Vous êtes sur la page 1sur 25

ARTICLE IN PRESS

International Journal of Impact Engineering 30 (2004) 1055–1079

Design of an aluminium-based vehicle platform for front


impact safety
A. Deba,*, M.S. Mahendrakumara, C. Chavana, J. Karvea,
D. Blankenburgb, S. Storenb
a
Centre for Product Design & Manufacturing, Indian Institute of Science, Bangalore 560012, India
b
Department of Machine Design and Materials Technology, NTNU, N-7491 Trondheim, Norway
Received in revised form 5 April 2004; accepted 6 April 2004

Available online 26 June 2004

Abstract

The current paper examines the design of an aluminium-intensive small car platform for desirable front
impact safety performance. A space frame-type architecture comprised of extruded aluminium members
with welded joints is considered for inherent structural rigidity, and low investment in terms of tooling. A
finite element model of the vehicle is employed for crash analysis using the explicit code LS-DYNA.
Confidence in analysis is established at the component level by benchmarking finite element models of
welded joints against experimental data, and axial crushing of aluminium tubes against published numerical
results and theoretical prediction. A numerical design of experiments is conducted for arriving at a front-
end design that will yield desirable safety performance during impact against a rigid barrier at 30 mph
(FMVSS 208 condition). For comparable new car assessment program performance at a higher speed of
35 mph, a lumped parameter idealization is used to identify the principal design changes that may be
necessary. The current approach of component level testing combined with finite element and lumped
parameter-based simulations can be regarded as an effective and time-saving procedure in the crash safety
design of new vehicles.
r 2004 Published by Elsevier Ltd.

Keywords: Aluminium; Vehicle; Front impact

*Corresponding author. Tel.: +91-80-2293-2922; fax: +91-80-2360-1975.


E-mail address: adeb@cpdm.iisc.ernet.in (A. Deb).

0734-743X/$ - see front matter r 2004 Published by Elsevier Ltd.


doi:10.1016/j.ijimpeng.2004.04.016
ARTICLE IN PRESS

1056 A. Deb et al. / International Journal of Impact Engineering 30 (2004) 1055–1079

1. Introduction

In order to design a successful battery-powered electric vehicle, its weight should be as low as
possible. Even for conventional IC engine-based powertrain, savings in weight using lightweight
materials such as aluminium can lead to increased fuel economy and reduction in pollution. In
developing countries like India, air-pollution arising mainly from traffic exhaust emissions is
significantly affecting the health and well-being of the dwellers in urban areas. There is thus a need
for designing lightweight vehicle platforms without sacrificing market requirements. Aluminium
alloys stand out as attractive material for body construction being about one-third lighter than
steel but possessing comparable strength to weight ratio as steel. Aluminium also has superior
corossion-resistance and recyclability when compared to mild or high strength steel. Usage of
aluminium would, however, undoubtedly have a negative impact on cost; although it may be
possible to meet the cost objectives by using extruded members, optimizing design by using
computer-aided engineering, and minimizing expenses related to manufacturing. It is noted that
only a few aluminium-intensive vehicles are available in the world automotive market today. Audi
A8 is one such vehicle in which sheet-metal and casting-based body parts are used predominantly.
The space frame body configuration for this vehicle is shown in Fig. 1. Think City is a small
aluminium-intensive electric car sold in some European countries such as Norway. The upper
body of this car is made from aluminium extrusion members mated with a steel chassis. Think
City can accommodate two passengers only including the driver.
The vehicle body studied in the current paper is unique in terms of having an integrated upper
body and chassis made completely from extruded members some of which need to be bent using
secondary processes to achieve an aesthetic exterior form. The members are connected at joints
using inert gas metal arc welding. Non-structural body panels acting as vehicle skin are not
included in the vehicle frame analyzed. The design of the proposed platform involves various
considerations [1], however, the focus of the work reported here is on front impact safety. For
assessment of the current vehicle platform for full frontal impact, the mandatory Federal Motor
Vehicle Safety Standards (FMVSS 208) [2] regulation for new vehicles in USA is considered.
According to this requirement, a vehicle is crashed against a rigid barrier with a speed of 48 kph
(30 mph). This is somewhat equivalent to head-on collision between two identical vehicles, one
being stationary and the other approaching it with a speed of 96 kph. A scrutiny of the space
frame body discussed in the next section will reveal that welded joints and axial members such as

Fig. 1. Audi A8 aluminium space frame [10].


ARTICLE IN PRESS

A. Deb et al. / International Journal of Impact Engineering 30 (2004) 1055–1079 1057

front rails are principal components contributing to energy-absorption during a frontal collision.
As finite-element-based simulation using the explicit code LS-DYNA is employed for impact
analysis of the vehicle, it is important that the body components mentioned above are modeled
satisfactorily. The predictability of two separate finite element models of a welded T-joint is
verified at first against a physical test carried out by the authors. This is followed by comparison
of finite-element analysis of axial crushing of an extruded aluminium tube against simulation
results given in [3] and theoretical mean crush load prediction [7]. After establishing confidence in
the modeling of the components described, a finite-element model of the complete vehicle body
with wheels, axles, suspension and electric motor is used for full frontal impact safety evaluation.
Finally, a computationally inexpensive spring-mass model is employed for assessment of the
current platform for NCAP condition with a higher impact speed of 35 mph and necessary design
changes are suggested for improved safety.

2. A space frame-based vehicle concept

The work reported in the current paper is a subset of a design project that aimed at creating a
prototype for a robust and lightweight small car ‘‘platform’’ for the Indian market. As the project
was of investigative nature and not a commercial one, it was decided that emphasis would be
placed on usage of extruded aluminium components for building the body. Initial cost analysis
suggested that it would be economic to choose a space frame configuration for a fully aluminium -
based chassis and upper body. Commercially available wheels, axles, suspension and steering
systems, and electric motor made predominantly of steel were integrated with the vehicle frame.
Weightage was given to aesthetics and ergonomics in the design of the computer-based concept.
Body panels made of lightweight materials such as composites can be attached to the skeletal
frame; however, these are unlikely to contribute in a significant manner to body rigidity and
crashworthiness. The overall body dimensions for the vehicle were chosen based on Competitive
Vehicle Benchmarking, i.e. an analysis of vehicles already available in the relevant segment of the
target market. For example, the target length and width of the vehicle (referred to as the ‘‘CPDM
Model’’ in Figs. 2 and 3) were chosen as 3000 and 1400 mm, respectively. These dimensions are
higher than those of the nearest small vehicles in India, namely the Bajaj Autorickshaw and an

Fig. 2. Overall length of the present vehicle concept vs. those for competing vehicles.
ARTICLE IN PRESS

1058 A. Deb et al. / International Journal of Impact Engineering 30 (2004) 1055–1079

Fig. 3. Overall width of the present vehicle concept vs. those for competing vehicles.

Fig. 4. Aluminium extrusion-based space frame vehicle concept.

electric car named Reva, but are less than those for the popular Maruti 800 (Figs. 2 and 3). The
envisioned CPDM car thus finds a place between the ubiquitous three-wheeler commuter taxi, i.e.
the autorickshaw and the highest sold small retail car, i.e. Maruti 800.
The design began with a computer-aided design (CAD) model of the space frame architecture
(Fig. 4) for the concept vehicle. One may consider this basic design as realistic and substantive
when compared to production vehicles (as can be seen by comparing Figs. 2 and 4). The term
‘‘platform’’ used earlier is intended to signify that variants such as vans and sedans can be
generated without major modifications to the current design. A finite element model reflecting the
final concept is shown in Fig. 5. The model employs a nearly uniform mesh for the whole vehicle.
The body-in-white static and vibration frequency targets were met after making suitable changes
to the initial concept. Before undertaking the frontal impact analysis against a rigid barrier as per
the FMVSS 208 test set-up, validation of finite-element modeling of welded joints and axial
crushing of components is carried out as described in the next two sections.
ARTICLE IN PRESS

A. Deb et al. / International Journal of Impact Engineering 30 (2004) 1055–1079 1059

Fig. 5. Finite-element model of prototype space frame.

3. Behavior of welded aluminium joints

The joints between extruded members in the vehicle modeled in Fig. 5 are proposed to be made
with MIG welding. Truly speaking, weld fillets should be modeled with 3D elements [12] having
appropriate material properties, and the effect of heat-affected zone (HAZ) surrounding a weld
should be taken into account. However, it is difficult to know weld properties directly. Although
good correlation is obtained in [12] between experimental and finite-element results, details of
mesh size used or material properties applied to the weld fillets and HAZ are not given. As usage
of small 3D elements in every joint poses severe demands on computational time in a large crash
model and as the extruded aluminium components are represented with shell elements, it was
decided that two approaches will be studied for simulation of the failure of a welded T-joint under
bending load applied on one of its legs (see Fig. 6 below for the test set-up): (1) monolithic
connection between the two mating components (as shown in Fig. 7), and (2) shell elements
(shown in a dark shade in Fig. 8) with modified properties for modeling the weld zone. The
aluminium extrusions used in the test were made from SAPA 6060 T6 alloy. The cross-sectional
dimensions of each component were 100 mm  50 mm with a wall thickness of 3.4 mm.
The properties of aluminium (SAPA 6060 T6) used were obtained from the material supplier’s
manual [10] and are given below:
Young’s modulus ðEÞ : 70 GPa; ð1aÞ

Poisson’s ratio ðvÞ : 0:33; ð1bÞ

Ultimate tensile stress ðsu Þ : 0:17 GPa; ð1cÞ

Yield stress ðsy Þ : 0:14 GPa; ð1dÞ


ARTICLE IN PRESS

1060 A. Deb et al. / International Journal of Impact Engineering 30 (2004) 1055–1079

Fig. 6. Test set-up for studying load–deformation behavior of a welded T-joint.

Fig. 7. Model with monolithic joint.


ARTICLE IN PRESS

A. Deb et al. / International Journal of Impact Engineering 30 (2004) 1055–1079 1061

Fig. 8. Model with distinct shell element properties for the weld zone.

Failure strain ðef Þ : 8%; ð1eÞ

Density ðrÞ : 2:7E  06 kg=mm3 ; ð1fÞ

Tangent modulus ðET Þ : 385 GPa: ð1gÞ


It is noted that three tensile tests performed with coupons obtained from the walls of extruded
members of grade 6060 T6 confirmed the yield and tensile strengths as well failure strain assumed
above. In order to simulate quasi-static loading with minimal inertial forces, the rigid cylindrical
rod shown in Fig. 6 is moved downward with a uniform speed of 2 m/s. The automatic contact
algorithm with key word CONTACT AUTOMATIC GENERAL is chosen for analysis with LS-
DYNA. The material model used is of type 24 (defined by the key word MATERI-
AL PIECEWISE LINEAR PLASTICITY in LS-DYNA) pertaining to Von Mises yield
condition with isotropic strain hardening, and strain rate-dependent dynamic yield stress based
on Cowper and Symonds model. The progression of stress beyond yield (according to scaling
algorithm) for this model is given below:
"  p 1=p #
e’
s ¼ ðsy þ ET epeff Þ 1 þ eff ; ð2Þ
D
where,
epeff ; e’peff are the effective plastic strain and strain rate; respectively; and ð2aÞ

D and p are material constants: ð2bÞ


The aluminium alloy being considered here is known to be relatively insensitive to high strain
rates [8]. Hence in the current study, the term within square brackets in the right side of Eq. (2) is
taken as unity signifying that the scaling of static yield stress due to strain rate effect is not
applied.
ARTICLE IN PRESS

1062 A. Deb et al. / International Journal of Impact Engineering 30 (2004) 1055–1079

Fig. 9. A comparison of test and simulation results for bending of welded T-joint.

In the test set-up in Fig. 6, the vertical load applied on the horizontal leg of the welded T-joint is
tracked along with the downward displacement of the load applicator. The recorded load–
deformation behavior is given in Fig. 9. Initially, analysis has been done with the monolithic joint
model shown in Fig. 7 the result for which is superimposed with the test result in Fig. 9. It is
observed that the monolithic joint simulation yields a peak load that is about 30% higher than the
test peak load. After computing the area under a given force–deformation curve, it is seen that the
monolithic joint simulation also over-predicts energy absorption during the entire history of
deformation by 10.5%. The obtained results make it obvious that for better correlation between
simulation and test behaviors, elements with special properties in the joint area may be necessary.
It is not readily apparent, however, as to what values of material parameters for the shaded
elements in Fig. 8 will yield a load–deformation variation close to the test response. A study is
therefore carried out in the form of a design of experiments (DOE) to determine the weld material
properties that would ensure minimum variations in peak load and energy absorption with respect
to the experimental values. In order to achieve the stated optimization in weld simulation, the
following objective function Ow is defined:
Ow ¼ ðDPÞ2 þ ðDEÞ2 ; ð3Þ
subject to the following constraints:
jDPjp10%; ð3aÞ

jDEjp10%; ð3bÞ
where,
DP is the percent variation in peak load with respect to the test value of 13:2 kN; and
ð3cÞ

DE the percent variation in energy absorption with respect to the test value of 338 J:
ð3dÞ
ARTICLE IN PRESS

A. Deb et al. / International Journal of Impact Engineering 30 (2004) 1055–1079 1063

Table 1
Values of material parameters considered in weld simulation optimization
sy (GPa) ef (%) ET (GPa)
0.098 6 0.347
0.112 8 0.385
0.126 10 0.424

Fig. 10. Variation of objective function for weld optimization.

(Note: the absorbed energy is computed as the area under a given load–deformation curve until
the load drops to zero).
The material parameters varied were sy, ef and ET. Twenty-seven (3  3  3) cases were
analyzed with combinations of sy, ef and ET from values given in the Table 1.
It is noted that in Table 1, values of sy considered are 10–30% lower than the baseline value for
the current aluminium alloy given in (1d) in anticipation of the weakened strength of HAZ in the
weld region. The other two parameters (i.e ef and ET ) are taken as equal to, above and below their
corresponding baseline values given by (1e) and (1 g). The variation of the computed objective
function Ow for the twenty-seven DOE cases is shown in Fig. 10.
It is seen from Fig. 10 that the best correlation between test and weld element-based analysis is
achieved for case 14 for which Ow is a minimum. A summary of peak load and energy absorption
data showing highly improved correlation between test and weld element-based analysis (case 14)
is given in Table 2. The values of design variables for case 14 are: sy ¼ 0:098 GPa, ef ¼ 10% and
ET ¼ 0:385 GPa. It is observed that the weld yield strength in case 14 is 30% lower than that given
by (1d) for the tube alloy.
In the finite element models of Figs. 7 and 8, an average element size of 15 mm is assumed for
compatibility with the mesh density used in the full vehicle model (Fig. 5). A finer mesh for the
vehicle model would place severe burden on available computational resource and the
optimization study described later could not be carried out in a reasonable time frame. With
the current state of CAD, finite-element pre-processing, and data management technologies, the
ARTICLE IN PRESS

1064 A. Deb et al. / International Journal of Impact Engineering 30 (2004) 1055–1079

Table 2
A comparison of absorbed energy and peak load for simulation and test cases
Case Energy absorbed (J) Peak load (kN) Difference in Difference in peak
energy absorption load between
between analysis analysis and test
and test (%) (%)
Physical test 337.7 13.2 N/A N/A
Monolithic joint 377.3 18.9 10.5 30.2
analysis (Fig. 7)
Weld element analysis 349.8 14.0 3.9 6.4
(Fig. 8, DOE case 14)

Fig. 11. Failure pattern for monolithic joint simulation.

present procedure of ‘‘tuning’’ finite-element modeling for predicting test behavior is also followed
in designing production vehicles. The width of the weld zone (including a weld fillet of ap-
proximately 5 mm width all around the junction between the two rectangular members)
considered in Fig. 8 is 30 mm in the vertical leg and 45 mm in the cantilever member on which load
is applied. These weld zone widths are in the same range as the recommended width of HAZ given
in [11], viz. 25 mm (1 in), and the width of HAZ used in finite element models in [12], i.e. 38 mm. In
[10], the reductions in ultimate strength in weld fillet and in HAZ are reported as 25% and 40%,
respectively. In [13], the authors found a drop in yield strength of 40–50% in coupon specimens
cut from HAZ areas in welded plates. The current reduction of 30% in yield and ultimate
strengths for combined weld fillet and HAZ areas represented by shell elements in Fig. 8 is thus
consistent with the quoted findings in [10,13].
The deformed shapes yielded by the finite element runs after the cylindrical load-head has
moved downward by about 28 mm are shown in Figs. 11 and 12. It is seen that due to symmetrical
ARTICLE IN PRESS

A. Deb et al. / International Journal of Impact Engineering 30 (2004) 1055–1079 1065

Fig. 12. Failure pattern for joint with welded elements.

nature of tensile and compressive stresses in the joint exerted by the outstanding leg, failure takes
place at both the top and bottom vertical edges of the joint. Separation of the cantilever leg from
the joint also takes place at the top and bottom horizontal edges. In the tested joint component in
Fig. 6, failure in the weld was initiated at the top corners of the joint (i.e. in the tension zone) and
propagated to result in complete tearing along a horizontal direction in the HAZ as predicted in
the finite element model. In the compression zone, cracks developed in the vertical weld fillets;
however, no through crack or tearing of extrusion wall in the HAZ in the horizontal direction was
observed. The lack of symmetry in the failure of top and bottom horizontal edges in the tested
component can be attributed to the fact that the lower edge of the cantilever leg is supported in
compression on the vertical member which prevents complete failure in the bottom horizontal
weld fillet. In the finite element model, on the other hand, the load from the cantilever leg is
transferred through the weld elements only to the vertical member. In order to obtain a closer
correlation of predicted failure pattern with observed behavior, the weld fillets should be modeled
with 3D elements as remarked earlier to connect the two extrusion members while the latter
should not be directly connected to each other. The current approach of using modified shell
elements (Fig. 8) to represent the weld zone, however, leads to a load–deformation behavior (Fig.
9) that can be considered as sufficiently close to the experimental curve and hence can be adopted
in full vehicle simulation.

4. Axial crushing of an extruded aluminium component

A key mode of failure of the vehicle frame components in front impact is in the form of
progressive buckling of front rails. Axial collapse of steel and aluminium columns has been
previously studied by a number of investigators (notably, [3–8]). Usually, a limit state of plasticity
approach will analytically yield the peak and average values of load generated during axial
ARTICLE IN PRESS

1066 A. Deb et al. / International Journal of Impact Engineering 30 (2004) 1055–1079

crushing. However, to obtain the entire history of deformation, nonlinear finite-element method
with robust contact algorithms would be most useful. The information obtained here through LS-
DYNA-based analysis of a square aluminium tube will be compared against PAM-CRASH-based
results previously reported in [3] and theoretically predicted mean crush load [7] for the given tube
material and geometrical properties.
The problem considered is the progressive buckling of a square column of sides 80 mm,
thickness 1.88 mm, and length 245 mm as given in [3]. The column is assumed to be made
of AA 6063 T7 alloy for which the elasto-plastic properties are taken from [3]. The constitutive
model used in the present LS-DYNA-based simulation is MAT PIECEWISE LINEAR
PLASTICITY in which values of effective stresses corresponding to effective plastic strains
are defined. The finite-element model of the component with an element size of 5 mm is shown in
Fig. 13. It is fixed at the base and is subject to gradual compression with a rigid flat plate
positioned on the top. In the beginning, a small initial gap is maintained between the plate and the
column to avoid initial penetrations. The plate is applied a uniform velocity of 2 m/s along the
column axis giving rise to its progressive failure by formation of folds as shown in Fig. 14.

Fig. 13. Finite-element model for axial compression simulation.

Fig. 14. Side view of simulated tube collapse after full compression.
ARTICLE IN PRESS

A. Deb et al. / International Journal of Impact Engineering 30 (2004) 1055–1079 1067

Fig. 15. Axial load vs. displacement behavior of square column.

It is pointed out that the manner of load application on the extrusion component is different
from what was followed in [3], although this is not expected to cause a significant difference in
results. In [3], velocity boundary condition is applied to the top of the column and the velocity is
ramped during the first 50 ms until a steady-state velocity of 2 m/s is reached. In the present case,
load is applied with a plate as this approach would be perhaps closer to an actual physical
experiment.
The load–deformation curve for the square tube obtained from LS-DYNA analysis is given in
Fig. 15. It has been filtered with SAE 1000-9 in LS-POST. The peak load of 62 kN obtained from
the present analysis agrees excellently with the value of 63 kN reported in [3]. The mean crush load
of 20 kN (for a total axial deformation of 180 mm after which the tube bottoms out) obtained here
is about 18% higher compared to a value of 16.9 kN reported in [3]. However, the current mean
load agrees excellently with a value of 20.6 kN yielded by the following theoretical relation due to
Abramowicz and Jones [7]:
Fmean ¼ 13:06s0 b1=3
m h
5=3
; ð4Þ
where
1 
s0 ¼ sy þ su ; ð4aÞ
2
sy ¼ 0:08694 GPa ½3 ; ð4bÞ

su ¼ 0:171 GPa ½3 ; ð4cÞ

bm ¼ b  h ðmean width of square tubeÞ; ð4dÞ

b ¼ 80 mm ðwidth of square tubeÞ; and ð4eÞ

h ¼ 1:8 mm ðthickness of square tubeÞ: ð4fÞ


ARTICLE IN PRESS

1068 A. Deb et al. / International Journal of Impact Engineering 30 (2004) 1055–1079

It may be noted that a mesh size sensitivity study was carried out by considering two coarser mesh
configurations, however, the 5 mm size mesh (Fig. 13) yielded most satisfactory correlation with
the theoretically predicted mean load given by (4) above.

5. Frontal impact simulation against a rigid barrier

The results presented in the previous sections on bending of a welded T-joint and progressive
axial collapse of an aluminium tube lead to a fair degree of confidence on finite-element modeling
procedures followed here. Using the same general approach for components, one could perform a
finite-element-based assessment of frontal impact safety of the full vehicle concept. The finite-
element model employed for impact simulation against a rigid barrier with a speed of 48 kph
conforming to the mandatory FMVSS 208 regulation in USA is shown in Fig. 5. The aluminium
members of the vehicle frame are assumed to be made of SAPA 6060 T6 alloy. All members are of
thickness 3.0 mm except the front fore-rails (comprising crush box and mid-rail shown in Fig. 16)
which have a thickness of 2.0 mm. The front fore-rails, which are the principal load-carrying and
energy-absorbing members in front impact, have cross-sectional dimensions of 100 mm  50 mm.
The mass of the vehicle is 620 kg including power train (i.e. a 72 V DC motor with front wheel
drive transmission system and a pack of twelve 6 V batteries).
The full vehicle front impact simulation is initially carried out (a) with all frame joints being
considered as monolithic, and (b) using weld elements (as in Fig. 17) at joints with optimized
properties obtained earlier. Overall vehicle deceleration responses obtained for cases (a) and (b)
when the vehicle is impacted against a rigid wall with an initial forward speed (in the negative x
direction in Fig. 17) of 48 kph (13.33 m/s) are shown in Fig. 18 (after processing with SAE 60
filter). This deceleration history can be looked upon as an average for the entire vehicle and is

Fig. 16. Vehicle front structure with segments of rail.


ARTICLE IN PRESS

A. Deb et al. / International Journal of Impact Engineering 30 (2004) 1055–1079 1069

Fig. 17. Full vehicle finite-element model with all joints represented with weld elements.

Fig. 18. Acceleration histories for front impact simulation cases (a) and (b).

directly obtained in the LS-DYNA post-processor. Alternatively, vehicle response (sometimes


referred to as front ‘‘crash pulse’’) could be studied at locations such as B-pillar-and-rocker joint
or under the rear seat (Fig. 22). The responses in Fig. 18 have a desirable square-type shape
indicating efficient energy absorption by the front structure. An important outcome of the current
analyses is that the peak decelerations as seen in Fig. 18 are not significantly different. The peak
deceleration turns out to be lower and energy absorption higher (as per Figs. 18 and 19) when
joints are represented with weld elements because of increased overall compliance in the structure
when compared to monolithic joint modeling. It needs to be mentioned that full front impact
against a rigid barrier is primarily an axial phenomenon while the component joint study
ARTICLE IN PRESS

1070 A. Deb et al. / International Journal of Impact Engineering 30 (2004) 1055–1079

Fig. 19. Energy plots for front impact simulation cases (a) and (b) (Note. KE—kinetic energy; IE—internal energy;
TE—total energy).

conducted in an earlier section corresponded to bending loading in which shear force carrying
capacity of the weld section is almost completely lost after failure of a few elements. This may
explain the general agreement between welded and monolithic joint representation results in full
frontal impact simulations as carried out here.
Snap-shots of vehicle deformation along with close-up views of the front structure are shown in
Fig. 20. It is seen that until 5 ms, no significant deformation in the front of the vehicle has
taken place. However by 12 ms, a major part of the crush boxes for both front rails has
collapsed. Referring to Fig. 18, the occurrence of first deceleration peak in both monolithic
and welded joint-based analyses at around 16–18 ms is likely to signify the complete collapse
of the left and right crush boxes (as observed in the deformation snap-shots). The occurrence of
another peak in Fig. 18 at 30 ms for either joint modeling approach indicates the formation of
one plastic hinge each in left and right mid-rail segments. The continuous drop in deceleration
from 45 ms (Fig. 18) corresponds to the beginning of rebound phase in the impact event under
consideration.
The majority of the elements in the front of the structure (Figs. 5 and 17) considered in the
previous analyses vary between 17 and 20 mm in size. This configuration is referred to as ‘‘uniform
vehicle mesh’’ for subsequent reference. The desired mesh size for predicting progressive buckling
of component rectangular tubes was earlier found to be 5 mm. In order to assess the effect of this
smaller element size, the leading portions of the vehicle front end including the fore-rails were re-
meshed with an average of 5 mm size elements and a complete vehicle front crash analysis was
carried out (Fig. 21). A comparison of the crash pulses between the nearly uniform and refined
front-end vehicle models at the rear of the vehicle (i.e. the zone in which NHTSA reports NCAP
test deceleration histories of new vehicles) is given in Fig. 22. It is observed that although there are
differences in the shapes of the acceleration curves, the magnitudes of peak and average
deceleration for the two cases are not significantly different. As these are the parameters that are
considered in the optimization of the vehicle design for front impact safety (to be described
later), the use of a coarser mesh is not likely to significantly affect the conclusions of the present
ARTICLE IN PRESS

A. Deb et al. / International Journal of Impact Engineering 30 (2004) 1055–1079 1071

Fig. 20. Vehicle snap-shots on the left and close-up views of front structure on the right during front impact simulation
against a rigid wall. (a) At analysis time = 5 ms (no significant deformation has taken place). (b) At analysis time =
12 ms (damage in bumper beam, collapse of a significant part of crush box, and plastic deformation in mid-rail). (c) At
analysis time = 30 ms (complete collapse of crush box, partial collapse of mid-rail, and severe damage in the curved
part of upper longitudinal member).

study. The uniform mesh vehicle model (with 74,000 shell elements) is highly advantageous
computationally as it requires 14 h less for analysis compared to the refined vehicle model (with
1,14,000 shell elements) on a personal computer equipped with an AMD Athlon XP 2000+
processor and Windows 2000 operating system.
ARTICLE IN PRESS

1072 A. Deb et al. / International Journal of Impact Engineering 30 (2004) 1055–1079

Fig. 21. Refined vehicle front end mesh.

Fig. 22. Comparison of decelerations at the middle of first cross-member under the rear seat.

6. Assessment of occupant injury

In order to ensure occupant safety, it is essential that the front passenger compartment
deformation is kept to a minimum following impact and that peak deceleration is within
reasonable limits. Changes in horizontal (L1 in Fig. 5) and diagonal (L2 in Fig. 5) front left door
apertures are shown in Fig. 23 for analysis cases (a) and (b). While change in L1 is insignificant,
that in L2 is also less than 10 mm for both cases. Thus the chance of leg injury at the given test
mode appears to be minimal. The peak vehicle decelerations in Fig. 18 are seen to be 40 g for case
(a) and 38 g for case (b). Test deceleration curves (after processing with SAE 60 filter) for a
number of body-on-frame type sport utility vehicles sold in USA are shown in Fig. 24. The
referred tests were carried out by National Highway Traffic Safety Administration (NHTSA) in
USA for NCAP rating for which the impact speed was 56 kph (15.6 m/s), i.e. 8 kph higher than
ARTICLE IN PRESS

A. Deb et al. / International Journal of Impact Engineering 30 (2004) 1055–1079 1073

Fig. 23. Changes in horizontal and diagonal front door apertures during impact.

Fig. 24. Responses obtained from NCAP tests for various sport utility vehicles.

that in the regulatory FMVSS 208 test. It is seen that peak deceleration for these tests is in the
range of 30–65 g. Since all of these vehicles perform well in terms of occupant injury criteria (i.e.
chest deceleration and Head Injury Criterion (HIC) and receive a minimum of 3-star rating, the
peak deceleration of 38–40 g obtained in the current simulations can be considered as acceptable
for the case of FMVSS 208 test configuration if the steering column and restraint systems (i.e. seat
belts and airbags) are properly designed. It may be possible to further reduce the peak
deceleration for the considered test mode by effecting changes in the crush box as demonstrated in
the next section.

7. Optimization of front impact performance

In order to achieve further reduction in deceleration level for an impact speed of 30 mph
(13.33 m/s), a procedure similar to that followed for weld property optimization is adopted. As
ARTICLE IN PRESS

1074 A. Deb et al. / International Journal of Impact Engineering 30 (2004) 1055–1079

Table 3
Combinations of values of front rail design variables for DOE study
lcb (mm) tcb (mm) Tmr (mm) Remarks
207 2 2 Baseline, case (a) in Section 5 as well as DOE case 12
207 2 2.5 DOE case 1
207 2 3 DOE case 3
207 2.5 2.5 DOE case 7
207 2.5 3 DOE case 16
207 3 3 DOE case 13
232 2 2 25 mm increase in baseline value of lcb, DOE case 11
232 2 2.5 25 mm increase in baseline value of lcb, DOE case 2
232 2 3 25 mm increase in baseline value of lcb, DOE case 15
257 2 2 50 mm increase in baseline value of lcb, DOE case 10
257 2 2.5 50 mm increase in baseline value of lcb, DOE case 17
257 2 3 50 mm increase in baseline value of lcb, DOE case 14
282 2 2 75 mm increase in baseline value of lcb, DOE case 9
282 2 2.5 75 mm increase in baseline value of lcb, DOE case 18
282 2 3 75 mm increase in baseline value of lcb, DOE case 4
307 2 2 100 mm increase in baseline value of lcb, DOE case 8
307 2 2.5 100 mm increase in baseline value of lcb, DOE case 5
307 2 3 100 mm increase in baseline value of lcb, DOE case 6

peak deceleration and door intrusion yielded by each of the simulation cases (a) and (b) did not
differ appreciably (Figs. 18 and 23), the monolithic joint representation is followed in the DOE-
type study to be carried out. The parameters varied for investigating possible reduction in
deceleration are the length (lcb) and wall thickness (tcb) of crush box, as well as the wall thickness
(tmr) of mid-rail. The crush box and the mid-rail segments of front rails are shown in Fig. 16.
Eighteen DOE cases are considered and the various combinations of values of lcb, tcb and tmr are
given in Table 3.
In order to assess the relative performance of design options, a time-average of the acceleration
response history is used here in conjunction with an intrusion-related parameter. For example, for
a DOE case (designated as case 10) in which the crush box length is increased by 50 mm compared
to the baseline, and crush box and mid-rail gages are both 2.0 mm, the filtered vehicle acceleration
and its time-average responses are shown in Fig. 25. The concept of average acceleration also
provides a convenient paradigm for application of the lumped parameter-based approach
discussed in the next section for further safety improvement at a higher impact speed. The goal of
obtaining an enhanced design in the vicinity of the baseline solution is assumed to be met with the
minimization of the following objective function:
Ov ¼ aavg þ d1 ; ð5Þ
where
aavg ¼ average deceleration in g’s ðas in Fig: 25 but with sign reversedÞ; ð5aÞ

d1 ¼ DL1 =L1  100 ¼ percentage reduction in door lateral dimension ðsee Fig: 5Þ: ð5bÞ
ARTICLE IN PRESS

A. Deb et al. / International Journal of Impact Engineering 30 (2004) 1055–1079 1075

Fig. 25. Analysis-based acceleration response for FMVSS 208-optimized case 10 and its time-average representation.

Fig. 26. Variation of objective function for front impact safety optimization.

The results of the DOE study are given in Fig. 26. It is seen from the variation of Ov in Fig. 26
that case 10 which has been described earlier yields the optimum safety solution. The maximum
change in vehicle weight for the various DOE options is negligible as only stretching of hollow
front rails up to a maximum of 100 mm is considered. Furthermore, it is noted that the optimum
safety solution for which the peak deceleration is 38 g is only 2 g lower than the peak deceleration
of 40 g obtained in the baseline monolithic joint-based response given in Fig. 18. It is noted that in
the DOE cases analyzed, the first lateral cross-member (located at the rear of the bumper beam)
is kept unchanged in position relative to the fire-wall. Hence, increasing the length of crush
box beyond 75 mm initiates its global buckling and does not lead to any advantage in safety
performance.
ARTICLE IN PRESS

1076 A. Deb et al. / International Journal of Impact Engineering 30 (2004) 1055–1079

8. Lumped parameter idealization for NCAP evaluation

As a logical next step in the front crash safety design of the current vehicle platform, an NCAP
evaluation is made with the help of a lumped parameter model (LPM) of the FMVSS 208-
optimized solution represented by case 10 in Fig. 26. The generic spring–mass model employed is
shown in Fig. 27 in which the vehicle is lumped as a single mass m and the front-end stiffness
properties of the vehicle are idealized by a nonlinear spring with force–deformation behavior
given in Fig. 28. The LPM of Fig. 27 has been previously applied [9] in the context of headform
safety impact design with closed-form dynamic solutions for an initial velocity of m given by v0.
The same computational procedure is also adopted here.

Fig. 27. A lumped parameter idealization of front impact.

Fig. 28. Behavior of idealized spring [9].


ARTICLE IN PRESS

A. Deb et al. / International Journal of Impact Engineering 30 (2004) 1055–1079 1077

In the idealized vehicle front end behavior represented in Figs. 27 and 28, the following
interpretations are made of the various parameters shown:
Fy ¼ average force sustained by the vehicle front structure; ð6aÞ

Ke ¼ elastic stiffness of the vehicle front; ð6bÞ


Ku ¼ stiffness characterization of the back-up structure after front end deformation
ð6cÞ
bottoms out sf  Ke ðwhere sf is an assumed stiffness factorÞ;
ls ¼ total length of the crushable front end
ðin the current case; sum of lengths of crush box and mid-railÞ and ð6dÞ

du ¼ maximum crush that can be realized ¼ cf  ls


ðwhere cf can be referred to as a compaction factor ½9 Þ: ð6eÞ
Before the LPM of Fig. 27 can be used for NCAP safety assessment of the current platform, its
applicability to FMVSS 208 simulation should be established. To this end, the following values of
pertinent spring–mass properties are assumed for representing the average acceleration version of
case 10 in Fig. 25 (noting that te, tm and tf in Fig. 25, respectively, denote the end of elastic phase,
beginning of unloading and end of impact event; and, the average deceleration is 26.3 g) :
m ¼ 620 kg ðfrom full vehicle finite element model of Fig: 5Þ; ð7aÞ

v0 ¼ 30 mph ¼ 13:33 mm=ms; ð7bÞ

du E383 mm; dynamic crush from finite element analysis of case 10; ð7cÞ

Fy Em  26:3g ¼ 160 kN; ð7dÞ

te E5 ms; tm E49 ms; tf E70 ms; ð7eÞ


tm
cf E ¼ 0:7; ð7fÞ
tf
te
de E du ¼ 27:4 mm; ð7gÞ
tf
Fy
Ke E ¼ 5:85 kN=mm; ð7hÞ
de
du
ls E ¼ 547 mm and ð7iÞ
cf
sf ¼ 0:15: ð7jÞ
Analysis using the current LPM requires as input the quantities given by Eqs. (7a), (7b), (7d), (7f)
and (7h)–(7j). It is noted that the value of sf in Eq. (7j) is chosen using a tri-and-error approach to
ensure that the current LPM is able to reproduce the equivalent uniform crash pulse in Fig. 25
without the bottoming out of the front structure. The acceleration history obtained after analysis
ARTICLE IN PRESS

1078 A. Deb et al. / International Journal of Impact Engineering 30 (2004) 1055–1079

Fig. 29. Acceleration responses yielded by LPM of Fig. 27.

is given in Fig. 29 and a constant deceleration of 26.3 g corresponding to phase (II) deformation in
Fig. 28 is obtained. Thus the LPM-based response exactly matches the average deceleration of
Fig. 25. Hence, the validity of the estimated input spring properties given above is grossly
established. The LPM is now re-analyzed for the NCAP test speed of 35 mph (i.e. v0 ¼ 15:56 mm/
ms). In this case, complete crush of the fore-rail comprising crush box and mid-rail takes place
resulting in a higher peak deceleration of 38 g and an average deceleration of approximately 32 g.
It is known from prior experience with the LPM for the case of headform impact safety (which is
really a scaled counterpart of vehicle front impact in terms of energy content) that the parameter ls
(and hence du via (6e)) plays the most significant role in reducing deceleration. After carrying out
several trial-and-error runs with the LPM, it was found that, if the front-end stiffness and load
carrying capacity of the vehicle are held as approximately constant, the dynamic crush ( i.e. du)
needs to be increased to 518 mm (from a previous value of 383 mm, (7c)) so that the uniform
deceleration of 26.3 g of case 10 for the lower test speed of 30 mph can be obtained. The pertinent
LPM-based acceleration response with cross marks is shown in Fig. 29. In terms of ls, the
requirement would be to increase the length of fore-rail to 740 mm (for cf ¼ 0:7). However, if the
crushability of front rail can be raised to more than 70% (cf > 0:7), a lower value of ls will suffice.
The directional design guidance provided by the LPM-based analysis has been verified in a finite
element model corresponding to case 10 in which the front fore-rails have been lengthened to
740 mm. There is scope for reducing this length by increasing crushability of mid-rail and aft-rail
shown in Fig. 16. In the absence of the valuable insight provided by the present LPM, it would
have been much more time consuming and expensive to obtain the same conclusions by applying
full vehicle finite-element analysis.

9. Conclusions

The current paper illustrates a systematic simulation-driven approach towards designing a


lightweight vehicle platform with aluminium extrusion-based components. A space frame-type
vehicle architecture is introduced and evaluated for front impact safety. Benchmarking of joint
analysis results against a physical test carried out in the present study, and of axial collapse of a
square aluminium column against published numerical results and theoretical prediction add
ARTICLE IN PRESS

A. Deb et al. / International Journal of Impact Engineering 30 (2004) 1055–1079 1079

credibility to subsequent full vehicle finite-element modeling and analysis. A finite-element-driven,


DOE-based optimization approach has been followed in deriving weld element properties that
would yield close correlation to test behavior. A similar approach has also been adopted for
optimization of full vehicle front impact crashworthiness for the 30 mph impact speed conforming
to the FMVSS 208 standard of NHTSA in USA. By applying an efficient LPM-based approach,
insight on necessary change to front end crush space is obtained for the tougher NCAP standard
corresponding to an impact speed of 35 mph. The present design approach effectively combines
component testing, detailed finite-element-based analysis, and lumped parameter idealization
leading to shortened design cycle and reduced development costs. Based on the results obtained,
the current vehicle design, which is lightweight and eco-friendly when compared to a steel bodied
vehicle of similar dimensions and has been prototyped, holds out promise in terms of
crashworthiness.

Acknowledgements

The authors would like to thank the Altech Program involving Hydro Aluminium (Norway),
IISc (Bangalore, India) and NTNU (Trondheim, Norway) for providing impetus to the present
effort. Thanks are also due to Prof. D.H. Sastry, Coordinator of the Altech Program at IISc, for
his interest and support, and to Prof. T. Welo, Department of Machine Design and Materials
Technology, NTNU, Trondheim for his valuable advice.

References

[1] Deb A. Transforming a lecture-oriented course in the design of automotive systems into a unique design
experience. ICEE 2002, Manchester, UK, August 18–21, 2002.
[2] http://www.nhtsa.dot.gov/cars/rules/standards/safstan2.htm.
[3] Santosa S, Wierzbicki T. Crash behavior of box columns filled with aluminium honeycomb or foam. Comput
Struct 1998;68:343–67.
[4] Hanssen AG, Langseth M, Hopperstad OS. Static crushing of square aluminium extrusions with aluminium foam
filler. Int J Mech Sci 1999;41:967–93.
[5] White MD, Jones N. Experimental quasi-static axial crushing of top-hat and double-hat thin-walled sections. Int J
Mech Sci 1999;41:179–208.
[6] Gupta NK, Velmurugan R. Consideration of internal folding and non-symmetric fold formation in axi-symmetric
axial collapse of round tubes. Int J Solids Struct 1997;34:2611–33.
[7] Abramowicz W, Jones N. Dynamic progressive buckling of circular and square tubes. Int J Impact Eng
1986;4:234–70.
[8] Langseth M, Hopperstad OS. Static and dynamic axial crushing of square thin-walled aluminium extrusions. Int J
Impact Eng 1996;18:949–68.
[9] Deb A, Chou CC, Barbat SD. An analytical study on headform impact protection space for a rigid target. SAE
Paper No. 2000-01-0608.
[10] Aluminium extrusion manual, SAPA, www.sapagroup.com.
[11] Sharp ML. Behavior and design of aluminium structures. New York: McGraw-Hill Education; 1993.
[12] Chan TK, Porter Goff RFD. Welded aluminium alloy connections: test results and BS8118. Thin-Walled Struct
2000;36:265–87.
[13] Aalberg A, Langseth M, Larsen PK. Stiffened aluminium panels subjected to axial compression. Thin-Walled
Struct 2001;39:861–85.

Vous aimerez peut-être aussi