponding crash
pulse criteria.
Figure 3.1: Approach of correlation analysis
Figure 3.2 shows a generic example of the resulting correlation diagrams. The dots
represent N pairs of dummy values and crash pulse criterion values. Further, regression curves
are calculated for each combination of crash pulse criterion and dummy value. In a first step
three regression curves are build by polynomial 1st, 2nd and 3rd order least square regression.
Figure 3.2: Example of correlation diagram with regression curve
As an assessment of correlation quality, the root mean square according to the regression
curve and N value pairs is calculated by
7
J. Metzger, L. Kübler, S. Gargallo: Characterization and Evaluation of Frontal Crash Pulses for USNCAP 2011
RMSij =
∑
N
k =1
(Yik − f ij ( X jk )) 2
N
,
(3.1)
where
Yik : Injury parameter i for pulse k,
X jk : Crash pulse criterion j for pulse k,
fij : Regression function.
The lower the RMS value, the higher the quality of correlation. Two conditions have to be
satisfied for an ideal crash pulse criterion
• RMSij = 0,
• fij monotonically increasing or decreasing.
Out of the three calculated regression curves of different order consequently the regression
curve fij which is monotonically increasing or decreasing in the overall pulse range and has the
lowest RMSij is used for further analysis.
3.3 Correlation Analysis
In order to compare correlation to all criteria, in the following RMS values of the
correlation between each specific dummy value and all evaluated pulse criteria are visualized
for the observed vehicles. Figure 3.3 shows the correlation of HIC15 versus pulse criteria for
the 50th perc. male dummy on the left and for the 5th perc. female dummy on the right side.
The lower the RMS value the better the correlation.
HIC15 - 50th perc. dummy
HIC15 - 5th perc. dummy
140
180
120
160
140
100
120
80
100
60
80
60
40
Vehicle 3
0
OLC++
OLC
Tv=0
SM35
SM25
smax
Vehicle 2
OLC++
OLC
Tv=0
SM35
SM25
smax
OLC++
OLC
Tv=0
SM35
SM25
smax
Vehicle 1
20
OLC++
OLC
Tv=0
SM35
SM25
smax
OLC++
OLC
Tv=0
SM35
SM25
smax
0
OLC++
OLC
Tv=0
SM35
SM25
smax
40
20
Vehicle 1
Vehicle 2
Vehicle 3
Figure 3.3: RMS of correlation HIC15 vs. pulse criteria for three vehicles,
50th perc. dummy (left) / 5th perc. dummy (right).
For the 50th perc. dummy OLC++ shows the best correlation for all three vehicles. For the
5th perc. dummy over all three vehicles Tv=0, OLC++ and smax show the best correlation to
HIC15.
Regarding chest acceleration, Figure 3.4, OLC++ shows the best correlation for 50th perc.
dummy, even though SM35/25 have a slightly better correlation in the case of vehicle 3. Here
the criterion smax gives worst results for both dummies in almost all vehicles. Again, there is
no obvious best criterion for the 5th perc. dummy regarding chest acceleration. OLC, SM35 and
OLC++ are the pulse criteria with the lowest RMS values.
8
J. Metzger, L. Kübler, S. Gargallo: Characterization and Evaluation of Frontal Crash Pulses for USNCAP 2011
3
3
2
2
1
1
0
Vehicle 1
Vehicle 2
Vehicle 3
0
OLC++
OLC
Tv=0
SM35
SM25
smax
4
OLC++
OLC
Tv=0
SM35
SM25
smax
4
OLC++
OLC
Tv=0
SM35
SM25
smax
5
OLC++
OLC
Tv=0
SM35
SM25
smax
5
OLC++
OLC
Tv=0
SM35
SM25
smax
Chest acceleration - 5th perc. dummy
6
OLC++
OLC
Tv=0
SM35
SM25
smax
Chest acceleration - 50th perc. dummy
6
Vehicle 1
Vehicle 2
Vehicle 3
Figure 3.4: RMS of correlation chest acceleration [g] vs. pulse criteria for three vehicles,
50th perc. dummy (left) / 5th perc. dummy (right).
Analogous trends follow for chest deflection, femur forces, neck compression and tension.
In Figure 3.5 exemplary chest deflection and femur forces are shown.
Chest deflection - 50th perc. dummy
Chest deflection - 5th perc. dummy
2,5
1,4
1,2
2,0
1,0
1,5
0,8
1,0
0,6
Vehicle 3
0,0
Femur forces - 50th perc. dummy
OLC++
OLC
Tv=0
SM35
SM25
smax
Vehicle 2
OLC++
OLC
Tv=0
SM35
SM25
smax
OLC++
OLC
Tv=0
SM35
SM25
smax
Vehicle 1
0,2
OLC++
OLC
Tv=0
SM35
SM25
smax
OLC++
OLC
Tv=0
SM35
SM25
smax
0,0
OLC++
OLC
Tv=0
SM35
SM25
smax
0,4
0,5
Vehicle 1
Vehicle 2
Vehicle 3
Femur forces - 5th perc. dummy
1,6
0,7
1,4
0,6
1,2
0,5
1,0
0,4
0,8
0,3
0,6
Vehicle 2
Vehicle 3
0
OLC++
OLC
Tv=0
SM35
SM25
smax
OLC++
OLC
Tv=0
SM35
SM25
smax
Vehicle 1
0,1
OLC++
OLC
Tv=0
SM35
SM25
smax
OLC++
OLC
Tv=0
SM35
SM25
smax
0,0
OLC++
OLC
Tv=0
SM35
SM25
smax
0,2
OLC++
OLC
Tv=0
SM35
SM25
smax
0,2
0,4
Vehicle 1
Vehicle 2
Vehicle 3
Figure 3.5: RMS of correlation chest deflection [mm] and femur forces [kN] vs. pulse criteria for three
vehicles, 50th perc. dummy (left) / 5th perc. dummy (right).
In Figure 3.6 exemplary the best correlation to one of the Nij values is shown for each
dummy. RMS values of correlations between NTF and pulse criteria are visualized for the 50th
perc. male dummy and NCE to pulse criteria for the 5th perc. female dummy. Regarding the
correlation to NTF, OLC++ again turns out to be the best pulse criterion to assess a dummy
value for 50th perc. dummy. SM25,35 show benefits for 5th perc. female.
9
J. Metzger, L. Kübler, S. Gargallo: Characterization and Evaluation of Frontal Crash Pulses for USNCAP 2011
NTF - 50th perc. dummy
NCE - 5th perc. dummy
0,040
0,09
0,035
0,08
0,030
0,07
0,06
0,025
0,05
0,020
0,04
0,015
0,03
0,010
Vehicle 3
0
OLC++
OLC
Tv=0
SM35
SM25
smax
Vehicle 2
OLC++
OLC
Tv=0
SM35
SM25
smax
OLC++
OLC
Tv=0
SM35
SM25
smax
Vehicle 1
0,01
OLC++
OLC
Tv=0
SM35
SM25
smax
OLC++
OLC
Tv=0
SM35
SM25
smax
0,000
0,02
OLC++
OLC
Tv=0
SM35
SM25
smax
0,005
Vehicle 1
Vehicle 2
Vehicle 3
Figure 3.6: RMS of correlation best Nij vs. pulse criteria for three vehicles,
50th perc. dummy (left) / 5th perc. dummy (right).
In summary it follows for 50th perc. male dummy: Among all evaluated pulse criteria and
vehicles the OLC++ criterion gives lowest RMS values.
For 5th perc. female dummy it is not possible to identify one criterion which is the best over
all vehicles. OLC++, OLC, Tv=0 and SM35 show comparable results.
In the next step, exemplary the quality of OLC++ correlation is visualized by several
correlation diagrams, as defined in section 3.1, see Figure 3.2. Each of the following figures
comprises four correlation diagrams. In the top row the best correlation diagram for the 50th
perc. dummy over all three vehicles is placed on the left side. The worst correlation diagram is
placed on the right. Thus the range of OLC++ quality over the three exemplary vehicles is
visualized in addition to the RMS values. The correlation diagrams for the 5th perc. dummy
are placed in the same order, but in the lower row.
In Figure 3.7 results for HIC15 over OLC++ are illustrated.
Figure 3.7: Best (left) and worst (right) correlation of HIC15 vs. OLC++ over all three vehicles for 50th perc.
dummy (top) / 5th perc. dummy (bottom).
10
J. Metzger, L. Kübler, S. Gargallo: Characterization and Evaluation of Frontal Crash Pulses for USNCAP 2011
The quality of the correlation between HIC15 and OLC++ for the 50th perc. dummy is very
good with some limitations for the driver model. Correlation for the 5th perc. dummy is worse
than for 50th. Even if reasonable correlation is found for vehicles 2 and 3, results for vehicle 3
indicate the need of an improved criterion for 5th perc. dummy.
Figure 3.8 shows the correlation results for chest acceleration.
Figure 3.8: Best (left) and worst (right) correlation of chest acceleration vs. OLC++ over all three vehicles for
50th perc. dummy (top) / 5th perc. dummy (bottom).
OLC++ shows sound correlation to chest acceleration for both 50th perc. and 5th perc.
dummy. However, improvement for 5th perc. would be beneficial, comparing results for driver
model in the lower right.
Even though OLC++ has not been developed with respect to chest deflection, a good
correlation is found in Figure 3.9 for the 50th perc. dummy. Also for 5th perc. dummy
correlation looks promising.
Analogous results are found for femur forces in Figure 3.10. For the 5th perc. dummy
correlation is better than the RMS value would indicate, since the approx. bi-linear correlation
characteristic is not covered by the applied regression schemes. Over the full pulse range out
of the TRW database two effects occur for the 5th perc. dummy, compare diagram in lower
right. First almost no contact femur to IP occurs. Starting from a specific pulse severity,
contact occurs with increasing amplitudes.
11
J. Metzger, L. Kübler, S. Gargallo: Characterization and Evaluation of Frontal Crash Pulses for USNCAP 2011
Figure 3.9: Best (left) and worst (right) correlation of chest deflection vs. OLC++ over all three vehicles for
50th perc. dummy (top) / 5th perc. dummy (bottom).
Figure 3.10: Best (left) and worst (right) correlation of femur forces vs. OLC++ over all three vehicles for
50th perc. dummy (top) / 5th perc. dummy (bottom).
As discussed in section 2 regarding Nij good correlation is expected if discontinuities do
not appear. This is for example the case in a wide range of NTF values for 50th perc. male,
compare Figure 3.11. For 5th perc. dummy it can be seen that correlation gets worse with
increasing discontinuities in the lower right.
12
J. Metzger, L. Kübler, S. Gargallo: Characterization and Evaluation of Frontal Crash Pulses for USNCAP 2011
Figure 3.11: Best (left) and worst (right) correlation of NTF vs. OLC++ over all three vehicles for 50th perc.
dummy (top) / 5th perc. dummy (bottom).
Analogous behavior can be observed for NCE in Figure 3.12.
Figure 3.12: Best (left) and worst (right) correlation of NCE vs. OLC++over all three vehicles for 50th perc.
dummy (top) / 5th perc. dummy (bottom).
13
J. Metzger, L. Kübler, S. Gargallo: Characterization and Evaluation of Frontal Crash Pulses for USNCAP 2011
In summary, even if OLC++ shows good correlation to many dummy values, it can be said
that no criterion gives sufficient correlation to all of the values relevant for the USNCAP 2011
rating, especially with respect to 5th perc. female.
4
DEVELOPMENT OF NEW CRITERIA
In chapter 2 it has been identified that separate criteria for each dummy value might be
beneficial to cover all effects. At least separate criteria for femur and neck and one combined
criterion for HIC15, chest deflection and chest acceleration came out to be favorable. The
analysis of correlation for existing criteria in chapter 3 supports this conclusion.
Consequently TRW is currently working on separate criteria for specific dummy values.
Two approaches are investigated in parallel. Based on signal attributes, e.g. TV=0
characteristic values of pulses are extracted. Additionally, the development of simplified
mechanical models representing dynamics of dummy or dummy regions excited by pulses in a
generic environment is continued.
Once separate criteria for particular dummy values are established, it is investigated
whether a combination by a weighted criteria approach gives sufficient overall correlation. If
this compromise is not acceptable, a Pareto approach will be applied in order to cover the
multi-criteria assessment task. Here all information remains available and a combination per
user preferences or focused body region is possible.
5
CONCLUSIONS AND OUTLOOK
With the changing USNCAP rating, complexity of crash pulse assessment increases due to
the higher number of applied dummy values, further body regions and additional dummies.
The target of this investigation was to assess whether existing pulse criteria like OLC++ are
also applicable for the new rating and to understand, if modifications are recommended to
cover the mechanisms driving the additional dummy values.
At first correlation between dummy values were analyzed to check whether it is generally
possible to establish a generic crash pulse criterion, or if separate crash pulse criteria are
required for some dummy values, or even if some mechanisms can not be covered by pulse
assessment at all. Different vehicle types were taken into account and it was observed if
differences between driver and passenger appear.
It was found that HIC15, chest deflection and chest acceleration can potentially be assessed
by one single crash pulse criterion. If some limitation in femur correlation is accepted, also
femur could be added. Also a generic indication of neck values could potentially be added to
such a criterion. However, that might require significant compromises.
This is also supported by looking at the correlation of dummy values with existing pulse
criteria. It was investigated to what extent pulse criteria derived for the current USNCAP
rating and other load-cases are capable to assess pulse severity with respect to the relevant
dummy values of the USNCAP 2011 rating.
Both the results from the dummy value correlation study and the investigation of
correlation of existing pulse criteria indicate that separate criteria for specific body regions,
covering the underlying physical effects for each dummy value, could be beneficial. Those
criteria could be combined either by using a weighted criteria approach with some limitations
in specific body regions or by applying a Pareto type approach that gives the user more
information but on the other hand increases complexity of standardization.
14
J. Metzger, L. Kübler, S. Gargallo: Characterization and Evaluation of Frontal Crash Pulses for USNCAP 2011
The investigation is on-going. Criteria are established via pulse signal attributes or
mechanical models approximating underlying physical effects for each injury parameter.
Those criteria can then be combined by weighted criteria or a Pareto approach. In a further
step threshold values could be defined for the new pulse criterion, that give an indication for a
pre-selection of restraint components depending on the specific vehicle pulse. The feasibility
of the approach and specific criteria combination for other load-cases like EuroNCAP would
then be a possible next step.
REFERENCES
[1] L. Kübler, S. Gargallo, K. Elsäßer: Characterization and Evaluation of Frontal Crash Pulses with
Respect to Occupant Safety, Proceedings Airbag 2008 – 9th International Symposium and
Exhibition on Sophisticated Car Occupant Safety Systems, 2008.
[2] L. Kübler, S. Gargallo, K. Elsäßer: Bewertungskriterien zur Auslegung von
Insassenschutzsystemen, ATZ Automobiltechnische Zeitschrift, 111, 06/2009, S. 426-433, 2009.
[3] TNO Automotive Safety Solutions: MADYMO Theory Manual, Release 7.2, 2010.
[4] M. Huang: Vehicle Crash Mechanics. Florida: CRC Press LLC, 2002.
15