JET ENGINE ROTOR FRAGMENT IMPACT ON COMPOSITE

JET ENGINE ROTOR FRAGMENT IMPACT ON COMPOSITE
JET ENGINE ROTOR FRAGMENT IMPACT ON COMPOSITE PANELS
Matti Loikkanen
David Powell
Journal of Structural Mechanics
Vol. 40, No 4, 2007, pp. 80 - 94
ABSTRACT
Ballistic impact on composite panels is studied in this paper. Spherical projectiles ware shot
against composite plates, impact damage was observed, and the initial and exit speeds were
measured. Explicit finite element program LS-DYNA, with some enhancements was used
to simulate the impact event. An analysis process has been developed and the simulated
results were compared with the tests results. Each ply in the panels was modeled with 8node solid elements with material model allowing complete three dimensional progressive
failure. Fracture mechanics based contact definition was modeled between plies with full
delamination capability between each ply.
INTRODUCTION
Modern jet airplane is a remarkably safe transportation machine. A two engine aircraft can
now fly non-stop virtually from any airport to any other airport in the world. However,
because the engine rotors spin at high speed and are designed for the minimum weight, the
blades occasionally break with catastrophic consequences. The worst accident caused by
rotor failure happened at Sioux City, Iowa in 1989. A DC-10 tail engine had a massive inflight failure, the main rotor disk failed, the pieces cut all the hydraulic lines, the aircraft
was rendered uncontrollable and crashed killing 123 people. The failure was caused by a
small manufacturing defect in the disk. Several other accidents have occurred over the
years, but the good news is that fatal accidents are getting less frequent. To put this in
perspective, the probability of a massive engine failure on any given flight is about 1.e-09.
On the other hand, the probability of hitting the jackpot on the Washington State Lottery is
about 1.0e-07. The improved aviation safety is due to better maintenance, better
manufacturing quality and better designs. The aviation authorities impose strict design
requirements and the aircraft companies are making a great effort to make the airplanes
safe. The electrical and hydraulic lines have to be adequately protected, the fuel tanks have
to be able to stand a specified impact without leaking and a fragment from one engine
cannot disable the other engine, etc.
80
Over the past several years The Boeing Company in cooperation with the Federal Aviation
Administration (FAA) and University California, Berkeley (UCB) and other research
facilities has conducted material testing and numerical computations on ballistic impact on
various airplane structures. FAA has funded and administered the work, UCB performed
mainly the testing and Boeing the computations. The purpose has been to develop
computational methods to analyze a high speed fragment impacting on the chosen targets.
Material testing had two aspects: small coupon tests to determine the material properties
needed as input to the computer program and ballistic testing where a projectile was shot
against a target specimen and penetration was measured. Then finite element (FE) models
were created and analyzed and the computational results compared with the test results.
Once a reliable computational tool is available then it is much cheaper to use it than
laboratory testing. In addition, computer analysis yields much more information than a test
and makes it possible to understand the underlying physics better, which in turn makes the
design process more accurate and faster. Commercially available explicit finite element
program LS-DYNA, developed by Livermore Software Technology Corporation (LSTC),
was chosen as the computational tool [2, 3].
Fig. 1. Massive jet engine failure.
81
BALLISTIC TESTING
Test Setup: One of many Boeing composite materials was used in the study [4]. Flat panels
with three different thicknesses, 0.06 in, 0.12 in and 0.24 in were tested. The nominal ply
thickness is about 0.0075 in and so the panels had 8, 16, and 32 plies, respectively. The
stacking sequence, for the 16 ply panel, starting from the bottom is: 45, 90, -45, 0, 45, 90, 45, 0, (center) 0, -45, 90, 45, 0, -45,90, 45 degrees. I.e. these are the ply rotations from the
reference axis. The stacking is symmetric about the center plane. In addition, the 4-ply (0, 45, 90, 45)-sequence repeats itself around the center plane. The 32-ply panel has four 4-ply
stacks both above and below the center plane in the same pattern. The overall panel with all
three thicknesses with this stacking sequence has almost isotropic bending and in-plane
stiffness properties (therefore called “quasi-isotropic”) and is a common design. Due to the
symmetric lay-up, the bending and in-plane stiffness are also uncoupled. The total size of
each panel was 12x12 inches and when mounted onto a steel test frame, the clear test area
was 10x10 inches. Several panels for each of the three thicknesses were tested to get some
statistical reliability. The mounting steel frame was stiff enough so that the plate boundary
was considered clamped. It takes 1 to 3 milliseconds for the projectile to penetrate the plate
depending on the impact velocity and thickness. This is long enough time for waves to
travel from impact point to the plate boundary and back before the event is over. This
means that the reflecting waves interfere with the penetration. The finite element model has
clamped boundaries and accounts for the reflecting waves. This was considered acceptable
although not ideal.
A nitrogen gas gun with half-inch diameter spherical steel projectiles was used in all the
tests shots. The spheres were one half inch in diameter and weighed 0.018 pounds (8.2
grams). The gun used industrial grade compressed nitrogen with a maximum pressure of
1500 psi. The gun barrel was approximately 52 in long. After 39 in of smooth barrel, there
were slits in the barrel downstream that relieved the pressure behind the projectile. A
regulator controlling the pressure could be set for any value between 25 and 1500 psi. The
solenoid valve that released the pressure was controlled by an electronic control box and
triggered from the adjacent room. The gun was able to propel the projectiles up to 1000 ft/s
speed. This was high enough for complete penetration with the 32-ply panels. In this study
all the shots were perpendicular and at the center of the plate. The advantage of using gas
gun over powder gun is that the pressure and the projectile velocity can be controlled more
82
precisely. The impact and exit (residual, after penetration) velocities were measured. Each
shot was videotaped with a high speed camera and the damage and hole to the panels were
photographed. Two light beams in the front of the panels were used to measure the impact
velocity while the exit velocity was determined from the high speed video.
Ballistic Limit: The impact and exit velocities are usually plotted on an x-y scale: impact
velocity on the horizontal axis and the exit velocity on the vertical axis. A typical ballistic
event is shown in Fig. 3 [1]. The shape of this plot is approximately the same regardless of
ballistic conditions, material target thickness, etc. The point where the exit velocity
becomes greater than zero is called the ballistic limit. After the limit, first there is steep rise
in the exit velocity plot and then it flattens out to a 45-degree slope and stays approximately
at constant slope indefinitely. To illustrate further, 45-degree reference line is drawn, i.e.
this has the same impact and exit velocities. The difference between the 45-degree
reference line and the real impact/exit velocity plot represents the velocity (and kinetic
energy) that the projectile looses in the penetration. The sharp rise in the plot comes from
material ductility and then at higher velocities the projectile velocity loss is practically
constant.
Exit Velocity
Idealized Ballistic Event
ref. line
exit speed
Impact Velocity
Fig. 2. Ballistic impact characteristics.
83
Test Results: Typical entry and exit holes for the spherical projectiles are shown in Fig. 3.
The entry side hole (on the left) is always cleaner than the exit side hole. Exit side always
has visible delamination. Inspection shows that it is mostly on the last few plies and extends
several inches away from the hole mainly in the fiber direction. However, it was impossible
with the available equipment to quantify exactly how much delamination there was
between each ply inside the plate and how much energy was dissipated in total
delamination.
The impact and exit velocities were plotted for all the test points for the three plate
thicknesses and are shown in Fig. 4. The thinnest, 8-ply panel (blue squares) has the lowest
ballistic limit, about 200 ft/sec, and takes the least amount energy for penetration, the 16ply ballistic limit is about 300 ft/sec (red triangles) and the 32-ply panel ballistic limit is
450 ft/sec. The dots form the same overall shape as shown in Fig. 2. The rise after the
ballistic limit is relatively low since composites lack ductility compared to aluminum, for
example
The initial kinetic energy vs. dissipated energy is plotted in Fig. 5. It shows that the
dissipated energy is approximately constant with each plate thickness over the velocity
range considered here. In Fig 6 the absorbed energy is plotted per one ply. The thicker plate
gives some advantage. Is seen that in the dissipated energy per ply increases somewhat with
the increased number of plies.
84
Fig. 3. Damage from spherical projectile impact; entry hole on the left, exit hole on the
right.
Coupon Tests: Most of the required material data [3] have been tested in the Boeing
Materials Labs and was available in the Boeing database. Fig 7 shows a typical forcedeflection curve for one ply from a static tension test. First there is a linear elastic region
and then a smooth round transition and then finally a sudden drop with the complete failure.
The shape would be very similar for compression and shear also. Tension in fiber direction
is by far the strongest direction. Unlike with metals, there is no long plastic region before
failure and the total failure energy is low.
800
700
8 Ply Composite Panel
16 Ply Composite Panel
Residual Velocity (ft/s)
600
32 Ply Composite Panel
Reference - Vin = Vout
500
400
300
200
100
0
0
100
200
300
400
500
600
Initial Velocity (ft/s)
Fig. 4. Ballistic velocities.
85
700
800
900
120
8-Ply Composite Panel
16-Ply Composite Panel
32-Ply Composite Panel
Energy Absorbed (J)
100
Reference - %100 Absorbed
80
60
40
20
0
0
50
100
150
200
Initial Energy (J)
Fig. 5. Absorbed energy.
COMPUTATIONS
Description of the Event: Fragment impact and penetration into a laminated composite
plate is difficult to compute correctly. As the projectile pushes against the plate, the contact
force between them builds up, the plies break one by one and the shear and normal stresses
between the plies will separate them from each others (plies delaminate). The deformation
around the projectile is localized and therefore it is necessary to consider a three
dimensional stress and strain state. Fiber is brittle with high tension strength but has almost
86
no strength in the other directions. Resin is viscoplastic with ductile failure and heavily rate
dependent. Since there is practically no plastic deformation before failure, the energy
required to fail the material is small. This makes composites weak under transverse impact
loading. The plies fail mainly in shear and compression. The hydrostatic component has a
significant part in the failure, but is not well captured in the available material models. The
resin and fibers break at different strain levels, but is not accounted for in the present
material model. The available material models simply combine the fiber and resin into one
homogenous orthotropic material.
4
Energy Absorbed per Ply (J/ply)
3.5
3
2.5
2
1.5
1
8-Ply Composite Panel
16-Ply Composite Panel
0.5
32-Ply Composite Panel
0
0
2
4
6
8
10
Initial Energy per Ply (J/ply)
Fig. 6. Energy absorbed per ply.
87
12
14
90
80
70
60
50
40
30
20
10
0
0
0.05
0.1
0.15
0.2
0.25
Fig. 7. Force vs. deflection for one ply.
The ply delamination (debonding, decohesion) is caused by in-plane shear (Mode II) and
normal tension (Mode I) between the pies. The crack tip runs in the direction of least
resistance between two plies. The energy released in the process is measured by the
dynamic fracture toughness, which is determined by testing. Fracture toughness is different
between Mode I and Mode II and is considered in the present material models. It is also
different between the fiber and perpendicular to fiber directions, which is not considered in
the present models. The two shear directions are lumped into one.
Computational Models: In the explicit FE analysis the mass matrix is uncoupled which
allows the equations to be solved explicitly without assembly and decomposition of the
global matrices. The stability condition requires that the time step is smaller than the time it
takes for acoustic wave to travel across one element. Wave speed is related to the largest
eigenvalue in the model, which in turn comes from the smallest element in the system [6].
Typical time step in explicit FE runs is 0.5e-7 to 1.0e-07 seconds and analysis up to 5
milliseconds requires some 50000 steps. For the best accuracy and computer economy, the
finite element mesh needs to be as uniform and smooth as possible.
88
Figure 8 shows a typical FE mesh for the 32-ply plate and spherical projectile. Only one
half of the model was considered in the computer analyses, although the model is not
strictly symmetric.
Fig. 8. Top view of the FE mesh on the left, side view on the right.
One 8-node solid element was used through each ply. The element size under the projectile
was 0.05 by 0.05 by 0.0075 in (ply thickness is 0.0075 in) and all the elements had
rectangular shape. A typical model had about 200000 elements. The bonding between the
plies was modeled with “contact tiebreak surfaces” which has the facture mechanics based
delamination capability and requires fracture toughness input. The contact was defined
between each ply allowing them to delaminate from each other. The impact force between
the projectile and the plate was modeled with “eroding contact surfaces”. The contact
surface redefines itself after an element fails and is removed from the model, thus tracking
correctly the force between the projectile and the plate
“Composite progressive damage and failure” material model was chosen for the composite
ply material. This model requires the normal and shear stiffness in all three directions with
the corresponding Poisson’s Ratios. The material strength in each direction, 9 in total, is
required as input. The model uses the material strength as the start of progressive failure
and the progressive failure then follows an exponential curve with user defined slope (m).
The test data (Fig 7) should match closely with the exponential unloading curve shown in
Fig. 9
89
0.18
m= 0.1
0.15
m= 1.0
m= 2.0
Stress (Msi)
0.12
m= 4.0
m= 8.0
0.09
m= 32.0
0.06
0.03
0.00
0.00E+00
3.00E-03
6.00E-03
9.00E-03
1.20E-02
1.50E-02
1.80E-02
Strain
Fig. 9. Idealized ply stress-strain response.
Computational Results: Fig 10 shows the penetration and failure of the 32-ply plate for
initial velocity of 574 ft/sec.
Some elements fail and are deleted, some elements are pushed forward under the projectile
and all the plies delaminate around the hole. Fig 11 has a typical projectile velocity profile
during penetration. The exit velocity was the primary unknown in the computer
simulations. Fig 12 has the summary of both the computed and experimentally determined
exit velocities for the 32-ply, 16-ply, and 8-ply plates for all the shots.
The thicker the plate, the more errors have chance to accumulate and more difficult it is to
get an accurate simulation because the time integration requires more steps. Furthermore
when the impact velocity is close the ballistic limit the impact event takes longer, the
program has to take more time steps and the programs requires more time steps and the
results tend to be less accurate.
90
Fig. 10. Penetration and failure.
91
Fig. 11. .Projectile velocity profile.
4 Discussion
Jet engine rotor fragment impact on composite structures has been studied both
experimentally and computationally. The computer hardware and software have reached
maturity so that reliable ballistic simulations are now possible, provided that accurate
material data is available and the analyst has some experience with the software. The
available composite material models have deficiencies which are corrected presently.
Simulation of the ply failure and delamination require very specialized material data and
require special testing. The numerical simulations are computationally intensive. A typical
simulation run takes at least overnight. Nevertheless the computer simulations are
substantially less expensive and yield more information than laboratory testing.
92
Finite element analysis of material failure does not converge to anything when the mesh is
refined. Too coarse a mesh (elements too large) can grossly overestimate the energy
required to fail the material (i.e. all materials, metals, composites, etc). When the mesh is
refined (elements are made smaller) the required energy to fail the material decreases
without bounds. When the elements are too small, the energy is underestimated and the
projectile penetrates too easily. There does not seem to be any reliable way to determine the
optimum mesh density without actually tuning the mesh with known test results.
Researches are making efforts to minimize this effect, but so far no one is promising a
complete cure.
Meanwhile, the present state is mature enough so that it can be used confidently for
practical aircraft design work.
900
8-Ply Simulation
800
Residual Velocity (ft/s)
700
600
8-Ply Experiment
16-Ply Simulation
16-Ply Experiment
32-Ply Simulation
500
400
32-Ply Experiment
Residual Velocity =
Initial Velocity
300
200
100
0
100
200
300
400
500
600
700
-100
Initial Velocity (ft/s)
Fig. 12. Spherical projectile: computations vs. testing.
93
800
900
5 References
1. Goldsmith, Werner, “Review of Non-Ideal projectile impact on targets”, International
Journal of Impact Engineering, Vol. 22, 1999.
2. Hallquist, John, O. “LS-DYNA Theoretical Manual”, Livermore Software Technology
Corporation, Livermore, CA, March 2006.
3. Hallquist, John, O, et al, “LS-DYNA Keyword User’s Manual”, Livermore Software
Technology Corporation, Livermore, CA, May 2007.
4. Powell, David, Zohdi T, and Johnson, George, “Failure Characterization of Composite
Aircraft Materials Under Ballistic Impact”, University of California, Berkeley, Report to
FAA to be published in 2008.
5. Powell, David and Loikkanen, Matti, “Explicit Finite Element Modeling of Composite
Plates for Containment of Critical Aircraft Components from Jet Engine Debris”,
University of California, Berkeley and The Boeing Co, Report to FAA, to be published in
2008.
6. Irons, B. M. and Treharne C., A Bound Theorem for Eigenvalues and its Practical
Applications”, 2nd Conf. on Matrix methods in Structural Mechanics, Wright-Patterson Air
Force Base, Ohio, 1971.
Matti Loikkanen
Boeing Commercial Airplanes,
Propulsion Engineering Technology &
Research, Seattle, WA, USA
David Powell
University of California,
Mechanical Engineering Department,
Berkeley, CA, USA
94
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement