18th ESV Paper shapes.

Differences between analytical predictions

and observed experimental results are explored.
Crash Pulse Modeling for Vehicle Safety Research Time histories are examined for the same vehicle at
different collision test speeds in order to investigate
Michael S. Varat the rate sensitivity of the vehicle collision response.
Stein E. Husher The analytical determination of static crush, dynamic
KEVA Engineering crush and vehicle structural restitution is discussed.
United States of America This research will show that a vehicles pulse shape
Paper 501 can be modeled and scaled using readily available
crash test data fitted with closed form functions.
ABSTRACT
There are numerous factors that can effect the
Computer simulation, component testing, and sled characteristics of a crash pulse. Some of these
tests often require the generation of suitable, derived include the vehicle shape, vehicle structure, vehicle
acceleration time histories to define a collision event. mass, collision partner, crash mode, and amount of
These time histories have shape, amplitude, and engagement. Given this variability, the modeling
duration characteristics. Suitable, derived technique must be flexible enough to accommodate a
acceleration time histories should be based on a wide range of characteristics but robust enough to
particular vehicles response in a staged full scale give good correlation to the parameters modeled.
crash test. A staged crash test includes
instrumentation in order to measure acceleration time The presented methodology first examined a set of
histories, force time histories and other engineering predefined crash pulse shapes and then interpreted an
parameters. Analytical techniques are developed to actual test pulse to develop the parameters that best
derive acceleration time histories at different fit the given shape. The next step was to develop a
collision severities based on the measured methodology to scale those characteristics in order to
acceleration time history in a particular crash test. run analytical simulations or tests at different test
speeds.
BACKGROUND
PULSE MODELING DERIVATIONS
The dependence of occupant response on the
collision pulse shape is well studied. Agaram [1] This research focused on four well recognized and
conducted DYNA 3D and MADYMO studies and identified shapes: Haversine (sin2), sine, square wave,
concluded that different crush pulses with identical and symmetric triangular. The procedure is to use
average accelerations values yield different occupant the minimum amount of variables to define the
response. The Agaram study [1] used square, collision pulse. Additionally, the generalized model
triangle and half sine pulse models . is derived for a two vehicle collision. The parameters
considered are:
Simulation software has been developed and utilized
in the automotive safety field to develop and scale a = acceleration
collision pulses. The Structural Impact Simulation t = time
and Model Extraction (SISAME) program has been s = displacement
developed under contract with the U.S. Department
of Transportation to extract optimal lumped- P = peak acceleration
parameter structural impact models from actual or T = duration of impact
simulated vehicle crash event data. This program V1= initial velocity of vehicle 1
models a vehicle using a lumped parameter V2= initial velocity of vehicle 2
methodology. This methodology utilizes rigid masses V1= change in velocity of vehicle 1
connected by load paths consisting of zero mass V2= change in velocity of vehicle 2
elements. The SISAME program requires a T = Collision Pulse Time duration
significant modeling effort in order to arrive at scaled
collision pulses. The model inputs:
VO1 = Impact Velocity of Vehicle 1
INTRODUCTION VO2 = Impact Velocity of Vehicle 2
V1 = Velocity Change of Vehicle 1
This study presents closed form functions that are V2 = Velocity Change of Vehicle 2
applied and compared to observed experimental pulse Mutual crush = total crush to both Vehicles
Outputs from the model will include the acceleration, Model 1: Haversine (Sin2).
velocity, and displacement time histories. The acceleration is written in Equation 1.
Additionally, peak acceleration, average acceleration,
t
and impact duration are also output from the models.
a = P sin 2 (1).
The Haversine pulse, Figure 1, is used as an example T
of the model development. This pulse has been
widely used to represent frontal barrier pulses and
has been identified in previous research as being a
good representation of a frontal barrier collision pulse
[3].

Figure 12. Graphical representation: Sin2.

Figure 1. Haversine pulse. Integrating acceleration with respect to time yields

velocity.
The half period sine pulse, Figure 2, is also utilized.
This has been previously identified in collision pulse P 2 t
V= 2 t T sin + C1 (2).
modeling as a standard pulse to represent a frontal 4 T
barrier impact [3,1].
The integration constant C1 can be evaluated by the
initial conditions of t = 0, V = Vo to get C1 = Vo

Figure 2. Sine pulse. Substituting C1 into Equation 2 yields

2 t
The square wave, Figure 3, which represents the P
collision as a constant acceleration, is presented as V= 2 t T sin + Vo (3).
the simplest representation of a collision [3,1]. 4 T

At t = T, V = Vo + V. Substituting into Equation 3

yields peak acceleration.

Figure 3. Square wave pulse. 2 V

P= (4).
The symmetric triangular pulse shape, Figure 4, is T
modeled. This pulse shape has also been previously
identified as a useful collision pulse model [2]. Integrating Equation 3 yields displacement

2 t
T 2 P cos + 2
2
t ( P t + 4Vo )
T (5).
S= + C2
Tr a

8 2
Figure 4. Triangular pulse.
At t = 0, s = 0. Solving for C2 yields
The development of the Haversine model is presented
here as an example. The sine, triangular, and square T 2 P
C2 = (6).
wave collision pulse models are derived in the
8 2
Appendix.
Substituting C2 into Equation 5 yields the
displacement as a function of time
 T 2 P 2 t t
S= cos 1 + (Pt + 4Vo ) (7).
8 2 T 4

The final parameter to be determined is the duration

of contact. At t1 = t2 = T, mutual crush = absolute
value of s1-s2. Solving Equation 7 for impact
duration results in the following determination of
duration.

mutual crush
T= (8). Figure 5. Force versus deflection compact four
1
abs (V1 V2 ) + (Vo Vo2 )


door sedan.
2
1

The development of the above pulse shape (and those

in the appendix) cover a wide range of characteristic
shapes to approximate a crash pulse. However, the
demonstrated methodology can be applied to most
integrable functions which allows the user to
customize this methodology to other pulse shapes.

MODEL APPLICATIONS

After the pulse shape has been selected and the

various characteristics of the function have been
derived, there are a number of factors that arise in the
Figure 6. Crush at separation as input into the
application of the derived model. When fitting a
pulse models acceleration versus time.
model to actual crash test data it is simple to
determine the V from the instrumentation; however,
the crush is not as easily quantified and various
difficulties arise as to exactly what crush value to
use.

A careful review of the derivation of the model

reveals that the use of the crush at separation of the
collision partners is required. The crush at separation
is easily determined from a derived force deflection
chart as seen in Figure 5. This separation crush was
input into the model for the sample test and the four
models were developed. Shown in Figure 6 are the Figure 7. Crush at separation as input into the
resulting four models plotted with the actual pulse models velocity versus time.
acceleration time history. The data is then integrated
and the resulting velocity time histories are shown in
Figure 7. This is integrated again and the resulting
displacement time histories are shown in Figure 8.

Figure 8. Crush at separation as input into the

pulse models displacement versus time.
STRUCTURAL RESPONSE restitution characteristics have a significant influence
on the post impact structural deformation data when
Analysis of the acceleration and force data generated considering what is occurring in the dynamic
provides some insight into the structural response environment. The force deflection of a mid size four
characteristics of the vehicle. For this analysis door sedan full scale barrier test is plotted and shown
restitution and crush rate sensitivity are considered. in Figure 11. The figure shows the maximum
Depicted in Figure 9 is an absorbed energy parameter dynamic structural deformation is about 750 mm. At
[4] for the frontal barrier tests conducted on a front the moment of separation from the barrier the
engine, front wheel drive midsize hatchback at 56, 64 deflection of the vehicle is about 675 mm. The
and 77 kph. This graph indicates that there is good measured static residual crush is about 490 mm. This
correlation with energy absorption as velocity difference between crush at separation and residual
increases. crush has been observed in numerous other frontal
barrier tests and indicates that vehicle structures
continue to restore after separation from the barrier
has occurred. The rate of restoration is not sufficient
to maintain contact with the load cell barrier face.

Figure 9. Comparison of EAF versus deflection

derived from a front engine front wheel drive
Figure 11. Force versus deflection midsize four
midsize hatchback tested at different velocities.
door sedan.
Figure 10 is an energy absorption graph of a small
pickup truck undergoing frontal barrier impacts at 24
VELOCITY SENSITIVITY
and 56 kph. There is similar structural response of
these 2 tests except in the very elastic initial 100 mm
An investigation was performed to analyze whether a
of deformation. It is apparent that the energy
particular vehicle is well modeled by a characteristic
absorption characteristics follow the same path; that
shape at different test speeds. If a vehicle can be
is similar amounts of deformation require similar
represented by a characteristic shape, then that shape
amounts of energy irrespective of test speed.
can be determined from a compliance test and readily
scaled to a different speed. While this concept has
not been generalized it has been observed in
numerous tests that have been analyzed. Two of
those studies are presented for discussion here.

Acceleration data from two rigid frontal barrier crash

tests of the same model full size van were studied and
shown in Figures 12 and 14. Overlaid on these
acceleration time histories is a sine model using the
Figure 10. Comparison of EAF versus deflection presented methodology. Examination of the resulting
derived from tests at different velocities small velocity time histories (Figures 13 and 15)
pickup truck. demonstrates clearly that the sine model represents
the vehicle dynamics. Even though these tests differ
RESTITUTION CHARACTERISTICS by a factor of 4 in energy, the characteristic pulse
shape is the same
After the maximum mutual crush has occurred the
impacting vehicle structures begin to rebound. The
 A second analysis is performed with frontal barrier
crash tests of a small four door sedan at 40, 48 and 57
kph - Figures 16-18. The sine model, again, best
represents this vehicles collision time history at all
three test speeds. In the course of this research, other
vehicles were studied with similar results. These
results indicate that determined characteristic shape
can be scaled to different speeds for similar impact
conditions.

Figure 12. Acceleration versus time for a full size

van at 25 kph. Sine fit.

Figure 16. Acceleration versus time for small four

door sedan at 40 kph. Sine fit.

Figure 13. Velocity Versus Time for a full size van

at 25 kph. Sine fit.
1 0

-5
Acceleration (g's)

-1 0

-1 5

-2 0

-2 5

-3 0
S in M o d e l
-3 5
T N 1 9
-4 0
0 .0 0 0 .0 2 0 .0 4
T im e
0 .0 6
(s )
0 .0 8 0 .1 0 Figure 17. Acceleration versus time for small four
Figure 14. Acceleration versus time for a full size door sedan at 48 kph. Sine fit.
van at 48 kph. Sine fit.

Figure 18. Acceleration versus time for small four

door sedan at 57 kph. Sine fit.

Figure 15. Velocity versus time for a full size van

at 48 kph. Sine fit.
 the acceleration and the velocity time histories as
shown in Figures 22 and 23. The full engagement
test is best represented by the sine model. In contrast
the offset deformable test is best represented by the
triangular model. Consequently, impact mode clearly
effects the pulse shape. Additional work must be
done in order to generalize a method that will account
for the differences observed in two different crash
modes.

Figure 19. Velocity versus time for small four

door sedan at 40 kph. Sine fit.

MODELING DIFFERENT COLLISION MODES

The generated crash pulses have been compared to

full engagement barrier testing. In order to study the
effect of impact configuration, a comparison was
made between a full engagement barrier test and an
offset deformable barrier test. Figures 20 and 21
show the sine fit to the full engagement test Figure 22. Acceleration versus time for a mid size
acceleration data. four door sedan 40% offset impact at 64 kph.
Triangular fit.

Figure 20. Acceleration versus time for a mid size

four door sedan full frontal impact at 56 kph.
Sine fit. Figure 23. Velocity versus time for a mid size four
door sedan 40% offset impact at 64 kph.
Triangular fit.

CONCLUSIONS

The derived collision pulse models have

been shown to represent the response of a
vehicle undergoing an impact.

The applied technique is easily expandable

to other geometric pulse shapes.
Figure 21. Velocity versus time for a mid size four
door sedan full frontal impact at 56 kph. Sine fit. The derived models have been shown to be
scalable to other impact severities, but care
Utilizing the same modeling technique, it was must be exercised in the application
determined that for the offset test a triangular pulse
was a better model. The triangular pulse is fitted to
 Structural response of the vehicle is [5] Varat, M., Husher S., Vehicle Impact Response
dependant on many characteristics which Analysis Through the Use Of Accelerometer Data, ,
include vehicle and impact mode SAE Technical Paper 2000-01-0850, Society of
parameters. Automotive Engineers, Warrendale, PA, 2000

Different crash modes with the same vehicle [6] Wooley R. L., Non-Linear Damage Analysis in
can exhibit different collision pulse shapes. Accident Reconstruction, SAE Technical Paper
2001-01-0504, Society of Automotive Engineers,
Inputs to the collision pulse models must be Warrendale, PA, 2001
carefully considered. Impact speed, V, and
crush data require careful analysis.
APPENDIX Sinusoidal, Square, and Triangular
For the vehicles studied, some vehicle Pulse Derivations
structural rebound takes place after
separation from the barrier. This Model 2: Sinusoidal Pulse.
phenomenon prevents the structural rebound The model evaluated is a sine model to represent the
from being directly measured by either load acceleration pulse.
cell or accelerometer instrumentation.
a = P sin( ) (2-1).
Care must be exercised when applying the t
data in this present research to a specific =
vehicle. T (2-2).

CONTACT where theta is chosen with boundary conditions that

Questions and comments are welcome and should be at t = 0, a = 0 and at t = T, a = 0.
addressed to the authors at:

KEVA Engineering, LLC

601 Daily Dr. Suite 225
Camarillo, California 93010
www.kevaeng.com
Figure A. Graphical Representation: Sine.
REFERENCES
[1] Agaram, V., Xu, L., Wu, J., Kostyniuk, G., Integrating this acceleration yields the relationship
Nusholtz, G. Comaprison of Frontal Crashes in for the velocity.
Terms of Average Acceleration, SAE Technical t
Paper 2000-01-0880, Society of Automotive T P cos
T
Engineers, Warrendale, PA, 2000. V = + C1
(2-3).
[2] Brach, R.M., Mechanical Impact Dynamics, John
Wiley and Sons, Inc., New York, NY, 1991. The integration constant C1 can be evaluated using
the initial condition: at t = 0, V = V0.
[3] Breed, D.S., Castelli, V., Sanders, W.T., A New T P
Automobile Crash Sensor Tester, SAE Technical C1 = + Vo
Paper 910655, Society of Automotive Engineers, (2-4).
Warrendale, PA, 1991.
Substituting C1 into Equation 2-3
t
[4] Kerkhoff, J.F., Husher, S.E., Varat, M.S.,
Busenga, A.M., Hamilton, K., An Investigation into T P cos
T T P
Vehicle Frontal Impact Stiffness, BEV, and Repeated V= + + Vo
Testing for Reconstruction, SAE Technical Paper (2-5).
930899, Society of Automotive Engineers,
Warrendale, PA, 1993. The peak acceleration (P) can be found by solving
Equation 2-5 knowing that at t = T, V = V0 + V.
 V At t=0, V = Vo. Solving for C1
P=
2 T (2-6). C1 = Vo
(3-3).
The integration of velocity yields displacement.
Substituting C1 into Equation 3-2
t
T 2 P sin
T T P
S= + t V + + C2 V = Pt + Vo
2 (2-7). (3-4).

C2 is solved by knowing that at t = 0, s = 0, therefore Solving for P at t = T, V = V0 +V. The peak

acceleration can now be determined. Note that the
C2 = 0 (2-8).
peak acceleration equals the average acceleration in
this model.
Substituting C2 into Equation 2-7 gives displacement
V
P=
t T
T 2 P sin (3-5).
T T P
S= + t V +
2 (2-9).
Integrate velocity to get displacement

1 2
The acceleration, velocity, and displacement time S= Pt + V0t + C2
histories all require the collision duration. The 2 (3-6).
collision duration is determined through the
following additional boundary condition: at t1 = t2 = At t = 0, S = 0. C2 can now be determined.
T, mutual crush = s1-s2. Substituting these conditions
into the displacement time equation, yields the C2 = 0 (3-7).
following equation for duration, T.
Substituting C2 into Equation 3-6
mutual crush
T=
V V

( )
abs Vo1 Vo2 + 1 2

S=
1 2
Pt + Vot
2 2 (2-10). 2 (3-8).

Model 3: Square Pulse. At t1 = t2 = T, mutual crush = absolute value of S1-S2.

The third derived model is that of a simple square Solving for duration yields
wave. For this relationship, peak and average
accelerations are equal throughout the entire collision mutual crush
T=
(V1 V2 ) + (Vo )
duration. 1
abs Vo2
a = P 0t T 2
1

(3-1). (3-9).

Model 4: Triangular Pulse

The final derived model is that of a triangular wave.

2P T

a= t 0 t
T 2
Figure B. Graphical Representation: Square.
a = 2P 1
t + 2P T <t T
The integration of acceleration with respect to time T 2
(4-1).
yields velocity

V = Pt + C1 (3-2).
 The preceding derivation must now be carried out for
1
t T
2 (4-10).

Figure C. Graph. Representation: Triangular. Velocity is found by integrating Equation 4-9 with
respect to time
The analysis of the triangular pulse is going to be
performed in two parts. First, the time up to t < T P 2
is analyzed. For this time period, velocity is found V= t + 2Pt+ C1
by the integration of Equation 20 with respect to T (4-11).
time. C1 is arrived at by solving Equation 4-11 at time
P 2 equal to one half of the duration.
V= t + C1 P T
T (4-2). V= + Vo
4 (4-12)
Using the known conditions, V = Vo at t = 0, C1 can
be determined Setting Equation 4-11 equal to Equation 4-12 allows
C1 = Vo (4-3).
for C1 to be solved for.
1
C1 = Vo P T
Substituting C1 into Equation 4-3 2 (4-13)
P 1
V = t 2 + Vo For t T Substituting C1 into Equation 4-11 gives velocity
T 2 (4-4). P 2 1 1
V= t + 2 Pt + Vo P T For t T
T 2 2 (4-14).
Since the assigned triangular acceleration pulse is
symmetrical about the midpoint, at time t = T, the
change in velocity will equal one half of the total Integration of Equation 4-14 yields displacement
P 3 1
change in velocity. S= t + Pt 2 + Vo t P T t + C 2
T 1 3T 2 (4-15).
At t = V = abs (Vo V )
2 2 (4-5). By setting Equation 4-9 equal to Equation 4-15 at
time equal to half the duration, C2 can be determined.
Assigning these boundary conditions allows the
PT 2
solution for the peak acceleration C2 =
2 V 12 (4-16).
P=
T (4-6).
Substituting C2 into Equation 4-15 results in:
Integrating Equation 4-4 with respect to time yields P 3 1 PT 2 1
S= t + Pt 2 + Vo t P T t + For t T
displacement 3T 2 12 2 (4-17).
P
S = t 3 + Vot + C 2
3 T (4-7). Combining the two derived relationships for
displacements up to the collision mid point and then
At t = 0, s = 0. This allows the determination of C2. after the collision midpoint, the duration can be
determined. At time t = T, the mutual crush =
C2 = 0 (4-8) absolute value of s1 s2. Algebraically solving for
duration of impact yields the following:
Substituting C2 into Equation 4-7
P 1 mutual crush
S = t 3 + V ot For t T T=
3 T
(V1 V2 ) + (Vo )
2 (4-9). 2
abs Vo2
3
1
(4-19).