Professional Documents
Culture Documents
Inwardscrushingpg210 191
Inwardscrushingpg210 191
A Thesis
of
Drexel University
by
Deepak Siromani
of
Doctor of Philosophy
December 2013
ii
© Copyright 2013
Deepak Siromani. All Rights Reserved
iii
DEDICATIONS
ACKNOWLEDGEMENTS
help, and support of many. First and foremost, I would like to express my sincere
gratitude to my two advisors, Professors Jonathan Awerbuch and Tein-Min Tan of the
Department of Mechanical Engineering and Mechanics. This dissertation would not have
been possible without their guidance, their constant support, and their encouragement
Lamberson, Matthew McCarthy, and Frank Moon. I appreciate their willingness to serve
I am deeply grateful to Dr. Alan Byar of The Boeing Co., Seattle, Washington,
who developed the original Boeing 737 model which was the basis for my extended
For the experimental work on the composite stanchions, credit is due to the
several undergraduate senior students who were instrumental in the process of fabricating
the specimens and performing the experiments, as part of their senior design projects.
They are: Gary Henderson, Doug Mikita, Kevin Mirarchi, Ryan Park, John Smolko,
Benjamin Cheng, Michael DeLuca, Daniel Donegan, Patrick Giberson and Christopher
Mucerino. I also wish to acknowledge the help of Mr. Douglas Ludin of The Boeing Co.,
Ridley Park, Pennsylvania, for providing the test materials and facilitating the fabrication
providing travel funds to various conferences, and the National Science Foundation
is deeply appreciated. Specifically, I am very grateful to Dr. Eli Fromm for giving me the
opportunity to take a leadership role in this course sequence which undoubtedly enhanced
excellent undergraduate and graduate education. The partial support provided by the
MEM Department, through its teaching assistantship program, was also highly valuable. I
am grateful for my colleagues and friends in the MEM department and in our research
group, especially Chris Swin, Reewanshu Chadha and Andrew Bergan, for their help
with numerous issues, for the thoughtful discussions and suggestions along the way and
Most importantly of all, I would like to thank my family for always supporting
me, encouraging me, and believing in me through all the major decisions I have had to
make so far. Without them, I would not be the person I am today. Last, but not the least, I
would like to thank all my friends for their help and support over the years, that has
TABLE OF CONTENTS
Vita ............................................................................................................................255
xii
LIST OF TABLES
Table 3-1: A comparison of the finite element modeling details between the original
and modified section models ...................................................................... 47
Table 7-3: Results of the parametric study showing the effect of DFAILC in MAT54
on the peak load, crush and SEA of the circular tube. ............................. 154
Table 7-4: Results of the parametric study showing the effect of SOFT in MAT54 on
the peak load, crush and SEA of the circular tube. .................................. 154
Table 7-5: Results of the parametric study showing the effect of DFAILC in MAT54
on the peak load, crush and SEA of the C-channel with a chamfer. ........ 157
Table 7-6: Results of the parametric study showing the effect of SOFT in MAT54 on
the peak load, crush and SEA of the C-channel with a chamfer. ............. 157
Table 7-7: Results of the parametric study showing the effect of DFAILC in MAT54
on the peak load, crush and SEA of the C-channel with a steeple. .......... 158
Table 7-8: Results of the parametric study showing the effect of SOFT in MAT54 on
the peak load, crush and SEA of the C-channel with a steeple. ............... 158
Table 7-9: Correlation between the experimental and simulation results for each cross-
section with (a) a chamfer trigger, and (b) a steeple trigger..................... 158
Table 7-10: Hourglass control options investigated for under-integrated elements .... 161
Table 7-11: Tiebreak input parameters for Option 8 and 11 ....................................... 169
LIST OF FIGURES
Figure 2-1: Vertical drop test of a Boeing 737 fuselage section, conducted at the FAA
WJH Technical Center [13,20] ................................................................... 18
Figure 2-2: Boeing 737 fuselage section (a) before lifting the test article to conduct the
drop test, and (b) after impact [13,20] ........................................................ 19
Figure 2-3: Finite element model of the 10-ft long Boeing 737 fuselage section
[13,20,21] ................................................................................................... 20
Figure 2-4: Detailed finite element model of the frame and cargo door of the Boeing
737 fuselage section [13,20,21] .................................................................. 21
Figure 2-5: Heath-Tecna bin in test article (left) and its finite element model (right)
[13,20] ........................................................................................................ 22
Figure 2-6: Hitco bin in test article (left) and its finite element model (right) [13,20] . 22
Figure 2-7: MAT82 material model used for the aluminum alloys [21] ...................... 23
Figure 2-8: Deformation of frames, as viewed from top front side of the model, for four
different coefficients of friction (t = 200ms) .............................................. 25
Figure 2-9: Detailed comparison of deformation on the lower LHS of the frame (t =
200ms) ........................................................................................................ 26
Figure 2-10: Comparison of acceleration-time histories at the four FS 452 Seat Track
locations for different coefficients of friction ............................................ 28
Figure 2-11: Dissipation of kinetic energy by the entire fuselage, the frames, and the
luggage during the impact duration for a friction coefficient of 0.3 between
the fuselage and impact surface.................................................................. 30
Figure 2-12: Comparison of energy dissipated by (a) frames; and (b) luggage, for four
different coefficients of friction ................................................................. 30
Figure 2-13: Comparison of baseline deformation with and without luggage (coefficient
of friction = 1.0) ......................................................................................... 32
xv
Figure 2-14: Effect of luggage on the acceleration-time histories at two FS 452 seat
track locations............................................................................................. 32
Figure 2-15: Side views of the deformed fuselage section for six different angles of
impact at 200ms.......................................................................................... 33
Figure 2-16: Deformation of frames for six angles of impact as viewed from top front
side of the model at 200 ms ........................................................................ 34
Figure 2-18: Dissipation of kinetic energy by the entire fuselage, the frames, and the
luggage during the impact duration for impact angles of 15° and 90° ....... 36
Figure 2-19: Comparison of energy dissipated by (a) frames; and (b) luggage for six
different angles of impact ........................................................................... 37
Figure 3-2: The original and modified B737 section models. The masses of the
subcomponents that were excluded from the original model were assigned
to the nodes to which the subcomponents are connected. .......................... 47
Figure 3-3: A comparison of the deformation of the original and modified section
models at selected stages of a rigid-surface impact ................................... 48
Figure 3-4: A comparison of seat track acceleration-time histories at FS 418 for rigid
surface impact............................................................................................. 49
Figure 3-5: Setup of the original section model for water impact, including a layer of
air on top of the water surface .................................................................... 50
Figure 3-6: A comparison of the deformed fuselage at 200ms of the original and
modified section models under water impact conditions. .......................... 52
Figure 3-7: A comparison of damage progressions in the fine meshed model and the
coarse meshed model. Failures in coarse meshed model occurred more
abruptly, causing sudden transmission of load to surrounding elements ... 52
Figure 3-8: A comparison of seat track acceleration-time histories at FS 418 for water
impact ......................................................................................................... 53
Figure 3-9: The 30-m long full-length model with a cargo door, under-floor luggage,
and simple nose and tail cones. FWD (FS 300), MID (FS 480), and AFT
(FS 700) indicate the locations of accelerometer elements. ....................... 56
xvi
Figure 3-10: Crash landing scenario used in this study: The fuselage has a 110 knots
approach speed, a 9.27° approach angle, and a 3° pitch angle. .................. 57
Figure 3-11: Deformation of the full-length fuselage model crash landing onto a rigid
surface. ....................................................................................................... 59
Figure 3-12: Deformation of the full-length fuselage model crash landing onto soil. .... 59
Figure 3-13: Deformation of the full-length fuselage model crash landing onto water. . 60
Figure 3-14: The overall deformation and contours of effective plastic strain of the
frames at selected time steps resulted from crash landings on three types of
terrains. ....................................................................................................... 62
Figure 3-15: Example of (a) the formation of plastic hinges at t = 52 ms at FS 800, near
the aft section, (b) frames impacting the passenger floor at t = 152 ms., and
(c) the occurrence of local frame buckling at 200ms, FS 200. ................... 63
Figure 3-16: Seat tracks acceleration-time histories resulting from a crash landing of a
full-length fuselage on a rigid surface. The magnitude of the peak
acceleration pulses progressively increased from the AFT to FWD
locations...................................................................................................... 64
Figure 3-17: Seat track acceleration-time histories comparison at aft and forward
locations resulting from a crash landing of a full-length fuselage on rigid,
soil and water terrains................................................................................. 65
Figure 3-18: Energy dissipation for a crash landing of a full-length fuselage on a rigid
surface. Frames and skins dissipated most of the energy. .......................... 66
Figure 3-19: Energy dissipations for crash landing on soil and water surfaces. Frames
and skin dissipated most of the energy. ...................................................... 67
Figure 3-20: Energy dissipation by (a) frames, and (b) skin for crash landing on rigid,
soil and water surfaces. .............................................................................. 68
Figure 3-21: The overall deformation and contours of effective plastic strain of the
frames at t = 152 ms resulted vertical impact on three types of terrains .... 69
Figure 3-22: Seat tracks acceleration-time histories for full-length fuselage vertical
impact on a rigid surface. ........................................................................... 70
Figure 3-23: Seat track acceleration-time history comparison at aft and forward locations
resulting from vertical impact of a full-length fuselage on rigid, soil, and
water terrains. ............................................................................................. 71
Figure 3-24: Energy dissipation for vertical impact on rigid surface. Frames and skin
dissipated most of the energies. .................................................................. 72
xvii
Figure 3-25: Energy dissipation for vertical impact on soil and water. For both cases the
frames and skin dissipated most of the energies. ....................................... 72
Figure 3-26: Energy dissipation by (a) frames, and (b) skin for vertical impact on rigid,
soil and water surfaces. .............................................................................. 73
Figure 3-27: A comparison of the deformed fuselage with luggage at 200ms under rigid
and water impact conditions. ...................................................................... 74
Figure 3-28: Energy dissipation for rigid and water impact conditions, showing
significant energy absorbed by the luggage. .............................................. 75
Figure 3-29: Seat track acceleration-time histories showing the effect of luggage for the
case of vertical rigid surface impact. .......................................................... 76
Figure 3-30: Seat track acceleration-time histories for effect of luggage in the case of
vertical water surface impact. ..................................................................... 76
Figure 4-2: (a) B737 fuselage section showing spring elements as energy absorbers,
and (b) a close-up view of a spring element showing the rigid connectors 81
Figure 4-3: Comparison of the deformation of the B737 fuselage section models with
luggage, without luggage, and with spring elements as energy absorbers at
three time intervals. .................................................................................... 83
Figure 4-4: Comparison of the acceleration-time histories at two FS 452 seat track
locations: (a) LHS outer seat track, and (b) RHS outer seat track. ............ 84
Figure 4-5: Comparison of the energy dissipated by the luggage and the spring
elements ...................................................................................................... 84
Figure 5-1: Failure trigger mechanisms: (a) chamfered-end; (b) Inward-splaying crush-
cap; and (c) Outward-splaying crush-cap. .................................................. 93
Figure 5-3: A comparison of initial peak loads and sustained crush loads for Group A
specimens tested in Phase I and II. ............................................................. 98
Figure 5-4: A comparison of SEA for Group A specimens tested in Phase I and II..... 99
Figure 5-7: A comparison of (a) initial peak loads and sustained crush loads, and (b)
SEA of Group A specimens with a combined, chamfered-end and inward-
folding crush-cap, trigger mechanism. ..................................................... 102
Figure 5-9: A comparison of (a) initial peak loads and sustained crush loads, and (b)
SEA of Group A and Group B specimens with chamfered-ends and
combined trigger mechanisms. ................................................................. 104
buckled strips; and e) Post-test view of the crushed end of the specimen,
showing the separation of the fiber strips from the tightly packed core. . 109
Figure 5-16: Global axial strain field of a chamfered specimen. The load increased
gradually, reaching the peak level (23.8 kN) when the entire chamfer
crushed completely, followed by an abrupt load drop, from the peak level
to 17.0 kN, Hot spots representing matrix splitting can be seen clearly to
initiate from the crushed region and propagate upwards at 17.0 kN load
level. ......................................................................................................... 113
Figure 5-17: Global hoop strain field of a chamfered specimen. The load increased
gradually, reaching the peak level (23.8 kN) when the entire chamfer
crushed completely, followed by an abrupt load drop, from the peak level
to 17.0 kN, Hot spots representing matrix splitting can be seen clearly to
initiate from the crushed region and propagate upwards at 17.0 kN load
level. ......................................................................................................... 113
Figure 5-18: Local hoop strain field of a chamfered specimen recorded using a high-
speed DIC system. Hot spots, representing matrix splitting, initiated at 15.9
kN and continued to grow as the load increased. ..................................... 114
Figure 6-1: Open-cross-sections with the two failure trigger mechanisms, all having the
same cross-sectional area and attached to a potted base to ensure stability.
.................................................................................................................. 121
the average of three tests (except for the angle-stiffener with a chamfer
trigger which represents the average of two tests). .................................. 124
Figure 6-3: A comparison of the (a) peak load and crush load, and (b) SEA for each
cross-section having a chamfer trigger and steeple failure triggers (numbers
indicate average of three specimens except for the angle stiffener with a
chamfer trigger which represents the average of two tests). .................... 125
Figure 6-5: Axial strain field throughout the web of a chamfered C-channel specimen.
.................................................................................................................. 128
Figure 6-6: Lateral strain field of a chamfered C-channel specimen showing sites of
matrix split initiation. ............................................................................... 128
Figure 6-8: Axial strain field of a steeple C-channel specimen showing steeple
crushing. ................................................................................................... 130
Figure 6-9: Lateral strain field of a steeple C-channel specimen showing sites of
initiation of matrix splitting. ..................................................................... 130
Figure 6-11: Axial strain field of a chamfered angle-stiffener specimen. .................... 132
Figure 6-12: Lateral strain field of a chamfered angle-stiffener specimen showing site of
initiation of matrix splitting. ..................................................................... 132
xxi
Figure 6-14: Axial strain field of a steeple angle-stiffener specimen. ............................ 134
Figure 6-15: Lateral strain field of a steeple angle--stiffener showing site of initiation of
matrix splitting. ........................................................................................ 134
Figure 6-17: Axial strain field of a chamfered hat-stiffener specimen. ........................ 136
Figure 6-18: Lateral strain field of a chamfered hat-stiffener specimen, showing sites of
initiation of matrix splitting. ..................................................................... 136
Figure 6-20: Axial strain field of a hat-stiffener specimen with a steeple trigger. ....... 138
Figure 6-21: Lateral strain field of a hat-stiffener specimen with a steeple trigger. ..... 138
Figure 6-22: Effect of the ratio of curved surfaces to the total perimeter on the SEA of
specimens with various cross-sections ..................................................... 139
Figure 6-23: Final deformation (after load removal) of: a) C-channel, b) angle-stiffener,
c) hat-stiffener, and d) Circular tube (from [81, Chapter 5]). .................. 140
Figure 7-1: Test specimens cross-sectional dimension (all dimensions in mm). ........ 148
Figure 7-3: Representations of the trigger mechanism in the single-layer finite element
models of (a) circular tube with a chamfer, and (b) C-channel with a
steeple. ...................................................................................................... 150
Figure 7-4: Penetration of the platen by tube elements according to the load-penetration
curve in the contact definition. ................................................................. 152
Figure 7-5: Final load-penetration curve for the contact definition used in single-layer
model. ....................................................................................................... 152
Figure 7-6: Unfiltered vs. filtered load-displacement curves obtained from the single-
layer chamfered model. An SAE 1000 Hz filter was used to obtain the
filtered data. .............................................................................................. 153
Figure 7-9: Representations of: (a) circular tube with a chamfer and (b) C-channel with
a steeple failure trigger mechanisms in the finite element models........... 160
Figure 7-10: Time-scaling investigation showing: a) loading rates used, and b) the
resulting load-crosshead displacements. .................................................. 163
Figure 7-12: A comparison of: (a) the unfiltered and filtered load-crosshead displacement
curves from a multi-layer simulation for a tube with a chamfer failure
trigger, and (b) the filtered simulation results and the experimental load-
crosshead displacement comparison for a tube with a chamfer failure
trigger. ...................................................................................................... 173
Figure 7-13: A comparison of: (a) the unfiltered and filtered load-crosshead displacement
curves from a multi-layer simulation for a C-channel with a chamfer failure
trigger, and (b) the filtered simulation results and the experimental load-
crosshead displacement comparison for a C-channel with a chamfer failure
trigger. ...................................................................................................... 174
Figure 7-14: Tiebreak investigation showing the comparison between: (a) the load-
crosshead displacement curves for the chamfered tube, and b) the final
deformation of the tubes. .......................................................................... 175
xxiii
Figure 7-15: Tiebreak investigation showing the comparison between: (a) the load-
crosshead displacement curves for the chamfered tube, and b) the
deformation of the steeple C-channel at ~23 mm of crush. ..................... 175
Figure 7-16: Material model investigation showing the comparison between: (a) the load-
crosshead displacement curves for the chamfered tube, and b) the final
deformation of the tubes. .......................................................................... 177
Figure 8-1: Failure trigger mechanisms: (a) a chamfered end, (b) an inward-folding
crush cap, and (c) an outward-splaying crush cap.................................... 185
Figure 8-2: (a) Experimental load-crosshead displacement curves, and (b) SEA of tubes
having a chamfer and combined failure triggers. The combined failure
trigger yielded a higher peak load, sustained crush load and SEA. ......... 185
Figure 8-3: Representations of the failure trigger mechanisms in the finite element
models. The chamfered end of the tube was modeled by staggering
element layers while the crush cap was modeled using rigid shell elements.
.................................................................................................................. 187
Figure 8-4: Comparison of the experimental and simulated deformation of the tube
with a chamfered trigger mechanism at various loading stages. .............. 194
Figure 8-9: Comparison of simulated and experimental (a) peak loads and sustained
crush loads, and (b) SEA for the tubes with a chamfer trigger mechanism
and a combined trigger mechanism. ......................................................... 197
Figure 8-10: Comparison of simulated and experimental (DIC) local hoop strain fields
on a tube with a chamfer trigger mechanism. .......................................... 199
Figure 9-1: Open-cross-sections with two failure trigger mechanisms (chamfers and
steeples), all having the same cross-sectional area and attached to a potted
base to ensure stability during the 50.8 mm crush displacement. ............ 206
xxiv
Figure 9-2: Test specimens cross-sectional dimension (all dimensions in mm) ......... 207
Figure 9-4: A comparison of: (a) initial peak load and crush load; and (b) SEA, for
each cross-section having a chamfer and steeple failure triggers (numbers
indicate average of three specimens except for the angle stiffener with a
chamfer trigger which represents the average of two tests). .................... 208
Figure 9-5: Representations of the C-Channel cross-section having: (a) chamfer and (b)
steeple failure trigger mechanisms in the finite element models. ............ 210
Figure 9-7: Comparison of the experimental and simulated deformation of the angle-
stiffener with a chamfered trigger mechanism at selected loading stages.
.................................................................................................................. 217
Figure 9-8: Comparison of the experimental and simulated deformation of the hat-
stiffener with a chamfered trigger mechanism at selected loading stages.
.................................................................................................................. 218
Figure 9-10: Comparison of the experimental and simulated deformation of the angle-
stiffener with a steeple trigger mechanism at selected loading stages. .... 219
Figure 9-11: Comparison of the experimental and simulated deformation of the hat-
stiffener with a steeple trigger mechanism at selected loading stages. .... 220
Figure 9-16: Comparison of simulated and experimental peak loads and sustained crush
loads for the specimens with (a) a chamfer, and (b) a steeple failure trigger
mechanism. ............................................................................................... 224
Figure 9-17: Comparison of simulated and experimental SEA for the specimens with (a)
a chamfer, and (b) a steeple failure trigger mechanism. .......................... 224
Figure 9-18: Comparison of simulated and experimental (DIC) axial strain fields in a C-
channel with a chamfer failure trigger mechanism. ................................. 226
Figure 9-19: Comparison of simulated and experimental (DIC) lateral strain fields in a
C-channel with a chamfer failure trigger mechanism. ............................. 226
Figure 9-20: Comparison of simulated and experimental (DIC) of the axial strain fields
in a C-channel with a steeple trigger mechanism. .................................... 227
Figure 9-21: Comparison of simulated and experimental (DIC) of the lateral strain fields
in a C-channel with a steeple trigger mechanism. .................................... 227
xxvi
ABSTRACT
Crashworthy Design and Analysis of Aircraft Structures
Deepak Siromani
Advisors:
Prof. Jonathan Awerbuch and Prof. Tein-Min Tan
developed previously for simulating the drop test of a 3m long Boeing 737 fuselage
section was used to evaluate the effect of the friction coefficient between the fuselage and
the ground, and of the aircraft’s angle of impact on the dynamic response of the structure.
scenarios on different terrains (i.e., rigid, soil, and water). The results from these studies
highlighted the importance of the subfloor structure and its components in the energy
absorption process during a crash landing, as well as the need for energy absorbing
devices, integrated with the subfloor structure, to mitigate the impact energy.
subfloor structure of aircraft or rotorcraft. First, tubes of circular cross sections with
chamfered-ends and steeple failure trigger mechanisms were investigated. The optimal
xxvii
configuration that resulted in the lowest initial peak load while providing the highest
Finite element models were developed to simulate the crushing behavior of the
methodologies. First, an existing single-layer approach was utilized that required careful
calibration of key parameter values used in defining the contact/penetration behavior and
material failure. This approach predicted the initial failure peak load and the load-
crosshead displacement curve but provided no insight into the failure process. Next, a
laminate configuration, element size and formulation, contact definitions, time step
control, delamination interface, and material model. This approach captured the failure
process and predicted the sustained crush load quite accurately. Such modeling could
CHAPTER 1: INTRODUCTION
between October 2001 and September 2009, the U.S. military lost 375 rotorcraft causing
496 fatalities. Only 19 percent of losses were due to combat hostile action whereas the
remaining 81 percent were due to mishaps. A high percent of these loses were deemed
survivable as most of the fatalities resulted from the crash. Rotorcraft crashworthiness,
which includes airframe crashworthiness and occupant protection, was identified as a key
crashworthiness is not only limited to rotorcraft or small aircraft as, over the past decade,
several aircraft crashes have occurred in which narrow and wide-body transport airplanes
were forced to perform emergency landings either during takeoff or landing. Some of
these involved the collapse of the landing gear, causing the aircraft fuselage to impact the
ground resulting in severe, but survivable damage [2-6]. In each of these examples, a
majority of the occupants survived the crash landings that caused severe damage to the
aircraft.
During a crash, the aircraft structure must be able to absorb the kinetic energy and
limit the forces and decelerations that are transmitted to the occupants, through the seat
and restraint system, to tolerable levels. The airframe must survive with minimal
collapse, to the extent that a livable volume is provided to the occupants through the
fatalities and reduce the amount of structural damage to the airframe and payload in the
2
event of a severe, but survivable, crash [7]. The larger the energy absorbed by the
While the history of crashworthiness studies can be traced back to 1910 [7], the
first modern approach to defining crashworthiness requirements for rotorcraft and small
fixed-wing aircraft was undertaken by the U.S. Army in the 1960’s. Studies conducted as
part of this program determined that the vertical load transmitted to the occupant during a
crash was a key factor in causing injury to the lumbar spine [7]. Thus, the attenuation of
this load has long been recognized as a vital component of a crashworthy design.
To address this design issue, there is a need for experimental data on the impact
response of aircraft under survivable crash conditions. Over the past few decades, there
have been various research programs on aircraft crashworthiness conducted by the U.S.
Department of Defense [8,9], NASA [8] and the FAA [8,9] to determine the dynamic and
impact responses of aircraft structures and the survivability of the occupants under severe
but survivable crash conditions. A summary of the impact tests performed at NASA
Langley’s Impact Dynamics Research Facility (IDRF) from 1975 until 2003 is provided
in [8]. These tests ranged from a full-scale crash test of a CH-47 Chinook helicopter to an
external fuel qualification test of a UH-60 Black Hawk helicopter. The IDRF facility has
the ability to perform full-scale crash tests of light aircraft and rotorcraft for a wide range
of combined forward and vertical velocity conditions. For example, a crash test of a
Sikorsky prototype helicopter, with a forward velocity of 9.6 m/s and a vertical velocity
of 11.6 m/s, was performed [10]. The FAA has performed impact tests on some larger
[12], a Lear Fan fuselage [11], and two Boeing 737 fuselage sections [13,14].
3
impacting hard surfaces, no such studies have been reported on full transport aircraft
fuselages impacting other terrains, such as soil and water. Yet, there have been cases
where the aircraft have crash-landed onto such surfaces [4,5]. Additionally, it has been
reported that more than 80% of rotorcraft crashes occur on non-rigid terrains and that
helicopters designed for hard surface impact do not perform well when impacting onto
water and soft-soil [15]. The US Army Research Lab conducted a series of vertical drop
tests of composite helicopter fuselages on concrete [16], soft-soil [17], and water [18].
The dynamic response of each case was evaluated and compared with the other cases
test of a metallic helicopter subfloor structure onto water was conducted and reported in
[19].
The prohibitively high cost of conducting such impact tests calls for the use of a
reliable computational model that could replicate as closely as possible the actual impact
conditions. Several of the test programs described previously have also included the
development of computational models to simulate the tests. The FAA’s several civil
aircraft research and certification programs used analytical methods at multiple levels,
from seat occupant models and simulations to support certification tests, to full-scale
impact tests of commuter-class aircraft, [9]. The level of computational analysis ranges
from the use of an early crash analysis program, KRASH, to simulate crash tests and
investigate airplane accidents, to the use of finite element programs such as MSC.Dytran
and LS-DYNA to simulate a drop test of an ATR42-300 and two drop tests of 3 m long
Boeing 737 fuselage sections, [14,20,21]. Further, for the drop tests conducted by the US
4
Army Research Lab and the European Union crashworthiness program discussed above,
non-linear finite element models were developed in LS-DYNA to simulate the drop tests,
[18]. These validated models can be used to study a variety of impact conditions and a
wide range of design alternatives, thus eliminating the need for repeated testing. Finite
element models have also been developed to analyze the deformation and energy
One common result from the drop tests and analysis performed in the various
studies discussed above is that the subfloor structure is a critical component in absorbing
impact energy during a crash landing. Thus optimizing the energy absorption capability
of the subfloor structure and its components could aid in improving the overall
the use of Deployable Energy Absorbers (DEAs) fitted on an MD-500 helicopter for a
full-scale crash test [22]. The DEAs consist of a composite honeycomb structure,
fabricated using fiber reinforced fabrics, which can be deployed externally to absorb
energy in the event of a crash. While test results showed that DEAs offer a high level of
energy absorption due to the large crushing zone available, their resulting parasitic
as adding floor beams, stanchions and stiffeners, to enable increased energy absorption.
Such structural members should add minimal weight while being able to sustain regular
5
flight load conditions and maximizing energy absorption in the event of a crash. For
small aircraft and rotorcraft, this can prove challenging due to the space constraints in the
Composite materials are now widely used in the aerospace industry and are
considered viable candidates for developing light weight energy absorbing devices due to
their high strength-to-weight ratio and their high specific energy absorption (SEA)
important that the initial peak load be kept low to minimize the G forces while the
sustained crush load is kept high as the damage progresses to maximize the energy
members are shown in Figure 1-1. An unmodified member would typically crush with a
high initial peak load, resulting in a catastrophic failure (e.g., local wall buckling),
followed by a low sustained crush load as the damage progresses. Since the SEA is
directly proportional to the area under the curve, such a member would yield a low SEA
as well. Thus, in order to prevent catastrophic failure at the beginning of the crushing
process, a failure trigger mechanism must be used. Several different failure trigger
mechanism options have been studied in order to determine the optimal design to
attenuate this initial peak load [23-32]. Failure trigger mechanisms can be classified as
internal failure triggers (e.g., chamfer, steeple, and ply drop-off) and external failure
triggers (e.g., crush-caps and plug initiators), which are added at one end of the
stanchion. These failure trigger mechanisms are capable of significantly reducing the
initial peak load, similar to the “energy absorbing member” curve in Figure 1-1.
6
Figure 1-1: An example of representative load-crosshead displacement curves for a unmodified composite
member compared to one optimized for energy absorption.
In order to maximize the SEA, the sustained crush load must be increased. This is
heavily dependent on the material system used. The most common composite materials
investigated for energy absorption purposes have been graphite/epoxy (Gr/Ep) [23-29,33-
The laminate stacking sequence has also been shown to have a great influence on the
SEA and a limited number of studies have been performed to determine the optimal lay-
up for the laminates [25,28]. Various cross-sectional geometries have been investigated
to determine their effect on the energy absorbing capacity. These include tubes with
sections, hourglass [52] and cone [43,43,52,55] shaped tubes, angle-stiffeners [28,42], C-
channels [28,42], flat plates [28,37,38], sandwich panels [49-51], foam-filled blocks [56],
hat stiffeners [39], and sine-wave members [29,36]. Finally, the type of failure trigger
mechanism used can also have an effect on the SEA. A few studies have specifically
investigated the effect of external failure trigger mechanisms on the SEA of tubular
7
members [31,32]. Hence, an optimal combination of these factors can yield a higher
sustained crush load and SEA, similar to the “energy absorbing member” curve in Figure
1-1.
The high cost of conducting experimental studies, however, presents a need for
ABAQUS and PAM-CRASH are typically used to simulate the behavior of a wide
these codes, they all contain similar libraries of material models, contact definitions and
other modeling options required to simulate the behavior of composite structures. Thus
Due to their complex failure mechanisms, the energy absorption and crushing
process of composite materials are not easily predicted using numerical methods. There
have been several attempts to develop explicit finite element models, with varying
degrees of success, for circular tubes [57-60], square tubes [59-63], C-channels [61,62],
angle-stiffeners [61,62], and hat-stiffeners [67]. Several modeling approaches have been
utilized that can primarily be differentiated based on their techniques of representing the
laminate, the use of different types of material models, and their method for representing
Composite structures can be modeled using either solid or shell elements; it has
been well-established that shell element models require less computation time and are
more widely used to model the axial crushing of composite members. The laminate can
8
be modeled using either a single layer or multiple layers of shell elements. In the ‘single-
layer’ model the laminate is modeled by using a single layer of shell elements, with each
ply being represented by an integration point in the thickness direction. In the ‘multi-
layer’ model, the laminate is modeled as multiple layers of shell elements, each layer may
represent either a single ply or a group of plies, and the layers are tied together using a
tiebreak contact definition or a cohesive zone model. The single-layer approach has the
advantage of requiring a short computation time, but is not capable of replicating the
time and is sensitive to the selection of material models and delamination interface
definitions. The tiebreak contact definition and cohesive zone model allow for the
has been established to correctly account for the energy absorbed due to delamination
during the crushing process. Further, the selection of material models range from
define the different types of material models also vary significantly and typically involve
experimental data. While various material models have been used to simulate the
9
crushing process1;
The previously developed finite element model [20,21] of a 3 m long Boeing 737
fuselage section was used to investigate the effect of the coefficient of friction and the
angle of impact between the fuselage and the surface. The results from this study
1
The experimental work was conducted by two senior design teams, as explain in the Acknowledgement
section of this Dissertation. The data reduction and analysis were conducted under my supervision. All ten
members of the Senior Design teams became co-authors in four of the conference proceedings
publications and presentations and in the two corresponding peer review publications.
10
representative, narrow-body transport aircraft fuselage onto rigid, soil and water terrains.
The dynamic response of the fuselage structure, including the deformation, acceleration-
time histories, and the energy absorption of various components for the B737 fuselage
section and the full-length fuselage model are analyzed and presented. Chapter 2
discusses the parametric studies on the impact event of the 3 m long fuselage section.
Chapter 3 presents the results of the impact event of the full-length fuselage at the three
crashworthiness of the B737 fuselage, a simple example using spring elements to model
the stanchions in the subfloor structure of the B737 fuselage section is presented in
Chapter 4.
different cross-sections and different failure trigger mechanisms were investigated for use
along with material to manufacture new specimens were provided by The Boeing Co.,
Ridley Park, PA. These specimens were used to investigate the various types of failure
caps). The most effective option was determined by the lowest initial peak load and
highest SEA achieved. New cylindrical specimens were fabricated and were tested with
the previously determined optimal trigger mechanism. The experimental procedure and
This study was extended to include three different cross-sectional geometries that
are more prevalent in the aerospace industry: C-channels, hat-stiffeners, and angle-
stiffeners. Two failure trigger mechanisms were investigated, chamfered-ends and steeple
triggers, and the most effective option was determined. The experimental procedures and
Finite element models were developed to simulate the experimental results: the
load-displacement data, SEA and failure process. Two different modeling approaches,
parameter values used in defining the contact/penetration behavior and material failure.
determining the most effective laminate configuration, element size and formulation,
contact definitions, time step control, delamination interface, and material model. The
is used to simulate the crushing behavior of the specimens tested in Chapter 5 and 6,
Chapter 10, along with key conclusions, scientific contributions, and recommendation for
future work.
This dissertation is presented in a “journal article style” format, where the main
includes its own abstract and introduction sections, and may reiterate some information
12
presented in other chapters. These chapters have been submitted for publication to
2.1. ABSTRACT
A finite element model developed previously for simulating the drop test of a 10
feet long Boeing 737 fuselage section was used in this study to evaluate the effect of the
friction coefficient between the fuselage and the ground and of the aircraft’s angle of
impact parameters represent different terrains (e.g., rigid, soil, grass, water surfaces) and
aircraft crash landing scenarios, respectively. The dynamic responses of the fuselage
structure under four coefficients of friction and six impact angles were simulated. The
on the fuselage, and the energy dissipation of key structural components were studied.
Results indicate that both the friction between the fuselage and the impact platform and
the angle of impact markedly affect the deformation of the entire structure, the
acceleration- time histories, and the energy dissipation by the frame and under-floor
luggage. In all cases, the frames and the luggage absorbed most of the impact energy.
2.2. INTRODUCTION
Over the past decade, several aircraft crashes have occurred in which narrow and
wide-body airplanes were forced to perform emergency landings either during takeoff or
2
Siromani, D., Awerbuch, J. and Tan, T-M., “Crashworthiness Analysis of a Boeing 737 Fuselage Section:
A Parametric Study on the Effects of Angle of Impact and Friction”, Submitted to: International Journal
of Crashworthiness.
14
landing. Some of these involved the collapse of the landing gear, causing the aircraft
fuselage to impact the ground resulting in severe, but survivable damage [2-6]. In each of
these examples, most of the occupants survived the crash landings that caused severe
damage to the aircraft. These examples illustrate the importance of understanding the
dynamic response of an aircraft structure to various impact conditions and its capacity to
absorb impact energy to protect its occupants. The type of impact terrain and the angle of
impact play a significant role in the amount of damage sustained by the aircraft, and
consequently, the energy dissipated by the structure and the impact forces transmitted to
the passengers. These five incidents are prime examples of the importance of aviation
opposed to aviation safety, which deals with accident prevention. During a crash, the
aircraft structure must be able to absorb the kinetic energy and limit the forces and
decelerations that are transmitted to the occupants, through the seat and restraint system,
to tolerable levels. The airframe must survive with minimal collapse, to the extent that a
livable volume is provided to the occupants through the crash sequence. Crashworthiness
studies aim at minimizing the likelihood of injuries or fatalities and reduce the amount of
structural damage to the airframe and payload in the event of a severe, but survivable,
crash [7]. The larger the energy absorbed by the structure the lower the propensity of
injuries or fatalities.
While full-scale crash tests of vehicles are routinely being conducted in the
expensive and extremely time consuming. Also, it is not feasible to perform repeated
15
tests to study various impact scenarios and design modifications. Further, while crash
tests provide voluminous and valuable data, it is not possible to study the effect and
separately. The available procedures to quantify the energy dissipated by the structure
and its individual components are quite limited. Hence, the use of analytical methods,
coupled with crash tests, is necessary to address these issues. Experimental data are
essential to validate the simulation results and understand the strengths and limitations of
the model. A validated model can then be used to study a variety of impact conditions
and a wide range of design alternatives, thus eliminating the need for repeated testing.
In the latter half of the 20th century, there were various research programs in
aircraft crashworthiness conducted by the U.S. Department of Defense [8,9], NASA [8]
and the FAA [9]. A summary of the impact tests performed at NASA Langley’s Impact
Dynamics Research Facility (IDRF) from 1975 until 2003 is provided in [8]. These tests
ranged from a full-scale crash test of a CH-47 Chinook helicopter in 1975 to an external
fuel qualification test of a UH-60 Black Hawk helicopter in 1999. The IDRF facility had
the ability to perform full-scale crash tests of light aircraft and rotorcraft for a wide range
of combined forward and vertical velocity conditions. For example, a crash test was
performed of a Sikorsky prototype helicopter in 1999 with a forward velocity of 9.6 m/s
A summary review of the FAA’s several civil aircraft research and certification
programs using analytical methods is provided in [9]. These research programs range
from seat occupant models and simulations to support certification tests, to full-scale
16
transport aircraft (Boeing 737) [13,14]. The review also describes the use of an early
crash analysis program, KRASH, to simulate crash tests and investigate airplane
accidents. The use of finite element programs such as MSC.Dytran and LS-DYNA to
simulate a drop test of an ATR42-300 and two drop tests of 3 m (10 ft) long Boeing 737
A vertical drop test of a B737 fuselage section was conducted in November 2000
at the FAA William J. Hughes Technical Center [13,20]. The 10ft long fuselage section
was outfitted with seats, mannequins, anthropomorphic test dummies, overhead stowage
bins and under-floor luggage. It was instrumented with strain gages, accelerometer, and
tracks), frames, and overhead stowage bins, as well as the overall structural deformation.
The primary goal of this drop test was to characterize the dynamic response of the two
developed using LS-DYNA, based on hand measured dimensions of the actual test
article, to simulate the impact event and the consequent dynamic structural response of
the fuselage [20]. Results of the simulation compared very well with the experimental
histories. Subsequently, the model was further refined, based on actual design dimensions
In this study, the refined model was used to perform a detailed parametric
analysis. The parameters studied are the friction coefficients between the fuselage and the
impact surface and the angle of impact. For each case, deformation characteristics of the
17
fuselage and its key components, the acceleration-time histories (at various locations in
the fuselage), and the energy dissipation of key structural components, were computed
and compared. The purpose of the friction parametric study was to investigate how the
fuselage will react to various impact terrains, which is necessary for further expanding
the model to more realistic impact conditions. The impact angle parametric study
provided an insight into the structural response of the aircraft to different emergency
As mentioned earlier, in November, 2000 a drop test of a 3 m (10 ft) long Boeing
737 fuselage section was conducted at the FAA William J. Hughes Technical Center
[13,20]. The section was dropped from a 4.3 m (14 ft) height (Figure 2-1), resulting in a
9.14 m/s (30 ft/s) initial impact velocity, which was chosen to represent a severe but
survivable impact.
The test article included seven frames that were spaced 0.5 m (20 in) apart. The
location of each frame on the actual fuselage is identified by a Fuselage Station (FS)
number, which is the distance, in inches, from a reference point located in the front of the
aircraft. The frames of the test article were located from FS 380 to FS 500. Note that the
left- and right-hand sides of the aircraft are with respect to the pilot, when looking at the
floor beam was added at each end of the fuselage section at FS 380 and FS 500. The
added under-floor beam at FS 380 can be seen in Figure 2-2(a). The under-floor
compartment was filled with actual passenger luggage, Figure 2-2(a). The test article also
18
included a cargo door; the forward edge of the door was in line with FS 420, while the aft
edge of the frame ended between FS 460 and FS 480. The frames of the door were
reinforced with short beams, which were connected to the nearest forward and aft
fuselage frames. The fuselage frames, connecting to the upper edge of the doorframe,
Figure 2-1: Vertical drop test of a Boeing 737 fuselage section, conducted at the FAA WJH Technical
Center [13,20]
Three rows of triple-occupant seats were installed on each side of the fuselage
section. The seats were attached to four longitudinal seat tracks, two on each side, and
19
were occupied by twelve mannequins and six anthropomorphic test dummies (ATD’s),
Two different types of overhead stowage bin were installed on the fuselage
section, one on each side, Figure 2-2(a). The details of the bins and their connections to
the fuselage structure are described in the following section. These overhead stowage
bins were loaded with wooden blocks to simulate luggage. The fuselage, its structural
The fuselage and its components were instrumented with numerous strain gauges
and accelerometers mounted strategically at various locations of the fuselage and its
components, as described in detail in [13,20]. The recorded data were used to correlate
with the simulation results at corresponding locations. Figure 2-2(b) shows the front view
of the final deformed state of the fuselage section after the drop test. The entire impact
(a) (b)
Figure 2-2: Boeing 737 fuselage section (a) before lifting the test article to conduct the drop test, and (b)
after impact [13,20]
20
The finite element model, developed in [20] to simulate the drop test of the 2000
drop test of the Boeing 737 fuselage section [13], was based on hand-measurements of
the test article. This model has since been updated with accurate geometry and material
properties [21]. The updated model was used in this study. Its key features are
summarized as follows:
The key components of the model include the fuselage skin, frames, floor, under-
floor beams, luggage, overhead stowage bins, camera mounts, and the cargo door. The
test dummies and mannequins were not modeled, but their masses were distributed on the
seat surfaces. The model of the test section is shown in Figure 2-3.
Figure 2-3: Finite element model of the 10-ft long Boeing 737 fuselage section [13,20,21]
As mentioned earlier, the test article included a cargo door; with reinforcements
to connect the doorframe to the surrounding fuselage frames, Figure 2-4. It will be shown
later that these reinforcements, which provided an added stiffness on the right-hand side
21
of the fuselage, had a major effect on the overall dynamic response of the entire structure.
The cargo door made the fuselage asymmetric, causing the fuselage to deform differently
on the left-hand side than on the right-hand side, and differently in the forward than in the
aft.
Figure 2-4: Detailed finite element model of the frame and cargo door of the Boeing 737 fuselage section
[13,20,21]
One of the primary goals of the drop test was to determine the responses of the
two different types of overhead stowage bin to severe but survivable impact conditions
[13,20]. The primary difference between the two stowage bins was in their support
systems. Figure 2-5 shows the Heath Tecna bin, installed on the right-hand side, and the
brackets on the side and vertical struts attached to longitudinal channels on top. Figure
2-6 shows the Hitco bin, installed on the left-hand side, and the corresponding model. It
was connected to fuselage frames by vertical and horizontal links to the side and tie rods
Figure 2-5: Heath-Tecna bin in test article (left) and its finite element model (right) [13,20]
Figure 2-6: Hitco bin in test article (left) and its finite element model (right) [13,20]
The fuselage section, which included the skin, frames, floor, floor beams, cargo
door, and camera mounts, was modeled using shell elements. The seat frames were
modeled with beam elements, and seat surfaces with shell elements. All supports for the
23
overhead stowage bins were modeled using beam elements, and the bins themselves were
The material properties used in the finite element model were based on standard
materials used in a Boeing 737 fuselage. The skin panels were assigned Aluminum 2024-
T3 properties and all other structural members were assigned Aluminum 7075-T6
properties. Material model MAT82 in LS-DYNA was used to represent each aluminum
alloy. This material model allows the element to fail in tension only. At the onset of
failure the properties are gradually softened at a preselected rate, gradually reducing the
stresses in the element. The element is deleted once the stresses reduce to zero. In
compression, the elements are allowed to continue to deform according the stress-strain
Figure 2-7: MAT82 material model used for the aluminum alloys [21]
significant: it directly affects the extent of crushing during impact, the intensity of the
24
acceleration pulses transmitted to the test dummies, and the energy dissipated during the
impact event. The luggage was modeled using solid elements with properties of viscous
foam. A parametric study was conducted in [20] and results were compared to available
experimental data to determine the appropriate foam properties (nonlinear stiffness and
viscous damper, to represent the increasing stiffness during crushing and to simulate
The final model consisted of 128,960 nodes, 107,212 shell elements, 13,824 solid
elements, and 3,953 beam elements (compared with 68,007 nodes, 53,407 shell elements,
13,824 solid elements, and 691 beam elements of the model developed in [2]). The total
weight of the model was 4,069 kg (8,970 lbs), as compared to the total weight of 4,023
kg (8,870 lbs) of the test article. This small difference in weight did not have any
The model presented in [20] and modified in [21] was used to perform this
parametric study. The parameters of interest are the coefficient of friction between the
fuselage and surface, and the angle of impact. Four issues are discussed below for each
case: i) the dynamic structural response, ii) the acceleration time-histories at selected
locations, iii) the energy dissipation by key structural components; and iv) the effect of
the luggage on the structural deformation. The case of vertical drop with friction
coefficient of 1.0, investigated in [20], is referred herein as the ‘baseline’ case. All
simulations were performed up to 200 ms after impact, since it was determined that all
i) Fuselage Deformation:
The simulations were performed under the vertical drop condition with friction
coefficient values between the fuselage and the impacted platform of 0.3, 0.5, 0.8, and
1.0 (baseline case). The deformed fuselage for the four cases, at the end of the simulation
(200 ms), is shown in Figure 2-8. On the cargo door side (right-hand side of the aircraft,
from pilot’s position) there is no noticeable sliding between the fuselage and the platform
in all four cases. This is due to the added stiffness provided by the cargo doorframe
reinforcements, limiting the frame deformation on the right-hand side, which is discussed
in detail in [20]. The left-hand side of the fuselage, however, shows a marked effect of
the friction coefficient. As expected, the most sliding occurs for the lowest friction
coefficient of 0.3, Figure 2-8a. The frames tend to slide and buckle toward the cargo
section of the fuselage. This effect is progressively diminished with increasing friction
coefficients of 0.5 and 0.8, Figure 2-8b and Figure 2-8c, respectively, with no sliding
Figure 2-8: Deformation of frames, as viewed from top front side of the model, for four different
coefficients of friction (t = 200ms)
26
A detailed view of the deformation that occurs on the left-hand side of the
fuselage is shown for each of the four cases in Figure 2-9. The figure highlights the
location and degree of frame plastic deformation and buckling. At the lowest friction
coefficient (of 0.3), the buckling of all seven frames occurs at the bottom of the fuselage,
along its left-hand side. This buckling is caused by the frames sliding laterally along the
bottom and toward the center of the fuselage, parallel to the impacted surface. For the
higher friction coefficients (of 0.5 and 0.8), the amount of siding is reduced resulting in
the frames buckling, along the left-hand side, further away from the center of the
fuselage. In all cases, the buckling is more severe in the aft section. This is a result of the
location of the stiffened cargo doorframe, located towards the front to the fuselage
section [20]. For the baseline case, where sliding is restricted, the buckling of the frames
occur primarily along the sidewalls, yielding fracture of all seven frames. In all cases, all
frames in contact with the surface are highly deformed. This plastic deformation extends
Figure 2-9: Detailed comparison of deformation on the lower LHS of the frame (t = 200ms)
27
Accelerometers were installed on the seat tracks and overhead storage bins of the
test article to record the acceleration-time histories [13,20]. Accelerometer elements were
included in the finite element model at the corresponding locations. Figure 2-10 shows
representative acceleration-time histories for the seat track accelerometers at FS 452. The
plots show the simulation results for each friction coefficient, plotted along with the
experimental results for comparison. The simulations results for all four cases, at each
location on the right-hand side of the fuselage, are very similar and agree very well with
the experimental data in terms of magnitude and timing of the acceleration pulses. This is
due to the negligible effect of varying the coefficient of friction on the right-hand side
(cargo door side) of the fuselage, as discussed in the previous section. The results on the
left-hand side show that the accelerations fluctuate more significantly for the lower
friction coefficients. Further, while the friction coefficient of 0.3 yields slightly higher
peak accelerations, no particular trend on the effect of friction is noticed. This is due to
the complex structure, which results in a multifaceted dynamic structural response and
It should be noted that the outer seat tracks are located closer to the frames and
therefore more accurately represent the effect of the impact being transmitted from the
frames to the under-floor beams. The inner seat tracks, closer to the center, are more
influenced by oscillations of the floor beams. For this reason, the outer seat tracks are
subjected to higher accelerations compared with the inner seat tracks, and mainly on the
right-hand side, near the doorframes. The magnitudes of the right-hand side acceleration
pulses are typically higher by 7 g’s to 8 g’s than the acceleration pulses on left-hand side.
28
This is due to the fact that the right-hand side response is influenced by the higher
stiffness of the cargo doorframe reinforcements, as discussed above and in further detail
in [20]. Similar behavior and agreement were recorded at all other seat track locations
(not shown here), however, with different peak accelerations and their timing.
Figure 2-10: Comparison of acceleration-time histories at the four FS 452 Seat Track locations for different
coefficients of friction
The effect of the friction coefficients on the energy dissipation is shown in Figure
2-11. The figure shows the energy dissipated by the entire fuselage, the frames, and the
luggage, for a friction coefficient of 0.3. The results are normalized with respect to the
initial kinetic energy of the simulation, i.e., the energy before the fuselage section
impacts the ground, at 0 ms. In all four cases, the luggage and the frame accounted for
29
most of the energy dissipated, approximately two-thirds of the initial kinetic energy. The
remaining energy was dissipated by the many other structural components such as the
skin, floor, floor beams, overhead bins, and camera mounts, etc. A comparison of the
energy dissipated only by the frames and luggage, for all four friction coefficients is
shown in Figure 2-12. These values were normalized with respect to the final energy
dissipated by the respective components in the baseline case (friction coefficient of 1.0).
It can be seen that varying the friction coefficient affects the energy absorbed by the
frames only marginally, Figure 2-12a. At the lowest friction coefficient the energy
dissipated by the frames, primarily starting at 75ms after impact, is between 5-10% lower
compared with the baseline case. This is due to the fact that for the lower friction
coefficients the left-hand side frames initially slide under the fuselage instead of
plastically deforming. On the other hand, the luggage dissipates higher energy when the
friction coefficient is lower, by nearly 20% compared with the baseline case. This is due
the greater extent of luggage crushing with the lower friction coefficients as a result of
the frames sliding and buckling beneath the fuselage. Finally, the slight fluctuations seen
in Figure 2-11 and Figure 2-12 results from the sequential events occurring during the
impact and crush processes. These details could be analyzed via the simulation; however,
Figure 2-11: Dissipation of kinetic energy by the entire fuselage, the frames, and the luggage during the
impact duration for a friction coefficient of 0.3 between the fuselage and impact surface
Figure 2-12: Comparison of energy dissipated by (a) frames; and (b) luggage, for four different coefficients
of friction
31
To further investigate the effect of the under floor luggage on energy dissipation,
a simulation of the case with friction coefficient of 1.0 with luggage removed was
performed and results are compared with that of the baseline simulation (which included
Figure 2-13, indicates that as a result of removing the luggage the lower frames
on the left-hand side of the fuselage fracture and subsequently impact the under floor
beams (not visible due to the forward beam reinforcement placed beneath the forward
floor beam). The damping effect of the luggage affects also the upper section of the
fuselage including skin and frames. For example, at 200ms, the left-hand side frames
deform plastically, and even buckle, much more severely without the luggage. Further,
removing the luggage increases the effect of the stiffened cargo doorframe on the overall
asymmetric deformation of the entire fuselage. This sequence of events causes the seat
track on the left-hand side to experience a much higher g-load; approximately 9 g’s
higher than that of the baseline case in the left-hand side outer track seat, Figure 2-14.
The right-hand side, on the other hand, experiences approximately the same amount of g-
load as the baseline case, although the arrival time of the pulse is delayed by
approximately 100ms. Similar behavior was recorded at all other seat track locations,
Figure 2-13: Comparison of baseline deformation with and without luggage (coefficient of friction = 1.0)
Figure 2-14: Effect of luggage on the acceleration-time histories at two FS 452 seat track locations
Five different angles of impact were analyzed and compared with the vertical
impact simulation. The five angles studied were 15°, 30°, 45°, 60°, 75°, and 90° from the
horizontal. For all six cases, the vertical component of the velocity was the same as in the
vertical drop test (9.14 m/s (30 ft/s). Thus, the horizontal velocity components, added to
33
the model to simulate the different angles of impact, were 34.1 m/s, 15.8 m/s, 9.14 m/s,
5.3 m/s, and 2.4 m/s, respectively. A friction coefficient of 0.5 was used for all six
i) Fuselage Deformation:
The post impact structural response of the frames was compared for all six angles
of impact. Particular attention was placed on the bottom section of the fuselage, where
contact occurs between the fuselage and platform. The deformed states at the end of the
simulations (200 ms) for the six angles of impact are shown in Figure 2-15. As expected,
for the smaller impact angles the forward section of the fuselage experiences the highest
deformation and most severe frame bucking, Figure 2-16. Also, for the smaller angles,
the deformation of the right-hand side of the frame is less prominent than that recorded
on the left-hand side, showing again the effect of the doorframe discussed earlier [13,20].
Figure 2-15: Side views of the deformed fuselage section for six different angles of impact at 200ms
34
Figure 2-16: Deformation of frames for six angles of impact as viewed from top front side of the model at
200 ms
The acceleration-time histories for the seat track accelerometers at FS 452 were
compared with the test data. The locations of the accelerometer elements on the seat
tracks were shown previously in [13,20]. Acceleration-time histories for each impact
angle are shown in Figure 2-17 for the right- and left-hand sides’ inner and outer seat
tracks. The simulation results for the right-hand side seat tracks, with the stiff cargo
doorframe located underneath, are quite similar for all angles in terms of peak
acceleration and timing of the first peak acceleration. The timing of the subsequent peak
acceleration marginally depends upon the angle of impact. The results for the left-hand
side seat track, however, vary significantly for different angle of impact. The varying
horizontal velocity causes the frames on the left-hand side to crush to a different extent
35
for each angle, as they are now subjected to horizontal forces as well. These horizontal
forces are generated by the friction between the fuselage and the impact surface.
Figure 2-17: Comparison of acceleration-time histories at four FS 452 Seat Track locations for different
angles of impact
The energy dissipation by the entire fuselage, the frames, and the luggage during
the impact event is shown in Figure 2-18 for two representative angles of impact. Similar
results were obtained for all other angles of impact (not shown here). The kinetic energy
and the energy dissipated during the impact event shown the Figure 2-18 have been
normalized with respect to the respective initial kinetic energy. Due to the large different
horizontal velocity components, the amount of initial kinetic energy changes significantly
36
in each case, with 90° impact being the lowest and 15° impact angle the highest, as
expected. As before, the frames and luggage account for most of the energy dissipated for
all angles of impact. All six cases show the same trend in the amount of energy dissipated
by the frames and luggage. Compared to the significant increase in kinetic energy (due to
the large increase in the horizontal component of velocity to yield the desired angle of
impact), the total energy dissipated by the fuselage structure barely changes between the
six angles. The remaining kinetic energy is due to the horizontal velocity component that
is reduced (by 36% and 98% for the 15° and 75° cases, respectively) by the friction
between the fuselage section and the surface by the end of the simulation.
Figure 2-18: Dissipation of kinetic energy by the entire fuselage, the frames, and the luggage during the
impact duration for impact angles of 15° and 90°
A comparison of the energy dissipated by the frames and luggage for all six cases
is shown in Figure 2-19. These plots are normalized with respect to the final energy
dissipated by the respective structural components in the vertical (90°) impact case. The
37
results show that the lower the impact angle the larger the energy dissipated by the
frames and the under-floor luggage (by 24% and 50%, respectively), relative to the
vertical impact. As mentioned earlier, the frames on the left-hand side of the fuselage
tend to crush very differently for the different angles of impact. For lower impact angles,
the left-hand side frames – towards the front of the fuselage – were crushed to a much
greater extent than for the higher impact angles, Figures 18 and 19. This caused the
luggage in those areas to be crushed to a greater extent, thus, increasing the energy
dissipated. A comparison of the final deformation of the luggage for impact angles of 15°
Figure 2-19: Comparison of energy dissipated by (a) frames; and (b) luggage for six different angles of
impact
38
The objective of this study was to use the validated Boeing 737 model, developed
earlier in [20] and updated in [21], to investigate the effects of two key impact
parameters, namely, the friction between the fuselage and the impacted surface and the
angle of impact, on the dynamic response of the fuselage section in terms of the structural
The results show that the coefficient of friction greatly affects the deformation of
the entire structure, primarily in term of the degree of lateral sliding between the bottom
of the fuselage and the impacted platform. As expected, the lower the friction coefficient
the higher the lateral sliding. Consequently, the lower friction yields higher plastic
deceleration affects the occupants the most), are all quite similar and are marginally
affected by the degree of friction. Because of the significant effect of the reinforcement
of the cargo doorframe on the right-hand side of the fuselage, as discussed in detail in
[20], the lateral sliding occurs mostly on the left-hand side. The resulting acceleration-
39
time histories along the left-hand side seat tracks fluctuated more significantly for lower
coefficients of friction, but the peak accelerations varied by no more than 2g’s in all
cases. The peak accelerations on the right-hand side are much higher, particular along the
outer seat tracks (closer to the cargo door), by up to 7 g’s, as was the case in [20],
independent of the friction coefficient. While the friction coefficient of 0.3 yields slightly
higher peak accelerations, no particular trend on the effect of friction is noticed. The
friction coefficient affects the amount of energy dissipated by the luggage only when it is
low enough to enable significant lateral sliding that causes added cargo crushing. In each
case, the luggage and frame together account for most of the energy dissipated.
The angle of impact study ranged from a vertical drop (90°) to a 15° angle of
descent. As with the friction coefficient study, the effect of the cargo doorframe
reinforcements was evident as the seat track acceleration-time histories on the right-hand
side were minimally affected by the change in angle of impact. The higher angles of
impact reduce the peak accelerations by up to 7g’s, primarily along the outer right-hand
side seat tracks. The acceleration-time histories for the left-hand side fluctuated more
significantly due to the varying degree of deformation of the frames on that side caused
by different horizontal component of velocity in each case. The peak accelerations are
much higher on the right-hand side, particular along the outer seat tracks (closer to the
cargo door), by up to 7 g’s, as was the case in [20]. The luggage dissipates more energy
for the lower impact angles due to the greater extent of deformation of the bottom frames
and the subsequent crushing of the luggage. As in the friction study, the luggage and
frame together account for most of the energy dissipated for all impact angles. These
40
results serve as a precursor to study the impact of full-length fuselage at various terrains
3.1. ABSTRACT
transport aircraft was developed to simulate crash-landing on different terrains (i.e., rigid,
soil, and water). To emphasize the effect of terrain on the dynamic structural
deformation, vertical impact of the full-length fuselage was also studied. The model was
constructed by expanding, to the full length, a previously developed 3-m long B737
section model that has been validated by experimental data from a drop test, excluding
the wings and replacing the structural details of the nose cone and tail section with a
simple shell model. Results include deformation patterns of the fuselage structure,
dynamic responses of the fuselage was also investigated for the case of vertical impact.
The results quantified the effect of the different terrains on the structural deformation and
acceleration-time histories.
3.2. INTRODUCTION
Several aircraft crashes have occurred over the past decade in which narrow- and
wide-body airplanes were forced to perform emergency landings. Some of these incidents
involved the collapse of the landing gear, causing the aircraft fuselage to impact the
3
Siromani, D., Awerbuch, J. and Tan, T-M., “Multi-terrain Crashworthiness Simulations of the Fuselage of
a Narrow-body Transport Aircraft”, Submitted to: International Journal of Crashworthiness.
42
ground, resulting in severe but survivable damage [3]. While considerable research has
been performed on testing and analysis of an aircraft impacting hard surfaces [8-
14,20,21], no such studies have been reported on full transport aircraft fuselages
impacting other terrains, such as soil and water. Yet, there have been cases where the
aircraft have crash-landed onto such surfaces [4,5]. Furthermore, it has been reported that
more than 80% of rotorcraft crashes occur on non-rigid terrains and that helicopters
designed for hard surface impact do not perform well when impacting onto water and
soft-soil [15].
The US Army Research Lab has conducted a series of vertical drop tests of
composite helicopter fuselage on concrete [16], soft-soil [17], and water [18]. The
dynamic response of each case was evaluated and compared [18]. For vertical impact
onto a concrete surface, the subfloor structure crushed and absorbed most of the kinetic
energy. For the soft-soil and water impacts tests, the deformation of the impacted media
dissipated a large part of the kinetic energy. The unsupported portion of the skin
underneath the subfloor structure was subjected to large in-plane membrane forces,
causing the skin to either fail in the case of water impact, or deform plastically in the case
of soft-soil impact. For all three cases, non-linear finite element models were developed
to simulate the drop tests. All three models reviewed in [18] compared well with the
experimental data.
a metallic helicopter subfloor structure onto water was conducted and reported in [19]. A
finite element model was also developed to simulate this drop test. The model proved
useful in analyzing the initial stages of impact, when most of the structural deformation
43
and energy absorption occurred. However, there was a poor correlation with experiments
at the final stages of impact. The authors concluded two primary areas for potential
failure is critical to load transmission to other energy absorbing components; and ii)
In an earlier study, a vertical drop test of a 3-meter long Boeing 737 fuselage
section was conducted at the FAA William J. Hughes Technical Center [13,20]. The test
article was outfitted with seats, mannequins, anthropomorphic test dummies, two
overhead stowage bins and under-floor luggage. The primary goal of this drop test was to
characterize the dynamic response of the two overhead stowage bins. The test section was
instrumented with strain gages and accelerometers to record the strain- and acceleration-
time histories at various locations on the floor (along seat tracks), frames, and overhead
stowage bins. High-speed cameras were used to capture the overall structural responses.
developed, using LS-DYNA, to simulate the drop test [20]. The model was constructed
based on the hand-measured dimensions of the actual test article. Simulation results, in
terms of the overall structural responses and the acceleration-time histories, were
compared with the experimental data, showing excellent agreement. Subsequently, the
model was further refined based on actual design dimensions of the various structural
components, including a more realistic connection of the overhead bins to the airframe,
etc. [21].
parametric study [72, Chapter 2] on the effects of the friction coefficients between the
44
impacting surfaces and of the impacting angle of obliquity. For each case, the
fuselage, and the energy dissipation of key structural components were analyzed.
The purpose of the friction parametric study was to investigate the effect of
various friction coefficients, which are associated with different terrains, on the structural
response of the aircraft. The study illustrated how the fuselage reacts to various impact
surfaces, which is necessary for further expanding the model to realistic impact
conditions.
The angle of impact parametric study provided an insight into the structural
response of the aircraft to different angles of impact, which better represents actual
aircraft crash landing scenarios. However, results of this study also revealed that the 3-m
long section could not realistically represent a full-length fuselage, because the former
This model was further expanded to study the impact response of a full-length
fuselage impacting three different terrains, namely: rigid surfaces, soil, and water, under
and to use the model to study the effects of crash landing on different terrains (i.e., rigid
surface, soft soil, and water) on the deformation of the fuselage, acceleration-time
histories along the seat tracks, and energy dissipation of various subcomponents. The
full-length model was to be constructed by expanding the original 3-m long section
45
model (hereafter referred to as the ‘original’ section model) developed in [20]. However,
a preliminary vertical water impact simulation conducted using the 3-m long original
section model showed that it required approximately 25 hours of CPU time using 64
parallel processors. This indicates that a full-length fuselage, having the same level of
modeling details as the original section model, would not be a viable option
structural details, was developed. The new model is henceforth referred to as the
‘modified’ section model. It is noted that both the original and modified section models
used type 16 fully integrated shell elements in LS-DYNA. For the material modeling, LS-
DYNA material type 82 (MAT 82) with a linear strain softening behavior after the failure
initiation was used to model aluminum alloys 2024-T3 and 7075-T6, which were used for
The criterion used for building the modified section model was to reduce the
number of elements while keeping the deformation of the subfloor structure and the seat
track acceleration-time histories similar to those of the original section model. The
original section model consisted of approximately 126,000 elements, which included the
various subcomponents, such as two overhead stowage bins, passenger seats, stringer
clips, rivets, and camera mounts. Detailed modeling of these subcomponents was deemed
unnecessary because of their marginal effect on the dynamic response of the fuselage
structure and acceleration-time histories along the seat tracks. Yet, the masses of the
excluded subcomponents were assigned to the nodes to which they were attached. For the
remaining fuselage structure, a mesh sensitivity study was conducted with the aim of
keeping the overall impact response and the acceleration-time histories (at strategically
selected locations throughout the fuselage section) similar to those of the original model.
For example, the mesh for the upper frames and skin did not have to be as detailed since
those sections experienced mostly elastic deformation throughout the impact event and
had little bearing on the overall response of the structure [20]. On the other hand, the
lower frames and skin experienced significant plastic deformation and failures (i.e.,
buckling and crushing), thus required a relatively fine mesh to achieve the desired
accuracy. The resulting number of nodes, elements and total mass of the modified section
model is compared with the original section model in Table 3-1 (test article total mass
was 4,023 kg). Both the original and the modified section models are shown in Figure
3-2. It is noted that the simulations using the modified section model required
47
approximately one-fifth of the computation time needed for the original section model
(i.e., 20 min. vs. 100 min. on a 64 CPU cluster for rigid surface impact).
Table 3-1: A comparison of the finite element modeling details between the original and modified
section models
Nodes Shell Elements Solid Elements Beam Elements Total Mass
(a) B737 Original Section Model (b) B737 Modified Section Model
Figure 3-2: The original and modified B737 section models. The masses of the subcomponents that were
excluded from the original model were assigned to the nodes to which the subcomponents are
connected.
Two vertical impact simulations were conducted to validate the reliability of the
modified section model: a) a rigid surface impact and b) a water impact. Simulation
results, in terms of deformation and acceleration-time histories, were compared with the
selected stages of impact on a rigid surface is shown in Figure 3-3. It should be recalled
that the original section model has been validated against the actual drop test data in
terms of the structural deformation and the acceleration-time and force-time histories of
the fuselage structure and its individual components [20]. The comparison shown in
Figure 3-3 indicates that the modified section model was able to capture the key
structural deformation such as the tilting of the fuselage to the left, and the excessive
failure of the lower left-hand side frames. The particular deformation characteristics
resulted primarily from the presence of the stiff cargo doorframes on the right-hand side,
as discussed in detail in [20]. Note that the left- and right-hand sides of the aircraft are
with respect to the pilot’s position, when looking at the front of the fuselage section.
(a) Original Section
Model
(b) Modified Section
Model
t = 32 ms t = 64 ms t = 96 ms t = 128 ms
Figure 3-3: A comparison of the deformation of the original and modified section models at selected
stages of a rigid-surface impact
49
seat tracks at Frame Station 418 (FS 418, i.e., 418 inches from a reference point located
in the front of the aircraft is shown in Figure 3-4. A good agreement has been established
between the results from the two models, as well as with the experimental data reported
in detail in [20,21]. It should be noted that in the drop test the accelerometers were
installed along the seat tracks and overhead storage bins of the test article to record the
acceleration-time histories [13]. In this study, accelerometer elements were placed only at
Figure 3-4: A comparison of seat track acceleration-time histories at FS 418 for rigid surface impact.
50
The setup of the original section model for water impact simulations is shown in
Figure 3-5. A layer of air, situated on top of the water, was included so that the wave
propagation in the water during the impact event could be visualized. There are two
different methods that have been used in modeling water, namely, the Arbitrary
method. Both methods have been shown to work well, but the SPH method is more
computationally intensive than the ALE method [15]. Hence, for this study, the ALE
Figure 3-5: Setup of the original section model for water impact, including a layer of air on top of the
water surface
which have a number of parameters that need to be determined experimentally. For this
study the Gruneisen equation-of-state was selected due to its successful use in similar
water impact studies [15]. The interaction between the fluid and the structure was
accomplished by using a penalty method. For the fluid-structure interaction to occur, the
51
Lagrangian mesh (for the fuselage structure) must spatially overlap with the Eulerian
mesh (for water and air). This spatial intersection is searched for continuously to
determine when the interaction between the water and structure will occur [73].
initial impact is shown in Figure 3-6. In both cases, the lower frames deformed and
buckled into the luggage in the subfloor region of the fuselage section. However, the
location of the buckling was different in the two models. In the original section model the
buckling occurred toward the center of the fuselage, whereas in the modified model it
As expected, the structural damage that occurred in the regions, modeled by the
finer mesh, tended to progress more gradually compared to the regions with modeled by a
coarser mesh; the latter tends to fail rapidly, causing abrupt transmission of the load to
the surrounding elements. Such a dynamic response is expected to lead to a more severe
failure. An example of the effect of mesh refinement is shown in Figure 3-7. The
modified section model, which had a coarser overall mesh compared to the original
model, was initially developed and validated with the original model for rigid surface
impact. The mesh in the modified section model was refined at selective regions to
achieve a better agreement with the deformation of the original section model,
particularly at the lower frames where deformation was most severe. When the same
modified section model was used for water impact, damage would initiate at regions in
the lower frames where a coarser mesh was used, resulting in deformation that was not at
selected locations along the seat track were very similar, Figure 3-8. The difference in
magnitude of the peak acceleration between the two models, other than that at the outer-
left location, was less than 2 g’s. The timing of the peak acceleration of the modified
section model was behind that of the original section model by approximately 10-20 ms
as the result of the delayed occurrence of buckling at the lower frames. It is noted that the
water impact simulations using the modified section model required approximately one-
53
tenth of the computation time needed for the original section model. (i.e., 2.5 hrs. vs. 25
Figure 3-8: A comparison of seat track acceleration-time histories at FS 418 for water impact
The above comparisons indicate that the simulation of the modified section model
demonstrated a high level of agreement for the rigid surface impact (Figure 3-4) and an
acceptable level of agreement for the water impact (Figure 3-8). In other words, the
computationally efficient model produces similar results to the original section model;
54
accordingly, this modified section model was adopted to construct the full length fuselage
model.
approximately 30 m long (i.e., B737-200) was used for this study. The primary section of
the fuselage, excluding the nose cone and the tail cone, was approximately 16 m long.
This primary section was assumed to be of constant cross-section and of the same
structural details as the original section model, which was constructed by replicating the
modified 3 m long section model. The lengths of the nose and tail cones were assumed to
the nose cone, tail cone, and wings are unavailable in the open literature and considered
mostly as manufacturer propriety. Obviously, these three major components, and the
accompanying complex substructures, greatly affect the impact response of the full-
length fuselage. Once the details on the geometry, material, and weight of each
component are made available, the current model could integrate the complete aircraft
structure. However, since the purpose of this study is to evaluate the effect of different
histories) of an aircraft fuselage, the use of the simplified full-length model was deemed
appropriate. Since in this study only results from the primary fuselage section were of
interest, for simplicity, the nose cone and tail section are represented by two shell element
models, in the general shapes of the B737 (attached to both ends of the primary section),
55
containing no substructures or tail stabilizers, Figure 3-9. It is noted that the lack of
substructure support for the nose cone would cause it to undergo excessive deformation
in certain impact scenarios. The two ends of the primary fuselage section were designated
as FS 200 (200 inches from the nose) and FS 820, respectively. A cargo door, similar to
that in the 3 m long section model, was placed from FS 340 to FS 420 in the full-length
model. In addition, the landing gear was not included in the model either as the focus of
this study was to investigate the crashworthiness of the fuselage structure under
emergency landing conditions in which the landing gears could be out of order or have
collapsed [3]. As with the nose cone and tail section, the effect of the landing gear’s
It was noted in the previous study on the 3-m long B737 fuselage section [20,72,
Chapter 2] that the luggage played an important role in energy dissipation during the
impact event. In order to investigate the effect of under-floor luggage on the full-length
fuselage, the luggage model developed in the 3 m section model [20] was also replicated
to fill the full-length fuselage model from FS 200 to FS 820, excluding the nose and tail
cones, Figure 3-9. The total mass of the full-length fuselage model with the simplified
nose and tail cones, and without wings, landing gears and luggage, was 14,758 kg. With
the luggage included, the total mass of the full-length model was 22,733 kg. The mass of
the nose cone and tail section was 245 kg and 657 kg, respectively.
Accelerometer elements were placed along the inner and outer seat tracks on both
sides of the passenger floor at a number of fuselage stations throughout the entire
fuselage to record the dynamic responses during an impact event. For the purpose of this
paper, the focus is placed on those accelerometers located at FS 300, FS 480 and FS 700,
56
denoted by forward (FWD), middle (MID), and aft (AFT), respectively, Figure 3-9. Of
the four accelerometers at each fuselage station location, only the results from the one
that had the highest peak acceleration are presented and compared.
Cargo Door
Figure 3-9: The 30-m long full-length model with a cargo door, under-floor luggage, and simple nose and
tail cones. FWD (FS 300), MID (FS 480), and AFT (FS 700) indicate the locations of
accelerometer elements.
Crash landings on three different types of terrains, namely, rigid surfaces (e.g., a
concrete runway), soil, and water, were simulated. For all simulations, a 110 knot (203.7
km/hr, or 56.6 m/s) approach speed was selected which is at the lower end of landing
speeds of typical transport aircraft. The same vertical component of velocity (9.14 m/s)
as the vertical drop test of the 3 m section was used [20]. This resulted in an initial
horizontal velocity component of 56 m/s and a 9.3° approach angle. It should be noted
that the 9.3° approach angle is much larger than the typical approach angle of 3° for a
normal landing [74]. In addition, the fuselage was given a 3° pitch angle, but with no yaw
Since the landing gear was not included in the model, all landing scenarios
studied herein were belly landings or gear-up landings. Further, the luggage, which has
been shown to absorb a substantial amount of impact energy [20,72, Chapter 2], was not
included in the crash landing simulations. Therefore, the results presented herein
Figure 3-10: Crash landing scenario used in this study: The fuselage has a 110 knots approach speed, a
9.27° approach angle, and a 3° pitch angle.
The effect of the luggage stored in the under-floor luggage bay was excluded
from the crash landing simulation. To demonstrate the effect of the luggage, a separate
study was conducted on vertical impact simulations, with and without under-floor
luggage, onto a rigid surface and water. The effect of the luggage was quantified by
comparing the deformation of the frames, the acceleration-time histories along the seat
tracks, and the energy dissipation of the various subcomponents of the fuselage with and
without luggage.
For the rigid surface landing, a friction coefficient of 0.5 between the fuselage and
the rigid surface was used [68]. The detailed discussion on the effect of friction
LS-DYNA offers a number of different material models for soil. The general
viscous foam material models, which allow the user the flexibility of using experimental
58
load-displacement curves, have been successfully used to model soil [17]. In addition,
there are numerous material models for specific ‘geomaterials’. One of the more common
models used to represent soil is *Mat_Soil_And_Foam, which was selected for this study.
As part of an effort to model Crew Exploration Vehicles subjected to soil impact [76],
NASA Langley Research Center (LaRC) required soil property inputs for their use of LS-
samples from NASA LaRC and Kennedy Space Center and conducted a series of tests to
derive the required parameters that best represent the material and loading conditions for
different types of soil such as unwashed sand, low density dry sand, high density in-situ
moisture sand, and high density flooded sand [77]. For this study, the unwashed soil
properties were used for the material model since the other three types are typically only
found in coastal regions. The soil was modeled using solid elements with an eroding
contact definition between the fuselage and the soil. Any severely distorted elements
would be assumed to have failed and be removed from the simulation. A friction
coefficient of 0.3 between the fuselage and the soil was used. For the water landing
simulations, the ALE method described earlier was used to model the fuselage structure
Figure 3-11 shows the deformation of the full-length fuselage model, at three
selected time steps, as a result of crash landing onto a rigid surface, i.e., (a) at t = 0 ms
when the aft end of the primary fuselage section contacted the ground, (b) at t = 80 ms
when the forward end of the primary fuselage section contacted the ground, and (c) at t =
59
200 ms when the simulation terminated. Due to the lack of substructure support, the
lower portion of the nose cone and tail section underwent excessive deformation
immediately after the forward section of the primary fuselage contacted the ground at t =
80 ms and eventually separated from the primary fuselage section, Figure 3-11(c).
Similar results for crash landings on soil and water are shown in Figure 3-12 and Figure
3-13, respectively. Unlike the case of rigid surface landing in which no deformation
occurred to the landing surface, both the soil and water underwent significant
below.
The overall deformation and contours of effective plastic strain of the frames, at
selected time steps, resulted from crash landings on three types of terrains are shown in
Figure 3-14 in a perspective view from the front of the fuselage. The occurrences of the
i) At 12 ms: Initial plastic deformation occurred at the lower frames of the aft
section, which came in contact with the terrains the earliest, due to the 3° pitch
angle.
contact with the terrains. Plastic hinges formed first in the aft sections followed by
3-15(a). As seen in the Figure, the extent and magnitude of the plastic strain
varies form frame-to-frame, depending upon the time of impact. The overall
appearance of the effective plastic strain field is quite similar in all three terrains;
yet, the rigid surface impact yields the largest plastic strain and the water impact
the least.
iii) At 92 ms: Plastic deformation extended throughout the lower frames of the entire
fuselage length. The bottoms of the aft frames have flattened out and failure (i.e.,
61
element deletion) has occurred. Note the similarity in the deformation between all
iv) At 152 ms: The bottoms of all lower frames have flattened out and most frames -
from the mid to the aft of the fuselage - have come in contact with the passenger
3-15(b). All frames in the aft section have undergone severe plastic deformation,
v) At 200 ms: All lower frames have been completely crushed and have come in
Figure 3-15(c). Frame buckling occurs both in the middle and side of the fuselage.
The extent and location of the buckling depends, to a great extent, on the friction
at this stage at any of the three terrains in any of the frames throughout the
fuselage length. The fuselage continued to move forward at the termination of the
simulations.
It is noted that the overall deformation of the frames caused by crash landings on
all three types of terrains was very similar at the end of the simulations. This can be
attributed to the fact that during the crash landings the fuselage moves forward
a) 12ms
b) 52ms
c) 92ms
d) 152ms
e) 200ms
Figure 3-14: The overall deformation and contours of effective plastic strain of the frames at selected time
steps resulted from crash landings on three types of terrains.
63
The acceleration-time histories along the seat tracks on the passenger floor at
forward, middle and aft locations of the fuselage resulted from crash landings onto a rigid
surface are shown in Figure 3-16. The magnitudes of the peak acceleration pulses at the
three locations are significantly different due to the different times and velocity at which
the three sections of the fuselage impacted the surface. Due to the 3 pitch angle the aft
section impacted the surface first, resulting in an earlier peak pulse (at 125 ms), followed
by the middle (at 150 ms), and then in the forward regions (at 190 ms). Furthermore, the
3 pitch angle also resulted in the forward location being approximately 0.8 m higher
than the aft location at the time of initial impact, causing the forward section to accelerate
downwards and impact the surface at a velocity of which the vertical component was
20% higher than that when the aft section impacted the surface. Consequently, the
magnitudes of the acceleration pulses were progressively higher when proceeding from
aft to forward locations. The highest acceleration, of 40 g’s, occurred at the forward
location at approximately 190 ms after the initial impact. Beyond 200 ms, the vertical
velocity rapidly subsided, but the fuselage continues to move forward with a horizontal
64
velocity of 52.4 m/s, i.e., only 7.5% lower than the initial velocity. Thus, only
approximately 15% of the initial kinetic energy has dissipated as discussed in further
detail below.
Figure 3-16: Seat tracks acceleration-time histories resulting from a crash landing of a full-length fuselage
on a rigid surface. The magnitude of the peak acceleration pulses progressively increased
from the AFT to FWD locations.
The simulation yielded similar results for crash landing on soil and water terrains.
For comparison, the acceleration-time histories at aft and forward locations, for all three
terrains are shown in Figure 3-17. Results at the aft locations, Figure 3-17(a), show that
the magnitudes of the first peak accelerations and their arrival times varied slightly for
the three terrains: 9.5 g’s at 65 ms for rigid surface landing, 6 g’s at 53 ms for soil surface
landing, and 8 g’s at 60 ms for water surface landing,. The subsequent peak accelerations
and their arrival times varied further, with the maximum peak accelerations for all three
The acceleration-time histories at the forward location for all three terrains were
nearly identical, Figure 3-17(b), with initial peak accelerations of 10 g’s, 13 g’s, and 6
65
g’s, for the rigid, soil, and water terrains, respectively, all occurring nearly at the same
time (110 ms). The peak acceleration for all three terrains is approximately 40 g’s, all
Figure 3-17: Seat track acceleration-time histories comparison at aft and forward locations resulting from a
crash landing of a full-length fuselage on rigid, soil and water terrains.
The normalized energy dissipated during the first 200 ms of the crash landing on a
rigid surface is shown in Figure 3-18. The kinetic energy was normalized with respect to
the initial kinetic energy (at 0 ms), and the energy dissipated was normalized with respect
to the total energy dissipated at 200 ms. Due to the large horizontal component of the
approach speed, only 15.4% of the initial kinetic energy was dissipated at 200 ms as the
fuselage was still moving forward at a relatively high speed. Approximately 19% of the
energy dissipated, or 2.9% of the initial kinetic energy, was absorbed by the fuselage
structure, with the frames and the skin absorbing approximately 11.5% and 7.5%,
respectively. The landing surface, being rigid, did not absorb any energy through
66
deformation; however, friction between the fuselage and the rigid surface accounted for
The energies dissipated during the first 200 ms of the crash landings on soil and
water are shown in Figure 3-19. Due to the lower friction, the initial kinetic energy
dissipated by the end of the simulation (200 ms) in each case was less than the rigid
surface case, with soil and water dissipating approximately 11% and 6% of the initial
kinetic energy, respectively. In the soil impact case, approximately 30% of the dissipated
energy, or 3.3% of the initial kinetic energy, was absorbed by the fuselage structure, as
compared to 70% and 4.2%, respectively, in the water impact case. It is noted that the soil
and water absorbed approximately 4.3% and 1.75%, respectively, of the dissipated
energy.
Figure 3-18: Energy dissipation for a crash landing of a full-length fuselage on a rigid surface. Frames and
skins dissipated most of the energy.
67
Figure 3-19: Energy dissipations for crash landing on soil and water surfaces. Frames and skin dissipated
most of the energy.
Figure 3-20 compares the energy dissipated by the frames and the skin for all
three terrains, normalized with respect to the energy dissipated in the rigid surface case at
200 ms. The rigid and soil terrains result in similar energy dissipation by the frames and
skin, with the soil being approximately 8% lower. In the water impact case, the energy
dissipated by the frames and skin was approximately 20% and 33% higher than the rigid
impact case, respectively. This indicates that the frame and skin in the water impact case
underwent higher plastic deformation than the rigid and soil cases. Apparently, the
continuous contact pressure applied by water over the entire subfloor structure (frames
and skin) has a greater effect on the overall deformation than the rigid and soil surfaces,
whose primary points of contact for load transfer are the frames.
68
Figure 3-20: Energy dissipation by (a) frames, and (b) skin for crash landing on rigid, soil and water
surfaces.
It has been shown that only a small amount of the initial kinetic energy dissipates
in a crash landing event, as the fuselage continued moving forward with a relatively high
velocity. In order to gain a better understanding on how the energy dissipates by the
various subcomponents, simulations were conducted for the case of vertical impacts of
the full-length fuselage, on all three terrains. For all three simulations, the fuselage was
dropped onto the surfaces with no pitch angle and with the same initial vertical velocity
Figure 3-21 shows the effective plastic strain of the fuselage without luggage at
152 ms after the initial impact onto the three terrains. The plastic deformation in the
frames and the skin has been completed at this stage and the upper portion of the frames
69
began to reverberate elastically. The reverberation was not as noticeable in the soil and
water impact cases because these terrains deformed under the impact loading enabling the
fuselage to further sink into these two terrains. The asymmetrical deformation, caused by
the stiff cargo doorframe [20,72, Chapter 2], was most prominent in the water impact
case where the severe deformation and failure occurs along the more compliant lower
left-hand side frames. No specific conclusions could be drawn regarding the behavior of
the nose cone and the tails sections because of the lack of sufficient structural details.
Figure 3-21: The overall deformation and contours of effective plastic strain of the frames at t = 152 ms
resulted vertical impact on three types of terrains
The acceleration-time histories at the forward, middle and aft locations on the
passenger floor resulted from vertical impact of the full-length fuselage, without luggage,
onto a rigid surface are shown in Figure 3-22. The magnitudes of the peak acceleration
pulses are quite similar at the forward and aft locations, with the maximum acceleration
being approximately 19 g’s, occurring at 190 ms at the aft location. Furthermore, the
peak acceleration at the aft location occurred approximately 40 ms later than that at the
forward location. It should be noted that the tail section, which is approximately 168%
heavier than the nose cone caused the aft frames to crush further, thus delaying the arrival
70
of the peak acceleration pulse. The addition of the nose and tail cones also resulted in
higher peak acceleration pulses at the forward and aft locations as compared with the
middle location.
Similar results were obtained for vertical impact on soil and water surfaces. For
soil impact, the peak acceleration at the aft section was approximately 22 g’s (at at 190
ms), whereas for water impact it was approximately 13 g’s (at 100 ms), Figure 3-23. As
expected, the acceleration pulses in water impact were mostly lower than in rigid and soil
impacts due to the large deformation of the water. The acceleration pulses in soil impact
were higher than in rigid and water impacts due to the different coefficient of friction
between the fuselage and the impact terrain used in this study (0.3 for soil and 0.5 for
rigid surface). A similar observation, on the effect of friction between the fuselage and
impact terrain, was made in a previous study [72, Chapter 2]. It should be noted that the
above discussion is merely to offer a preliminary glance on the significance of nose cone
tail section and on the overall impact response of the full-length fuselage. It should be
expected that the effect of the actual nose cone and tail section would be quite different.
Figure 3-22: Seat tracks acceleration-time histories for full-length fuselage vertical impact on a rigid
surface.
71
Figure 3-23: Seat track acceleration-time history comparison at aft and forward locations resulting from
vertical impact of a full-length fuselage on rigid, soil, and water terrains.
The energy dissipated during the first 200 ms of the vertical impact on a rigid
surface is shown in Figure 3-24. All energies are normalized with respect to the initial
kinetic energy (at 0 ms). More than 90% of the initial kinetic energy was dissipated
within the first 150 ms after initial impact. Of the total dissipated energy, more than 90%
was absorbed by the frames and skin to form plastic deformation. By 150 ms, all plastic
deformation was completed and the structure underwent an elastic reverberation for the
The energy dissipation plots for soil and water impacts are shown in Figure 3-25.
The plastic deformation in the frames and the skin dissipate most of the energy. For the
water impact case, a negligible amount of energy was dissipated by the water, whereas
for the soil impact case the soil deformed and absorbed nearly 8% of the dissipated
energy.
72
Figure 3-24: Energy dissipation for vertical impact on rigid surface. Frames and skin dissipated most of the
energies.
Figure 3-25: Energy dissipation for vertical impact on soil and water. For both cases the frames and skin
dissipated most of the energies.
Figure 3-26 compares the energy dissipated by the frames and the skin for all
three terrains, normalized with respect to the final energy dissipated in the rigid surface
impact at 200 ms. The rigid and soil terrains resulted in similar energy dissipation by the
frames, with the soil being only 6% higher. In the water impact case, the energy
73
dissipated by the frames was 36% lower than the rigid impact case. The soil and water
terrains resulted in similar dissipation by the skin, which were approximately 20% and
26% lower than the rigid terrain, respectively. The rate of energy absorption by the skin
Figure 3-26: Energy dissipation by (a) frames, and (b) skin for vertical impact on rigid, soil and water
surfaces.
It was noted in a previous study [20,72, Chapter 2] that the luggage played an
important role in energy dissipation during the impact event. In fact, for vertical impact
of the single-section fuselage, the luggage dissipates approximately one third of the total
initial kinetic energy. This issue was further investigated herein by conducting vertical
impact simulations of the full-length fuselage model on the rigid and water terrains with
luggage contained in the cargo bay. The luggage was modeled in the same way as that in
[20,72, Chapter 2]. For both simulations, the fuselage was dropped onto the surfaces with
74
no pitch angle and with the same initial vertical velocity of 9.14 m/s as the crash landing
simulations. Figure 3-27 shows a comparison of the deformed fuselage at 200 ms for
both, rigid and water, impact conditions. As expected, soil impact (not shown here)
should yield intermediate deformation behavior relative to these two extreme cases.
Clearly, the presence of luggage in the cargo bay prevented the lower frames from
deforming upwards and contacting the floor beams. In terms of energy dissipation, results
show that for both terrains the luggage dissipated approximately 45% of the total kinetic
.
With Luggage Without Luggage
(a) Rigid impact
(b) Water impact
Figure 3-27: A comparison of the deformed fuselage with luggage at 200ms under rigid and water impact
conditions.
75
Figure 3-28: Energy dissipation for rigid and water impact conditions, showing significant energy absorbed
by the luggage.
The effect of the luggage on the acceleration-time histories at the forward and aft
locations of the fuselage is shown in Figure 3-29 for the case of rigid surface impact.
Results show that the presence of the luggage reduced the acceleration pulses by more
Figure 3-29: Seat track acceleration-time histories showing the effect of luggage for the case of vertical
rigid surface impact.
For water impact, the luggage was not as effective in lowering the maximum peak
acceleration. For example, the presence of luggage increased the first peak acceleration at
the forward location from 8 g’s to 12 g’s, Figure 3-30. Overall, the luggage acted like a
damper - reducing the amplitude of subsequent oscillations, yielding nearly a single long
duration pulse.
Figure 3-30: Seat track acceleration-time histories for effect of luggage in the case of vertical water surface
impact.
77
excluding wings, and landing gears, was developed to simulate crash-landing on different
terrains, such as rigid surfaces, soils, and water. The full-length model consisted of a
primary fuselage section, a nose cone, and a tail cone, the latter two represented by
3-m long B737 single-section model, which has been validated against drop test
experimental data, was first simplified to enable a computationally efficient model for the
impact simulations on rigid and water surfaces was conducted to ensure that the
simplified section model was capable of replicating the deformed configuration and
acceleration-time histories similar to that of the original section model. The results
simulations, while requiring one-fifth of the computation time. The full-length fuselage
model was then constructed by expanding a simplified section model which was
employed to conduct a series of crash landing simulations on a rigid, soil, and water
terrains.
Simulation results showed that the peak accelerations for all three terrains were
very similar, with the highest peak acceleration being 40 g’s and occurring at the forward
section. The aft section typically showed lower peak acceleration pulses, arriving at
different times for each terrain. In all three cases the frames absorbed more energy than
the skin and, in the soil and water impact cases, the terrain also deformed, dissipating a
portion of the initial kinetic impact energy. The effect of luggage on the dynamic
78
response of the fuselage structure, impacting the three terrains, was also investigated by
conducting vertical impact simulations. The luggage played an important role as the
major energy absorber, as well as acting as a damping mechanism that rapidly diminished
4.1. INTRODUCTION
737 fuselage section was used to study the effect of friction and the angle of impact
between the fuselage section and a rigid surface. This model was extended in Chapter 3
to model a full-length narrow-body transport aircraft fuselage which was used to study
the dynamic response during impact at various terrains (rigid, soil and water). Both
scenarios. One common result was the significant effect of the under-floor luggage on the
overall energy absorption process and, specifically, in reducing the forces (accelerations)
transmitted to the occupants. Hence, integrating energy absorbing mechanisms into the
existing subfloor structures could be a viable option to mitigate the impact energy. In this
chapter, a simple example on the effect of using energy absorbing structural members on
the Boeing 737 fuselage section is presented. The setup of the model is discussed first,
followed by a comparison of the results with that of the original B737 fuselage section
models with and without under-floor luggage. This brief and preliminary study serves as
peak loads, energy absorption capacity, and failure processes of various composite-made
stanchions.
80
Complete modeling details of the B737 fuselage section model were described in
Chapter 2. For the present study, the luggage was removed from the subfloor region of
the model and replaced with energy absorbing stanchions. However, integrating realistic
models of stanchions into the subfloor structure of the B737 model requires additional
information, such as connections and joints to the surrounding frames, that is not readily
available. Therefore, the stanchions were modeled using spring elements. It should be
simplification of the actual geometry, affecting the accuracy of the results. Hence, the
purpose of this study is merely conceptual: to test the concept of using stanchions to
decrease the decelerations on the occupant. Further, the results presented herein should be
viewed considering the following assumptions: i) the load will always be transmitted
axially, where no bending or twisting of the stanchions is permitted; ii) the sustained
crush load on the stanchion remains constant throughout the entire impact event; and iii)
the stanchions remain connected to the nodes, including during rebound of the fuselage.
In this study, discrete spring elements with an inelastic spring material model
(MAT S08) [73] were used to model the stanchions. MAT S08 material model requires a
user-defined load-displacement curve for both tension and compression. A typical load-
A total of 14 spring elements were used in the 3 m long section model; two at
each frame, seven on the left- and seven on the right-hand sides, corresponding with their
81
location shown in Figure 2-4. The springs were connected between nodes located at the
top of the floor beams and the bottom of the lower frames, and were aligned vertically,
Figure 4-2(a). By connecting the spring elements directly to nodes in the beams and the
frames, the forces in the springs could be transmitted only through these nodes, yielding
distortion and severe failure of the elements surrounding these nodes, rigid elements were
35
MAT S08 (Spring)
30
25
20
Load (kN)
15
10
5
0
0 10 20 30 40 50
Displacement (mm)
Figure 4-1: Load-displacement curve used in MAT S08
Figure 4-2: (a) B737 fuselage section showing spring elements as energy absorbers, and (b) a close-up
view of a spring element showing the rigid connectors
82
The simulation results of the B737 model with spring elements are compared with
that of the B737 models with and without luggage, shown in Chapter 2. The structural
deformation of the fuselage at three different time intervals shows that the spring
elements substantially reduced the deformation of the subfloor structure, as well as the
overall deformation of the frames, Figure 4-3. The passenger floor displaces vertically
down by approximately 300 mm in the spring model, compared to nearly 450 mm in the
model with luggage. That is, compared with the luggage, the spring elements provide a
stiffer crushing response. It is worth noting that a stanchion could provide only local-
support to the passenger floor, while the effect of the luggage is distributed nearly
Figure 4-4(a), shows that the spring model results in an acceleration peak that is
marginally lower than that of the model without luggage but higher than the model with
luggage. This is expected since the spring selected herein exhibited higher stiffness
compared to the luggage and were located in discrete points under the floor. On the right-
hand side, Figure 4-4(b), all three models result in very similar acceleration peaks. At
both the left- and right-had side locations, the spring model results in earliest arrival of
the acceleration peak. Similar to the case with luggage, the spring model typically results
in a single acceleration pulse with a longer duration, compared to the several short pulses
resulting from the impact with no luggage. Similar behavior, but with varying peak
accelerations and timing, was recorded at all other seat track locations (not shown here).
83
Figure 4-3: Comparison of the deformation of the B737 fuselage section models with luggage, without
luggage, and with spring elements as energy absorbers at three time intervals.
Figure 4-5 shows the comparison between the energy dissipated by the luggage
and the spring elements, normalized with respect to the total energy dissipated by the
luggage. It can be seen that the spring elements absorb approximately half the amount of
energy absorbed by the luggage. Further, the spring elements only absorb energy up to
approximately 75 ms after the initial impact, while the luggage continues to absorb
84
energy till the end of the simulation. This indicates that the spring elements stopped
25 25
(a) (b)
20 20
15 (a) 15 (b)
Acceleration (G's)
Acceleration (G's)
10 10
5 5
0 0
-5 With Luggage -5 With Luggage
Without Luggage Without Luggage
-10 -10
With Spring Elements With Spring Elements
-15 -15
0 100 50150 200 0 50 100 150 200
Time (msec) Time (msec)
Figure 4-4: Comparison of the acceleration-time histories at two FS 452 seat track locations: (a) LHS
outer seat track, and (b) RHS outer seat track.
1.00
With Luggage
Spring Elements
Normalized Energy Dissipated
0.80
0.60
0.40
0.20
0.00
100 150 0 200 50
Time (msec)
Figure 4-5: Comparison of the energy dissipated by the luggage and the spring elements
85
This brief study demonstrated the feasibility of using energy absorbing members
(i.e. stanchions) to dissipate the impact energy during a crash event. The deformation
comparison shows that stanchions, modeled as spring elements, are able to prevent
excessive failure of the frames, similar to that recorded for the case without luggage.
However, the crushing response of the spring elements selected in this study is slightly
stiffer than that of the luggage. The energy dissipation comparison reveals that the spring
elements do not crush for the entire duration of the impact event, compared to the case
with the luggage, which crushes till the end of the simulation. This obviously affects the
amount of impact energy dissipated by the spring elements, which ultimately absorb
approximately 50% of the energy absorbed by the luggage. In terms of the acceleration-
time histories, the spring elements result in a lower acceleration peak than the case with
luggage on the left-hand side, and a comparable acceleration peak to the other two cases
elements to model the energy absorbing members, there are other factors that could
significantly affect the results of the simulation. For example, the location of the springs
was arbitrarily assigned, based on the alignment of pre-existing nodes in the beams and
the frames. Relocating the springs, either outward or inward, could alter the dynamic
response of the structure. Further, the number of springs used needs to be taken into
consideration as well. In this study, two stanchions per frame were modeled; however,
increasing or decreasing the number of stanchions could significantly affect the results. It
should be noted that the luggage described in Chapters 2 and 3 was distributed nearly
86
uniformly under the floor. Finally, the stanchions were modeled using a representative
could have a very different effect on the dynamic response of the fuselage structure.
spring elements to represent the stanchions. These results are essentially the foundation
that motivated the further research into the energy absorption capabilities of
finite element models capable of simulating the crushing process of these members,
discussed in Chapters 7 to 9.
87
5.1. ABSTRACT
the optimal configuration that would result in the lowest initial peak load while providing
the highest possible specific energy absorption (SEA). The chamfer failure trigger proved
to be the most effective at lowering the initial peak load while yielding a high SEA. The
terms of decreasing the initial peak load and increasing the SEA. These results were
significantly affected by the corner radii of the crush-caps: the smaller the radius the
higher the initial peak load and the SEA. It was determined that combining a chamfered
tube with an inward-folding crush-cap yielded the lowest initial peak load and the highest
SEA.
4
Siromani, D., Henderson, G., Mikita, D., Mirarchi, K., Park, R., Smolko, J., Awerbuch, J. and Tan, T.,
“An Experimental Study on the Effect of Failure Trigger Mechanisms on the Energy Absorption
Capability of CFRP Tubes under Axial Compression,” Submitted to: Composites Part A: Applied Science
and Manufacturing.
88
5.2. INTRODUCTION
occupant protection, has been identified as a key area of focus to improve survivability in
the event of a crash [1]. Studies have shown that the vertical load transmitted to the
occupants during a crash is a key factor in causing injury to the lumbar spine [78]. Thus,
the mitigation of this impact load has long been recognized as a vital consideration of a
the use of Deployable Energy Absorbers (DEAs) fitted on an MD-500 helicopter for a
full-scale crash test [22]. The DEAs consist of a composite honeycomb structure,
fabricated using fiber reinforced fabrics, which can be deployed externally to absorb
energy in the event of a crash. While test results showed that DEAs offer a high level of
energy absorption due to the large crushing zone available, the resulting parasitic weight,
as floor beams, stanchions and stiffeners, to provide for higher energy absorption. Such
members must sustain normal flight load conditions while reducing initial peak loads
upon impact and maximizing energy absorption during a crash. For small aircraft and
rotorcraft, this can prove challenging due to the space constraints in the subfloor
compartment [23].
structural components due to their high strength-to-weight ratio and their high specific
89
materials for crashworthy applications in the aircraft and automotive industries. For
simplicity, cost effectiveness, and for the purpose of understanding the fundamental
crushing process of composite stanchions, most studies were conducted under quasi-static
similar laminate configurations Gr/Ep absorbed the highest energy while K/Ep absorbed
the least. However, all three material systems exhibited higher energy absorption
capabilities than conventional metallic structures, such as aluminum and steel [33,53].
Simply, the extensive damage that occurs during the crushing process of brittle
higher energy than during the elastic/plastic deformation of metals. In most cases
(depending upon laminate configuration and fabrication processes) the more brittle the
material system the higher the energy absorbed under compression [35].
The effect of fiber orientation for all three material systems have been
investigated using [0/±θ]n laminates [25], showing that the highest energy absorption was
achieved using a [0/±15]n laminate. Similarly, it was shown that increasing the number
[28].
effect on the energy absorbing capacity. These include tubes with circular [23,25-27,33-
90
35,45,48,52-54], square [42,47,52] and hexagonal [52] cross-sections, hourglass [52] and
cone [43,43,52,55] shaped tubes, angle-stiffeners [28,42], C-channels [28,42], flat plates
[28,37,38], sandwich panels [49], foam-filled blocks [56], hat stiffeners [39], and sine-
wave members [29,36]. Tubular specimens have been used extensively due to their self-
supporting capability that does not require end fixtures. Open cross-section specimens, on
the other hand, need supporting devices to maintain stability during the crushing process,
e.g., [28,42], which add complexities to proper load introduction. Results have shown
that circular tubes can achieve SEA values well over 100 kJ/kg [23], which is much
higher than any other cross-sections investigated, e.g., [28,37-39,42]. However, when
integrated into a beam structure, designed for a rotorcraft subfloor, their energy
absorption capacity is significantly reduced [23]. Possible causes for this unexpected
result were being examined and were not discussed in detail by the authors in [23].
diameter-to-thickness (D/t) ratio [25]. The results obtained with [±45°] graphite/epoxy
tubes indicate that a reduction in tube D/t ratio results in an increase in energy absorption.
For tubes with the same D/t ratio, the SEA decreases with increasing tube diameter.
It has been well established that introducing an effective failure initiation site,
which triggers progressive failure, is crucial for reducing the initial peak crush load and
composite tubes fail abruptly and catastrophically, with a high initial peak load followed
by a very low energy absorption capacity. Failure trigger mechanisms can be classified as
internal triggers (e.g., chamfer, steeple, and ply drop-off) and external triggers (e.g.,
crush-caps and plug initiators), which are added at one end of the stanchion.
91
Numerous studies have shown that chamfer failure triggers are very effective at
reducing the initial peak crush load while maintaining high sustained crush loads and
SEA [23-28]. The crush behaviors of thin-walled hollow square [42,52] and circular
tubes [52] with chamfer failure triggers have been compared against those with steeple
failure triggers. Results showed that the for square tubes steeple failure trigger was
significantly more effective than the chamfer failure trigger at maintaining a higher
sustained crush load, but the opposite was recorded for circular tubes. The effect of
external failure trigger mechanisms, using crush-caps, on the initial peak load and the
SEA of graphite/epoxy cylinders were studied in [31,32]. In [31] two different types of
crush-caps were studied, forcing fibers to splay outward and inward, respectively. Each
type of crush-cap was fabricated with two different corner radii (3 mm and 5 mm) used to
cause stress concentrations at the end of the tube. The results indicate that the inward-
folding crush-caps increased material interaction and provided higher SEA. Also, in both
cases, the crush-cap with the smaller corner radius provided a higher SEA. A “load-
the addition of an outer-curvature constrainer designed to increase the SEA. The results
indicated that the smaller curvature radius of the load-control attachment yielded a higher
SEA.
combined (chamfered-end and crush-cap) failure trigger mechanisms. The effect of the
corner radii of the crush-caps on peak load, SEA and crush behavior of the tubes is also
92
evaluated. Results of this experimental study have been used to validate the numerical
unidirectional tape pre-preg with a 12 K tow and a 180° C curing resin (350° F), provided
Pennsylvania. All specimens were 101.6 mm (4.0 in) long with an inner diameter of 29.4
mm (1.16 in) and a wall thickness of 1.47 mm (0.058 in), resulting in a D/t ratio of 20.
Two different sets of tubular specimens were tested. The first set consisted of cylindrical
tubes donated by The Boeing Co. The lay-up consisted of nine plies of unidirectional
material oriented in 0° and ±15° directions. The exact lay-up remains company
proprietary. The second set was prepared in house with a lay-up of [+15/-15/+15/03/-
15/+15/-15]. This lay-up was selected based on the work reported in [25].
one end of the tubular specimen, Figure 5-1(a). Alternatively, failure could be triggered
by attaching a crush-cap at one end of the tube. In this study, two different types of crush-
caps were studied, each causing the crushed material to fold either inward (Figure 5-1(b))
or splay outward (Figure 5-1(c)). The effect of the corner radius of the crush-caps on the
peak load, SEA, and crush behavior was investigated. Based on these results, a combined
93
failure trigger, i.e., chamfered specimens with inward-folding crush-caps, were also
studied.
All tests were conducted under quasi-static axial compression conditions. The
tests were carried out on an Instron Testing Machine, Model 5800R. Loading was applied
under stroke control mode at a displacement rate of 0.127 mm/s (0.3 in/min). All
Utilizing a 101.6 mm (4.0 in) long specimen, the 50.8 mm (2 in) crushing length
provided sufficient crush data from which the amount of energy absorbed, and the
A digital image correlation (DIC) system (an ARAMIS 4M system from GOM)
was used to capture the full deformation and strain fields during loading. The system is
based on two charged coupled device (CCD) cameras, having 4 megapixels resolution,
and is capable of measuring strains ranging from 0.01% to over 100%. Between 150 -
94
220 images were taken during each test, at one- second intervals. High definition videos
Test specimens were divided into two groups. Group A consists of 37 specimens
made from Boeing-supplied tubes, which were tested in three Phases: In Phase I the
the effect of the two crush-cap failure trigger mechanisms (inward-folding and outward-
splaying), each having three different corner radii, on the crushing process was studied
using flat-end specimens. In Phase III, the effect of combining the chamfer with crush-
cap trigger was examined. Detailed test matrix is given in Table 5-1.
two phases. In Phase I, baseline data were established by testing chamfered specimens. In
Phase II, the optimal failure trigger mechanism, based on the results of the Group A tests,
cap, was investigated. Three specimens were tested for each failure trigger scenario and a
single flat-ended specimen was tested in Phase I of Group A. Detailed test matrix is given
in Table 5-2.
II Inward
Chamfer 1.58 mm (1/16 in) 3
Folding
∫ ( )
(1)
P(δ) is the corresponding load, and ρ and A are the mass density and cross-sectional area,
respectively. Accordingly, the SEA was obtained for each test by dividing the area of the
load-displacement curve by the total mass of the crushed section of the specimen.
96
and crush-cap failure trigger mechanisms, and the combination of the two mechanisms.
All specimens were made from The Boeing Co. supplied tubes.
specimen and three chamfered specimens are shown in Figure 5-2. Results of the three
introduction. Clearly, in the absence of a failure trigger mechanism, the crushing of the
flat-ended specimen initiated with a very high initial peak load (75 kN), followed by an
instant drop, by nearly 15 folds, in the load, yielding a low sustained crush load, Figure
5-3 and Figure 5-4. This crushing behavior resulted in a low SEA value. The results of
the chamfered specimens, on the other hand, show a much lower average initial peak load
(27 kN) and a slower and smaller load drop, yielding a slightly higher average sustained
crush load and a 16% higher SEA, Figure 5-3 and Figure 5-4.
The deformation characteristics and failure process of the flat-ended and chamfer-
ended tubes (discussed below) show that during the crushing process the outer two plies
splayed outward and the inner seven plies folded inward and compacted into the hollow
tube. Clearly, the peak load, the sustain crush load, the corresponding SEA, and the crush
behavior could be greatly affected if all plies are forced to either fold inward or splay
97
outward. Therefore, in the Phase II of this study, crush-caps were used as failure triggers
to initiate failure in flat-ended specimens. Two types of crush-caps were used, causing
crush-cap was manufactured with three different corner radii, Table 5-1. Three tests were
performed per crush-cap. Results indicate that the load-crosshead displacements curves
were highly reproducible. Thus, the average values from the three specimens for each
curves are shown in Figure 5-5. It can be seen that the inward-splaying crush-caps
reduced the peak loads significantly, to 52%, 27%, and 23% of that of the flat-end
specimen for corner radii of 2.4 mm, 4.0 mm, and 5.6 mm, respectively, Figure 5-3. The
crush-cap with the smallest corner radius (2.38 mm) also yielded much higher sustained
crush load (1.4 times) and SEA (1.4 times) than the flat-end specimen, Figure 5-4. The
crush-caps with the two larger corner radii (4.0 mm and 5.6 mm), on the other hand, did
not yield favorable results in terms of lowering the peak load and increasing the SEA as
outward-splaying crush-caps with three different corner radii are shown in Figure 5-5.
The outward-splaying crush-caps reduced the initial peak loads to 80%, 12%, and 13% of
that of the flat-end specimen for corner radii of 2.4 mm, 4.0 mm, and 5.6 mm,
sustained crush load or SEA, Figure 5-3 and Figure 5-4. The crush-cap with the smallest
98
corner radius (2.4 mm) yielded the highest sustained crush load and SEA. This corner
radius, however, yielded a sustain crush load and SEA approximately 53% of the flat-
Figure 5-2: A comparison of load-crosshead displacement curves of a flat-end specimen and three
chamfered specimens from Group A.
Figure 5-3: A comparison of initial peak loads and sustained crush loads for Group A specimens tested in
Phase I and II.
99
Figure 5-4: A comparison of SEA for Group A specimens tested in Phase I and II.
Figure 5-5: A comparison of averaged load-crosshead displacement curves for flat-ended specimens from
Group A attached to (a) three inward-folding crush-caps, and (b) three outward-splaying
crush-caps with different corner radii.
Summary of Crush-cap Trigger Mechanisms: The results shown in Figure 5-3 and
Figure 5-4 indicate that the smallest corner radius (2.4 mm) yielded the highest sustained
crush loads and SEA in both types of crush-caps. The inward- folding crush-cap yielded
the highest SEA (120 kJ/kg) and sustained crush load (27 kN). It also resulted in a
relatively higher initial peak load (39 kN) as compared to that of chamfered specimen (19
kN), but was still 48% lower than that of the flat-end specimen (75 kN).
100
Comparing the results discussed above with those reported in [31], shows a
similar trend where smaller radii crush-caps yield higher SEA for both types of crush-
caps. Further, both studies show that the inward-folding crush-caps yield higher SEA the
outward-splaying crush-caps.
Based on the results of Phase II study, it was established that the chamfered end
was most effective at reducing the initial peak load while the inward splaying crush-cap
having the smallest corner radius resulted in a higher sustained crush load and SEA. In
the initial peak load and increasing the sustained crush load and SEA. The effect of the
corner radius of the crush-cap was further characterized, particularly at the smaller radius
range, using four additional radii along with the original smallest radius (2.4 mm), Table
5-1.
Figure 5-6 shows the results of the combined failure trigger mechanism with five
Results indicate that the crush-cap corner radius affected the initial peak load
significantly. The larger the corner radius the lower the initial peak load. The initial peak
load reduced from 45 kN for the sharp corner (0.05mm) to 13 kN for the largest radius
(3.2 mm), Figure 5-7(a). The SEA initially increases with the corner radius to a
maximum of 157 kJ/kg for 1.6 mm corner radius before dropping to as low as 37 kJ/kg
for the largest radius (3.2 mm), Figure 5-7(b). The sustained crush load shows a similar
trend since it is directly proportional to the SEA, Figure 5-7(a). It should be noted that the
101
maximum SEA occurs when the corner radius is similar to tube’s wall thickness (1.47
mm). This suggests that the tube wall thickness-to-the corner radius ratio is directly
Comparing the results of the 2.4 mm corner radius crush-cap with corresponding
Phase II results, it can be seen that the chamfer helped to reduce the initial peak load from
39 kN to 23 kN, but was not effective in increasing the sustained crush load (17 kN vs. 27
kN) or SEA (78 kJ/kg vs. 120 kJ/kg). The smaller corner radius (1.6 mm) crush-cap, on
the other hand, yielded an SEA (157 kJ/kg) twice as high as that of the 2.4 mm corner
radius crush-cap, but also resulted in an initial peak load (37 kN) that was approximately
61% higher. These results guided the test program of Group B specimens, as discussed
below.
Figure 5-6: A comparison of averaged load-displacement data of Group A specimens with a combined,
chamfered-end and inward-folding crush-cap, trigger mechanism.
102
Figure 5-7: A comparison of (a) initial peak loads and sustained crush loads, and (b) SEA of Group A
specimens with a combined, chamfered-end and inward-folding crush-cap, trigger
mechanism.
Results of Group A specimens showed that the chamfer failure trigger is very
effective at reducing the initial peak load while maintaining a high sustained crush load
and SEA. The inward-folding crush-cap, with a small corner radius, was not as effective
as the chamfer failure trigger at reducing the initial peak load, but was quite effective at
maintaining a higher sustained crush load and SEA. A combination of the two failure
1.6 mm, yielded a low initial peak load while maintaining a high sustained crush load and
SEA, proved to be the optimal combination for this particular set up. This optimal failure
trigger configuration was used to study the crush behavior of Group B specimens,
manufactured in-house at Drexel, for which the lay-up and material properties were
known. Two phases of study were conducted with Phase I for establishing baseline data
103
using chamfered specimens and Phase II using chamfered specimens combined with the
inward-folding crush-cap.
Tests were conducted with three chamfered specimens to establish the baseline
data. Results are shown in Figure 5-8 and Figure 5-9, along with data from the Group A
chamfered specimens for comparison. It can be seen that Group B specimens have a
slightly lower average initial peak load than that of Group A specimens (24 kN vs. 27
kN) and yielded higher sustained crush load (28 kN vs. 22 kN) and SEA (126 kJ/kg vs.
101 kJ/kg). While the two specimen groups were of the same material, the slight changes
in laminate stacking sequence and curing processes yielded the above differences.
crush-caps with a corner radius of 1.6 mm were tested. Results are shown in Figure 5-8
and Figure 5-9, together with data from the corresponding Group A (Phase III) test
results for comparison. The results show similar trend as in Phase I: lower initial peak
load (by 19%) and slightly higher SEA (by 7%) than the corresponding results obtained
sequence and fabrication process. Compared with the baseline data, obtained in Phase I
of Group B specimens, the combined failure trigger yielded a 25% higher initial peak
load (30 kN vs. 24 kN) and a 33% higher SEA (168 kJ/kg vs. 126 kJ/kg).
end and crush-cap (with a 1.6 mm corner radius) failure trigger was more effective in
104
increasing the sustained crush load and SEA, with a moderate level of initial peak load. It
should be noted that the added parasitic weight of the crush-caps could be mitigated by
Figure 5-8: A comparison of averaged load-crosshead displacement curves of Group A and Group B
specimens with chamfered-ends and combined trigger mechanisms.
Figure 5-9: A comparison of (a) initial peak loads and sustained crush loads, and (b) SEA of Group A and
Group B specimens with chamfered-ends and combined trigger mechanisms.
105
specimen, post-crushing loading, and unloading, for each specimen was recorded using a
high definition digital video camera. Photographs of the specimen were also taken before,
during, and after each test. The discussion below is divided according to the type of test
i) Flat-Ended Specimen
The crush process of Group A flat-ended specimen, shown in Figure 5-10, shows
upon first collapse), 10 mm (during the crush process) and 40 mm (near the end of the
manifested by an abrupt local wall and plies buckling and crushing at the lower end of the
specimen that was in contact with the lower support platen of the testing machine, Figure
5-10(b), which corresponds to the abrupt load drop seen in Figure 2. This sudden fracture
was accompanied with significant audible level noise and a large amount of Gr/Ep dust.
splitting was observed along the outer 0° plies and propagated along the length of the
specimen with increasing crosshead displacement, Figure 5-10(d). These matrix splits
were nearly evenly spaced along the circumference of the specimen, forming
approximately 40 narrow, 1.0-4.0 mm wide, Gr/Ep strips. Two of the nine outer plies
splayed outward, sliding along the lower rigid support platen of the testing machine as
106
of the specimen revealed that approximately seven inner plies folded inward and
the photographs in Figure 5-11 and Figure 5-12, respectively, showing the deformation
mm (intermediate level of the crush process) and 40 mm (near the end of the 50.8 mm
crushed gradually, with no apparent local wall and/or plies buckling. The matrix splits
appear nearly evenly spaced along the circumference, Figure 5-11(b). These matrix splits
initiated at the chamfered region and progressed, with increasing crosshead displacement,
along its length, Figure 5-11(c), forming narrow (approximately 1.0-4.0 mm wide)
occurred primarily along the outer two-three plies with six-seven inner plies folded
The failure process of the Group B chamfered specimens, Figure 5-12, is nearly
identical. Comparing Figure 5-11 and Figure 5-12 at the same crosshead displacement
shows little difference in the failure process of the two groups of specimens, with matrix
splitting occurring immediately at the completion of the chamfer crushing. The matrix
splitting, however, progressed more rapidly in Group A, Figure 5-11(c) compared with
Group B, Figure 5-12(c). Significant matrix splitting occurred also with the Group B
specimens, yielding similar narrow laminate strips, Figure 5-12(d). The apparent twisted
lamina segments in Figure 5-12(d) and Figure 5-12(e) are due to matrix splitting along
the outer 15° plies. The difference in the post-failure appearance after load removal of the
two Groups, Figure 5-11(e) and Figure 5-12(e), is attributed to the difference in laminate
stacking sequence. Detailed examination of the post-test Group B specimens indicates the
outermost -15°/+15°/-15° ply group in the Group B specimens delaminated along the
inner neighboring 0° ply group. As a consequence, the ply group of the outer three plies
splayed outward, while the inner six plies (two ply groups) folded and compacted inward
into the inner space of the hollow tube, in a manner similar to Group A specimens.
108
in Figure 5-13. Significant matrix splitting and delamination of the outer two plies
significant buckling of the thin Gr/Ep strips along the outer 0° ply, having a high
109
slenderness ratio, occurred, Figure 5-13(c) and Figure 5-13(d). The buckled region
progressed from the bottom, along the specimen length, corresponding to the extension of
the matrix splitting which formed long and narrow (1.0-4.0 mm-wide) Gr/Ep strips. Upon
load removal the buckled strips separated and scattered away from the core cylinder,
clearly exposing the inward-folding and the compacted nature of the crushed material,
Figure 5-13(e).
Post-test examination of the crushed end of the Group A specimens showed that
the inward-folding crush-caps forced the inner plies to fold toward the center of the tube.
This resulted in a large amount of material interactions and a tightly packed core. As
loading continued, the Gr/Ep material was forced into the packed core of the specimen,
Significant matrix splitting and delamination of the all plies occurred, forming the typical
increases, the plies are forced to splay outward, Figure 5-14(c) and Figure 5-14(d), which
prevented fibers packing at the core, thus reducing fiber interactions, Figure 5-14(e). This
resulted in a much lower SEA when compared to the specimens having inward-folding
crush-caps.
It was shown that the optimal energy absorption could be accomplished with a
corner radius, Figure 5-4. The typical failure process of the combined failure trigger
specimen is shown in Figure 5-15. The deformation and crushing observed are similar to
111
that of the flat-end specimens with inward-folding crush-caps, Figure 5-13, where all
plies fold into and compact inside the hollow tube. However, the chamfered-end
specimen crushed gradually without buckling of the outer plies, as is the case with the
flat-ended specimens, Figure 5-13. Since the entire folding and crushing process occurs
The strain fields of six chamfered tubes were monitored via a Digital Image
Correlation (DIC) system. Three of the specimens were recorded using a low-speed DIC
system and three were recorded using a high-speed DIC system. In this section,
representative results from the lows speed and high speed DIC images are presented.
The global fringe patterns, showing the axial and hoop strain fields, respectively,
Figure 5-16. The axial strain field indicates that failure occurs near or at the maximum
112
load level of 23.8 kN, upon chamfer crushing, resulted in an abrupt load drop from 23.8
kN to 17.0 kN. At this stage, the entire crush region at the bottom of the specimen was
experiencing an axial strain of approximately 1%. Several ‘hot spots’ can be seen in the
crush region at 17.0 kN in both figures. These hot spots corresponded to initiation sites of
matrix splitting along the fibers, which was confirmed, at later stages of crosshead
displacements, via the recorded videos. These local failures occurred nearly
simultaneously with the load drop. Similar results were observed in all other chamfered
specimens.
Figure 5-18 shows the local strain fields, taken with a high-speed DIC system, at
eight different load stages (from 14.9 kN to 25.8 kN) in a 10 mm x 10 mm region in the
crush zone just above the chamfer region. The chamfered zone is the dark region beneath
the photogrammetry images in the figure. The first hot spots, representing matrix splits,
initiated just above the chamfer between 14.9 kN and 15.9 kN, well before they were
captured with the low speed DIC system, Figure 5-16. This indicates that these matrix
splits initiated before the chamfer was completely crushed, i.e., before the load drop
(Figure 5-2). With increasing crosshead displacement the extent and number of these
splits increases with spacing nearly identical to those observed in the post-test
examinations of the specimens, e.g. Figure 5-12, which ultimately generated the 1.0-4.0
5.5. CONCLUSIONS
quasi-static compression. Tests were conducted with flat-ended, chamfer-ended, and with
was to determine the most effective approach to decrease the initial peak crush load while
increasing the sustained crush load and SEA. Two groups of specimens were tested, both
having similar lay-ups and manufacturing procedures: Group A specimens were provided
by The Boeing Co. and Group B specimens were fabricated in-house at Drexel.
It was determined that the chamfer failure trigger was most effective at reducing
the initial peak load while maintaining a high-sustained crush load and high specific
energy absorption (SEA). The inward-folding failure trigger approach was not as
effective at reducing the initial peak load, but was more effective than using a chamfer
for maintaining a high-sustained crush load and SEA. On the other hand, the outward-
115
splaying crush-cap was ineffective at reducing the initial peak load or in maintaining a
high-sustained crush load and SEA. These key results highlight the importance of forcing
inward-folding deformation and the consequent material crushing. Further, the results
obtained with different corner crush-cap radii showed that the smaller radii provide better
trigger mechanism was use to further study the effect of corner radius of the crush-caps.
Results showed that the crush-cap with a 1.6 mm (1/16 in) corner radius yielded a
moderate level of initial peak load and the highest sustained crush load and SEA (60%
lower and 82% higher, respectively, than that of a flat-end specimen without crush-caps).
These quantitative results depend on laminate configuration and tube geometry (length,
diameter, and wall thickness). Results obtained from this study have been used to
validate a finite element model for simulating the quasi-static crushing process of
composite tubes. Results of the simulation study are reported in [79, Chapter 8].
116
6.1. ABSTRACT
The effect of two failure trigger mechanisms (chamfers and steeples) on initial peak
loads, crush loads, and failure progression and crushing was studied. Strain fields
(measured via digital image correlation system) captured the damage initiation and
progression throughout the crushing process. Results showed that the steeple trigger is
more effective at reducing the initial peak loads of the C-channel and hat-stiffeners, while
the chamfer trigger yields a lower initial peak load for the angle-stiffener. The angle- and
hat-stiffeners absorb similarly high specific energy absorption (SEA) while the C-
6.2. INTRODUCTION
survivability in the event of a crash [1]. Several studies, investigating the dynamic
response of aircraft structures and the survivability of occupants, have shown that the
5
Siromani, D., Cheng, B., DeLuca, M., Donegan, D., Giberson, P., Mucerino, C., Awerbuch, J. and Tan, T.,
“Experimental Investigation on the Energy Absorption Capacity During Crushing of Axially Loaded
Thin-Walled Gr/Ep Members,” Submitted to: Journal of Composite Materials.
117
2-4]. Hence, integrating energy absorbing structural members into the existing subfloor
structural components due to their high strength-to-weight ratio and their high specific
groups have investigated the energy absorbing characteristics of composite materials for
crashworthy applications in the aircraft and automotive industries. For simplicity, cost
effectiveness, and for the purpose of understanding the fundamental crushing process of
composite stanchions, most studies were conducted under quasi-static axial compression
loading conditions. The most common composite materials investigated have been
52], and Kevlar/epoxy (K/Ep) e.g., [24,25,33-38], studying different thin-walled cross
hexagonal and hourglass tubes [52], cones [43,43,52,55], open cross sections [28, 39,42],
flat plates [28,37,38], and sandwiched panels [49-51]. Results showed that in similar
laminate configurations Gr/Ep absorbed the highest energy while K/Ep absorbed the
least. However, all three material systems exhibited higher energy absorption capabilities
than conventional metallic structures, such as aluminum and steel [33,53]. Simply, the
extensive damage that occurs during the crushing process of brittle composites, due to the
combination of multiple failure mechanisms, absorbs a much higher energy than during
configuration and fabrication processes) the more brittle the material system the higher
The effect of fiber orientation for all three material systems have been
investigated using [0/±θ]n laminates [25], showing that the highest energy absorption was
achieved using a [0/±15]n laminate. Similarly, it was shown that increasing the number of
0° plies in [452/0n/452]s laminates increased the energy absorption of the members [28].
compression [81, Chapter 5]. Emphasis was placed on determining the optimal failure
trigger mechanisms, including chamfers, crush-caps (forcing inward folding and outward
splaying of the entire tube wall), and their combinations, to reduce the initial peak load
and increase SEA. It was determined that the chamfer trigger was most effective at
reducing the initial peak load. The inward-folding failure trigger approach was not as
effective at reducing the initial peak load, but was more effective than the chamfer failure
trigger at maintaining a high sustained crush load and high SEA. On the other hand, the
outward-splaying failure trigger was ineffective at reducing the peak crush load and in
maintaining a high sustained crush load and SEA. These key results highlighted the
absorption mechanism. Further, the results obtained with different crush cap radii showed
that the smaller radii provide higher sustained crush loads and SEA. Based on these
results, a series of tests was performed combining both failure trigger mechanisms; i.e.,
tests, five crush-caps with different corner radii were tested to determine the optimal
119
configuration. Results showed that the combined failure trigger mechanism yielded a
relatively low initial peak load and the highest sustained crush load and SEA.
However, circular tubes are not typically used in the aerospace industry and
cannot easily be integrated into a rotorcraft subfloor structure [23]. A limited number of
studies have been performed on thin-walled members of other cross-sections, often used
in the aerospace industry, such as angle-stiffeners and C-channels e.g., [28,42], and hat-
stiffeners e.g., [39]. The effect of the radius of curvature of angle-stiffeners and C-
channels was studied in [42] for plain-weaved Gr/Ep, showing that the larger the ratio of
the corner radius to the perimeter of the specimen the higher the SEA. Results have
shown that these thin-walled cross-sections yielded SEA values between 30 and 85 kJ/kg
[28,39,42]. The available data, however, are not sufficient to rank the different cross-
sections in terms of initial peak loads, sustained crush load, and SEA.
component to reduce the initial peak load and maximize energy absorption. Failure
triggering could be classified as internal triggers (e.g., chamfer, steeple, and ply drop-off)
or external triggers (e.g., crush caps and plugs). Numerous studies have shown that
chamfer triggers are very effective at reducing the initial peak crush load while
maintaining sustained crush loads, e.g., [23-28]. The crushing behaviors of hollow square
[42,52] and circular tubes [52] with chamfer triggers have been compared against those
with steeple triggers. Results showed that the steeple trigger was significantly more
effective than the chamfer trigger at maintaining a higher sustained crush load for square
The objective of this study is to experimentally investigate the failure process and
cross-sections having two failure triggering mechanisms. These results were used to
angle-stiffener, and hat-shaped stiffener. The specimens were made of Hexcel IM7/8552
Graphite/Epoxy unidirectional tape pre-preg with a 12 K tow and a 180° C curing resin
(350° F),, provided by the Structures Technology group of The Boeing Company at
Ridley Park, Pennsylvania. Lay-up and dimensions are listed in Table 6-1. To ensure
stability during compressive loading, the specimens were supported by a 25.4 mm (1.0
in.) thick potted base, Figure 6-1. The cross-sectional areas were the same for all cross-
sections. As a result, the wall thickness of the angle-stiffener was twice the thickness (and
Two failure trigger mechanisms were studied: a chamfer trigger and a steeple
trigger. Figure 6-1 shows the three cross-sections with a 45° chamfer and a 15° steeple
Figure 6-1: Open-cross-sections with the two failure trigger mechanisms, all having the same cross-
sectional area and attached to a potted base to ensure stability.
All tests were conducted under quasi-static axial compression conditions. The
tests were carried out on an Instron Testing Machine, Model 5800R. Loading was applied
under stroke control mode at a displacement rate of 3.81 mm/min (0.15 in/min). All
101.6 mm (4.0 in) long specimen, the 50.8 mm (2.0 in) stroke provided sufficient crush
data from which the amount of energy absorbed could be calculated and compare with
the simulation results [80, Chapter 9]. Three specimens were tested for each cross
section and each failure trigger, for a total of 18 specimens Table 6-1.
A digital image correlation (DIC) system (an ARAMIS 4M system from GOM)
was used to capture the full deformation and strain fields during loading. The system is
based on two charged coupled device (CCD) cameras, having 5 megapixels resolution,
and is capable of measuring strains ranging from 0.01% to over 100%. Between 150 -
220 images were taken during each test, at one- second intervals.
∫ ( )
(1)
where δCR is the crushing length, 0 ≤ δ ≤ δCR is the crosshead displacement, P(δ)
is the corresponding load, and ρ and A are the mass density and cross-sectional area,
respectively. Accordingly, the SEA was obtained for each test by simply calculating the
for each cross-section are shown in Figure 6-2(a). All three cross-sections reach similar
initial peak loads, of approximately 32 kN, at which point the chamfer has fully crushed,
followed by a sudden load drop, reaching the lower sustained crush load for the
remainder of the crush displacement. The C-channel experienced the largest load drop
caused by the partial flange buckling and crack formation along the two corners.
The angle- and hat-stiffeners experienced load drops less severe than exhibited by
the C-channel. Also in these two cases the load drop is attributed to partial wall buckling
The average load-displacement curves of the three steeple specimens for each
cross-section are shown in Figure 6-2(b). Unlike the sudden load drop exhibited by the
chamfered specimens, here the load gradually increases to its highest value. This load
increase corresponds to the gradual crushing of the steeple and the simultaneous increase
in the contact area with the compression platen of the testing machine. In all three cases,
the highest load occurs when the compression platen comes in full contact with the entire
cross-sectional area. The crosshead displacement at the highest load, Figure 6-2(b), is a
direct result of the height of the steeple: 10 mm for the C- channels, 6 mm for the angle-
stiffeners, and 12 mm for the hat-stiffeners. The subsequent load drop is relatively small
with the sustained load staying relatively constant for the C-channels and angle-stiffeners,
but gradually decreasing (marginally) in the hat-stiffener case for the remainder of the
50.8 mm crush displacement. The peak load for C-channel is the lowest among the three
124
cross-sections, and it occurs later into the crushing process (at approximately 10 mm of
the crosshead displacement) when the crosshead comes into contact with the two flanges
of the C-channel. The angle-stiffener exhibited the highest peak load of the three cross-
sections, which also occurred earlier into the crushing process. The lower subsequent
sustained crush load is due to significant crack propagation at the specimens’ corners.
The hat-stiffener exhibited an initial peak load which is 20% lower than the angle-
stiffeners’, exhibiting a marginal load drop and very similar sustained load.
Figure 6-2: A comparison of load-crosshead displacement curves for each cross-section having a: (a)
chamfer trigger, and (b) steeple trigger. Each curve represents the average of three tests
(except for the angle-stiffener with a chamfer trigger which represents the average of two
tests).
Figure 6-3(a) shows the comparison between the initial peak load and sustained
crush load for all three cross-sections with both trigger mechanisms. For the C-channels
and the hat-stiffeners, the steeple trigger is more effective at reducing the initial peak load
by 36% and 12%, respectively. However, for the angle-stiffener, the steeple trigger
resulted in an initial peak load that is 19% higher than the chamfer failure trigger. For all
three cross-sections, the steeple failure trigger resulted in sustained crush loads that were
between 10% and 20% higher than the chamfer failure trigger. This, in turn, yielded SEA
125
values for the steeple specimens that were 5%, 21%, and 4% higher than the C-channel,
6-3(b). For both trigger mechanisms, the angle- and hat-stiffeners yielded similarly high
Figure 6-3: A comparison of the (a) peak load and crush load, and (b) SEA for each cross-section having
a chamfer trigger and steeple failure triggers (numbers indicate average of three specimens
except for the angle stiffener with a chamfer trigger which represents the average of two
tests).
The failure process (recorded via the video cameras) and the axial and lateral
strain fields (recorded via the DIC) are discussed below for the three cross-sections and
i) C-Channel
typical chamfered C-channel is shown in Figure 6-4. The chamfered specimen began to
crush gradually with partial inward buckling of the flanges and outward buckling of the
web towards the mid-length of the specimen, Figure 6-4(b). After the chamfer was
126
completely crushed, several matrix splits appeared at the corners of the specimen that
relieved the outward buckling of the web and caused it to bend and splay outwards,
Figure 6-4(c). The matrix splits, which initiated at the corners, progressed along the
length of the specimen with many more matrix splits occurring along the flanges with
increasing displacement, Figure 6-4(d). The extension of the matrix splits reduced the
buckling, increasing the bending and the outward and inward ply splaying. As expected,
initial failure at some random site (post-chamfer crushing) causes an asymmetric crush
process. The web tends to bend and splay outwards further on the left side than on the
right side of the specimen, Figure 6-4(b) to Figure 6-4(d). The outer plies of the web
stayed relatively intact throughout the crushing process but experienced excessive
half of the plies in the web and flanges splayed inwards while the remaining plies splayed
The axial and lateral strain fields for the chamfered C-channel are shown in
Figure 6-5 and Figure 6-6, respectively, at four representative load levels. Note that the
127
DIC images shown in the figure are shortly after the completion of chamfer crushing,
thus the crosshead displacement at this stage is less than 2.5 mm, Figure 6-2(a). The
compressive axial strain concentration occurs first at the web/flange corners, as is evident
by the hot-spots located at the top left and top right hand sides of the C-channel, Figure
6-5. Beyond the peak load, the compressive axial strain increases throughout the web as a
result of the chamfer crushing, followed by extensive crushing, bending, and partial
buckling at the top of the flange. The non-uniform strain field in the specimen is a result
The DIC images indicate high tensile lateral strains throughout the web, increasing up to
the peak load (35 kN), Figure 6-6. This is due to the outward buckling that occurs at the
center of the web. Initially, small lateral tensile strains are observed in the flange/web
corners. After the chamfer is completely crushed, and the load drops (e.g., to 25.8 kN),
high lateral tensile strains occurred along the web/flange corners, causing the matrix
splits. At the same time, the high lateral strains in the web are completely reduced due to
the matrix splits at the corners, relieving the outward buckling in the web. The initiation
of these matrix splits is clearly manifested by the hot-spots seen along these corners,
extending throughout the length of the specimen with further crosshead displacement.
These matrix splits extend along the entire specimen, enabling further bending and
buckling of the web and flanges, yielding significant reduction of the lateral strain at the
center of the web. Thereafter, the lateral strain field is affected by the bending and
outward splaying of the web, as evident in the photographs of Figure 6-4. The non-
uniform lateral strain field is a result of the asymmetric deformation of the specimen and
the effect of the formation of matrix splits, which relieved some of lateral strain. Viewing
128
the sequence of the many strain contour recorded (not shown here) provides a vivid
Steeple End: The steeple specimen began to crush gradually with matrix splits
appearing along the top surface of the flanges and the web with increasing crosshead
the flanges was observed and only minor outward buckling of the web occurred, which
was relieved almost immediately upon the steeple crushing. After the steeple was fully
crushed, the specimen continues to crush progressively (Figure 6-7(c)) with the
129
remainder of the crushing process being very similar to the chamfered specimen. The
only noticeable difference was the larger number of matrix splits that were observed in
the web of the steeple specimen, Figure 6-7(b), Figure 6-7(c) and Figure 6-7(d).
Figure 6-8 and Figure 6-9 show the axial and lateral strain fields for the steeple C-
channel, respectively; up to shortly before the completion of the steeple crushing, thus the
crosshead displacement at this stage is less than 7.5 mm, Figure 6-2(b). As discussed
earlier, Figure 6-7, steeple crushing begins immediately upon load application, e.g., at 2.4
kN, resulting in the high compressive axial strain at the tip of the steeple, Figure 6-8.
With further steeple crushing, the high axial compressive strain propagates towards the
remainder of the steeple. The lateral tensile strain begins to form at both the tip of the
steeple and in the web of the specimen. Some outward buckling was observed in the web,
which resulted in the formation of moderate lateral strains in this region. As the steeple
continues to crush, high lateral strains occurred at the top of the specimen, while the
lateral tensile strain in the web reduces due the decrease in buckling of the web.
130
ii) Angle-Stiffener
is shown in Figure 6-10 at four selected stages of crosshead displacement. Similar to the
matrix splits, appearing along the corner along with partial inward buckling of the legs,
Figure 6-10(b). After the chamfer was completely crushed, large cracks appeared along
the corner of the specimen that progressed with increasing load, Figure 6-10(c) and
Figure 6-10(d). As with the C-channel, the legs initially buckled, however, with the
occurrence of matrix splitting (first in the corner and later across the legs) the
131
asymmetric failure process is manifested by the left leg buckling significantly more than
the right leg, Figure 6-10(b) and Figure 6-10(c). Examination of the post-test specimens
revealed that approximately half the plies splayed inwards while the remainder splayed
The axial and lateral strain fields for the chamfered angle-stiffener are shown in
Figure 6-11 and Figure 6-12 respectively, at five selected load levels, up to
crushed, compressive axial strain develops throughout both legs caused by inward
buckling. As the load is increased to 32.3 kN, the left leg developed higher compressive
axial strain due to excessive buckling, Figure 6-11, as discussed earlier. Once the chamfer
is fully crushed and the peak load is reached, the high strain transferred to the top of the
specimen and initiated the progressive crushing of the two legs, as manifested by the
The lateral strain fields in Figure 6-12 shows high tensile strains at the top corner,
initiating matrix splits at that location. Higher strains are observed on the left leg due to
the excessive buckling on that side, as discussed earlier. As the crushing progresses, these
cracks extend along the corner, reducing the tensile strain in the legs as the load reached
the peak load value of 33.7 kN. The somewhat non-uniform lateral strain field seen at the
top of the specimen is due to the asymmetric deformation of the specimen and the effect
of the formation of matrix splits which relieved some of the lateral strain.
Steeple End: In the case of the steeple angle-stiffener, since the steeple is located
at the corner of the specimen, the matrix splits appear at the corner immediately upon
loading, Figure 6-13(a) and Figure 6-13(b). With the completion of steeple crushing, the
remaining crushing process becomes very similar to that of the chamfered angle-stiffener
specimen, Figure 6-13(c) to Figure 6-13(e). The only noticeable difference is the greater
extent of damage that occurred at the corner of the steeple specimen, Figure 6-13(e),
which is caused by the early damage initiation that occurs immediately upon loading.
Consequently, little if any buckling of the leg occurs: the primary deformation is leg
Figure 6-14 and Figure 6-15 show the axial and lateral strain fields, respectively,
for steeple angle-stiffeners at very low crosshead displacements of less than 2.5 mm.
Similar to the C-channel specimen, it can be seen that at a very low load of 0.37 kN, high
compressive strain developed at the tip of the steeple, Figure 6-14. As the displacement
increases, the high strain propagates towards the intact regions of the steeple. Similar
134
behavior was observed for the lateral tensile strains at the top of the steeple in Figure
6-15. As the steeple crushed further, lateral strains increased moderately along the entire
length of the corner, spreading out into the legs. This indicated the start of the legs
bending and splaying outwards. Both figures show relatively uniform strain fields.
iii) Hat-Stiffener
appeared along the corners of the chamfer with partial buckling of the flanges as it started
135
to crush gradually, Figure 6-16(a). After the chamfer was completely crushed, large
cracks developed along the outer plies of the entire specimen, which progressed along
their length with increasing crosshead displacement, Figure 6-16(c) and Figure 6-16(d).
The massive accumulation of matrix splits limited the degree of buckling, causing
primarily bending and outward and inward splaying of the plies. Post-test examination of
the specimens (after load removal) revealed that approximately half the plies splayed
The axial and lateral strain fields for the chamfered hat-stiffener are shown in
Figure 6-17 and Figure 6-18 respectively, at four selected load levels, up to
approximately 3.0 mm crosshead displacement, well beyond the peak load, Figure 6-2(a).
As the specimen is crushed, axial strain developed along the right side of the web, and
built up along the right side of the specimen after the peak load of 32.5 kN was reached,
Figure 6-17. This was followed by extensive crushing and bending at the top of the
specimen.
136
The lateral strain developed on the left and middle faces and spread to the right
face when the peak load was reached, Figure 6-18. After the chamfer is completely
crushed and the peak load was reached, high tensile strains developed along the corners
causing the formation and progression of axial cracks. These matrix splits extend along
the entire specimen, enabling further bending and splaying of the web and flanges.
Steeple End: Similar to the previous two cases, the steeple specimen began to
crush gradually with matrix splits occurring throughout the entire top surface in contact
with the compression platen, Figure 6-19(a) and Figure 6-19(b). Unlike the chamfered
137
specimen, no buckling of the flanges was observed. After the steeple was fully crushed,
the specimen continued to crush progressively with the remainder of the crushing process
being very similar to the chamfered specimen, Figure 6-19(c) to Figure 6-19(e).
Figure 6-20 and Figure 6-21 show the axial and lateral strain fields, respectively,
for the steeple hat-stiffener, at four selected load levels, up to approximately 2.2 mm
crosshead displacement, well below the peak load, Figure 6-2(b). As expected,
subsequent to steeple crushing high axial strain developed at the top of the middle web,
beginning at the two corners, Figure 6-20. At the same time moderate axial strains build
up in the web indicating the beginning of outward bending of the middle web. As the
steeple crushes, very high axial strains were observed at the top of the web, with the
The lateral tensile strains develop at the top corners of the web before progressing
to the center of the web, Figure 6-21. As the load increases, the strains in the web
increased and started to move towards the steeple. As the steeple is crushed further, high
strains are visible at the top of the web only. This indicates minor outward buckling at the
138
middle of the web which resulted in outward bending and splaying once the steeple was
completely crushed.
SEA using small and large thin-walled angle-stiffeners and C-channels, and hollow
square tubes, all fabricated with T700/2510 plane weave carbon fiber/epoxy prepreg
having a lay-up of [0/90]4S. The authors attempted to identify the influence of the curved
corners by varying the ratio of the curved surfaces to the total perimeter of the specimens.
(1)
where r is the radius of the corner, n is the fraction of a circle representing the
corners, and P is the total length of the cross-section perimeter. The data recorded in this
study are plotted together with the data reported in [42] in Figure 6-22. The results show
that both sets of data abide by the same linear relations proposed in [42], suggesting that
the more curved a specimen is the larger its energy absorbing capacity.
The C-channel with the lowest yields the lowest SEA, while the hat-stiffener
with the highest yields the highest SEA. Further, a previous study performed in [81,
Chapter 5] showed that circular tubes, fabricated from the same material but with a
different lay-up, can absorb significantly higher energy (approximately 126 kJ/kg). The
tube data ( =100%), fall on the same line on Figure 6-22 (not shown here for clarity).
Hence, this further confirms that the SEA is greatly dependent on the cross-sectional
Figure 6-22: Effect of the ratio of curved surfaces to the total perimeter on the SEA of specimens with
various cross-sections
140
To further investigate this trend in SEA, the final deformation of the C-channel,
angle-stiffener, hat-stiffener, and circular tube (from [81, Chapter 5]), after load removal,
are compared in Figure 6-23, in increasing order of SEA from (a) to (d). The C-channel,
Figure 6-23(a), has the least amount of damage to the inward-splaying plies as compared
to the other cross-sections. The inner plies of the flanges are severely damage, however,
the web stayed relatively intact with a few major matrix splits. On the other hand, the
angle- and hat-stiffeners show a much greater amount of damage and interaction between
the inward-splaying plies, Figure 6-23(b) and Figure 6-23(c), respectively. The circular
tube shows that most of the material (7 of 9 plies) is severely damaged and has been
compacted into the center of the tube, Figure 6-23(d). Hence, increasing interaction
amongst material that is crushed and forcing it towards the inside of the cross-sections
greatly increases the SEA of the specimen. Cross-sections that contained a longer
‘curved’ segments (i.e. have a larger ) promote the desired material interaction and
displacement towards the inner sides of the curved sections, thus increasing the overall
energy absorbed.
6.6. CONCLUSION
angle-stiffeners and hat-stiffeners) and two types of failure trigger mechanisms (chamfer
and steeple) were investigated to determine which approach is best in terms of decreasing
the initial peak load while increasing the sustained crush load and SEA. The results show
that the angle- and hat-stiffeners were the most effective at absorbing energy, having very
similar SEA values. The C-channel, on the other hand, has a much lower SEA value. The
steeple failure trigger was able to provide a lower initial peak load than the chamfer
failure trigger for the C-channel and hat-stiffener. For angle-stiffeners, both failure
The analysis of the failure processes and DIC strain fields highlighted several
important differences between the three cross-sections and the two failure trigger
mechanisms. During chamfer crushing, significant buckling occurs in the web and
flanges of the C-channel, and in the legs of the angle-stiffener. The hat-stiffener exhibited
some minor buckling in the flanges. Upon complete crushing of the chamfer, failure
would typically initiate at the corners of the cross section, in the form of matrix splits and
axial cracks due to the high local lateral strains. The cracks continue to grow in the
corners, along the length of the specimen, resulting in bending and splaying of the plies
in both, the inward and outward, directions. With the steeple specimens, failure that
initiates at the tip of the steeple causes the bending and splaying of the plies in both,
inward and outward, directions. The C-channel and hat-stiffeners exhibit some minor
web and flange buckling during the crushing process, however, no buckling occurs in the
the work done in [42], showed that the cross-sections that had a larger ratio of curved
sections resulted in a higher SEA due to the greater extent of damage and interaction
amongst the inward-splaying plies. Additional studies are warranted to determine the
effect of constituents, laminate configuration (where the optimal orientation of the off
axis plies depends on cross section size and geometry), wall thickness, and cross-
7.1. ABSTRACT
approach, were employed and results were compared with experiments. By carefully
behavior and material model response, the single-layer approach is capable of predicting
the initial failure peak load and the load-crosshead displacement curve, but provided no
insight into the failure process. For the multi-layer approach, a comprehensive
investigation was performed to determine the most effective and accurate method to
model the crushing behavior of the composite members. This included determining the
most efficient way of modeling the laminate and its interfaces, and the most effective
element size and formulation, contact definitions, time-step control, and material model.
The multi-layer approach captured the failure process and predicted the sustained crush
7.2. INTRODUCTION
absorbing devices that can be integrated into the subfloor structure of rotorcraft to
6
Siromani, D., Awerbuch, J. and Tan, T-M., “Modeling Methodologies for Simulating the Axial Crushing
Behavior of CFRP Members,” Submitted to: Finite Elements in Analysis and Design.
144
increase its energy absorption capability and improve the crashworthiness of the structure
[23]. The high specific energy absorption (SEA) of composites is a result of the
matrix splitting, fiber-matrix debonding, and delamination, e.g. [24-28]. As part of this
study, a two-part experimental study was conducted to investigate the energy absorption
compression. The first part of the study focused on the crushing behavior of circular
tubes. Emphasis was placed on determining the optimal failure trigger mechanisms,
including chamfered-ends, two types of crush-caps (forcing inward folding and outward
splaying, respectively, of the material), and their combinations, to reduce the initial peak
load and to increase SEA [81, Chapter 5]. The second part of the study focused on
investigating the crush behavior of structural members having open cross-sections that
are more prevalent in the aerospace industry, including C-channels, angle-stiffeners, and
hat-stiffeners, with two different failure trigger mechanisms, namely, chamfered-ends and
Due to the high cost of conducting experimental studies, there is a need for
composite members. Such models could serve as design tools to optimize the energy
absorption capacity of composite members when subjected to crush loading. There have
been several attempts to develop explicit finite element models, with varying degrees of
success, for circular tubes [57-60], hollow square tubes [59-63], C-channels [61,62],
using either solid or shell elements. Using shell element models requires less computation
145
time and they are more widely used to model the axial crushing of composite members.
The laminate can be modeled using either a single layer or multiple layers of shell
elements. In the ‘single-layer’ model the laminate is modeled by using a single layer of
shell elements, with each ply being represented by an integration point in the thickness
direction. In the ‘multi-layer’ model, the laminate is modeled using multiple layers of
shell elements, each layer may include either a single ply, or a group of plies, and the
layers are tied together using a tiebreak contact definition. Grouping plies together can
The single-layer modeling approach has been used in a number of recent studies
studying the crushing of thin-walled square tubes [58], it was shown that the single-layer
approach was capable of accurately capturing the local wall buckling and unstable
collapse. However, the approach could not depict the progressive failure process. By
carefully adjusting the values of key material and numerical parameters in the material
model (e.g., the failure strain parameters in MAT 54) and the contact characteristic
between the end of the test specimen and the loading fixture, the single-layer model was
able to yield good correlation with experimental load-displacement curves for the cross-
sections studied in [61,62]. However, the single-layer representation cannot depict the
approximately capturing the progressive failure process for tubes undergoing progressive
crushing e.g., [57-60]. However, this approach has not always yielded good correlation
hollow tube was simulated in [57] using LS-DYNA with MAT 54 material model. A
good agreement with the experimental load-displacement was achieved, but the
simulation showed significant local buckling of the tube, rather than the brittle failure
carbon fiber reinforced plastic (CFRP) tube was performed in [58]. The multi-layer finite
element model was able to accurately predict the peak load, but the overall SEA was
underestimated by 33%. The difference in SEA was attributed to the model’s inability to
reproduce the formation of debris wedged between the fronds of the tube’s wall due to
excessive element deletion. In [59,60], the simulations of the crushing behavior of hollow
glass-polyester circular and square tubes were performed using the multi-layer approach;
results did not agree well with the experimental load-displacement curves, deformation,
and failure behavior due to the model’s inability to reproduce the axial matrix cracks
the element edges in the axial direction of the tubes were introduced to simulate the
mechanics model (CODAM) was used in studying the behavior of braided composite
tubes [64], with multiple layers of shell elements and a tiebreak contact definition to
capture delamination. A “debris wedge” model formed between the delaminated surfaces
of neighboring plies during the crushing process was incorporated to improve the
accuracy of the simulation. While this approach can induce ply splaying in the
involving multiple failure processes. Further, it was noted in [66] that some of the
which are not readily available. As a result, these parameters have to be obtained by
woven carbon fiber epoxy prepreg made hat-stiffener was performed using multi-layer
approach in [67]. It was shown that increasing Mode I and Mode II energy release rates
for the cohesive interface yielded better correlation with experimental results than those
The primary focus of this paper is on the finite element modeling and simulations
performed to capture the crushing behavior of circular tubes and open cross-section
members. Two different modeling approaches using LS-DYNA were employed, namely,
the single-layer approach and the multi-layer approach. For each approach, the setup of
the model is discussed in detail followed by examples of the results obtained from the
SEA, and crush behavior. Details of the experimental results have been reported in [81,
Chapter 5] and [82, Chapter 6] and their comparison with multi-layer models results have
been reported in [79, Chapter 8] and [80, Chapter 9], respectively. This paper provides
The test specimens were fabricated using Hexcel IM7/8552 Graphite/Epoxy pre-
preg. All specimens were 101.6 mm long and the open cross-section specimens were
fabricated with the same cross-sectional area. Lay-up sequences and specimen
dimensions are listed in Table 7-1 and shown in Figure 7-1. The circular tubes were self-
supported while the open cross-section specimens were supported by a 25.4 mm thick
potted base, Figure 7-2, to ensure stability. All specimens included a failure trigger
148
mechanism to initiate progressive failure during the crushing process. The circular tubes
were tested with chamfers and crush-caps [81, Chapter 5], while the open cross-sections
were tested with chamfer and steeple ends [82, Chapter 6].
All tests were carried out under quasi-static axial compression at a crosshead
50.8 mm, which provided sufficient crush data to calculate the SEA. The full-field
deformation and strain fields on the surfaces of selected specimens were also recorded
using Digital Image Correlation (DIC). The experimental results for the circular tubes are
described in [81, Chapter 5] and the open cross-sections in [82, Chapter 6].
The single-layer modeling approach used in [61,62] was adopted for this study. A
single layer of shell elements was used to represent the laminate. Individual plies were
modeled using through-the-thickness integration points, with one integration point per
ply in the thickness direction. Each integration point was assigned a ply thickness and an
orientation corresponding to the stacking sequence listed in Table 7-1. The 45° chamfer
was approximated by reducing the number of plies in the first row of elements to three
plies, Figure 7-3, with the height of the first row being the same as that of the chamfer.
The steeple trigger was modeled by making two 15° cuts to the top of the model, Figure
7-3, to realistically represent the test specimen. All specimens were modeled using
rectangular elements of size 1.27 x 1.27 mm, resulting in approximately 6,400 elements
for the chamfered tube model, 5,500 and 4,800 elements for the chamfered and steeple C-
channel, respectively, and a similar number of elements for the other two open cross-
sections.
For the single-layer models, the four-node Belytschko-Tsa shell element (type 2
in LS-DYNA [73]) with single integration point and hourglass control (type 1, hourglass
coefficient = 0.1) to suppress zero energy modes was used. This is a computationally
150
efficient element formation that yields accurate results as long as no zero energy
Figure 7-3: Representations of the trigger mechanism in the single-layer finite element models of (a)
circular tube with a chamfer, and (b) C-channel with a steeple.
behavior of the specimens. MAT54 is a progressive failure model that uses the Chang-
Chang failure criterion to determine failure of each ply (associated with an integration
point) [73]. This model allows the user to create a local material coordinate system to
specify the orientation of each ply. There are 21 parameters in MAT54 that need to be
specified; 15 of which are physical parameters and six are numerical parameters [73]. Of
the 15 physical parameters, 10 parameters are material constants the values of which
XT XC YT YC SL
2326.2 MPa 1200.1 MPa 62.3 MPa 199.8 MPa 92.3 MPa
The remaining five physical parameters are the tensile and compressive failure
strains (element deletion strains) in the fiber direction (DFAILT and DFAILC), the matrix
and shear failure strains (DFAILM and DFAILS), and the effective failure strain (EFS).
The six numerical parameters can be adjusted to yield desired material behavior. Based
on an extensive parametric study, it was concluded that of these six parameters, the crash-
front element softening parameter (SOFT) is of key importance to this study. This
Details are discussed in the next section, along with the results.
A fixed boundary condition was assigned to the nodes along the flat end of all
finite element models. For single-layer models, only one contact definition between the
loading platen and the chamfer/steeple end of the specimen is needed. The contact
definition used to model this interaction was contact rigid nodes to rigid body, following
the approach outlined in [61,62]. The advantage of using this contact definition is that it
distributes the contact force over multiple rows of elements, thus preventing the load
from dropping to zero after the deletion of each row of elements. This is accomplished by
allowing the nodes of the tube to penetrate into the platen by a specified distance, Figure
152
7-4. As the nodes penetrate, they are resisted by the forces calculated based on a user
Platen
Penetration
Figure 7-4: Penetration of the platen by tube elements according to the load-penetration curve in the
contact definition (Figure 7-5).
The initial load-penetration curve used in this contact definition was estimated
from the compressive failure load and cross-sectional area of the element. It was
subsequently fine-tuned to ensure that at least two rows of elements penetrated the platen
prior to any element deletion. Figure 7-5 shows the selected load-penetration curve used.
4
Contact Load (kN)
0
0 1
2 3 4
Penetration (mm)
Figure 7-5: Final load-penetration curve for the contact definition used in single-layer model.
153
The modified load-penetration curve prevented a sudden drop of the load to zero,
but still resulted in a very noisy load-displacement behavior that resembled a saw-tooth
pattern, Figure 7-6. Various filtering schemes were investigated and it was determined
that the SAE 1000 Hz filter resulted in a reasonable load-displacement curve, Figure 7-6.
120
Unfiltered
100 Filtered (SAE 1000Hz)
80
Load (kN)
60
40
20
0
0 10 20 30
Crosshead Displacement (mm)
Figure 7-6: Unfiltered vs. filtered load-displacement curves obtained from the single-layer chamfered
circular tube model. The SAE 1000 Hz filter was used to obtain the filtered data.
i) Circular Tubes
As mentioned in the previous section, there are five physical parameters (failure
strains) and six numerical parameters in MAT54 whose values need to be determined
the physical parameter DFAILC (fiber compression failure strain) had the greatest effect
on the value of initial peak load while the numerical parameter SOFT (crash-front
element softening parameter) had the greatest effect on the value of sustained crush load,
which determined the value of SEA. By adjusting these two parameters it is possible to
obtain simulation results that agree with experiment. Table 7-3 and Table 7-4 show the
154
effect of varying these two parameters on the initial peak load, sustained crush load and
SEA. Parameters DFAILT, DFAILM and DFAILS were found to have a marginal effect
on the results and were kept constant at arbitrarily selected values of 0.02, 0.02 and 0.03,
respectively. It was found that simulations with DFAILC = -0.0075 and SOFT = 0.22
yielded the load-crosshead displacement behavior for chamfered tubes that matched very
well with experimental data, as shown in Figure 7-7. The SEA obtained from the
simulation (129 kJ/kg) also compared well with the experimental SEA (127 kJ/kg).
Table 7-3: Results of the parametric study showing the effect of DFAILC in MAT54 on the peak load,
crush and SEA of the circular tube.
Peak Load Crush Load SEA
DFAILC
(Num/Exp %) (Num/Exp %) (Num/Exp %)
-0.0100 158.7% 162.5% 161.3%
-0.0050 61.4% 83.1% 83.9%
-0.0060 83.2% 83.1% 83.9%
-0.0075 103.3% 83.1% 83.9%
-0.0080 105.1% 83.1% 83.9%
Table 7-4: Results of the parametric study showing the effect of SOFT in MAT54 on the peak load, crush
and SEA of the circular tube.
Peak Load Crush Load SEA
SOFT
(Num/Exp %) (Num/Exp %) (Num/Exp %)
0.22 103.3% 104.7% 101.5%
0.25 103.3% 118.2% 120.1%
0.28 103.3% 131.2% 132.5%
0.30 103.3% 158.5% 160.3%
fitting the MAT 54 parameters to the experimental data. Further, it could not replicate
the deformation and failure processes of the composite structure, such as matrix splitting,
155
this approach could not be used to simulate the crushing behavior of tubes having crush-
cap failure trigger mechanism becuase the contact definition could not capture the effect
of crush-caps.
50
Experiment
40 Simulation
Load (kN)
30
20
10
0
0
10 20 30 40 50
Crosshead Displacement (mm)
Figure 7-7: Single-layer simulation vs. experimental load-crosshead displacement curve for the circular
tubes having a chamfer trigger.
MAT 54 parameters could be used to simulate the crushing behavior of the three open
was conducted to determine the required parameters for the C-channel specimen. Similar
to the case of the circular tubes, it was found that DFAILC and SOFT were the only two
parameters that have noticeable effects on the crushing behavior. By adjusting the values
of the DFAILC and SOFT parameters it was possible to obtain simulation results that
correlated well with the experimental data for the C-channel with a chamfer trigger,
Figure 7-8(a). Parameters DFAILT, DFAILM and DFAILS were kept constant at values of
0.02, 0.013 and 0.03, respectively. Table 7-5 and Table 7-6 show the effect of varying
156
these two parameters on the initial peak load, sustained crush load and SEA. The optimal
values obtained for DFAILC and SOFT were -0.008 and 0.10, respectively.
In [61,62] it was found that the by applying the same DFAIL parameters to all
cross-sectional geometries and by varying SOFT, good correlation between the simulated
between the SEA and SOFT was observed, which could be related to the shape of the
cross-section [61,62]. A similar approach was used in this study: the DFAIL parameters
determined from the C-Channel parametric study were applied to the angle and hat
stiffener models. The SOFT parameter was then varied to match the sustained crush load
and SEA in the simulations to the experimental results. Good correlation was obtained
with the crush load and SEA, however, the initial peak loads were not predicted
accurately, Table 7-9. This indicated that the DFAILC parameter needed to be modified
The reason for the differing results obtained in this study as compared to those
reported in [61,62] is a direct consequence of the contact definition used in the single-
layer approach. As explained earlier, the load applied to the nodes is directly proportional
to their penetration distance into the loading platen. Therefore, for cross-sections with the
same area, as was the case for this study, the load applied by the contact definition will
always be the same, regardless of the shape of the cross-section. In other words, a
different set of SOFT and DFAILC values are needed for different cross-sectional
geometries in order to accurately predict the peak load, sustained crush load, and SEA.
157
Table 7-5: Results of the parametric study showing the effect of DFAILC in MAT54 on the peak load,
crush and SEA of the C-channel with a chamfer.
Peak Load Crush Load SEA
DFAILC
(Num/Exp %) (Num/Exp %) (Num/Exp %)
-0.013 73.1% 158.5% 145.6%
-0.002 77.7% 174.4% 160.9%
-0.004 90.1% 197.5% 182.5%
-0.006 96.2% 208.6% 192.7%
-0.008 102.3% 212.8% 197.0%
Table 7-6: Results of the parametric study showing the effect of SOFT in MAT54 on the peak load,
crush and SEA of the C-channel with a chamfer.
Peak Load Crush Load SEA
SOFT
(Num/Exp %) (Num/Exp %) (Num/Exp %)
0.05 104.5% 50.6% 54.2%
0.10 105.1% 108.7% 106.1%
0.15 103.1% 158.5% 149.4%
0.20 100.8% 194.2% 180.5%
A similar parametric study was performed for the C-channel having a steeple
trigger, Table 7-7 and Table 7-8. For this case, it was not possible to establish a similar
level of correlation as with the chamfered specimens. The reason is that the steeple lacks
the weakened first row of elements (present in the chamfered case) to take advantage of
the SOFT parameter. Figure 7-8(b) shows a comparison of the simulation and
experimental load-displacement curves. This model was very sensitive to minor changes
to the DFAILC and SOFT parameters, often resulting in unstable collapse. When the C-
channel stiffener parameters were applied to the hat and angle stiffener models, poor
correlation was obtained, Table 7-9. Hence, similar to the chamfered case, al cross-
sections would require a separate set of parameters in order to obtain accurate and stable
results.
158
Table 7-7: Results of the parametric study showing the effect of DFAILC in MAT54 on the peak load,
crush and SEA of the C-channel with a steeple.
Peak Load Crush Load SEA
DFAILC
(Num/Exp %) (Num/Exp %) (Num/Exp %)
-0.013 89.7% 104.8% 104.2%
-0.005 71.3% 83.0% 81.8%
-0.010 81.6% 100.0% 99.4%
-0.015 91.5% 106.6% 106.0%
-0.020 94.9% 111.5% 110.8%
Table 7-8: Results of the parametric study showing the effect of SOFT in MAT54 on the peak load, crush
and SEA of the C-channel with a steeple.
Peak Load Crush Load SEA
SOFT
(Num/Exp %) (Num/Exp %) (Num/Exp %)
0.05 18.7% 16.9% 17.4%
0.10 42.4% 45.4% 45.2%
0.15 56.7Z% 63.9% 63.2%
0.20 65.6% 75.6% 74.6%
(a) (b)
40 40
Experiment Experiment
Simulation Simulation
30 30
Load (kN)
Load (kN)
20 20
10 10
0 0
0 10 20 30 0 10 20 30
Crosshead Displacement (mm) Crosshead Displacement (mm)
Figure 7-8: Single-layer simulation vs. experimental load-crosshead displacement curve for the C-
channels having (a) a chamfer trigger, and (b) a steeple trigger.
Table 7-9: Correlation between the experimental and simulation results for each cross-section with (a) a
chamfer trigger, and (b) a steeple trigger.
(a) (b)
Peak Crush Peak Crush
SEA SEA
Specimen Load Load Specimen Load Load
[Error %] [Error %]
[Error %] [Error %] [Error %] [Error %]
C-Channel 3.9% 6.4% 7.2% C-Channel 21.9% 17.8% 0.4%
Angle Angle
20.1% 5.1% 5.4% 66.7% 58.1% 64.0%
Stiffener Stiffener
Hat Stiffener 20.5% 3.1% 1.4% Hat Stiffener 30.3% 24.6% 33.3%
159
In the multi-layer approach the laminate was divided into multiple layers of shell
elements, each layer may contain multiple plies to improve computational efficiency, and
layers were tied together using a tiebreak contact definition. All layers contain an equal
number of plies to maintain equal spacing between the individual layers. The circular
by three layers of shell elements, with the inner [+15/-15/+15] plies, the middle [03]
plies, and the outer [-15/+15/-15] plies represented by the inner, middle, and outer layers,
and were modeled by five layers of shell elements, with each layer representing two plies
(i.e., layer 1: [02], layer 2: [+45/-45], layer 3: [02], layer 4: [-45/+45], and layer 5: [02]).
It is noted that angle stiffeners are twice as thick as the other open cross-sections (to
maintain the same cross-sectional area in all three open cross-sections) and were modeled
Each ply was represented by a single integration point through the thickness of
the shell element layer. The 45° chamfer was modeled by staggering the length of each
shell layer such that the inner layer’s first row of elements represented the start of the
chamfer, and the outer layer represented the end of the chamfer, Figure 7-9(a). The
steeple trigger was modeled by making two 15° cuts to the one end of the model, Figure
7-9(b).
160
(a) (b)
Figure 7-9: Representations of: (a) circular tube with a chamfer and (b) C-channel with a steeple failure
trigger mechanisms in the finite element models.
A mesh density study performed on the circular tube concluded that a maximum
element size of 0.635 x 0.635 mm was required in order to accurately simulate the
initiation and progress of damage. The same element size was used for the curved
surfaces in all cross-sections (i.e., tubes and corners of open cross-section models). For
the flat sections (i.e., web and flanges) in the models, it was determined that an element
size of 1.27 x 1.27 mm was sufficient. This resulted in the tube model having
approximately 77,000 elements and the open cross-section models having approximately
55,000 elements. It should be noted that for the steeple specimens the auto-meshing
process resulted in an asymmetric mesh, which may have affected the crushing process.
single-integration point (type 2 in LS-DYNA, [73]) was chosen to model the specimens.
Various hourglass control options, with both viscous and stiffness formulations, Table
7-10, were investigated to suppress zero energy deformation modes that accompany the
hourglass deformation modes. Therefore, either the mesh would need to be refined to
further investigate the use of this element formulation, or the fully-integrated elements
161
would need to be used. Since both options would results in a significant increase in
computation time, to ensure accurate results, fully-integrated elements (type 16, [73])
setup. The circular tubes did not require any boundary conditions as they were placed
standing upright between the loading platen and the base of the testing machine, with
corresponding contact definitions. The open cross-sections required the nodes along the
flat end of the models to be fixed in all degrees of freedom to simulate the potted base
The loading platen of the testing machine was modeled as a rigid surface and the
interaction between the loading platen and the specimens was modeled using a surface to
surface contact definition (contact automatic surface to surface). The contact between the
base and the circular tube was defined using a “rigid wall” contact definition
(rigidwall_planar), which does not allow node penetration. For the simulation of the
chamfered tube with a crush cap, the crush cap was assumed to be rigid and was defined
using the same surface-to-surface contact definition as the platen-tube contact. To prevent
self-penetration of any of the shell layers, a single surface contact definition (contact
162
automatic single surface) was added to all layers. A friction coefficient of 0.3 was used
between the specimens and the Instron loading platen and base.
) was used for all specimens. However, in the simulation, this low loading rate
would result in a prohibitively long computation time due to the small time-step required
by explicit time integration codes. There are two methods to address this issue, namely,
time-scaling method and mass-scaling method [85,89]. Both methods were evaluated in
i) Time-scaling
In the time-scaling method the load is applied at a much higher rate, thus reducing
the total simulation time. However, in a quasi-static simulation, it is important that the
load is applied in a manner that would yield a minimal inertial effect on the results, i.e.,
the ratio of the kinetic energy to the internal energy must be reasonably small [57,85].
Four different loading rate functions were investigated, Figure 7-10(a). These include 1)
a sinusoidal loading rate function to slowly ramp up the loading rate from 0 m/s to 2.5
m/s [85], 2) a constant loading rate of 0.65 m/s [57], 3) a step function, from a 0.65 m/s
rate for the first 0.01 seconds to a 2.5 m/s rate thereafter [67], and 4) a step function, from
a 1.3 m/s rate for the first 0.02 seconds to a 2.5 m/s thereafter [67]. For all cases the ratios
of the kinetic energy to the internal energy were less than 10% upon initial contact and
less than 5% throughout the remainder of the crushing process, thus indicating that the
inertial effects are minimal. Comparing the load-displacement curves among the four
163
loading functions show that the results were not sensitive to the first three loading rate
functions, Figure 7-10(b). The fourth function, however, appears to have a more
noticeable effect on the load-displacement curve. Hence, the third function (step function
increasing the loading rate from 0.65 m/s to 2.5 m/s) was selected as it yielded the least
2
Load (kN) 25
1.5 20
15
1
10
0.5
5
0 0
0.010 0.02 0 5 10
Time (sec) Crosshead Displacement (mm)
(a) (b)
Figure 7-10: Time-scaling investigation showing: a) loading rates used, and b) the resulting load-crosshead
displacements.
ii) Mass-scaling
a non-physical mass to selective elements in the structure. There are two options
up the masses of the selective parts; and 2) automatic mass-scaling with a user specified
minimum time-step. Both options were investigated in this study. For the first option a
scaling factor of 100 was used (for comparison, the mass was scaled by a factor of 1000
164
in [57,58]), and for the second option the minimum time-step was set to 4x10-8 sec,
which was more than double the average time-step through the crushing process of the
un-scaled simulation. Figure 7-11 shows the load-crosshead displacement curves of the
two mass-scaled simulations compared to the un-scaled simulation. Results of the first
option (with a scaling factor of 100) is seen to be significantly different from the un-
scaled simulation, indicating a substantial effect of the added nonphysical mass. Results
of the second option are very similar to that with no mass-scaling and the time-step was
affected only after 3 mm of crushing, yielding a slightly lower crush load from that point
forward. While both options resulted in shorter computation times (10x and 2x faster than
the un-scaled simulation), the adverse effect of the added mass could not be ignored.
Therefore, mass-scaling was not used in any of the simulations in this study.
No Mass Scaling
Mass x100
Automatic Mass Scaling (dt = 4e-8 sec)
140
120
100
Load (kN)
80
60
40
20
0
0 1
2 3 4 5 6 7 8 9 10
Crosshead Displacement (mm)
Figure 7-11: Mass-scaling investigation showing the resulting load-crosshead displacements.
simulation of delamination at the interface between adjacent shell element layers. Two
165
tiebreak formulations were investigated for the purpose of this study; namely, tiebreaks
zone formulation [86] (option 8 and 11 in LS-DYNA, respectively [73]). The former,
which requires interlaminar normal and shear strengths and a critical distance to interface
failure as input parameters, has been used to model delamination in crush simulations
e.g., [63,64]. However, the optimal selection of the critical failure distance parameter has
not been thoroughly studied in open literature. The latter, the formulation of which is
similar to that of cohesive zone elements, requires the standard input parameters, such as
interlaminar normal and shear strengths, fracture toughness under pure Mode I and Mode
II loading, interfacial stiffness for normal and shear modes, and a parameter for the
power-law or the Benzeggagh-Kenane law (B-K law) that describes crack propagation
[73]. This formulation has been used to model delamination in ballistic impact [86], but
not for crush simulations. A description of the model setup using each formulation is as
follows:
i) Option 8
In this tiebreak formulation, damage initiates when the stresses on the interface
| | | |
( ) ( ) (1)
in which and are the normal and shear stresses acting at the interface, and
NFLS and SFLS are the normal and shear strengths of the tie, respectively. Once the
damage has initiated, the two surfaces begin to separate and the interfacial stresses are
then scaled down as a linear function of the separation distance. The critical distance,
166
denoted by CCRIT, at which failure occurs (i.e., deletion of tiebreak and advancing of
(2)
where:
√ ( ) | | (3)
and is the energy released due to the failure of the tiebreak interface. A
sensitivity study was conducted, to determine the relative effect(s) of Mode I and Mode II
on the tiebreak failure process, using the circular tubes. It was determined that for
simulating the crushing of composite tubes, Mode II fracture is the dominant mode of
failure during the tie-break failure process. Thus, to simplify the simulations, a pure
(4)
where is the Mode II critical energy release rate, the value of which
was obtained from [87] for IM7/8552, as shown in Table 7-11. Any lower values
will cause the ties to break prematurely, yielding a high rate of delamination progression.
and typically requires a very fine mesh. To improve the computational efficiency, the
methodologies discussed in [88] were adopted. The proposed solution involves lowering
the interlaminar strengths whilst keeping the fracture toughness constant in order to adapt
167
the length of the cohesive zone for a given mesh size. While this approach is intended to
be used with cohesive zone models, it can also be applied to the tiebreak formulation
used here since the tiebreaks follow a traction-separation law similar to that used in
cohesive zone formulations. The required interfacial strength ̅ can be calculated from
[88]:
̅ √ (5)
energy release rate, is the desired number of elements in the cohesive zone, and is
the mesh size in the direction of the delamination progression. The minimum number of
elements required for the cohesive zone has not been well established. Various studies
have used anywhere from two elements to 10 elements [88]. Although no cohesive
elements are required for the tiebreak formulation applied herein, the same concept used
for cohesive elements has been applied to the tiebreaks. A separate sensitivity study
conducted here concluded that five elements were sufficient to simulate the propagation
of delamination. Hence, with , Equation (5) was used to solve for the new NFLS
and SFLS values. The new SFLS was substituted into equation (4) to calculate the
CCRIT value.
Since the laminate of the open cross-sections was modeled as five and ten layers
of shell elements with each layer representing two plies, the tiebreak contact was defined
only between these shell layers, rather than between individual plies. However, in reality,
delamination could occur along any, if not all, ply interfaces during specimen crushing,
as was observed experimentally [81,82, Chapters 5 and 6]. In order to account for the
168
energy dissipated by these additional for delamination interfaces, CCRIT was scaled by
the ratio of the number of ply interfaces to the number of tiebreak interfaces ,
defined as:
(6)
[81,82, Chapters 5 and 6]) that delamination occurred among all plies. The values used
and calculated in Equations (4), (5) and (6) are listed in Table 7-11. It should be noted
that was calculated using the smaller element size (0.635 mm) for the circular
tube and the larger element size (1.27 mm) for the open cross-section specimens
ii) Option 11
For this tiebreak formulation the critical failure distance, CCRIT, is not required
because the energy release rates, and , are directly input into the simulation.
However, these energy release rates still need to be scaled by the same ratio that CCRIT
was scaled by, Equation 6, in order to account for the delamination occurring at the
interfaces that are not accounted for in this model. Also, the new values calculated for
NFLS and SFLS using Equation 5 were required to address mesh size dependency
discussed earlier. Further, this formulation requires tiebreak interfacial stiffness, , for
normal and shear modes, which can be calculated from [88] as:
(7)
169
is the thickness of the adjacent sub-laminate, in this case, the shell element layer (i.e.,
the thickness of three plies for the circular tube and two plies for the open cross-sections).
The resulting interfacial stiffness for each cross-section is listed in Table 7-11. Further,
simple DCB simulations were performed with this tiebreak formulation and it was found
that the simulations were very sensitive to the contact viscous damping coefficient and
the part stiffness damping coefficient [73,89]. A stable solution was achieved by using a
viscous damping coefficient of 1.0 and a part stiffness damping coefficient of 0.02. A
comparison of the results obtained from the two tiebreak formulations is discussed after
In the multi-layer modeling approach, two material models, MAT54 and MAT58,
progressive failure model that uses the Chang-Chang failure criterion to determine the
failure of each ply (associated with an integration point) [73]. This material model has
been studied extensively to model the crushing process of brittle composites e.g., [57,58].
Lubliner-Taylor (MLT) damage model and the Hashin failure criteria [73]. This model
allows for stiffness degradation based on the accumulation of damage such as micro-
cracks and cavities. MAT58 has been studied extensively to model the crushing process
of braided composites [66]. The material properties used to define both materials models
are listed in Table 7-2. A description of the model setup using each material model is as
follows:
i) MAT54
that need to be determined parametrically, most importantly the tensile and compressive
failure strains in the fiber direction, the matrix and shear failure strains, and the effective
failure strain. A parametric study was conducted to determine the optimal values of the
unknown parameters for the multi-layer modeling approach. It was found that the only
MAT54 parameter that needed to be adjusted was DFAILM, which is the failure strain in
the matrix direction [73]. Adjusting the value of DFAILM enabled matrix splitting to
occur, as observed in the experiments [81,82, Chapters 5 and 6]. DFAILM values ranging
from 5% to 30% were analyzed and it was determined that, for the chamfered tube, a
value of 10% strain yielded the most accurate representation of the crushing process. This
DFAILM value was used for the elements in the corner regions of the open cross-section
specimens as well since they had the same element size as that of the circular tube model.
To determine the optimal DFAILM value for the flat surface regions where elements are
twice the size of the corner elements, a separate parametric study was conducted. It was
found that a DFAILM value of 5% strain would provide the most accurate representation
ii) MAT58
The MAT58 material model uses the maximum effective strain criterion to
determine failure. An integration point (or ply) fails and ceases to carry any load if the
effective strain computed at that point reaches the maximum effective strain, defined by
the input parameter ERODS. All plies in an element must fail before the element is
deleted. A parametric study was performed to determine an ERODS values using the
chamfered tube having three layers of shell elements. It was found that, even with
ERODS values as high as 50% strain, the high compressive axial and hoop strains in the
elements representing the tube chamfer typically caused premature element deletion,
resulting in a catastrophic failure in the chamfer. Ultimately, it was determined that the
ERODS values for the inner two shell layers needed to be set to a much higher value (i.e.,
100%) to prevent catastrophic failure of the chamfer. The ERODS value for the outer
shell layer was set to 25% in order to enable matrix splitting to occur.
Further, MAT58 allows the user to specify a stress limit factor, SLIMXX, which
scales down the stress in any direction that has reached its respective strength, thus
achieving an elastoplastic like behavior in the damaged elements. A SLIMXX value of 1.0
gives an elastic-perfectly plastic like behavior until the failure strain (ERODS) is reached,
which is recommended for compression and shear [73]. In tension, a smaller SLIMXX
value is recommended and a value of 0.2 was selected for this study. Finally, MAT58
also allows the user to specify the strain at the strength for each direction. In this study,
this strain in each direction was calculated by dividing the strength by the elastic
modulus, effectively assuming a linearly elastic behavior before the strength was reached.
172
investigate the various modeling options and parameters discussed in the previous
section. In some cases, the options/parameters in question were evaluated using the C-
channel with a steeple failure trigger in order to determine their effect on a different
cross-section and failure trigger mechanism. The final modeling methodology was then
The results obtained from the two tiebreak options and the two material models,
detail. It is noted that the simulation results typically were very noisy and filtering
i) Filtering
The load-crosshead displacement curve of the chamfered tube shows three load
spikes, regardless of the tiebreak or material model used, occurring within the first 2 mm
of the crosshead displacement, Figure 7-12(a). These three load spikes correspond to the
crushing of the three shell element layers in the chamfered region. With the current three-
layer model, the curve could be smoothed using various data filtering schemes. Clearly,
the ‘smoothness’ of the curve depends on the degree of data filtering, noting that
excessive filtering will reduce the utility of the results. A range of filter frequencies
raging from SAE 180 Hz to SAE 1,000 Hz were studied and results indicated that SAE
300 Hz filter provided the best continuous initial load-displacement behavior that also
compared well with the experimental data, Figure 7-10(b). Similarly, the unfiltered
173
simulation results from the C-channel chamfer, show five load spikes, correspond to the
crushing of the five shell element layers in the chamfered region, Figure 7-13(a). It was
found that in this particular case an SAE 600 Hz filter provided the most accurate results,
Figure 7-13(b). It is noted that by modeling each individual ply as a separate shell
element layer, (hence with a much higher computational cost), the number of load spikes
would increase to nine (or ten for the C-channel) and the magnitude of each spike would
For the C-channel with a steeple failure trigger the load-crosshead displacement
curves did not require any filtering as the steeple geometry could be modeled accurately
Load (kN)
30 30
20 20
10 10
0 0
0 5 10 15 20 0 5 10 15 20 25
Crosshead Displacement (mm) Crosshead Displacement (mm)
(a) (b)
Figure 7-12: A comparison of: (a) the unfiltered and filtered load-crosshead displacement curves from a
multi-layer simulation for a tube with a chamfer failure trigger, and (b) the filtered simulation
results and the experimental load-crosshead displacement comparison for a tube with a
chamfer failure trigger.
174
Load (kN)
30 30
Load (kN)
20 20
10 10
0 0
50 10 15 20 25 0 5 10 15 20 25 30
Crosshead Displacement (mm) Crosshead Displacement (mm)
(a) (b)
Figure 7-13: A comparison of: (a) the unfiltered and filtered load-crosshead displacement curves from a
multi-layer simulation for a C-channel with a chamfer failure trigger, and (b) the filtered
simulation results and the experimental load-crosshead displacement comparison for a C-
channel with a chamfer failure trigger.
chamfered tube for each tiebreak formulation show very similar results in terms of the
initial peak load and sustained crush load, Figure 7-14(a). This indicated that either
option was capable of simulating delamination in the models and the computational cost
was the same for both options. Further, comparing the final deformation of the tubes with
the test specimen also showed that both options give similar results in terms of the extent
The tiebreak options were further investigated in the simulations of the crushing
of the C-channel with a steeple trigger. In this case, the initial peak load was very similar
for both options, but the sustained crush load predicted by option 11 was approximately
30% of that predicted by option 8, Figure 7-15(a). Comparing the deformation of the
simulation results with the experiments (at ~25 mm of crush) showed that option 11
caused more buckling in the web than option 8, yielding a lower sustained crush load,
175
the coupon level, i.e. correlating the simulation with experimental DCB and MMB tests.
50 Option 8
Option 11
40 Experiment (Avg)
Load (kN)
30
Experiment
20
10
0
0 2 4 6 8 10 12
Displacement (mm)
Option 8 Option 11
(a) (b)
Figure 7-14: Tiebreak investigation showing the comparison between: (a) the load-crosshead displacement
curves for the chamfered tube, and b) the final deformation of the tubes.
20 Option 8
18 Option 11
16
14
Load (kN)
12
10
Experiment
8
6
4
2
0
0 5 10 15 20 25
Crosshead Displacement (mm)
Option 8 Option 11
(a) (b)
Figure 7-15: Tiebreak investigation showing the comparison between: (a) the load-crosshead displacement
curves for the steeple C-channel, and b) the deformation of the steeple C-channel at ~23 mm
of crush.
176
As noted earlier, option 11 was very sensitive to various damping coefficients and
it is not possible to parametrically analyze their effect at the structural level. This must be
performed at the coupon level which is beyond the scope of this study. Therefore,
tiebreak option 8 was selected for the remainder of this study as it yielded the most
accurate results.
simulations of the chamfered tube using material models MAT54 and MAT58,
respectively, are shown in Figure 7-16(a). While the initial peaks loads predicted by both
models are almost identical and very similar to the experimental peak load, the sustained
crush load predicted by MAT58 model is approximately 35% less than that predicted by
MAT54 model. Comparing the deformation shows that both models correctly predict the
direction of splaying of each layer. However, more element deletions in the outer layer
were predicted by the MAT58 model as compared to the MAT54 model. In the inner two
layers, barely any element deletion was visible in the MAT58 model due to ERODS
being set to 100%. Hence, despite using a much higher failure strain for the inward-
folding plies, MAT58 model was not capable of simulating the accurate sustained crush
load. This could be due to the effective failure strain calculation that takes into account
all strain components. In the MAT54 model, elements were only allowed to fail in the
matrix direction, similar to what was observed in the experiments. Hence, a possible
improvement to the MAT58 model could be an option of allowing the user to specify a
separate failure strain in each direction. For the purpose of this study, the MAT54 model
50 MAT54
MAT58
40 Experiment (Avg)
Load (kN)
30
Experiment
20
10
0
0 5 10 15 20
Displacement (mm)
MAT54 MAT58
(a) (b)
Figure 7-16: Material model investigation showing the comparison between: (a) the load-crosshead
displacement curves for the chamfered tube, and b) the final deformation of the tubes.
In this study, finite element models were developed using LS-DYNA to simulate
the crushing behavior of composite stanchions. Two approaches were employed to model
the crushing process, namely, a single-layer approach and a multi-layer approach. The
curve of the circular tube and the C-channel with a chamfer trigger. This approach
parameters required by material model 54 (MAT54) to correctly fit the simulation results
parameters, DFAILC and SOFT, play a key role in predicting the initial peak load and
sustained crush load, respectively. An attempt was made to find a common set of
parameters that could be used across the different cross-sections. Results indicated,
178
however, that each cross-section requires a separate DFAILC and SOFT definition. This
method proved to be even less successful when applied to specimens with a steeple
trigger mechanism due to the lack of the chamfer row of elements to effectively use the
SOFT parameters. Not only the simulations of steeple specimens were mostly unstable
numerically, but also it was not possible to find a common set of parameters to use
determine the most accurate method to model the crushing behavior of the composite
members. This included determining the most efficient way of modeling the laminate and
its interfaces, and the most effective element size and formulation, contact definitions,
time-step control, and material model. It was determined, in order to maintain consistent
spacing between each layer, the laminate needed to be divided into an equal number of
plies per shell element layer. Fully-integrated, four-node elements were used to generate
the mesh, with a finer mesh required for curved surfaces (circular tube and corners of
open cross-sections) that the flat surfaces (web and flanges of open cross-sections). The
quasi-static experimental loading rate was modeled using a two-step function to ramp us
the loading rate while minimizing the dynamic effects. In order to simulate delamination,
two tiebreak options were investigated and it was determined that option 8 (a simple
an energy based approach. Finally, two material models were investigated (MAT54 and
MAT58) and it was determined that MAT54 provided the best results in terms of the
performed in order to obtain the values of the unknown parameters. It was determined
179
that DFAILM was the only parameter that needed to be adjusted in order to obtain good
correlation with the experimental results. A parametric study was performed to determine
an optimal set of input parameters for the circular tube and the C-channel. These
parameters were then used across the remaining case studies [79,80, Chapters 8 and 9].
Overall, the results obtained were highly satisfactory with the approach being able to
accurately replicate the crushing process observed in the experiments. In most cases, the
8.1. ABSTRACT
Finite element models were developed using LS-DYNA to simulate the crushing
chamfered end with a crush cap failure trigger mechanisms. Laminated tubes were
modeled by using multiple layers of shell elements with tiebreak contact definitions
between layers. Each layer contained multiple plies. Delamination was simulated by
failure of the tiebreaks, which was determined using an energy-based approach. Material
model 54 (MAT54) in LS-DYNA was used to define the material ply properties. Results
were compared with experiments in terms of the load-displacement curve, specific energy
absorption (SEA), failure process, and surface strain distribution. The multi-layer
approach was capable of accurately predicting the sustained crush load, SEA, damage
8.2. INTRODUCTION
focus to improve occupant survivability in the event of a crash [1]. The subfloor structure
dissipating energy during the impact event. Several studies have been performed over the
7
Siromani, D., Awerbuch, J. and Tan, T., “Finite Element Modeling of the Crushing Behavior of Thin-
Walled CFRP Tubes under Axial Compression,” Submitted to: Composites Part B: Engineering.
181
past three decades to investigate the dynamic response of aircraft structures and the
of these studies showed the need for energy absorbing devices, integrated with the
subfloor structure, to mitigate the impact energy. Composite materials are considered as
possible candidates for such integrated energy absorbing devices due to their high
strength-to-weight ratio and their high specific energy absorption (SEA) capacity, e.g.
[24-28], resulting from their particular failure characteristics during the crushing process.
The maximum crush load and SEA depend greatly on the selected cross-sections, wall
axial compressive (crush) loads [81, Chapter 5]. Tubes of circular cross sections with
three types of failure triggering mechanisms, (i.e., a chamfered end, a flat end attached to
a crush cap, or a combined chamfered end and crush cap) to initiate progressive failure,
were investigated to identify the optimal configuration that would result in the lowest
initial peak load while providing the highest possible SEA. It was shown in [81, Chapter
5] that a combined chamfered end with an inward-folding crush cap yielded the highest
Due to the high cost of conducting experimental studies, there is a need for
members. There have been several attempts to develop explicit finite element models,
with varying degrees of success, for circular tubes [57-60], square tubes [59-63], C-
182
structures can be modeled using either solid or shell elements. In general, shell element
models require less computation time and are more widely used to model the axial
crushing of composite members. The laminate can be modeled using either a single layer
or multiple layers of shell elements. In the ‘single-layer’ model the laminate is modeled
by using a single layer of shell elements, with each ply being represented by an
integration point in the thickness direction, as was done for C-channels, angle-stiffeners,
and hollow square tubes in [61,62]. In the ‘multi-layer’ model, the laminate is modeled
by using multiple layers of shell elements, each layer may represent either a single ply or
a group of plies, and the layers are ‘tied’ together using a tiebreak contact definition.
The single-layer modeling approach has been used in a number of recent studies,
to simulate the crushing behavior of various composite members. For example, in the
case of thin-walled square tubes, this approach was capable of accurately capturing the
local wall buckling and unstable collapse; however, it could not depict the progressive
failure process [58]. By carefully adjusting the values of key material and numerical
parameters in the material model (e.g., eleven parameters in MAT 54), and the contact
definition between the end of the test specimen and the loading fixture, the single-layer
model was able to yield good correlation with experimental load-displacement curves for
appropriate for describing the failure and crush behavior. The utility of the single-layer
The multi-layer modeling approach, on the other hand, can better capture the
failure process for tubes undergoing progressive crushing e.g., [57-60]. However, this
183
approach has not always yielded good correlation with the experimental load-
displacement curves. The crush behavior of a hollow tube was simulated in [57] using
LS-DYNA with MAT 54. The simulation results agreed quite well with the experimental
load-displacement curves, but showed significant local buckling of the tube, instead of
brittle failure that was observed in the experiments. A simulation of progressive crushing
of a thin-walled square CFRP tube was performed in [58]. The finite element model was
able to accurately predict the peak load, but the overall SEA was underestimated by 33%.
In the experiment debris wedged between the fronds of the tube’s wall contributed to
energy absorption while in simulations, instead of forming debris, the crushed elements
were deleted and did not contribute to energy absorption. In [59,60] the simulations of the
crushing behavior of hollow circular and square tubes were conducted and compared with
experimental observations. It was shown that the model was not able to reproduce the
axial matrix splits observed experimentally. As a result, the simulation did not agree well
overcome the modeling deficiency, pre-defined seams were introduced along the element
edges in the axial direction of the tubes to simulate the propagation of axial cracks, which
The primary focus of this study is to use the multi-layer modelling methodology
developed in [90, Chapter 7] to simulate the crushing behavior of CFRP circular tubes
having two different failure trigger mechanism. The experimental work, reported in [81,
Chapter 5], is briefly summarized first, followed by a description of the finite element
Circular tube specimens, made of Hexcel IM7/8552, were tested in [81, Chapter
chamfered ends (Figure 5-1a), inward-folding crush caps (Figure 5-1b), outward splaying
crush caps (Figure 5-1c), and a combined chamfered end and inward-folding crash cap,
on increasing the SEA and reducing the initial peak crush load. The effect of corner
radius of the crush cap on the peak load and SEA was further investigated to determine
the optimal crush cap configuration. The experimental results obtained with a [+15/-
15/+15/03/-15/+15/-15] laminate, are used in this numerical study. The specimens were
101.6 mm long with an outer diameter of 32.3 mm and wall thickness of 1.47 mm. Two
failure triggering mechanisms have been modeled: i) a 45° chamfered end tube, Figure
5-1a; and ii) a chamfered end tube with an attached inward-folding crush cap having a
1.58 mm corner radius, Figure 5-1b [81, Chapter 5]. The latter was found to yield the
All tests were carried out under quasi-static axial compression at a crosshead
50.8 mm, which provided sufficient crush data to calculate the SEA. The full-field
deformation and strain on the surfaces of selected specimens were also recorded using a
Test results showed that the chamfer was very effective at reducing the initial
peak load; while the inward-folding crush-cap was most effective at increasing the
185
sustained crush load and SEA. Further, combining an edge chamfered tube with an
inward-folding crush-cap (‘combined’ failure trigger) yielded the highest SEA with a low
initial crush load. Figure 8-2 shows the average load-crosshead displacement curves and
SEA of tubes with the chamfered end and the combined failure trigger mechanisms.
(a) (b)
Chamfer-end Chamfer-end
50 150
Combined Combined
40
Load (kN)
SEA (kJ/kg)
30 100
20
10 50
0
100 20 30 40 50 0
Crosshead Displacement (mm) SEA
Figure 8-2: (a) Experimental load-crosshead displacement curves, and (b) SEA of tubes having a chamfer
and combined failure triggers. The combined failure trigger yielded a higher peak load,
sustained crush load and SEA.
186
LS-DYNA finite element code was used in this study. Four-node, fully integrated,
shell elements (Type 16) were used to model the tubes. A multi-layer approach,
developed in [90, Chapter 7], was used to model the specimens and MAT54 material
model was used to simulate the crushing behavior of composite tubes under axial
compressive load.
In the multi-layer approach the laminate is divided into multiple layers of shell
elements, each layer consists of either a single ply or multiple plies, and layers are tied
together using a tiebreak contact definition. To improve the computational efficiency, the
with the inner [+15/-15/+15] plies, the middle [03] plies, and the outer [-15/+15/-15] plies
represented by the inner, middle, and outer layers, respectively. The diameters of the
mid-planes of the inner, middle, and outer layers were 29.8 mm, 30.8 mm and 31.8 mm,
points corresponding to the three plies represented by that layer. An element size of 0.635
x 0.635 mm was used, resulting in the model having approximately 77,000 elements. The
resulting computation time for these models was approximately 30 hours on a cluster
with 24 cores.
The chamfer was modeled by staggering the length of each shell layer such that
the inner layer’s first row of elements represented the start of the chamfer, and the outer
layer’s first row of elements represented its end, Figure 8-3(a). The steel crush caps were
187
modeled using rigid shell elements, Figure 8-3(b). The inner corner radius of the crush
Figure 8-3: Representations of the failure trigger mechanisms in the finite element models. The
chamfered end of the tube was modeled by staggering element layers while the crush cap was
modeled using rigid shell elements.
The shell element layers were tied to each other using a tiebreak contact definition
(contact one way surface to surface tiebreak) with option 8 in LS-DYNA [73]. This
tiebreak formulation allowed for simulating delamination at the interface between shell
element layers. Damage initiates when the stresses on the interface satisfy the failure
criterion [73]:
| | | |
( ) ( ) (1)
in which and are the normal and shear stresses acting on the interface, and
NFLS and SFLS are the normal and shear strengths of the tie, respectively. Once the
damage has initiated, the two surfaces begin to separate and the interfacial stresses are
then scaled down as a linear function of the separation distance. The critical distance,
denoted by CCRIT, at which failure occurs (i.e., deletion of tiebreak and advancing of
(2)
188
where:
√ ( ) | | (3)
and is the energy released due to the failure of the interface. A sensitivity
study was conducted and it was determined that for simulating the crushing of composite
tubes, Mode II fracture was the dominant mode of failure during the tie-break failure
process. Thus, to simplify the simulations, a pure Mode II delamination was assumed.
Consequently, and , and the critical distance for failure is given by:
(4)
where is the Mode II critical energy release rate, the value of which
was obtained from [87] for IM7/8552, as shown in Table 8-1. A further investigation
showed that any lower values will cause the ties to break prematurely, yielding high
and typically requires a very fine mesh. To improve the computational efficiency, the
methodologies discussed in [88] were adopted. While these methodologies were intended
for use with cohesive zone models, they can be applied to the tiebreak formulation used
here as well since the tiebreaks follow a traction-separation law similar to that used in
cohesive zone formulations. The proposed solution involves lowering the interlaminar
strengths whilst keeping the fracture toughness constant in order to adapt the length of the
cohesive zone for a given mesh size. The required interfacial strength ̅ can be
calculated as [88]:
189
̅ √ (5)
energy release rate, is the desired number of elements in the cohesive zone, and is
the mesh size in the direction of the delamination progression. The minimum number of
elements needed in the cohesive zone has not been well established. Various studies have
used anywhere from two elements to 10 elements [88]. In this study, the results of a
sensitivity study concluded that five elements were sufficient to simulate the propagation
of delamination. Hence, with , equation (5) was used to solve for the new NFLS
and SFLS values. The new SFLS was substituted into equation (4) to calculate the CCRIT
value.
Since the 9-ply laminate of the tube was modeled as three layers of shell elements
with each layer representing three plies, the tiebreak contact was defined only between
these shell layers, rather than between individual plies. However, in reality, delamination
could occur along any, if not all, ply interfaces during specimen crushing, as was
observed experimentally [81, Chapter 5]. In order to account for the energy dissipated by
these additional delamination interfaces, CCRIT was scaled by the ratio of the number of
(6)
The values used and calculated in equations (4), (5) and (6) are given in Table
8-1. They were used to simulate the crushing of the tube with the chamfered and the
9.08 GPa 0.2 kJ/m2 1.33 kJ/m2 5 0.635 mm 22.5 MPa 57.3 MPa 0.046 mm 0.185 mm
Both the loading platen and the base of the testing machine were modeled as rigid
surfaces. The tubes were placed standing upright between the loading platen and the
base. The interaction between the loading platen and the tube was modeled using a
contact between the base and the tube was defined using rigidwall_planar, which does
not allow node penetration. For the simulation of the chamfered tube with a crush cap,
the crush cap was assumed to be rigid and was defined using the same surface-to-surface
The tiebreak contact definition between the shell layers not only facilitates the
simulation of delamination, but also prevents layers from penetrating each other after the
tiebreak has failed, as the contact definition would remain in effect. It should be noted
that since the crush cap forces the layers to deform inwards, the inner layer would
eventually come in contact with itself, i.e., with the inner wall of tube. Therefore, a single
surface contact definition (contact automatic single surface) was added to the inner shell
behavior of the specimens. MAT54 is a progressive failure model that uses the Chang-
Chang failure criterion [73] to determine failure of each ply (associated with an
integration point). This model allows the user to define a local material coordinate system
to specify the orientation of each ply. There are 21 parameters in MAT54 that need to be
specified, 15 of which are physically based and six are numerical parameters [73]. Out of
the 15 physical parameters, 10 parameters are material constants that were obtained from
The remaining five failure parameters are the tensile and compressive failure
strains in the fiber direction, the matrix and shear failure strains, and the effective failure
strain. The six numerical parameters were either estimated or set to their default values,
depending on the behavior required of the material model. A parametric study was
conducted to determine the optimal values of the unknown parameters. It was found that
the only MAT54 parameter that needed to be adjusted was DFAILM, which is the failure
strain in the matrix direction [90, Chapter 7]. Adjusting the value of DFAILM enabled
matrix splitting to occur, as observed in the experiments. DFAILM values ranging from
5% to 30% were analyzed and it was determined that, for the chamfered tube, a value of
192
10% strain yielded the most accurate representation of the crushing process. For the
combined failure trigger mechanism case there was no need for element deletion, hence
DFAILM was set to the default value which disabled the element deletion feature.
8.5.1. DEFORMATION
show that the modeling approach described above provided a good depiction of the
failure process. Figure 8-4(a) compares the experimental and simulated crushed end of
the chamfered tube at the completion of chamfer crushing. The simulation predicts that
the outer layer of shell elements (three plies) splayed outwards, while the other two shell
element layers (six plies) folded inwards. This result is very similar to that observed in
the experiments, where two plies splayed outwards and seven folded inwards [81,
Chapter 5]. The difference is attributed to the particular grouping of three plies per shell
than the actual number of cracks formed during the crushing process, Figure 8-4(b) to
Figure 8-4(d). While the simulation predicts some 12 matrix splits in the outer layer,
approximately 40 cracks - in the form of matrix splits progressing along the fiber
direction - were observed in experiment. Further, the simulation predicts that the matrix
splits occur through the entire thickness of the outer layer (three plies), while in
experiment matrix splits occurred, randomly and separately, across individual plies or ply
groups. These matrix splits generated 1.0-4.0 mm wide CFRP segments of single plies or
ply groups, which splay and extend progressively along the surface of the testing platen
193
with increasing crushing distance. The smaller number of matrix splits predicted by the
simulation is not sufficient to cause the formation of the numerous narrow ply/ply-group
segments. As a result, the outer three-ply segments are prevented from splaying outward
freely in the simulation. Instead, they fold upwards and slide along the outer tube wall,
progressively throughout the crushing distance of the tube. Furthermore, as the fibers in
the outer plies aligned in the +15° and -15° direction, the matrix splits extend, during the
continuous crosshead displacement, along the fibers directions and the ply/ply-group
tube, Figure 8-4(d) and Figure 4(e). The simulation also captured the twisting phenomena
in the outer three-ply layer, but not as extensively as observed experimentally. The
majority of the crushed material was folded inward, packed, crushed, and compressed
into the interior of the tube, Figure 8-4(e). It is noted here that during unloading, the two
It should be noted that matrix splitting, which in the current model is represented
by element deletion, could be better simulated by using a significantly finer mesh with
corresponding change in failure strain, by having the elements within each layer split
along their edges, etc. Further, the twisting of the ply/ply-group segments during the
outward splaying could be more accurately predicted if the mesh was aligned with the
fiber direction; this would require each ply to be modeled as a separate shell element
layer. Therefore, this model cannot predict matrix splitting per se, but the occurrence of
axial cracks in the tube’s wall. The added computational cost incurred by employing any
194
of these possible approaches, however, would have been prohibitively high for
For the case of combined failure trigger mechanism, the chamfered end of the
tube was attached to a rigid inward-folding crush cap, which forces the material to be
packed into the core of the tube. Figure 8-5 shows a comparison of the deformations
obtained from experiment and simulation. It can be seen that the simulation replicated the
crushed end quite accurately by forcing all plies to fold inward and pack into interior of
the tube.
(b) 5 mm of crushing
Experiment Simulation
Final deformation
Figure 8-5: A comparison of the experimental and simulated deformation of a tube with a combined
chamfer and inward-folding crush cap trigger mechanism.
chamfered tube, showing both the unfiltered (dotted line) results and the filtered (solid
line) results using SAE 300 Hz filter. The unfiltered plot shows three load spikes,
occurring within the first 2 mm of the crosshead displacement. These three load spikes
correspond to the crushing of the three shell element layers in the chamfered region. It is
noted that by modeling each individual ply as a separate shell element layer, (hence with
a much higher computational cost), the number of load spikes would increase to nine and
the magnitude of each spike would reduce significantly, rendering a much smoother load-
displacement curve. With the current three-layer model, the curve could be smoothed
using various data filtering schemes. Clearly, the ‘smoothness’ of the curve depends on
the degree of data filtering, noting that excessive filtering will reduce the utility of the
results. Applying a range of filters, between SAE 180 Hz to SAE 1,000 Hz, indicated
that SAE 300 Hz filter provided the best continuous initial load-displacement behavior
that also compared well with the experimental data, Figure 8-7. It should be noted that
this filtering process did not affect the average sustained crush load and the overall SEA,
196
Figure 8-7, as these values correlated very well with the experimental results regardless
of the filter applied. All simulation results presented herein are smoothed using the SAE
300 Hz filter.
50 Simulation unfiltered
Simulation filtered (SAE 300Hz)
40
30
Load (kN)
20
10
0
0 5 10 15 20
Crosshead Displacement (mm)
Figure 8-6: A comparison of the unfiltered and filtered load-crosshead displacement curves from a multi-
layer simulation for a tube with a chamfer failure trigger.
30
20
10
0
5 0 10 15 20 25
Crosshead Displacement (mm)
Figure 8-7: Load-crosshead displacement comparison between experiment and simulation for specimens
with a chamfer trigger mechanism.
Simulation results for the tube with the combined failure trigger mechanism are
shown in Figure 8-8, together with experimental data from three tests, showing very good
agreement. The simulation, however, predicts a slightly earlier initial peak load, with a
50
40
30
Load (kN)
20
Tube Combined Test 1
Tube Combined Test 2
10 Tube Combined Test 3
Simulation (Filtered - SAE 300Hz)
0
5 0
10 15 20 25
Crosshead Displacement (mm)
Figure 8-8: Load-crosshead displacement comparison between experiment and simulation for specimens
with a combined chamfer and inward-folding crush cap trigger mechanism.
loads, and SEA show good correlation, with the simulation predicting 8% - 20% higher
peak load and essentially identical crush load and SEA (Figure 8-9).
140
30 120
Load (kN)
SEA (kJ/kg)
100
20 80
60
10 40
20
0 0
PEAK LOAD CRUSH LOAD SEA
(a) (b)
Figure 8-9: Comparison of simulated and experimental (a) peak loads and sustained crush loads, and (b)
SEA for the tubes with a chamfer trigger mechanism and a combined trigger mechanism.
198
The finite element model was further validated by comparing the resulting strain
distribution in the chamfered tubes with the experimental measurements obtained from
the high-speed Digital Image Correlation (DIC) system [81, Chapter 5]. Figure 8-10
shows a comparison of the hoop strain distributions for a chamfered tube at four selected
loading stages prior to reaching the initial peak load. That is, these DIC images show the
strain field in the bottom 20 mm section of the tube, during the crushing of the chamfer.
For clarity, it should be noted that the bottom of each DIC image coincides with the top
of the chamfered region, (i.e., the chamfered region is the dark region beneath the DIC
image). Several “hot spots,” representing areas of high hoop tensile strain, initiated at the
top of chamfered region, and propagate upwards along the tube wall with increasing load.
It should be noted that these ‘hot spots’ initiated prior to the completion of chamfer
crushing and initiated at approximately 18.5 kN, i.e., at 71% of the peak load of 25.8 kN.
The predicted locations and strain magnitudes of these hot spots correlated well with the
DIC results. The axial extension of these hot spots eventually became matrix splits. Thus,
the finite element model is capable of accurately predicting the initiation of such matrix
cracks (or matrix splits). A further development of the model discussed earlier, such as
Simulation
Experiment
Two finite element models were developed using LS-DYNA to simulate the
crushing behavior of composite tubes with chamfer and combined failure trigger
mechanisms. A multi-layer shell element approach was used to model the deformation
and damage progression of the composite tubes. Each layer of shell element could
contain either a single ply or multiple plies. Layers were tied together using tiebreak
which was determined using an energy based approach. Material model 54 (MAT54) was
used to represent each ply, and it was found that only the matrix failure strain, DFAILM,
needed to be adjusted in order to obtain good correlations with the experimental results.
Simulation results showed, for both the chamfer and combined failure trigger cases, that
the failure processes, strain fields, peak load, sustained crush loads, and SEAs all
compared very well with the experimental results. The correlation of the peak load
depends, of course, on the filters used. The chamfered region would have to be modeled
200
in greater detail (e.g., modeling each ply separately) than the current approximation in
order to achieve better results. In both cases, filtering the simulation data, using SAE 300
Hz filter, resulted in a better agreement of the initial peak load with the experimental
data, without affecting the sustained crush load. Finally, the modeling methodology
developed in this study has been shown to be capable of capturing quite accurately the
overall crushing behavior of the tubes under axial compressive loading. The extensive
node splitting in the axial direction. Further, aligning the mesh with the lamina direction
could result in a better representation of the matrix cracking, resulting in a more accurate
deformation and crushing process. This will require modeling each ply as an individual
layer. All these, and other, approaches would add significant computational cost.
201
9.1. ABSTRACT
Finite element models were developed using LS-DYNA to simulate the crushing
multiple layers of shell elements with tiebreak contact definitions between layers. Each
layer contained multiple plies. Delamination was simulated by failure of the tiebreaks,
which was determined using the interlaminar fracture toughness values. Material model
54 (MAT54) in LS-DYNA was used to define the material ply properties. Results were
absorption (SEA), failure process, and surface strain distribution. The multi-layer
approach was capable of accurately predicting the sustained crush load, SEA, and
9.2. INTRODUCTION
focus to improve occupant survivability in the event of a crash [1]. The subfloor structure
8
Siromani, D., Awerbuch, J., Tan, T.-M., “Finite Element Modeling of the Crushing Behavior of Thin-
Walled Open Cross-Section CFRP Members under Axial Compression,” Submitted to: Journal of
Reinforced Plastics and Composites.
202
dissipating energy during the impact event. Several studies have been performed over the
past three decades to investigate the dynamic response of aircraft structures and the
of these studies showed the need for energy absorbing devices, integrated with the
subfloor structure, to mitigate the impact energy. Composite materials are considered as
possible candidates for such integrated energy absorbing devices due to their high
strength-to-weight ratio and their high specific energy absorption (SEA) capacity, e.g.
[24-28], resulting from their particular failure characteristics during the crushing process.
The maximum crush load and SEA depend greatly on the selected cross-sections, wall
[82, Chapter 6]. The effect of two failure trigger mechanisms (chamfers and steeples) on
initial peak loads, crush loads, and failure progression and crushing was studied. Results
showed that the steeple trigger is more effective than the C-channel and hat-stiffeners in
reducing the initial peak loads, while the chamfer trigger yields a lower initial peak load
for the angle-stiffener as compared with the other two cross-sections. The angle- and hat-
stiffeners absorb similarly high specific energy absorption (SEA) while the C-channels
Due to the relatively high cost of conducting repeated experimental studies (with
different materials and component geometries) and the need to have a design tool to
203
construct the optimal energy absorption sub-structures, there is a need for reliable
members. There have been several attempts to develop explicit finite element models
with varying degrees of success for circular tubes [57-60], hollow square tubes [59-63],
structures can be modeled using either solid or shell elements. It has been well-
established that shell element models require less computation time and are more widely
used to model the axial crushing of composite members. The laminate can be modeled
using either a single layer or multiple layers of shell elements. In the ‘single-layer’
model the laminate is modeled by using a single layer of shell elements, with each ply
model, the laminate is modeled as multiple layers of shell elements, each layer may
represent either a single ply or a group of plies, and the layers are tied together using a
The single-layer modeling approach has been used in a number of recent studies
studying the crushing of thin-walled square tubes [58], it was shown that the single-layer
approach was capable of accurately capturing the local wall buckling and unstable
collapse. However, the approach could not depict the progressive failure process. By
carefully adjusting the values of key material and numerical parameters in the material
model (e.g., the eleven parameters in MAT 54) and the contact characteristic between the
end of the test specimen and the loading fixture, the single-layer model was able to yield
describing the failure and crush behavior. The utility of the single-layer model is further
approximately capturing the failure process for tubes undergoing progressive crushing
e.g., [57-60]. However, this approach has not always yielded good correlation with the
simulated in [57] using LS-DYNA with MAT 54. A good agreement with the
local buckling of the tube, rather than the brittle failure observed experimentally. A
[58]. The multi-layer finite element model was able to accurately predict the peak load,
but the overall SEA was underestimated by 33%. The difference in SEA was attributed to
the model’s inability to reproduce the formation of debris wedged between the fronds of
the tube’s wall due to excessive element deletion. In [59,60], the simulations of the
crushing behavior of hollow circular and square tubes were performed using the multi-
layer approach; results did not agree well with the experimental load-displacement
curves, deformation, and failure behavior due to the model’s inability to reproduce the
axial matrix cracks observed experimentally. To overcome the modeling deficiency, pre-
defined seams along the element edges in the axial direction of the tubes were introduced
to simulate the propagation of axial cracks, which yielded better agreement. A continuum
damage mechanics model (CODAM) was used in [64], with multiple layers of shell
model formed between the delaminated surfaces of neighboring plies during the crushing
process was incorporated to improve the accuracy of the simulation. While this approach
can induce ply splaying in the experimentally-observed direction, it may not be generally
applicable to cases with different failure processes. Further, it was noted in [66] that some
processes, which are not readily available. As a result, these parameters have to be
hat-stiffener was performed using multi-layer approach in [67]. It was found that
increasing Mode I and Mode II energy release rates for the cohesive interface yielded
better correlation with experimental results than those obtained by using the conventional
The primary focus of this study is to adapt the multi-layer finite element modeling
methodology that was developed in [90, Chapter 7] to simulate the crushing behavior of
the thin-wall open cross-section members. The experimental work, reported in [82,
Chapter 6], is briefly summarized first, followed by a description of the finite element
models, the results of the numerical simulations, and comparison with experiments.
and hat-shaped stiffener) fabricated from Hexcel IM7/8552, were tested in [82, Chapter
and steeples (Figure 9-1), on increasing the SEA and reducing the initial peak load and
crush load. To ensure stability during the compressive loading, the specimens were
206
supported by a 25.4 mm (1.0 in.) thick potted base. Laminate lay-up sequences and
specimen dimensions are listed in Table 9-1 and shown in Figure 9-2. The cross-
sectional areas were the same for all cross-sections. As a result, the wall thickness of the
angle-stiffener was twice the thickness (i.e., twice number of plies) of the other two
Figure 9-1: Open-cross-sections with two failure trigger mechanisms (chamfers and steeples), all having
the same cross-sectional area and attached to a potted base to ensure stability during the 50.8
mm crush displacement.
All tests were carried out under quasi-static axial compression at a crosshead
50.8 mm, which provided sufficient crush data to calculate the SEA. The failure process
was recorded with still and video cameras and the full-field deformation and strain on the
surfaces of all specimens were recorded using a Digital Image Correlation (DIC) system.
The test results showed that the steeple trigger is more effective at reducing the
initial peak loads of the C-channel and hat-stiffeners, while the chamfer trigger yields a
lower initial peak load for the angle-stiffener. The angle- and hat-stiffeners absorb
similarly high specific energy absorption (SEA) while the C-channels absorb the least.
Figure 9-3 shows the average (of three test results) load-crosshead displacement curves
recorded for the three cross sections and two failure triggers. The corresponding
quantitative data on the initial peak load, the sustained crush load, and SEA are shown in
Figure 9-4. Additional results and discussion are provided in [82, Chapter 6].
208
Figure 9-3: A comparison of load-crosshead displacement curves for each cross-section having a: (a)
chamfer trigger, and (b) steeple trigger. Each curve represents the average of three tests
(except for the angle-stiffener with a chamfer trigger which represents the average of two
tests).
Figure 9-4: A comparison of: (a) initial peak load and crush load; and (b) SEA, for each cross-section
having a chamfer and steeple failure triggers (numbers indicate average of three specimens
except for the angle stiffener with a chamfer trigger which represents the average of two
tests).
behaviors of the open cross-section specimens. Specimens were modeled using four-
node, fully integrated, shell elements (Type 16) and the crushing behavior of the
composite members under axial compressive loads was modeled using MAT54.
209
In the multi-layer approach, the laminate is divided into multiple layers of shell
elements, where each layer consists of either a single ply or multiple plies, and layers are
laminate were modeled by five layers of shell elements with each layer representing two
plies (i.e., layer 1: [02], layer 2: [+45/-45], layer 3: [02], layer 4: [-45/+45], and layer 5:
[02]). The angle-stiffener specimen, which is twice as thick and made of the [02/+45/-
elements was assigned two through-thickness integration points corresponding to the two
plies represented by that layer. An element size of 1.27 x 1.27 mm was used for the flat
sections in the models (i.e., web and flanges). For curved sections (i.e., the corners of the
specimens), a refined element size of 0.635 x 0.635 mm was used for a more accurate
simulation of the damage progression in these regions. For all specimens the total number
of elements was approximately 55,000. The computation time needed to simulate the
The 45° chamfer was modeled by staggering the length of each shell layer such
that the inner layer’s first row of elements represented the start of the chamfer and the
outer layer represented the end of the chamfer, Figure 9-5(a). The steeple trigger was
modeled by making two 15° cuts to the top of the model, Figure 9-5(b).
210
(a) (b)
Figure 9-5: Representations of the C-Channel cross-section having: (a) chamfer and (b) steeple failure
trigger mechanisms in the finite element models.
The shell element layers were tied to each other using a tiebreak contact definition
(contact one way surface to surface tiebreak) with option 8 in LS-DYNA [73]. This
tiebreak formulation allowed for simulating delamination at the interface between shell
element layers. Damage initiates when the stresses on the interface satisfied the failure
criterion [73]:
| | | |
( ) ( ) (1)
in which and are the normal and shear stresses acting at the interface, and
NFLS and SFLS are the normal and shear strengths of the tie, respectively. Once the
damage has initiated, the two surfaces begin to separate and the interfacial stresses are
then scaled down as a linear function of the separation distance. The critical distance,
denoted by CCRIT, at which failure occurs (i.e., deletion of tiebreak and advancing of
(2)
211
where:
√ ( ) | | (3)
and is the energy released due to the failure of the interface. A sensitivity
composite tubes showed that Mode II fracture was the dominant mode of failure during
the tie-break failure process. To simplify the simulations, the same approach was adopted
given by:
(4)
where is the Mode II critical energy release rate, the value of which
was obtained from [87] for IM7/8552, as shown in Table 9-2. An investigation performed
in [90, Chapter 7] showed that any lower values would cause the ties to break
and typically requires a very fine mesh. To improve the computational efficiency, the
methodologies discussed in [88] were adopted. The proposed solution involves lowering
the interlaminar strengths whilst keeping the fracture toughness constant in order to adapt
the length of the cohesive zone for a given mesh size. While this approach was intended
for use with cohesive zone models, it can also be applied to the tiebreak formulation used
here since the tiebreaks follow a traction-separation law similar to that used in cohesive
zone formulations. The required interfacial strength ̅ can be calculated from [88]:
212
̅ √ (5)
energy release rate, is the desired number of elements in the cohesive zone, and is
the mesh size in the direction of the delamination progression. The minimum number of
elements needed in the cohesive zone has not been well established. Various studies have
used anywhere from two elements to 10 elements [88]. Although no cohesive elements
are required for the tiebreak formulation applied herein, the same concept used for
cohesive elements has been applied to the tiebreaks. A separate sensitivity study
concluded that five elements were sufficient to simulate the propagation of delamination.
Hence, with , Equation (5) was used to solve for the new NFLS and SFLS values.
The new SFLS was substituted into equation (4) to calculate the CCRIT value.
Since the laminate of the open cross-sections was modeled as either five or 10
layers of shell elements, with each layer representing two plies, the tiebreak contact was
defined only between these shell layers, rather than between individual plies. However,
delamination could occur along any, if not all, ply interfaces during specimen crushing,
as was observed experimentally [82, Chapter 6]. In order to account for the energy
dissipated by these additional delamination interfaces, CCRIT was scaled by the ratio of
the number of ply interfaces to the number of tiebreak interfaces , defined as:
(6)
Chapter 6]) that delamination occurred among all ten (20 for angle stiffeners) plies. The
values used and calculated in Equations (4), (5) and (6) are listed in Table 9-2. It should
213
be noted that was calculated using the larger element size (1.27 mm) used for the
flat surfaces and was used to simulate delamination in throughout the open cross-section
All boundary conditions used in the models were accurate representations of the
experimental setup. The open cross-sections required the nodes along the flat end of the
models to be fixed in all degrees of freedom to simulate the potted base used for support.
The loading platen of the testing machine was modeled as a rigid surface with a friction
coefficient of 0.3. The members were placed standing upright between the loading platen
and the base. The interaction between the loading platen and the specimen was modeled
The tiebreak contact definition between the shell layers not only facilitates the
simulation of delamination, but also prevents layers from penetrating each other after the
tiebreak has failed, as the contact definition would remain in effect. It should be noted
that the splaying of the inner and outer layers and their subsequent folding would force
them to come in contact with the un-deformed inner or outer wall of the specimen.
214
Therefore, a single surface contact definition (contact automatic single surface) was
behavior of the specimens. MAT54 is a progressive failure model that uses the Chang-
Chang failure criterion [73] to determine failure of each ply (associated with an
integration point). This model allows the user to define a local material coordinate system
to specify the orientation of each ply. There are 21 parameters in MAT54 that need to be
specified, 15 of which are physically based and six are numerical parameters [73]. Out of
the 15 physical parameters, 10 parameters are material constants that were obtained from
[83] and [84]. The values of these 10 parameters used in this study are shown in Table
9-3.
The remaining five failure parameters are the tensile and compressive failure
strains in the fiber direction, the matrix and shear failure strains, and the effective failure
strain. The six numerical parameters were either estimated or set to their default values,
depending on the behavior required of the material model. In [90, Chapter 7], a separate
parametric study was conducted to determine the optimal values of the unknown
parameters and it was found that the results are insensitive to a wide range of values of
five of these parameters. The only MAT54 parameter that needed to be adjusted was
215
DFAILM, which is the failure strain in the matrix direction [73]. Adjusting the value of
DFAILM enabled a steady matrix splitting to occur, as observed in the experiments. The
parametric study performed in [90, Chapter 7] resulted in a DFAILM value of 10% strain
that provided the most accurate representation of the crushing process of circular
composite tubes. This DFAILM value was used for the elements in the corner regions of
the open cross-section specimens since their size is the same as that of the circular tube
model [90, Chapter 7]. To determine the optimal DFAILM value for the flat surface,
regions where elements are twice the size of the corner elements, a separate parametric
study was conducted. It was found that a DFAILM value of 5% strain would provide the
most accurate representation of the crushing process. These values were used in the
9.5.1. DEFORMATION
displacements are shown in Figure 9-6, Figure 9-7, and Figure 9-8, respectively. It should
be noted that the difference in post-test configurations of experiment and simulation was
due to the fact that, in all cases, the simulations terminated at the maximum crosshead
displacement and the unloading process was not performed. For the C-channel, the
simulation accurately predicts the damage initiation as the chamfer crushes and the
buckling in the flanges and web with increasing crosshead displacement, Figure 9-6(b).
216
Similar behavior was recorded for the angle- and hat-stiffeners in Figure 9-7(b) and
As the crushing progresses, the simulation predicts the formation and growth of
matrix cracks at the corners and in the web of the C-channel, as well as outward splaying
of the outer plies, Figure 9-6(c) and Figure 9-6(d). However, the simulation was not able
to predict the number of matrix cracks formed during the crushing process, particularly in
the corners and flanges. These matrix cracks generated 5-10 mm wide CFRP segments of
single plies or ply groups, which splay inwards and outwards progressively along the
surface of the testing platen with increasing crushing distance. Similar behavior was
recorded for the angle- and hat-stiffeners in Figure 9-7(c) to Figure 9-7(d) and Figure
For the C-channel and hat-stiffener, the inability to form sufficient number of
matrix cracks resulted in some buckling of the web and flanges that was not observed in
the experiment, Figure 9-6(d) and Figure 9-6(e), and Figure 9-8(d) and Figure 9-8(e),
respectively. It should be noted that the experimental results, discussed in [82, Chapter
6], showed that the web of the C-channel and hat-stiffener had the tendency to buckle in
the beginning of the crushing process. However, the formation of matrix splits and cracks
at the corners caused the web to crush and splay inward or outward, and quickly relieved
the buckling.
Comparing the final deformed states of all three cross-sections shows that the
simulation was able to accurately capture the overall crushing behavior, including the
direction of splaying of each layer and the extent of damage, Figure 9-6(e), Figure 9-7(d),
and Figure 9-8(e), respectively. It is noted that during unloading in the experiment, the
217
segments).
Experiment
Simulation
(a) Pre-test (b) δ ~ 1.5mm (c) δ ~ 10mm (d) δ ~ 40mm (e) Post-test
Figure 9-6: Comparison of the experimental and simulated deformation of the C-channel with a
chamfered trigger mechanism at selected crosshead displacement stages.
Experiment
Simulation
(a) Pre-test (b) δ ~ 1.5mm (c) δ ~ 10mm (d) δ ~ 40mm (e) Post-test
Figure 9-7: Comparison of the experimental and simulated deformation of the angle-stiffener with a
chamfered trigger mechanism at selected loading stages.
218
Experiment
Simulation
(a) Pre-test (b) δ ~ 1.5mm (c) δ ~ 10mm (d) δ ~ 40mm (e) Post-test
Figure 9-8: Comparison of the experimental and simulated deformation of the hat-stiffener with a
chamfered trigger mechanism at selected loading stages.
displacements are shown in Figure 9-9, Figure 9-10, and Figure 9-11, respectively. As
was due to the fact that the simulation of unloading process was not performed. For all
three cross-sections, the simulation accurately predicts the crushing of the steeple, the
formation of several matrix cracks in the web and at the corners, and the inward/outward
splaying of the different layers. However, similar to the chamfer case, the lack of
sufficient number of matrix cracks formation in the simulation prevents the development
significant buckling in the web of the C-channel, Figure 9-9, minor buckling in the left
leg of the angle-stiffener, Figure 9-10, and minor buckling in the flanges of the hat-
Experiment
Simulation
(a) Pre-test (b) δ ~ 10mm (c) δ ~ 15mm (d) δ ~ 40mm (e) Post-test
Figure 9-9: Comparison of the experimental and simulated deformation of the C-channel with a steeple
trigger mechanism at selected loading stages.
Experiment
Simulation
(a) Pre-test (b) δ ~ 6mm (c) δ ~ 15mm (d) δ ~ 40mm (e) Post-test
Figure 9-10: Comparison of the experimental and simulated deformation of the angle-stiffener with a
steeple trigger mechanism at selected loading stages.
220
Experiment
Simulation
(a) Pre-test (b) δ ~ 12mm (c) δ ~ 20mm (d) δ ~ 40mm (e) Post-test
Figure 9-11: Comparison of the experimental and simulated deformation of the hat-stiffener with a steeple
trigger mechanism at selected loading stages.
channel with a chamfer failure trigger mechanism, showing both the unfiltered (dotted
line) results and the filtered (solid line) results. The unfiltered plot shows initial load
fluctuations, occurring within the first 2 mm of the crosshead displacement. These initial
load fluctuations, consisting of five load spikes, correspond to the crushing of the five
shell element layers in the chamfered region. The three higher load spikes correspond to
the crushing of the three [02] layers (layer 1, 3, and 5), whereas the [+45/-45] layers
(layer 2 and 4) result in lower load spikes that are visible only upon closer examination of
Figure 9-12. By modeling each individual ply as a separate shell element layer, (with a
much higher computational cost), the number of load spikes would increase to ten and the
magnitude of each spike would reduce significantly, rendering a much smoother load-
displacement curve. With the current five-layer model, the curve could be smoothed
using various data filtering schemes. It was found that in this particular case an SAE 600
Hz filter provided the most accurate results. It should be noted that in [90, Chapter 7] an
SAE 300 Hz was found to yield the best results for a three-layer finite element model.
221
This indicates that the degree of filtering required decreases with increasing in the
number of shell element layers used to model the laminate (i.e. the chamfered section).
50 Simulation unfiltered
Simulation filtered (SAE 600Hz)
40
Load (kN) 30
20
10
0
0 5 10 15 20 25
Crosshead Displacement (mm)
Figure 9-12: A comparison of unfiltered and filtered load-crosshead displacement curves from a multi-
layer simulation for a C-channel with a chamfer failure trigger.
stiffener, and hat-stiffener with chamfer and steeple failure trigger mechanisms is show in
Figure 9-13, Figure 9-14 and Figure 9-15, respectively. It should be noted that the load-
displacement data of the specimens with steeple triggers did not require any filtering as
the steeple geometry could be modeled accurately without any approximation. In most
cases, the simulation was able to predict quite accurately the load-crosshead displacement
behavior in terms of the initial peak load (within -1% to +8%) and sustained crush load
(within ±1%), Figure 9-16. The only exception is the angle-stiffener with a steeple
trigger, where the initial peak load is under predicted by 34%. Recall that in order to
properly simulate the initiation and steady progression of matrix splitting along the
size) with a higher DFAILM value of 10% failure strain were used to model the corner
222
regions. For C-channels and hat-stiffeners where the apex of the steeple is located far
away from the corners, this modeling strategy worked very well as the two failure
mechanisms are largely independent of each other. For the angle stiffener, on the other
hand, the apex of the steeple coincided with the corner of the cross-section. The
considerably weakened the material at the steeple apex, resulted in a much lower initial
peak load. Despite this discrepancy, the SEA predicted by all six simulations showed a
high level of agreement with the experimental results, Figure 9-17. It should be noted that
the data filtering process did not affect the average sustained crush load and the overall
SEA, as these values correlated very well with the experimental results, regardless of the
filter applied.
(a) (b)
50 C Chamfer Test 1 50 C Steeple Test 1
C Chamfer Test 2 C Steeple Test 2
40 C Chamfer Test 3 40 C Steeple Test 3
Simulation (Filtered - SAE 600Hz) Simulation (Unfiltered)
Load (kN)
Load (kN)
30 30
20 20
10 10
0 0
5 0 10 15 20 25 30 0 5 10 15 20 25 30
Crosshead Displacement (mm) Crosshead Displacement (mm)
Figure 9-13: Load-crosshead displacement comparison between experiment and simulation for C-channels
with (a) a chamfer, and (b) a steeple trigger mechanism.
223
(a) (b)
50 Angle Chamf. Test 1 50 Angle Steeple Test 1
Angle Chamf. Test 2 Angle Steeple Test 2
40 Simulation (Filtered - SAE 600Hz) 40 Angle Steeple Test 3
Simulation (Unfiltered)
Load (kN)
Load (kN)
30 30
20 20
10 10
0 0
50 10 15 20 25 30 0 5 10 15 20 25 30
Crosshead Displacement (mm) Crosshead Displacement (mm)
Figure 9-14: Load-crosshead displacement comparison between experiment and simulation for angle-
stiffeners with (a) a chamfer, and (b) a steeple trigger mechanism.
(a) (b)
50 Hat Chamfer Test 1 50 Hat Steeple Test 1
Hat Chamfer Test 2 Hat Steeple Test 2
40 Hat Chamfer Test 3 40 Hat Steeple Test 3
Simulation (Filtered - SAE 600Hz) Simulation (Unfiltered)
Load (kN)
Load (kN)
30 30
20 20
10 10
0 0
50 10 15 20 25 30 0 5 10 15 20 25 30
Crosshead Displacement (mm) Crosshead Displacement (mm)
Figure 9-15: Load-crosshead displacement comparison between experiment and simulation for hat-
stiffeners with (a) a chamfer, and (b) a steeple trigger mechanism.
224
(a) (b)
C Chamfer Experiment C Steeple Experiment
C Chamfer Simulation C Steeple Simulation
Angle Chamfer Experiment Angle Steeple Experiment
Angle Chamfer Simulation Angle Steeple Simulation
50 50 Hat Steeple Experiment
Hat Chamfer Experiment
Hat Chamfer Simulation Hat Steeple Simulation
40 40
Load (kN)
Load (kN)
30 30
20 20
10 10
0 0
PEAK LOAD CRUSH LOAD PEAK LOAD CRUSH LOAD
Figure 9-16: Comparison of simulated and experimental peak loads and sustained crush loads for the
specimens with (a) a chamfer, and (b) a steeple failure trigger mechanism.
(a) (b)
C Chamfer Experiment C Steeple Experiment
C Chamfer Simulation C Steeple Simulation
Angle Chamfer Experiment Angle Steeple Experiment
100 Angle Chamfer Simulation 100 Angle Steeple Simulation
90 Hat Chamfer Experiment Hat Steeple Experiment
Hat Chamfer Simulation 90
Hat Steeple Simulation
80 80
SEA (kJ/kg)
70
SEA (kJ/kg)
70
60 60
50 50
40 40
30 30
20 20
10 10
0 0
SEA SEA
Figure 9-17: Comparison of simulated and experimental SEA for the specimens with (a) a chamfer, and (b)
a steeple failure trigger mechanism.
225
The finite element models were further validated by comparing the resulting
strain distribution in the specimens with the experimentally measured axial and hoop
strain fields recorded with the Digital Image Correlation (DIC) system [82, Chapter 6].
Since similar results were recorded for all cross-sections, only the results recorded for the
C-channel are presented herein. Figure 9-18 shows a comparison of the axial strain
during the initial crush process of the chamfer: at two crosshead displacements prior to
reaching the initial peak load and the third after progressive crushing begins. The finite
element model is not able to accurately predict the axial strains during the initial stages of
crush due to the load fluctuations caused by the approximate chamfer model, as discussed
in the previous section. After the chamfer was fully crushed, however, the model was
able to predict similar high strain levels as observed in the experiment, at approximately
5 mm of crosshead displacement. The lateral strain fields, Figure 9-19, show that the
model was able to predict some of the high lateral tensile strains in the web of the C-
channel, from approximately 1.0 to 1.4 mm of crush. These high strains are due to some
minor outward buckling in the center of the web, discussed in more detail in [82, Chapter
6]. At 2.5 mm crosshead displacement, both simulation and experiment show high strains
Figure 9-20 shows a comparison of the axial strain distributions for a steeple C-
channel at three selected crosshead displacement stages. Results show that the simulation
can accurately predict the location and magnitude of the high axial strains in all three
226
stages. A similarly high level of correlation is observed in Figure 9-21 for the lateral
strain fields.
Simulation
Experiment
Simulation
Experiment
Finite element models were developed using LS-DYNA to simulate the crushing
behavior of open cross-section CFRP composite members with chamfer and steeple
228
failure trigger mechanisms. A multi-layer shell element approach was used to model the
laminates. Each layer of shell element could contain either a single ply or multiple plies.
Layers were tied together using tiebreak contact definitions. Delamination between layers
was simulated by tiebreak failure, which was determined using an energy-based criterion.
LS-DYNA Material model 54 (MAT54) was used to model the crushing behavior. It was
determined that only the matrix failure strain, DFAILM, was needed to be adjusted in
order to obtain good correlations with the experimental results. Simulation results
showed, for both the chamfer and steeple failure trigger cases, that the failure processes,
strain fields, peak load, sustained crush loads, and SEAs all compared very well with the
stage, which was caused by the modeling approximation of the chamfered region, could
be reduced by modeling the chamfer in greater details, such as modeling each ply
separately. In all chamfer cases, filtering the simulation data using an SAE 600 Hz filter
resulted in a better agreement of the initial peak load with the experimental load-
crosshead displacement data, without affecting the sustained crush load and SEA. For the
steeple members, no filtering was required as the steeple was modeled without any
has been shown to be capable of capturing the overall crushing behavior of the members
The investigation into aircraft crashworthiness began with the use of a previously
developed finite element model of a 3 m long Boeing 737 fuselage section, which has
been validated against drop test experimental data, to perform a parametric study on the
effect of the coefficient of friction and angle of impact between the fuselage and a rigid
surface. Four different friction coefficients and six different angles of impact were used
to determine their effect on the dynamic response of the fuselage section in terms of the
The results from the parametric study prompted the development of a full-length
landing on different terrains, including rigid surfaces, soils, and water. The 3 m long
B737 single-section model was first modified to a more computationally efficient model,
and a verification study, comprised of vertical impact simulations on rigid and water
surfaces, was conducted to ensure that the modified section model was capable of
the original section model. The results indicated that the modified model, while requiring
one-fifth of the computation time, was able to provide accurate simulation results. The
full-length fuselage model was then constructed by expanding the modified section
model. The full-length model consisted of a primary fuselage section, a nose cone, and a
tail section, the latter two represented by simple shell elements, excluding all structural
subcomponents. The landing gear and wings were excluded from the model due to the
230
lack of necessary design details. The full-length model was then employed to conduct a
absorbing structural members into the subfloor structure of the B737 single-section
model was presented. This study utilized spring elements to represent composite
member. The results were compared to the B737 simulations with and without luggage in
The major conclusions of this investigation are listed below. Additional details
i) The overall deformation of the B737 fuselage section was significantly affected
by varying the coefficient of friction and the angle of impact. The lower friction
coefficients and impact angles yielded higher plastic deformation and more
extensive buckling of the bottom frames, particularly on the left-hand side of the
fuselage.
ii) The cargo doorframe reinforcements on the right-hand side of the fuselage had a
iii) Regardless of the friction coefficient or the angle of impact, the peak acceleration
pulses were higher on the right-hand side than on the left-hand side due to the
iv) In all cases, the frames and luggage together accounted for most of the energy
dissipated. For lower friction coefficients and impact angles, the greater
deformation of the lower frames caused the luggage to crush more and, hence,
v) The crash landing simulations with the full-length model on rigid, soil and water
terrains yielded similar peak acceleration pulses, the highest occurring at the
vi) In all three cases the frames absorbed more energy than the skin and, in the soil
and water impact cases, the terrain also deformed, dissipating a portion of the
vii) The luggage played an important role as the major energy absorber, as well as
fuselage structure.
viii) Using spring elements in the subfloor region of the B737 single section model
ix) The results indicated that the stanchions could be used to improve the
histories and energy dissipation; however, further research is warranted into this
subject.
232
crashworthiness.
certification program.
iii) Aid in investigating actual aircraft crashes and study potential crash scenarios.
improving the energy absorbing capabilities of the aircraft subfloor structure. Hence, a
two-part study was performed to evaluate the energy absorbing capabilities of composite
members for use as stanchions in the subfloor structure. In the first part of the study,
composite tubes, made of graphite/epoxy laminates, were crushed under quasi-static axial
compression. The objective was to determine the most effective approach to decrease the
initial peak crush load while increasing the sustained crush load and SEA. Tests were
splaying crush-caps failure trigger mechanisms. The effect of the corner radius of the
crush-caps on the crushing process was also evaluated. Based on the results obtained
from the initial study, a new combined chamfered-end and inward-folding crush-cap
trigger mechanism was use to further study the effect of corner radius of the inward-
folding crush-caps.
233
The second part of this study included three different cross-sectional geometries
that are more prevalent in the aerospace industry: C-channels, angle-stiffeners and hat-
stiffeners. Two types of failure trigger mechanisms (chamfered-ends and steeples) were
investigated to determine which approach was most capable of decreasing the initial peak
A digital image correlation (DIC) system was used to monitor the axial and hoop
strain fields in the specimens during the crushing process. The DIC strain fields captured
the initiation and progression of damage, and highlighted several important differences
between the failures processes of the various cross-sections used in this study. Further,
these strain fields provide an additional point of validation for the finite element models.
The major conclusions of this investigation are listed below. Additional details
i) For the circular tubes, the chamfer failure trigger was most effective at reducing
the initial peak load while maintaining a high-sustained crush load and high SEA.
ii) The inward-folding failure trigger approach was not as effective at reducing the
initial peak load in the tubes, but was more effective than using a chamfer for
maintaining a high-sustained crush load and SEA. On the other hand, the
iii) A clear correlation between forcing more crushed material into the center of the
tube (using the inward-folding crush-caps), and an increase in SEA was observed.
234
Further, the results obtained with different corner crush-cap radii showed that the
smaller radii provide better results in terms of sustained crush loads and SEA.
mechanism yielded a moderate level initial peak load and the highest sustained
crush load and SEA as compared to any of the other failure triggers.
v) For the open cross-sections, the angle- and hat-stiffeners were the most effective
at absorbing energy, yielding very similar SEA values. The C-channel, on the
vi) The steeple failure trigger was able to provide a lower initial peak load than the
chamfer failure trigger for the C-channel and hat-stiffener. For angle-stiffeners,
vii) Cross-sections that had a larger ratio of curved sections typically resulted in a
higher SEA due to the greater extent of damage and interaction amongst the
inward-splaying plies.
The results from this investigation can be of significant use when designing energy
ii) Selection of a failure trigger mechanism design that is most appropriate for the
In this phase of the study, finite element models were developed using LS-DYNA
employed to model the crushing process, namely, a single-layer approach and a multi-
layer approach. The single-layer approach used one layer of shell elements to represent
this approach involved performing an extensive parametric study to obtain the values of
certain parameters required by material model 54 (MAT54) to correctly fit the simulation
results to the experimental load-crosshead displacement data. It was determined that two
parameters, DFAILC and SOFT, play a key role in predicting the initial peak load and
sustained crush load, respectively. An unsuccessful attempt was made to find a common
develop a methodology to model the crushing behavior of the composite members. This
included determining the most effective laminate configuration, element size and
model. First, it was determined that the laminate needed to be divided into an equal
number of plies per shell element layer, in order to maintain consistent spacing between
236
each layer. Fully-integrated, four-node elements were used to generate the mesh, with a
finer mesh required for curved surfaces (circular tube and corners of open cross-sections)
that the flat surfaces (web and flanges of open cross-sections). The quasi-static
experimental loading rate was modeled using a two-step function to ramp us the loading
rate while minimizing the dynamic effects. In order to simulate delamination, two
tiebreak options were investigated and it was determined that option 8 (a simple bilinear
energy based approach. Finally, two material models were investigated (MAT54 and
MAT58) and it was determined that MAT54 provided the better results in terms of the
performed in order to obtain the values of the unknown parameters. It was determined
that DFAILM was the only parameter that needed to be adjusted in order to obtain good
correlation with the experimental results. A parametric study was performed to determine
an optimal set of input parameters for the circular tube and the C-channel. These
The major conclusions of this investigation are listed below. Additional details
displacement curve of the circular tube and the C-channel with a chamfer trigger
However, it was not possible to predict the deformation and failure process due to
237
the inability of this model to form delamination between plies, matrix splitting,
etc.
ii) An attempt was made to find a common set of parameters that could be used
across the different cross-sections. However, it was found that each cross-section
iii) When applied to specimens with a steeple trigger mechanism, this method was
unsuccessful due to the lack of the chamfer row of elements to effectively use the
SOFT parameters. These models were mostly unstable and it was not possible to
iv) The multi-layer modeling methodology was capable of accurately capturing the
overall crushing behavior of the specimens under axial compressive loading. The
failure processes, strain fields, peak load, sustained crush loads, and SEAs
v) The chamfered region would have to be modeled in greater detail (e.g., modeling
each ply separately) than the current approximation in order to achieve better
initial peak loads. In all cases, filtering the simulation data resulted in a better
agreement of the initial peak load with the experimental data, without affecting
vi) An energy-based approach was used to determine the input parameters for the
scaling factor was introduced to account for ply interfaces that were not
vii) The element deletion criterion used has a significant effect on the material models
capability to correctly predict the SEA. Permitting only element deletion in the
matrix direction (i.e., matrix splitting) is key to the successful use of the material
model.
such as the most efficient way to model the laminate and its interfaces, and the most
effective element size and formulation, contact definitions, time-step control, and
material model. This modeling methodology has great potential for the following
applications:
selection.
ii) Support the design of crashworthy aircraft by incorporating these models into the
subfloor structure.
was performed, using a Boeing 737 fuselage section and a representative full-length
aircraft model. In order to further extend this work and improve the accuracy and
reliability of the results, realistic geometry and modeling details of the aircraft structure
239
are required. The current model does not take into account the effect of the landing gear,
wings, nose cone and tail section. These simplifications will undoubtedly have an effect
on the overall crashworthy response of the aircraft and will need to be addressed in the
future.
behavior of the composite members yielded excellent results for four different cross-
sections, three different failure trigger mechanisms, and two different lay-ups. However,
there are several issues that could be further investigated to improve the results. For
strain, or by allowing node splitting in the axial direction. Further, aligning the mesh
with the lamina direction could result in a better representation of the matrix cracking,
resulting in a more accurate deformation and crushing process. This will require
modeling each ply as an individual layer. These approaches, however, would add
the simple bi-linear traction-separation law worked better than the fundamentally more
accurate cohesive zone formulation. The primary difference between the two
formulations is that the cohesive zone formulation takes mode-mixity into account, thus
deserves further investigation and validation at the coupon level, particularly when using
a single contact interface to simulate delamination between multiple plies. Similarly, the
240
study into the optimal material model to simulate crushing showed that the simpler
progressive failure model (MAT54) was more effective than the continuum damage
mechanics based model (MAT58). The latter typically under-predicted the energy
absorbed which appeared to be due to the effective failure strain calculation performed to
account for ply failure. A possible solution to this could be to integrate separate failure
strain for fiber, matrix and shear directions, similar to the MAT54 implementation.
. The incorporation of the finite element models, describing the failure process
and energy absorption capacity of the various types and configurations of the stanchions
studied, into the model of the B737 fuselage warrants further investigation. Several issues
need to be addressed in order to successfully incorporate the stanchion models into the
subfloor structure. From a design perspective, this includes the type of cross-section of
required, failure trigger mechanism employed, etc. From a computational perspective, the
key issue is determining which modeling methodology can be used to represent the
static loading rate to study the crushing behavior of the composite members. Hence, the
finite element models were developed assuming a quasi-static loading rate. For
loading rate would have on the crushing behavior of these composite specimens. The
241
following sections briefly discuss: i) prior experimental results reported in the literature
on the effect of strain-rate on the compressive behavior of composite laminates; and ii)
the implications of including strain rate effects on the finite element methodology
presented in this dissertation and how the modeling could further be updated in the future
composite materials can broadly be classified into two categories: those performed to
study the material response (i.e., using a split-Hopkinson pressure bar) and those
performed to study the structural response (i.e., using tubular members). The former
provides and understanding of the materials response to different loading rates in terms of
its compressive modulus of elasticity, strength and ultimate strain; while the latter
includes the structural response to different loading rates in terms of crush loads and
The effect of strain rate on the material properties of carbon/epoxy laminates has
been reported in the literature (e.g. [91] to [96]), yielding mostly consistent results. For
laminates were characterized at dynamic strain rates of up to 118 s-1 in [91] and 110 s-1 in
[92]. Both studies reported an increased longitudinal compressive strength at the higher
strain rates compared to the static strength (up to 79% in [91] and 40% in [92]) but found
authors also investigated the transverse compression and in-plane shear properties of the
same composite materials in [93] and [94]. Both studies reported an increased transverse
242
compressive modulus (up to 37% in [93] and 12% in [94]) as well as the transverse
compressive strength (up to 100% in [93] and 45% in [94]). A similar trend was observed
for the in-plane shear modulus and strength in both studies. In [95] and [96], similar
trends for the compressive strengths were reported; however, [96] reported an increased
The quasi-static and dynamic crush tests on composite members reported in the
literature yielded mixed results. For example, in [24] and [98] no clear rate effect on the
SEA was observed during the crushing of graphite/epoxy [24,98], glass/epoxy [24,98]
and Kevlar/epoxy [98] tubes at quasi-static and dynamic loading rates (up to 8.5 m/s in
[24] and 7.6 m/s in [98]). The load-displacement curves and failure processes observed
were similar for different loading rates in both studies. Contrary to these two studies,
however, a clear rate effect was observed for the same materials in [34], reporting that the
SEA decreased by up to 30% for the dynamic loading rates (of 5.5 m/s), while the failure
processes were relatively similar. Similar trends were reported in [100] and [101] for
graphite/epoxy tubes tested under quasi-static and dynamic loading rates (of 7.5 m/s in
[100] and 8.5 m/s in [101]). On the other hand, the opposite trend was reported in a study
performed using graphite/epoxy and Kevlar/epoxy circular tubes [102], where the SEA
loading rates. Note that Kevlar, being a nylon material, is more rate dependent material.
Similar results were reported in [103] for square graphite/epoxy tubes tested at quasi-
static and dynamic loading rates (of 5.4 m/s). It was noted in [103] that the strain rate
increase in compressive longitudinal, transverse and shear strengths, and the transverse
and shear modulus of elasticity, but no change in the longitudinal modulus. Whereas, the
studies performed at the structural level showed mostly inconclusive results with respect
to the energy absorbed during crushing at different loading rates. The reasons for the
should be further noted that the crush tests were typically performed at much lower strain
rates (~20 to 30 s-1) than those conducted with coupon specimens (>100 s-1). Finally, the
the crushing behavior and energy absorption have not been studied under dynamic
loading. Therefore, there is a clear need to further investigate the crushing response of
this dissertation, delamination between the plies played a significant role in the energy
rate on the delamination process, specifically on Mode I and Mode II fracture toughness
values. Since these are matrix-dominated failures it should be expected that the higher
this approach may be adopted to simulate crushing under dynamic loading rates as well,
244
Firstly, the time-scaling functions used to decrease the computation time for
simulating quasi-static loading cases would become unnecessary for the dynamic loading
cases and the actual dynamic loading functions could be used. Second, the material
model would have to be modified to account for material rate dependency. This issue can
be addressed by simply replacing the current material properties with the corresponding
material model that is capable of predicting the material response at various loading rates.
The first approach can be implemented into the material model (MAT 54) used in this
study, in a manner similar to [103] where the compressive strength values were increased
in the dynamic model. The second approach would require replacing MAT 54 with rate-
dependent material models such as MAT 158 or MAT 161/162. There are, however,
many more input parameters required by these material models that would need to be
obtained parametrically, by calibrating them with test data. Therefore, if the required
dynamic test data and material properties are available, it will be possible to update the
example, in [104] the advantage of using strain rate dependent material properties (using
MAT 158) instead of quasi-static material properties (using MAT 58) to simulate the
addressed in a similar fashion. The current tiebreak and cohesive formulations require a
fixed set of input parameters that include, or can be calculated from, the Mode I and
245
Mode II interlaminar fracture toughness values. With known fracture toughness values
for a predetermined loading rate, the tiebreak parameters in the current modeling
LIST OF REFERENCES
4. "Airbus crashes in New York River." BBC News. 16 Jan. 2009. Available at:
http://news.bbc.co.uk/. (last confirmed on 4 April 2013).
5. “Turkish plane crashes in Amsterdam.” BBC News. 25 Jan. 2009. Available at:
http://news.bbc.co.uk/.(last confirmed on 4 April 2013).
6. “New York: Ten injured as plane crash lands at LaGuardia Airport.” BBC News.
23 Jul. 2013. <http://news.bbc.co.uk/>. (last confirmed on 8 Aug. 2013).
8. Jackson, K. E., Boitnott, R.L., Fasanella, E.L., Jones, L.E. and Lyle, K.H., “A
Summary of DOD-Sponsored Research Performed at NASA Langley's Impact
Dynamics Research Facility,” American Helicopter Society 60th Annual Forum,
Baltimore, MD, June 7-14 2004.
10. Jackson, K. E., Fasanella, E.L. and Lyle, K.H., “Crash Certification by Analysis –
Are We There Yet?,” American Helicopter Society 62nd Annual Forum, Phoenix,
AZ, May 9-11 2006.
11. Fasanella, E. L., Jackson, K. E., and Lyle, K. H., “Crash Simulation of an
Unmodified Lear Fan Fuselage Section Vertical Drop Test,” Proceedings of the
54th American Helicopter Society Annual Forum, Washington D.C., May 20-22,
1998.
247
12. Jackson, K. E., and Fasanella, E. L., “Crash simulation of a vertical drop test of a
commuter-class aircraft,” International Journal of Crashworthiness, Vol 10,
2005, pp. 173-182.
13. Abramowitz, A., Smith, T. G., Vu, T., Zvanya, J. R., "Vertical Drop Test of a
Narrow-Body Transport Fuselage Section with Overhead Stowage Bins,"
DOT/FAA/AR-01/100, September 2002.
14. Fasanella, E. L., and Jackson, K. E., “Crash Simulation of Vertical Drop Tests of
Two Boeing 737 Fuselage Sections”, U.S. Army Research Laboratory, Vehicle
Technology Center, Langley Research Center, March 2002.
15. Jackson, K. E. and Fuchs, T. T., “Comparison of ALE and SPH Simulations of
Vertical Drop Tests of a Composite Fuselage Section into Water,” 10th
International LS-DYNA Users Conference, Dearborn, MI, 8-10 June 2008.
16. Lyle, K. H., Bark, L. W., Jackson, K. E., "Evaluation of Test/Analysis Correlation
Methods for Crash Applications," Proceedings of the 57th American Helicopter
Society Annual Forum, Washington D.C., May 9-11, 2001.
17. Sareen, A. K., Sparks, C., Mullins, B. R., Fasanella, E. L., and Jackson, K. E.,
"Comparison of Soft Soil and Hard Surface Impact Performance of a Crashworthy
Composite Fuselage Concept," Proceedings of the AHS Forum 58, Montreal,
Canada, June 11-13, 2002.
18. Fasanella, E. L., Jackson, K. E. and Lyle, K.H., “Multi-Terrain Impact Testing
and Simulation of a Composite Energy Absorbing Fuselage Section,” American
Helicopter Society 60th Annual Forum, Baltimore, MD, June 8-10 2004.
19. Hughes, K., Campbell, J., Vignjevic, R., “Application of the finite element
method to predict the crashworthy response of a metallic helicopter under floor
structure onto water,” International Journal of Impact Engineering, Vol. 35,
2008, pp. 347 – 362.
20. Byar, A., “A Crashworthiness Study of a Boeing 737 Fuselage Section” Ph.D.
Dissertation, Drexel University, May 2003.
21. Byar, A. D., Ko, J. and Rassaian, M, “737 Fuselage Section Drop Test Simulation
Using LS-Dyna Finite Element Method,” International Crashworthiness
Conference, Kyoto, Japan, June 22-25 2008.
22. Jackson, K. E., Kellas, S., Littell, J. D., “Full-Scale Crash Test of an MD-500
Helicopter with Deployable Energy Absorbers,” Proceedings of the 66th
American Helicopter Society Annual Forum, Phoenix, AZ, May 11-13, 2010.
23. Ludin, D., Renninger, M., “The Development of a Floor Former Concept
Incorporating Energy Absorbing Composite Tubes,” Proceedings of the 65th
American Helicopter Society Annual Forum, Grapevine, TX, May 27-29, 2009.
248
25. Farley, G., “Effect of Specimen geometry on the Energy absorption of Composite
Materials,” Journal of Composite Materials, Vol. 20, 1986, pp. 390-400.
27. Thuis, H., Metz, V. H., “The influence of triggering configurations and laminate
lay-up on the failure mode of composite crush cylinders,” Journal of Composite
Structures, Vol. 28, 1994, pp. 131-137.
29. Feraboli, P., “Development of Corrugated Test Specimen for Composite Materials
Energy Absorption,” Journal of Composite Materials, Vol. 42, 2008, pp. 229-256.
30. Palanivelu, S., Paepegem, W., Degrieck, J., Kakogiannis, D., Ackeren, J.,
Hemelrijck, D., Wastiels, J., Vantomme, J., “Experimental study on the axial
crushing behavior of pultruded composite tubes,” Polymer Testing, Vol. 29, 2010,
pp. 224-234.
31. Yang, Y., Nakai, A., Uozumi, T., Hamada, H., “A Collapse Trigger Mechanism
for Progressive Crushing of FRP”, Proceedings of the American Society for
Composites: Twentieth Technical Conference, 2005.
32. Ueda, M., “Energy Absorption Capability of CFRP Tube using Load-Control
Attachment,” Proceedings of the American Society for Composites: Twenty
Fourth Technical Conference, Vol 4, 2009, pp. 2340-2351.
35. Kim, J., Yoon, H. and Shin, K., “A study on the crushing behavior of composite
circular tubes with different reinforcing fibers,” International Journal of
Composite Engineering, Vol. 38, 2011, pp. 198-207.
36. Hanagud, S., Craig, J. I., Sriram, P., Zhou, W., “Energy Absorption Behavior of
Graphite Epoxy Composite Sine Web,” Journal of Composite Materials, Vol. 23,
1989, pp. 448-460.
249
37. Jackson, K., Morton, J., Lavoie, J. A., Boitnott, R., “Scaling of Energy Absorbing
Composite Plates,” Journal of the American Helicopter Society, Vol. 39, 1994,
pp. 17-23.
38. Lavoie, J. A., Kellas, S., “Dynamic Crush Tests of Energy Absorbing Laminated
Composite Plates,” Journal of Composites: Part A, Vol. 27A, 1996, pp. 467-475.
39. Joosten, M. W., Dutton, S., Kelly, D., Thomson, R., “Experimental Evaluation of
the Crush Energy Absorption of Triggered Composite Sandwich Panels Under
Quasi-Static Edgewise Compressive Loading,” Journal of Composites: Part A,
Vol. 41, 2010, pp. 1099-1106.
40. Mamalis, A. G., Manolakos, D. E., Ioannidis, M. B., Papapostolou, D. P., “The
static and dynamic axial collapse of CFRP square composite tubes: Finite element
modeling,” Journal of Composite Structures, Vol. 74, 2006, pp. 213-225.
41. Hamada, H., Coppola, J. C., Hull, D., Maekawa, Z., and Sato, H., “Comparison of
energy absorption of cabon/epoxy and carbon/PEEK composite tubes,” Journal of
Composites: Part A, Vol. 23, 1992, pp. 245-252.
42. Feraboli, P., Wade, B., Deleo, F., Rassaian, M., “Crush Energy Absorption of
Composite Channel Section Specimens,” Journal of Composites: Part A, Vol. 40,
2009, pp. 1248-1256.
43. Mahdi, E., Sahari, B.B., Hamouda, A.M.S. and Khalid, Y.A., “An experimental
investigation into crushing behavior of filament-wound laminated cone-cone
intersection composite shell,” Journal of Composite Materials, Vol. 51, 2001, pp.
211-219.
44. Alkateb, M., Mahdi, B.B., Hamouda, A.M.S. and Hamdan, M.M., “On the energy
absorption capability of axially crushed composite elliptical cones,” Journal of
Composite Structures, Vol. 66, 2004, pp. 495-501.
46. Czaplicki, M. J., Robertson, R. E., Thornton, P. H., “Comparison of Bevel and
Tulip Triggered Pultruded Tubes for Energy Absorption,” Journal of Composite
Science and Technology, Vol. 40, 1991, pp. 31-46.
47. Jimenez, M. A., Miravete, A., Larrode, E., Revuelta, D., “Effect of Triggering
Geometry on Energy Absorption in Composite Profiles,” Journal of Composite
Structures, Vol. 48, 2000, pp. 107-111.
48. Song, H. W., Du, X. W., “Off Axis Crushing of GFRP Tubes,” Journal of
Composite science and Technology, Vol. 62, 2005, pp. 246-257.
250
49. Mamalis, A. G., Manolakos, D. E., Iannidis, M. B., Papapostolu, D. P., “On the
Crushing Response of Composite Sandwich Panels Subjected to Edgewise
Compression: Experimental,” Journal of Composite Structures, Vol. 71, 2005, pp.
2065-73.
50. Velecela, O., Found, M. S., Soutis, C., “Crushing Energy Absorption of GFRP
Sandwich Panels & Corresponding Monolithic Laminates,” Journal of
Composites: Part A, Vol. 38, 2005, pp. 1149-58.
51. Velecela, O., Soutis, C., “Prediction of Crushing Morphology of GRP Composite
Sandwich Panels Under Edgewise Compression,” Journal of Composites: Part B,
Vol. 38, 2009, pp. 914-923.
52. Palanivelu, S., Paepegem, W., Degrieck, J., Vantomme, J., Kakogiannis, D.,
Ackeren, J., Hemelrijck, D., Wastiels, J., “Crushing and energy absorption
performance of different geometrical shapes of small-scale glass/polyester
composite tubes under quasi-static loading conditions,” Journal of Composite
Structures, Vol. 89, 2009, pp. 399-407.
53. Thornton, P.H. and Jeryan, R.A., “Crash energy management in composite
automotive structures,” International Journal of Impact Engineering, Vol. 7,
1988, pp. 167-180.
54. Lau, S.T.W., Said, M.R. and Yaakob, M.Y., “On the effect of geometrical designs
and failure models in composite axial crushing: A literature review,” Journal of
Composite Structures, Vol. 94, 2012, pp. 803-812.
55. Meredith, J., Ebsworth, R., Coles, S.R., Wood, B.M. and Kirwan, K., “Natural
fiber composite energy absorption structures,” Journal of Composite Science and
Technology, Vol. 72, 2012, pp. 211-217.
56. Taher, S. T., Zahari, R., Ataollahi, A., Mustapha F. and Basri, “A double-cell
foam-filled composite block for efficient energy absorption under axial
compression,” Journal of Composite Structures, Vol. 93, 2011, pp. 992-1007.
57. Huang, J., Wang, X., “Numerical and experimental investigations on the axial
crushing response of composite tubes,” Journal of Composite Structures, Vol. 91,
2009, pp. 222-228.
58. Mamalis, A. G., Manolakos, D. E., Ioannidis, M. B., Papapostolou, D. P., “The
static and dynamic axial collapse of CFRP square composite tubes: Finite element
modeling,” Journal of Composite Structures, Vol. 74, 2006, pp. 213-225.
59. Palanivelu, S., Paepegem, W., Degrieck, J., Kakogiannis, D., Ackeren, J., and
Hemelrijck, D., “Parametric study of crushing parameters and failure patterns of
pultruded composite tubes using cohesive elements and seam, Part I: Central
delamination and triggering modeling”. Polymer Testing, Vol. 29, pp 729-741,
2010.
251
60. Palanivelu, S., Paepegem, W., Degrieck, J., Kakogiannis, D., Ackeren, J., and
Hemelrijck, D., “Parametric study of crushing parameters and failure patterns of
pultruded composite tubes using cohesive elements and seam, Part II: Multiple
delaminations and initial geometric imperfections”. Polymer Testing, Vol. 29, pp
803-814, 2010.
61. Deleo, F., Wade, B., Feraboli, P., Rassaian, M., “Crashworthiness of Composite
Structures: Experiment and Simulation,” Proceedings of the 50th AIAA
Structures, Structural Dynamics and Materials Conference, Palm Springs, CA,
May 4-7, 2009.
63. Xiao, X., Botkin, M., Johnson, N. L., “Axial crush simulations of braided carbon
tubes using MAT58 in LS-DYNA,” Thin-Walled Structures, Vol. 38, 2009, pp.
2247-59.
64. McGregor, C., Vaziri, R., Xiao, X., “Finite element modeling of the progressive
crushing of braided composite tubes under axial impact,” International Journal of
Impact Engineering, Vol. 37, 2010, pp. 662-72.
65. McGregor, C., Zobeiry, N., Vaziri, R., Poursartip, A., “A constitutive model for
progressive compressive failure of composites,” Journal of Composite Materials,
Vol. 42, 2008, pp. 2687-2716.
66. Xiao, X., “Modeling energy absorption with a damage mechanics based
composite material model,” Journal of Composite Materials, Vol. 43, 2009, pp.
427-444.
67. Joosten, M. W., Dutton, S., Kelly, D., Thomson, R., “Experimental and
Numerical Investigation of the Crushing Response of an Open Section Composite
Energy Absorption Element,” Journal of Composites Structures, 2010.
72. Siromani, D., Byar, A., Awerbuch, J. and Tan, T., “Crashworthiness Analysis of a
Boeing 737 Fuselage Section: A Parametric Study on the Effects of Angle of
Impact and Friction,” Submitted to: International Journal of Crashworthiness.
73. Hallquist, J. O., “LS-DYNA Keyword User’s Manual, Version 971, Volumes 1
and 2,” Livermore Software Technology Corporation, Livermore, CA, May 2007.
74. “Advisory Circular (AC) No. 700-028” Transport Canada. 8 Aug. 2013. Available
at: http://www.tc.gc.ca/.
75. “Boeing 737 family – Technical Information.” Boeing. 3 Apr. 2010. Available at:
http://www.boeing.com/commercial/737family/specs.html.
77. Thomas, M. A., Chitty, D. A., Gildea, M. L., and T’Kindt, C. M., “Constitutive
Soil Properties for Unwashed Sand and Kennedy Space Center,” NASA/CR-
2008-215334.
79. Siromani, D., Awerbuch, J., Tan, T.-M., “Finite Element Modeling of the
Crushing Behavior of Thin-Walled CFRP Tubes under Axial Compression,”
Submitted to: Composites Part B: Engineering.
80. Siromani, D., Awerbuch, J., Tan, T.-M., “Finite Element Modeling of the
Crushing Behavior of Thin-Walled Open Cross-Section CFRP Members under
Axial Compression,” Submitted to: Journal of Reinforced Plastics and
Composites.
81. Siromani, D., Henderson, G., Mikita, D., Mirarchi, K., Park, R., Smolko, J.,
Awerbuch, J. and Tan, T.-M., “An Experimental Study on the Effect of Failure
Trigger Mechanisms on the Energy Absorption Capability of CFRP Tubes under
Axial Compression,” Submitted to: Composites Part A: Applied Science and
Manufacturing.
82. Siromani, D., Cheng, B., DeLuca, M., Donegan, D., Giberson, P., Mucerino, C.,
Awerbuch, J. and Tan, T.-M., “Experimental Investigation on the Energy
Absorption Capacity During Crushing of Axially Loaded Thin-Walled Gr/Ep
Members,” Submitted to: Journal of Composite Materials.
83. Camanho, P., Maimi, P., and Davila, C., “Prediction of size effects in notched
laminates using continuum damage mechanics”. Composite Science and
Technology, Vol. 67, pp 2715-2727, 2007.
253
84. Krueger, R., Paris, I. L., O’Brien, T. K. and Minguel, P. J., “Fatigue Life
Methodology for Bonded Composite Skin/Stringer Configurations,” Technical
Report NASA/TM-20001-210842 ARL-TR-2432, 2001.
86. Loikkanen, M., Praveen, G., and Powell, D., “Simulation of Ballistic Impact on
Composite Panels,” 10th International LS-DYNA Users Conference, Dearborn,
MI, 2008.
87. Hansen, P. and Martin, R., “DCB, 4ENF and MMB delamination characterization
of S2/8552 and IM7/8552”. Final Technical Report US Army N68171-98-M-
5177. 1999.
88. Turona, A., Davila, C., Camanho, P. and Costa, J., “An engineering solution for
mesh size effects in the simulation of delamination using cohesive zone models”.
Engineering Fracture Mechanics. Vol. 74, pp 1665-1682, 2007.
90. Siromani, D., Awerbuch, J., Tan, T.-M., “Modeling Methodologies for Simulating
the Axial Crushing Behavior of CFRP Members,” Submitted to: Finite Elements
in Analysis and Design.
91. Hsiao, H. M. and Daniel, I. M., “Strain rate behavior of composite materials,”
Composites: Part B, Vol. 29B, 1998, pp. 521-533.
93. Hsiao, H. M., Daniel, I. M. and Cordes, R.D., “Strain rate effects on the
transverse compressive and shear behavior of unidirectional composites,” Journal
of Composite Materials, Vol. 33, 1999, pp. 1620-1642.
94. Koerber, H., Xavier, J. and Camanho, P.P., “High strain rate characterization of
unidirectional carbon-epoxy IM7-8552 in transverse compression and in-plane
shear using digital image correlation,” Mechanics of Materials, Vol. 42, 2010, pp.
1004-1019.
95. Bing, Q. and Sun, C.T., “Modeling and testing strain rate-dependent compressive
strength of carbon/epoxy composites,” Composites Science and Technology, Vol.
65, 2005, pp. 2481-2491.
96. Hosur, M.V., Alexander, J., Vaidya, U.K. and Jeelani, S., “High strain rate
compression response of carbon/epoxy laminate composites,” Composite
Structures, Vol. 52, 2001, pp. 405-417.
254
100. Schultz, M., Hyer, M. and Fuchs, H., “Static and dynamic energy-absorption
capacity of graphite-epoxy tubular specimens,” Mechanics of Composite
Materials and Structures, Vol. 8, 2001, pp. 231-247.
101. Jackson, A., Dutton, S., Gunnion, A.J. and Kelly, D., “Investigation into laminate
design of open carbon-fiber/epoxy sections by quasi-static and dynamic
crushing,” Composite Strutures, Vol. 93, 2011, pp. 2646-2654.
102. Farley, G., “The effect of crushing speed on the energy-absorption capability of
composite tubes,” Journal of Composite Materials, Vol. 25, 1991, pp. 1314-1329.
103. Mamalis, A. G., Manolakos, D. E., Ioannidis, M. B., Papapostolou, D. P., “The
static and dynamic axial collapse of CFRP square composite tubes: Finite element
modeling,” Journal of Composite Structures, Vol. 74, 2006, pp. 213-225.
104. Goldberg, R. K., Roberts, G. D., Blinzler, B. J., Kohlman, L. W. and Binienda,
W. K., “Characterization and Analysis of Triaxially Braided Polymer Composites
Under Static and Impact Loads,” Earth and Space 2012, pp. 1254-1263.
255
VITA
DEEPAK SIROMANI
EDUCATION
Ph.D., Mechanical Engineering Drexel University, Philadelphia, PA December 2013
Research: Crashworthy Design and Analysis of Aircraft Structures
Process Engineering Intern Rohm and Hass, Newark, DE October 2005 – March 2006
• Assisted with process improvement projects at different stages of production
• Implemented a defects tracking system and trained operators to use it
• Updated P&ID's to reflect recent changes
• Developed and revised Return to Service procedures and other operating procedures for various
equipment
September 2004 –
Validation Engineering Intern GlaxoSmithKline, Upper Merion, PA
March 2005
• Performed Validation Maintenance studies and Installation & Operational Qualifications
on production and non-production equipment including autoclaves, depyrogenation
ovens, vessels and freezers
• Generated protocols for Cleaning Validation on various filter skids
• Developed and revised Standard Operating Procedures for studies and operating
calibration equipment
RESEARCH INTERESTS
Crashworthiness of aircraft structures; Impact, penetration and crush analysis; Damage mechanics of
composite materials; Fluid-structure interaction (ALE and SPH methods); Material and damage modeling
TECHNICAL SKILLS
LS-DYNA, LS-PrePost, Hypermesh, ANSYS, ABAQUS, Creo Parametric (formerly Pro/Engineer),
AutoCAD, MatLab, Mathematica
TEACHING EXPERIENCE
Course Coordinator - Drexel University, Philadelphia, PA June 2009 – December 2012
Introduction to CAD
• Developed and managed a hybrid online and in-class course to introduce all first-year
engineering students to computer-aided design
• Created online lectures and tutorials, in-class and take-home assignments to assist students in learning
the basics of AutoCAD and Creo Parametric (formerly Pro/ENGINEER)
Teaching Fellow - Freshman Drexel University, Philadelphia, PA September 2008 – June 2013
Engineering Design Labs
• Responsible for running weekly labs and mentoring freshman design projects
• Assisted in the development of a reverse engineering lab module
SELECTED PUBLICATIONS
Siromani, D., Cheng, B., DeLuca, M., Donegan, D., Giberson, P., Murcerino, C., Awerbuch, J. and Tan,
T., “An Experimental and Numerical Study on the Energy Absorption Mechanisms of Axially Loaded
Graphite/Epoxy Members with Various Cross-sections,” Proceedings of the 2012 Aircraft Airworthiness
& Sustainment Conference, April 2-5, 2012, Baltimore, MD.
Siromani, D., Henderson, G., Mikita, D., Mirarchi, K., Park, R., Smolko, J., Ludin, D., Awerbuch, J. and
Tan, T., “Experimental and Numerical Crashworthiness Investigation into the Energy Absorption
Mechanisms of Axially Loaded CFRP Tubes,” Proceedings of the American Society for Composites 26th
Technical Conference, September 26-28, 2011, Montreal, QC, Canada.
Siromani, D., Awerbuch, J. and Tan, T., “Multi-terrain Crashworthiness Simulations of a Boeing 737
Fuselage Section,” Proceedings of the 2010 Aircraft Airworthiness & Sustainment Conference, May 10-
13, 2010, Austin, TX.
Siromani, D., Byar, A., Awerbuch, J. and Tan, T., “Crashworthiness Simulation of a Boeing 737
Fuselage Section: A Parametric Study on the Effects of Angle of Obliquity and Friction on its Dynamic
Response,” Proceedings of Aging Aircraft 2009, May 4-7, 2009, Kansas City, MO.
INVITED PRESENTATIONS
Siromani, D., Awerbuch, J. and Tan, T., “Multi-terrain Crashworthiness Simulations of a Narrow Body
Transport Fuselage under Vertical and Oblique Impact Conditions,” CMH-17 Crashworthiness Working
Group Meeting, November, 2011, Wichita, KS.
257
Siromani, D., Awerbuch, J. and Tan, T., “An Experimental and Numerical Study on Energy Absorption
Mechanisms of CFRP Members,” CMH-17 Crashworthiness Working Group Meeting, November, 2011,
Wichita, KS.