Vehicle Crash Can FEA Model Simulation and Validation With Experiment Data

Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

Nonlinear Engineering 2015; 4(2): 83–90

Thomas Bayley, Sekhar Reddy, and Mohammad Fard*

Vehicle Crash Can FEA Model Simulation and


Validation with Experiment Data
DOI 10.1515/nleng-2014-0030 teristics will be preserved. Since the vehicle collision is a
Received December 12, 2014; revised March 12, 2015; accepted May transient event involving highly non-linear loading, large
10, 2015. scale deformation and high volume of energy transfer, the
Abstract: Vehicle crash-can transfers the kinetic energy performance of the crash can has been an interesting area
into strain energy stored in the structure via a sequential for researchers. As the DANNER/TIC insurance ratings was
deformation wall crushing process. introduced into the automotive industry, consumers’ cost
The performance of the crash can FEA model and its abil- of ownership and purchase selection may be driven and
ity of predicting mechanical behaviours in a crash event is influenced in the Europe market and an effort to reduce
assessed and verified via the FEA analysis and correspond- the crash repair cost from the insurance company [1], vari-
ing experiment data. Compression tests are performed to ous geometries and configurations of crash can have been
validate the FEA simulation results. Wall tolerance sensi- studied and evaluated.
tivity is also investigated via FEA simulation in order to Wierzbicki introduced the self-consistency theory de-
examine if and how the manufacturing defect could im- scribing the crashing behaviour of a class of thin-walled
pose impact on crash can energy absorption capability. It structures in 1983 [2]. A basic folding mechanism was con-
is found that the performance of crash-can with manufac- structed to represent the main features of folds and wrin-
turing tolerances can be predicted by the FEA model de- kles on typical standard sheet metal structures. The re-
rived in this paper. Internal energy analysis indicates the searches on topology options have been done by Abbasi
wall thickness is tolerance insensitive and manufacturing et al. [3], Omar et al. [4] and Chen, and Wierzbicki [5]
defect can be disregarded under the manufacturing toler- about the energy absorption capability with various ge-
ance standards. ometry configurations. Lu & Yu [6] continued to study ge-
ometry options for energy absorption devices in conjunc-
Keywords: crash can, compression test, FEA, energy ab- tion with material alternatives. Various types of energy ab-
sorption, tolerance sensitivity sorbing tubes have been tested and studied with different
cross section areas, such as circular tubes, square tubes,
honeycomb tubes and taper tubes for automotive applica-
tions. Energy absorption capability of square tubes made
1 Introduction of dual phase and martensitic steels have been investi-
gated by Yan, Kantner, Zhu and Nadkarni [7]. An increased
Vehicle crash can is designed as an energy management energy absorbing performance has been reported compar-
device in the low speed (under 54 km/h) vehicle colli- ing with high strength low alloy steel (HSLA). An investiga-
sion. It is treated as a replaceable energy control device tion of extruded aluminium alloy AA5754 crash can under
to handle the impact energy and prevent the damage to different topologies has been carried out to study the spe-
the main rail. The kinetic energy possessed by the moving cific energy absorbing capability by Faruque and Saha [4].
vehicle will be converted into strain energy (internal en- Hexagon and octagon tubes are reported as the most ef-
ergy) and stored in the crash can via sequential deforma- ficient geometry followed by the square tube under sin-
tion behaviour. As a result, extra damage to the main rail gle cell structure. A comparative study of energy absorp-
will be prevented and high speed crashworthiness charac- tion of foam-filled and multi-cell square columns has been
conducted by Zhang, and Cheng [9]. The results showed
that a 50-100% higher energy absorption was observed
*Corresponding Author: Mohammad Fard: School of Aerospace, with multi-cell square columns comparing to foam-filled
Mechanical and Manufacturing Engineering, RMIT University, Mel-
ones. Abramowicz and Jones carried out a dynamic axial
bourne, Australia, E-mail: [email protected]
Thomas Bayley, Sekhar Reddy: School of Aerospace, Mechanical crushing on square tubes on a drop hammer test rig. Four
and Manufacturing Engineering, RMIT University, Melbourne, Aus- deformation modes governing the behaviour for different
tralia
84 | Thomas Bayley et al., Vehicle Crash Can FEA Model Simulation and Validation with Experiment Data

ranges of cross section are theoretically predicted. An in-


vestigation of stress/strain behaviour of aluminium alloys
(AA6 and AA7 family in T6 temper) at low to medium strain
rates has been done by Chen, Clausen, Hopperstad and
Langseth [8]. The results demonstrated the strain rate of
studied aluminium alloys are insensitive at low to medium
strain rates ranging from 0.1 to approximate 2000 s−1 .
The numerical process such as finite element method
is an invaluable tool to assess the crash behaviour and
the energy management. It also dramatically reduces the
cost of conducting physical test but delivers highly reli-
able results to predict the mechanical behaviour for ve- Fig. 1. Crash can and location.
hicle crashworthiness. Lee et al. investigated the high ve-
locity impact loading of thin walled structure by 3D FEA
method correlating to test. The simulation results yielded 4 Tension Test
a similar correlation to the test results. A non-linear FEA
analysis of quasi static axial collapse response of cylindri- The strain rate sensitivity of the crash-can material was ex-
cal tubes by multiple identical rings was carried out by amined using tensile test. Tensile test samples were made
Salehghaffari et al. They developed a global meta model of aluminium alloy AA6060 temper T6 (a commonly used
for mean crush force and energy absorption. alloy in the automotive industry) with a nominal chemical
In order to obtain accurate simulation result for the de- composition shown in Table 1. The samples were obtained
sign and development of crash can, a reliable FEA model directly from crash cans via cutting and machining per-
becomes crucial to the automotive industry. In this arti- formed by CNC lathe under the specification instruction
cle, a FEA model was built from crash can CAD file used from American Society for Testing and Materials (ASTM).
by current automotive industry with some reasonable sim- Fifteen test samples were prepared for the tensile test.
plifications. Physical axial loading test was conducted to Tensile test was performed in the room temperature
verify the simulation results for the purpose of validating to determine the strain rate sensitivity of crash can ma-
the accuracy of FEA model. The mechanical behaviours terial. Five test speeds were considered for obtaining the
of crash can was also investigated to demonstrate the strain rate sensitivity of the crash-can material as shown
crashing modes in different stages. Tolerance sensitivity in Table 2. Therefore, the fifteen test samples have been
of FEA model and crash can were studied with the aim of divided into five groups with three in each shown in Fig-
analysing manufacturing feasibility. ure 2. The thickness and width of each sample within the
gauge length were measured to obtain the cross-section
area shown in Table 2. The tensile test was carried out on
2 Method Instron 2716-020 testing machine as shown in Figure 2.
The load magnitude was then recorded against the strain.
The corresponding stress was calculated using load and
3 Crash Can Structure cross-section values. Therefore, the stress-strain curve was
drafted to analyse the strain rate sensitivity.
The automotive crash can which is used in this research is
shown in Figure 1. This crash can is a thin-walled structure
derived from square tube, which consist of two double-
layer and two single-layer side walls. On each double-layer
side, there is a connecting fin locating on the centre line
between the two walls shown in Figure 1(c). In vehicle, the
crash can is connected to the bumper beam (bumper rein-
force) as shown in Figure 1. The crash initiators are con-
sidered on both upper and lower sides of the crash can.
The crash initiators are used to generate initial sequential
deformation. Fig. 2. Tensile test samples.
Thomas Bayley et al., Vehicle Crash Can FEA Model Simulation and Validation with Experiment Data | 85

Table 1. Chemical composition of aluminium alloy AA6060-T6 (values in percentage).

Si Fe Mg Mn Cu Ti B Zn Al
0.44 0.22 0.48 0.016 0.003 0.007 0.001 0.01 Balance

Fig. 3. Tensile test platform.

Fig. 4. MTS 311.31s compression test system

5 Crash Can Compression Test


result, the range of the element size is set from 3 mm to
Crash can compression test was conducted to collect the 5 mm. Elastic plastic material type is applied to the crash
data for the validation of crash can simulation results. The can with properties shown in Table 3.
test was conducted on MTS 311.31S compression testing Two planes were created to comply with the test set-
machine as shown in Figure 3. The force magnitude was up shown in Figure 5(b). One plane is placed on top of the
recorded against the displacement. Two crash cans were crash can to manipulate the impactor and the other one
tested: one with 100 mm/s and the other with 150 mm/s. is attached to the bottom of crash can as the platform. The
area for both of them was set at 300 mm × 300 mm which is
bigger than the cross section area of the crash can so that
purely axial force is preserved at the interface. Rigid ma-
6 Crash Can FEA Modelling terial type is assigned to both of them and properties are
shown in Table 3. A mass element of 800 kg which is equal
The FEA model of crash can was developed based on the to half of the vehicle mass is allocated to the top plane to
mid-surface extracted from the CAD file in a global Carte- preserve the inertia effect as well as a prescribed velocity
sian coordinate system as shown in Figure 5. The general is defined to represent the crashing direction along the +X
modelling techniques, in automotive industry, were used direction.
to build up the crash can FEA model. According to the test condition, all six degrees of free-
The crash can body was meshed with Belytschko- dom at the interface of bottom plane and crash can must be
Lin-Tsay shell element which provides computational ef- clarified. Since the crash can is vertically placed on the test
ficiency while maintaining the result accuracy [10]. In or- platform without any side support, the bottom boundary
der to control the computing cost in a reasonably accept- conditions of the FEA model are detailed in Table 4 with
able scope with a relatively high accuracy of simulation respect to the coordinate system shown in Figure 5.
86 | Thomas Bayley et al., Vehicle Crash Can FEA Model Simulation and Validation with Experiment Data

Table 2. Tensile test groups.

SAMPLE Thickness (mm) Width (mm) Cross-section area


(mm2 )
1 1.726 8.972 15.4857
GROUP A (1MM/MIN) 2 2.503 8.975 22.4644
3 1.892 8.998 17.0242
1 1.859 8.945 16.6288
GROUP B (100MM/MIN) 2 1.910 8.976 17.1442
3 1.699 8.958 15.2196
1 1.835 8.975 16.4691
GROUP C (200MM/MIN) 2 1.806 8.963 16.1872
3 1.664 8.962 14.9128
1 1.716 8.955 15.3668
GROUP D (300MM/MIN) 2 1.812 8.990 16.2899
3 1.845 8.952 16.5164

Table 3. Property of material.

Type Density (kg/mm3 ) Yield strength (GPa) Young’s modulus Poison ratio
(GPa)
Aluminium 6060-T6 MAT 24 2.7×10−6 0.24 59 0.33
Steel MAT 20 7.8×10−6 n/a 210 0.33

The crash can body is divided into five parts for mea-
surement and a nominal definition of each part is clarified
in Figure 6. The measurement for each part was conducted
four times and all the values were recorded in Table 5. The
average value is assigned to each wall as thickness for sim-
ulation and the minimum and maximum values are used
for tolerance sensitivity investigation illustrated in a later
section. Since the connecting fin and inside wall join two
walls on double-layer side, a small continuous thickness
increase occurs at both ends. As the thickness change is
Fig. 5. FEA model of crash can. in a very small scale, the fin thickness is assumed to be
constant and the thickness along the corner line adopts
neighbour value.

Table 5. Wall thickness measurement (unit in mm).


Table 4. The boundary conditions of the bottom of the crash can for
FEA crash analysis (µ1 is static coefficient of friction; µ2 is dynamic
DW1 DW2 SW1 SW2 Connecting
coefficient of friction).
fin
1st Measurement 1.78 1.98 2.60 1.80 1.96
Degree of freedom Boundary condition
2nd Measurement 1.76 1.90 2.82 1.94 2.02
X translation free
3rd Measurement 1.72 1.98 2.56 1.88 2.04
Y translation µ1 = µ2 = 0.3
4th Measurement 1.84 1.98 2.70 1.92 1.98
Z translation µ1 = µ2 = 0.3
Average 1.775 1.960 2.670 1.885 2.00
X rotation µ1 = µ2 = 0.3
Minimum 1.72 1.90 2.56 1.80 1.96
Y rotation constrained
Maximum 1.84 1.98 2.82 1.94 2.04
Z rotation constrained
Thomas Bayley et al., Vehicle Crash Can FEA Model Simulation and Validation with Experiment Data | 87

Fig. 8. Aluminium alloy AA6060-T6 stress-strain curve under static


axial loading.

The maximum and minimum peak stress occurs at


257.824 MPa and 248.329 MPa under the speed of 1 mm/min
Fig. 6. Nominal definition of crash can side wall.
and 300 mm/min respectively. The stress difference can be
calculated as 3.7% indicating aluminium alloy AA6060-T6
is strain rate insensitive which comply with the result con-
7 Results cluded by Y. Chen et al. [8]. As a result, only static com-
pression test is needed to be conducted instead of various
7.1 Tensile Test Results dynamic tests in terms of the validation of FEA model ac-
curacy.
The tests were successfully done, so that all test sam- According to the 0.2% offset yield criterion, the yield
ples, after reasonable elongations, were broken within strength can be calculated as 240 MPa and the Young’s
the gauge length area under the axial loading (Figure 7). Modulus can be obtained as 59 GPa. They will be used as
The stress-strain curve of the test samples for 1, 100, 200, material properties together with the stress-strain curve in
and 300 mm/min velocities are shown in Figure 8. Each the simulation.
graph is an arithmetic mean of the three test runs.

7.2 Compression Test Results

The samples after the compression test are shown in Fig-


ure 9. It can be depicted that the crash initiators suc-
cessfully developed a sequential deformation pattern. The
crashing force was managed in a controlled manner and
was transferred progressively into the adjacent wall. Thus,
the bottom part which was not reached by crashing force
Fig. 7. Tensile sample after test.
was maintained in its original square shape. However, the
crash can body tears along the four corner lines, which in-
The maximum strain value reduces each time corre- dicates a reduced upward resistance force at the interface.
sponding to a higher test speed due to the inertia effect.
The maximum and minimum strain reaches 0.15335 and The force-time curve is represented in Figure 10(a).
0.9158 under the speed of 1 mm/min and 300 mm/min re- The maximum interface force occurs approximately 0.3 s
spectively. The yield strength and ultimate tensile strength earlier in case one than in case two due to a faster test
are both around 250 MPa. Hence, according to the charac- speed. However, the magnitude of both these forces are
teristics of aluminium family, it can be concluded that the nearly the same which were recorded as 142.878 kN and
stress-strain curve depicted above delivers confident and 144.372 kN for 150 mm/s and 100 mm/s respectively. The
accurate data reflecting the property of aluminium alloy variation patterns of interface force for both cases are sim-
AA6060-T6. ilar regardless of the test speeds which is illustrated in
88 | Thomas Bayley et al., Vehicle Crash Can FEA Model Simulation and Validation with Experiment Data

Case 1 (150 mm/s) crash can correlation


Case one simulation was conducted under the speed of
150 mm/s and force-displacement curve is plotted in Fig-
ure 11. The simulation result yields a good correlation
to test and predicts every mechanical behaviour of crash
can. The simulation precisely predicts each peak force at
21.88 mm, 70.03 mm, 80.97 mm, 102.86 mm and 116.72 mm
corresponding to each displacement in the test. The max-
imum interface force obtained at 70.03 mm is 142.961 kN
which is 0.06% higher comparing with the test value
142.878 kN.

Fig. 9. Crashing samples after compression test.

Figure 10(b). Moreover, the sequential deformation is re-


flected by the fluctuation of trend lines. The interface force
performs a more severe fluctuation at 100 mm/s than at
150 mm/s. Fig. 11. Force-displacement curve at the speed 150 mm/s.

Since the tearing occurred along the four corner lines,


the test result yields a lower force value than the simula-
tion result almost across the entire range. Before the in-
terface force reaches its maximum value, the structural in-
tegrity of the crash can is well maintained so that a good
correlation to test is preserved. However, interface force is
obtained at about 10% lower comparing with the simula-
tion result as severe tearing occurs subsequently.
Fig. 10. Force-displacement curve at the test speed of 100 mm/s and
150 mm/s.
Case 2 (100 mm/s) crash can correlation
Case two simulation was conducted under the speed of
100 mm/s and force-displacement curve is reported in
Figure 12. The correlation is little coarser than that in
7.3 Simulation Results case one, but the accuracy is still well preserved. Each
peak force is predicted at the specific location with an ac-
Crash can simulation is performed by the transient dy- ceptable displacement deviation. The maximum interface
namic nonlinear solver Ls-Dyna. The simulation results forces are 143.492 kN and 144.361 kN at 79.821 mm in the
were filtered by 10 Hz frequency filter so as to reduce the test and 75.503 mm in the simulation respectively.
high frequency ambient noise and provide a smooth curve Tearing also occurred along the four corner lines, but
for analysis. The crash initiators successfully generated the structural integrity of the crash can is better main-
the sequential deformation pattern indicated by the wave tained than the previous case due to the lower test speed.
shaped curves shown in the two figures below. The test result influenced by tearing starts to deflect the
simulation at 106.098 mm. A continuous increased force
difference between test and simulation was reported.
Thomas Bayley et al., Vehicle Crash Can FEA Model Simulation and Validation with Experiment Data | 89

Table 6. Internal energy analysis.

Max upper (J) Max lower (J) Max common (J) Max energy Max deviation
difference (J)
Case 1 22778.7 20877.1 21167.7 1611.0 7.6%
Case 2 18417.3 16801.7 17434.8 1615.6 9.3%

Fig. 13. Internal energy-time curves.

Fig. 12. Force-displacement curve at the speed 100 mm/s.


8 Conclusions
Tolerance sensitivity investigation Tensile test results reveal that aluminium alloy AA6060-
Due to the imperfection of manufacturing process, toler- T6 is strain rate in sensitive so that only static compres-
ance deviation will be introduced into the final product. sion test is needed in terms of verification of crash can
The wall thickness of the crash can will be influenced by FEA model. Young’s modulus is obtained as 59 GPa and
such manufacturing defect so that cross section area is al- yield strength is calculated as 240 MPa from the stress-
tered. Since the stress is determined by the external load strain curve. Compression test is conducted at two differ-
and cross section area, it becomes a variable quantity with ent speeds of 150 mm/s and 100 mm/s respectively. The
the cross section area. As a result, the strain energy (inter- test result indicates both of them reach a similar maxi-
nal energy) that the crash can is able to accumulate may mum force regardless of the test speed. However, tearing
shift subsequently. The internal energy analysis approach occurred along the four corner lines for both cases. Crash
will be adopted to determine the tolerance sensitivity in initiators successfully generate the sequential deforma-
the following section. tion in both compression test and simulation. The simu-
The internal energy with respect to time is shown in lation force-displacement curves are well correlated with
Figure 13. Both graphs indicate a significant fast energy the physical test data. So the crash can FEA model is func-
increase in a very short time. Since the kinetic energy pos- tional to predict the mechanical behaviours during crash-
sessed in case one is larger than that in case two, the max- ing event. Tolerance sensitivity investigation reveals the
imum internal energy reaches a higher value in case one. wall thickness is tolerance insensitive so that the manu-
The maximum energy difference and deviation are calcu- facturing defect introduced into the final product can be
lated with respect to the common case and listed in Ta- ignored.
ble 6. Considering the reasonable simplification of bound-
ary condition at the bottom of crash can and the GM manu-
facturing standard, the energy deviation is evaluated in an
acceptable extent. As a result, the wall thickness is treated References
as tolerance insensitive.
[1] A. Mamalis, D. Manolakos, A. Baldoukas, G. Viegelahn (1991),
Energy dissipation and associated failure modes when axially
loading polygonal thin-walled cylinders, Thin-Walled Struc-
tures, 12 (1991) 17-34.
[2] T. Wierzbicki (1983), On the crushing mechanics of thin-walled
structures, Journal of Mechanics, Vol. 50/727
90 | Thomas Bayley et al., Vehicle Crash Can FEA Model Simulation and Validation with Experiment Data

[3] M. Abbasi, A. Ghafari-Nazari, S. Reddy, M. Fard (2014), A new [8] Y. Chen, A.H. Clausen, O.S. Hopperstad, M. Langseth (2009),
approach for optimizing automotive crashworthiness: con- Stress-strain behaviour of aluminium alloys at a wide range of
current usage of ANFIS and Taguchi method, Structural and strain rates, International Journal of Solids and Structures 46
Multidisciplinary Optimization, Volume 49, Issue 3, pp 485- (2009), 3825-3835
499 [9] X. Zhang, G. Cheng (2007), A comparative study of energy ab-
[4] Omar Faruque, Nripen Saha (2008), Extruded aluminium crash sorption characteristics of foam-filled and multi-cell square
can topology for maximizing specific energy absorption, SAE columns, International Journal of Impact Engineering, 34
Technical Paper, 2008-01-1500 (2007) 1739-1752
[5] W. Chen, T. Wierzbicki (2001), Relative merits of single-cell, [10] O.S. Hopperstad, T. Berstad, H. Ilstad, O.-G. Lademo, M.
multi-cell and foam-filled thin-walled structures in energy Langseth (1998), Effects of the yield criterion on local defor-
absorption, Thin-Walled Structures, 39 (2001) 287-306 mations in numerical simulation of profile forming, Journal of
[6] Lu G, Yu Tongxi (2003), Energy absorption of structures and Materials Processing Technology, 80-81 (1998) 551-555
materials, Woodhead Publishing Limited, Cambridge, UK
[7] B. Yan, C. Kantner, H. Zhu, G. Nadkarni (2005), Evaluation of
crash performance of a hat section component using dual
phase and martensitic steels, SAE Technical Paper 2005-01-
0837

You might also like