Вы находитесь на странице: 1из 20

50th AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials Conference<br>17th

4 - 7 May 2009, Palm Springs, California

AIAA 2009-2451

Finite Element Analysis for Advanced Repair Solutions: A


Stress Analysis Study of Fastened and Bonded Fuselage Skin
Repairs
Domenico Furfari * and Nikolaus Ohrloff.
Airbus, Hamburg, Germany
Gebhard Schmidt
HIGH END Engineering GmbH, Hamburg, Germany

The advanced repair solutions investigated within this research programme, both
mechanically fastened and adhesively bonded repairs, could extend their in-service life
beyond the Design Service Goal (DSG) of the aircraft in which they are embodied. Beside the
test activity (discussed in previous publications), calculation methods (based on Finite
Element models) were developed to improve current methods to predict the fatigue behavior
of repairs and to optimize the design principles of the advanced repair solutions. The use of
internal doubler to reduce the secondary bending, different fastener type or additional
fastener row, as well as, the influence of fastener installation (Hi-Lok interference fit) are
reported. Other advanced repair solutions, such as bonded repair or the use of glare
doubler to repair aluminum skin, were also investigated and reported in this work. In this
paper different kind of Finite Element (FE) models used to simulate the tests are reported.
The stresses resulted from the FE analysis were compared with the ones measured
experimentally by strain gauges.

Nomenclature
s
k
r
rdepth
U0
F0
Smax
Smin
Sm
Sa

E
S70E
n

=
=
=
=
=
=
=
=
=
=
=
=
=
=
=

Static coefficient of friction


Kinematic coefficient of friction
Fastener radius
Fastener countersunk depth
Initial GAP opening
GAP preload
Maximum stress
Minimum stress
Mean stress
Stress amplitude
Strain
Stress
Youngs Modulus
Secant modulus at 0.7E slope
Ramber-Osgood shape factor

I. Introduction
A. Background of the Research

Research & Technologies, Structural Analysis Stress Methods, Kreetslag 10, 21129 Hamburg, Germany
Senior Expert Fatigue, Structural Analysis, Kreetslag 10, 21129 Hamburg, Germany

Stress Engineer, Stress and Methods, Georg-Heyken-Strae 4, 21147 Hamburg, Germany.


1
American Institute of Aeronautics and Astronautics

Copyright 2009 by Airbus. Published by the American Institute of Aeronautics and Astronautics, Inc., with permission.

N the frame of the project IARCAS (Improve and Assess Repair Capability of Aircraft Structures) fatigue
behavior of advanced repair solutions, both mechanically fastened and adhesively bonded, have been investigated.
As the time that the aircraft remains in service extends, the operational costs (such as maintenance inspections and
related repair actions), vital to keep the aircraft in service with the required levels of safety, increases 1. Airlines
operators repair their aircraft according to the repair solutions provided in the Structural Repair Manual (SRM).
Existing repair solution, contemplated in the SRM of each in-service aircraft, should be improved and developed to
extend the fatigue crack initiation and the crack growth lives (i.e. reduces the operation costs). In-service repairs on
a primary structure of an aircraft are usually performed by the airlines according to the repair solutions provided in
the Structural Repair Manual (SRM). Consequently, the aerospace industry has to enable a simple evaluation of
damages and to provide repair procedures that are safe, less time consuming, simple to perform, inexpensive and
durable as much as possible. As the reduction of downtimes is the highest priority of the operator, an increase of
allowable damage size could be a big improvement for the airliner to reduce this time penalty or the Aircraft On
Ground (AOG) situations that can also disturb the complete flight schedule and lead to even big delays in the air
traffic.
B. Fuselage skin doubler repairs
The standard fuselage skin repairs applied in the
aircraft, according to the SRM, require the use of the same
material for the repair patches (hereafter called doublers)
and the structural component to be repaired as well as
common rivet diameter in all the fasteners of the repair,
constant transverse and longitudinal fastener pitches and
fixed doubler shape (standard rectangular shape with
rounded corner of minimum 10 mm radius). An example
of SRM fuselage skin repair is shown in Figure 1.
The influence of fastener diameter, variable transverse
and longitudinal fastener pitches and optimized doubler
shape (e.g. circular shape) have been investigated in this
research program and the results of this activity are
included in 2, 3. Many other factors can influence the
fatigue life (inspection threshold) and inspection interval of
repairs. The more important are: size of cut-out, doubler
thickness, fastener type, edge margin (distance between
fastener and doubler edge), number of fastener rows, Figure 1. Example of SRM fuselage skin repair.
countersunk depth. The impact on inspection threshold and
inspection interval of these factors has been identified and assessed by Schmidt and Brandecker in previous work 4.
Beside such factors, which are manly related to design principles of repairs, it should be taken into account other
ones that could be referred herein as human factors. The work environment, where the repair is embodied in the
aircraft to keep it flying, is significantly different than the one during the manufacturing of the aircraft itself. The
repairs are installed manually, that is holes are hand-drilled as well as fastener are hand-driven. In such conditions,
reducing as much as possible the AOG time could lead to variation in the repair installation (in particular fastener
installation) such as different rivet interference, unsymmetrically installed rivet, under driven rivet installation (lack
of clamping force).
Recently, the Federal
Aviation
Administration
(FAA) and Delta Airlines
(DAL) jointly conducted a
three years project on tear
down
inspection
and
extended fatigue tests of a
retired Boeing 727 aircraft 5,
6
. The tear down inspection of
lap joints from the fuselage of
(a) differences in rivet interference
(b) unsymmetrically installed rivet
the retired aircraft revealed Figure 2. Sections of rivets in lap joint 7, 8. Deformation of sealant shows
evidence of variation in the variation in rivet interference (a); in (b) evident unsymmetrically installed rivet.
rivet
interference
and
2
American Institute of Aeronautics and Astronautics

unsymmetrically installed rivet 9, 10, two examples are shown in Figure 2.


Atre and Johnson investigated the influence of under-driven and over-driven rivets as well as the hole quality, on
the fatigue life of lap joints 11. The following conclusions can be drawn after this investigation: the fatigue life of the
joints increased with increasing rivet interference; under-driven rivets in lap joint have significantly less fatigue life
than over-driven rivets. On the other hand, Mller in his Ph.D. work demonstrated the importance of the squeeze
force on the fatigue life of riveted joints 11.

II. Fastened Repair FE 3D Models


C. Software Tool FE 3D Model
A software tool (named RE-JOINT: RivEted-JOINT), which consists of several MSC.Patran macros, was
developed to study the mechanical behavior of a flat lap/butt joint under static unidirectional loading and to create
parametrically the FE models of the advanced fastened repair solutions investigated during this activity. REJOINT reads the input file (.txt file), builds
the model (e.g. mesh, boundary conditions,
load cases, etc. etc.) and writes an analysis
file. In the input file the user defines the
general structure of the joint by choosing
sub-components from a library and
positioning them as a sort of puzzle to
build the final joint configuration. Figure 3 Figure 3. Example of sub-components from library defined in REshows examples of sub-components from JOINT tool.
the library defined in the RE-JOINT tool. The user can define the specific sizes of each sub-component and the xycoordinates of reference point to align the parts.
The results of the analysis are then imported and a post processing analysis performed. All macros were written
with the PCL language (Patran Command Language). Two main macros (i.e. main1.pcl and main2.pcl) call
subordinates macros specifically defined to build the FE model and for post processing analysis respectively. The
main PCL macros itself are called by an UNIX shell script (i.e. run_job). This script creates a subfolder, copies the
files in that folder and calls the macros. Once the FE model is created the Nastran analysis manager is called and
the analysis job is submitted to MSC.Nastran for processing. After the analysis has finished, the *.f06 file is
created and checked by Nastran. If errors occurre the script plots the relative errors information. The software tool
has been developed for UNIX environment but it has been also adapted for Windows environment.
Particular attention has been paied to develop fastened joint FE models which include the complete rivet
geometry (such as countersunk and formed head), contact definition (between skins and between fasteners and
skins), elastic-plastic material behavior for full non-linear analysis. The input file (.txt) can enable GAP elements in
all the contact regions of the components (i.e. between skins and between fasteners and skins). The macros called for
this purpose create automatically the local coordinate systems used by the GAP element. The following paragraphs
describe in more details the analysis models created by this tool.
D. Contact Definition with GAP Elements (Nastran FE-element)
The non-linear GAP elements as contact definition between the components were used. The GAP element
provides point-to-point contact characterization setting with different stiffness value in both perpendicular and
parallel direction of the contact. The GAP elements refer to a local coordinate system, which must be defined at the
contact nodes with x-axis coincident to the direction between the point A and B of the element (A is the point of the
GAP element connected to the first component and B is connected to the adjacent component).
The stiffness in the x-axis direction defines the open-close condition of the contact while the coefficient of
friction (static or kinetic) defines the characteristic of the contact in the plane of the contact. The default value for
the axial compression stiffness used in all the models was 1.000.000 N/mm. This value resulted convenient for all
the models presented in this document. Very high values of the stiffness could cause non-convergence problems and
low values of this parameter could cause inaccurate results as the components in contact could penetrate each other.
For instance, if the GAP elements are not used for the joint geometry shown in Figure 4, the contact pairs penetrate
each other causing inaccurate results. The GAP elements location is shown on the right side of the figure; in this
example only GAP elements between the plates were used while the contact between the fasteners and the plates
was simulated by merging the coincident nodes (equivalence). In general the closed GAP stiffness should not
exceed the stiffness of the adjacent contact area by 1000 times. The axial tension stiffness (opening stiffness) was set
to 0.001 N/mm.
3
American Institute of Aeronautics and Astronautics

When the GAP element is open


(elongation of the element) there is
Gap Elements
Doubler-Skin
no contact and then no friction.
When the gap element is closed,
there are three different conditions.
The first case is when the gap is
sliding (no friction, friction
coefficient: s=k=0). The second
case occurs when the gap element is
Skin-Support Doubler
sticking (s0). The third case
occurs when the gap element is
slipping (k0). The friction Figure 4. Gap elements location at surface contact (right). Joints
contact definition can be used to deformation if GAP elements are not used (left).
model the effect of the sealant in a
joint or the effect of bonding if hybrid rivet-bond joint want to be modeled.
RE-JOINT gives also a possibility to merge the coincident nodes of adjacent components. In such way the cpu
time is very much reduced but this
choice is to the detriment of the accuracy
of the stress distribution especially
256
locally at the fastener connections.
Figure 5 shows an example of the
maximum principal stress distribution of
a 3 rivets lap joint. A strip of the joint
(half rivet pitch, 11 mm) was modeled;
this makes the model valid to analyze
only solutions where the full transfer
load condition is respected (such as,
with the reference to 3 and 4, coupon Figure 5. Maximum principal stress distribution 3D FE model of base
specimens or the middle strip of large line solution with merging nodes (equivalence) at the contact area
repair solution). Both coupon specimen between rivets and plates.
geometry (2.0 mm skin thickness and 2.5
mm doubler thickness) and large/small flat panel geometry (1.6 mm skin thickness and 1.8 mm doubler thickness)
were modeled. The material used for these models was Aluminum AA2024-T351 for all the components (skin,
doubler and rivets) but is possible to select several different materials by the input file (a material database, included
in a macro named matprop.pcl, has been included in the software tool). A geometric non-linear solution (elastic
material behavior) was performed and the nodes at the contact region between the rivets and the plates were merged
(equivalence). The location of the maximum stress and the value (256 MPa) is also shown in the figure.
In this case the local stress
distribution is very much influenced
by the contact characteristic both in
terms of maximum stress location and
value.
Figure 6 shows the maximum
principal stress distribution of the
base line solution after non linear
analysis with linear material behavior.
The results shown in the figure refer to
MPa as units. The possible critical
locations resulted from this analysis
were the rivet hole at the doubler (skin Figure 6. Max principal stress distributions 3D FE model of base line
run-out) and the rivet hole at the skin solution with GAP elements at the contact area between rivets and
(doubler run-out) as shown in the plates.
figure a).
Comparing Figure 5 with Figure 6 it can be noted that the critical locations at the skin of the repair are positioned
90 degrees apart depending if contacts are modeled as merging common nodes or by means of GAP elements. The
4
American Institute of Aeronautics and Astronautics

more accurate model (with GAP elements at the contact surfaces) provides the correct critical location (which can
led to the typical failure mode: net section failure). Moreover, the stress peak is also influenced by the local fastener
model resulting in a complete wrong prediction (non conservative) of the local stress if the merge nodes technique is
used (260MPa for the merged model against the 460MPa for the GAP elements model).
E. Fastener Modeling
A 3D solid model of the rivet, including the countersunk and the formed head, is created automatically. The user
defines the radius r of the rivet in the input file and a macro, called by the main program, creates the complete
geometry and a solid mesh (countersunk and form head
included). The countersunk angle has a fixed value of
50. Figure 7 shows the geometry of the rivet
countersunk used in the model.
The depth is calculated by the following formula:

(900 500 )
rdepth = tan
1800

0.266 AF 2r

where, 0.266 is a basis correction factor; AF is an


Adjustment Factor and 2r is the fastener diameter.
The formula is valid for a nominal diameter between
0.094 and 0.188 inch. All diameters bigger than 0.188
are multiplied by 0.286 (= Basis Correction Factor x
Adjustment Factor).
This formula is optimized for NAS1097 and Hi- Figure 7. Geometric features used to model the
Lok fasteners. If other rivet types are used, rdepth fastener head.
should be re-calculated.
The value of rdepth was rounded to obtain a mesh compatible between adjacent components. The result depends
on the parameter relemheigth (defined in the PCLmacros). A small value of relemheigth leads to closer Table I. Depth of countersunk as function of fastener
diameter for NAS1097 rivet.
values of rdepth.
relemheight = 0,2
Table I shows the rounded values of the rdepth
correspondent to relemheight of 0.2. This value gives the
Real Re- Rounded
NAS1097
best approximation of the real rdepth value of the rivet
joint
Re-joint
Nom Diameter
B
Value for Value for
NAS1097.
D
( = rdepth)
rdepth
rdepth
GAP elements at the cylindrical and conical areas of
Inch
m
m
Inch
m
m
m
m
mm
countersunk rivets and between the formed head and the
0,60
0,094
2,388
0,021
0,533
0,533
plate are created automatically (if the GAP option is
0,80
0,125
3,175 0,029 0,726
0,726
enabled by the input file). At the conical part coordinate
1,00
0,156
3,962 0,037 0,945
0,944
systems for each radial line of gap elements are created.
1,20
0,188
4,775 0,046 1,168
1,172
The Patran commands creates the Cartesian GAP
1,60
0,250
6,350 0,060 1,524
1,524
coordinate
systems
which are orientated
perpendicular to the
contact area with their xaxis.
Figure 8 shows the
local coordinate systems
z
of the GAP elements at
Z
x
the
rivet
contacts.
R
Because the contact
definitions by means of
GAP elements imply a
node-to-node connection Figure 8. Local coordinate systems for GAP element locations at the rivet contacts.
it is required to have a The GAP elements are shown as red points in the pictures.
high
number
of

5
American Institute of Aeronautics and Astronautics

coordinate systems at the countersunk surface contact (a coordinate system at each radial line of nodes). On the
contrary a single cylindrical coordinate system was used to orient all the GAP elements at the cylindrical contact
surface of the rivet.
F. Interference Fit and Friction
The GAP elements are characterized by four parameters: s=coefficient of static friction, k=coefficient of
kinetic friction, U0=initial gap opening and F0=preload. It is possible to simulate the effect of the interference fit of
the fasteners in the joint, setting an initial value of gap opening or a preload. For instance, because U0 is the
separation of the gap element with the unit length, it can be used to simulate the interference fit. A negative U0
value can be used to simulate an interference fit between fasteners and holes. A positive initial gap is used to
simulate a clearance fit of the fasteners in the holes. The gap nodes are separated (along the gap axis) by the initial
gap value. About the preload F0, a positive axial force F0 indicates compression.
An example of application of the interference fit effects (for Hi-Lok fasteners) on the stress distribution for the 3
rivet rows lap joint model is reported in this paragraph. The interference conditions are modeled setting initial values
of the gap displacement as described above. Five values of increasing interference were modeled: 10m, 20m,
30m, 40m and 50m. The load
cases considered in the analysis
were: I) interference fit (a
fictitious external load of 1N was
applied in the skin only to satisfy
the boundary conditions) and II)
external load (100 MPa based on
the skin gross area) + interference
fit. The effect of the interference
fit on the stress distribution was
assessed in terms of Smax, Smin, Sm
(mean stress) and Sa (stress Figure 9. Location of GAP element for modeling the interference fit
condition.
amplitude) changing.
The initial gap opening parameter was applied only for the gap elements at the cylindrical contact between the
fastener and the sheets (skin and doubler). Figure 9 shows the gap elements with initial gap opening parameter U0
different than 0 to simulate the interference conditions (red nodes in the figure). Because only half of the joint was
modeled (symmetry) to simulate 20m interference required an initial gap opening value of 10m in the model.
Similarly 5m, 15m, 20m and 25m were
used as initial gap opening values
corresponding to interferences of 10m,
30m, 40m and 50m respectively.
The
maximum
principal
stress
distribution for the 3 rivet rows lap joint
solution with no interference fit (initial GAP
displacement 0 m) was already shown in
Figure 6. At the reference location,
corresponding to the element 19371 in the FE
model, the maximum principal stress was 459
MPa.
Figure 10 shows the maximum principal
stress distribution for the 3 rivet rows lap
joint solution with 20m interference fit and
no external load applied, only a fictitious load
of 1N applied to the gross section of the skin.
The very low external load applied was
necessary to satisfy the boundary conditions.
The stresses shown in the figure are in MPa. Figure 10. Maximum principal stress distribution for the 3
The maximum principal stress at the rivet rows lap joint solution with 20m interference fit. No
reference location (Elm 19371 of the FE external load applied. All stresses are shown in MPa.
model) resulted 181 MPa.
6
American Institute of Aeronautics and Astronautics

S max S min
and mean
2

S a=
stress

S + S min
Sm= max
.
2

Reference Stress (Elm 19371) (MPa)

Figure 11 shows the maximum


principal stress distribution for the 3 rivet
rows lap joint solution with 20m
interference fit. The external load applied
was 100 MPa based on the gross section of
the skin. The stresses shown in the figure
are in MPa. The maximum principal stress
at the reference location (Elm 19371 of the
FE model) resulted 370 MPa.
Comparing the maximum principal
stress at the reference location for the
solution without interference fit (initial
GAP displacement 0 m) and with
interference fit of 20 m resulted that the
maximum stress for the solution without
interference
fit
was
reduced
by
approximately 20%. Moreover the
minimum stress, when the external load is
removed (0 MPa if no interference fit is
considered), was increased if the Figure 11. Maximum principal stresses for the 3 rivet rows lap
interference fit is taken into account. The joint with 20m interference fit and the external load.
amplitude stress also decreased at values of interference fit considered.
Figure 12 shows the maximum principal stress distribution at the reference location of the 3 rivet rows lap joint
solution model for increasing
600
values of interference fit. The
stresses shown in the plot are
y = -0.0035x3 + 0.4675x2 - 12.285x + 459.56
500
always the maximum principal
R2 = 0.9995
stress at applied external load
(Smax), with no external load
400
applied (Smin) and the resulting
y = 9.2029x - 0.9048
amplitude stress
2
300

R = 0.9989

Smax (at Max Ext Load)


Smin (at Min Ext Load)
Amplitude
Mean Stress
Smin - Trendline
Smax - Trendline

200

100

0
0

10

15

20

25

30

35

40

45

50

The increasing value of the


Interference Fit (micron)
interference fit had as main
Figure
12.
Maximum
principal
stress distribution at the reference
consequence the reduction of the
location
of
the
3
rivet
rows
lap
joint
model for increasing values of
amplitude stress despite the
interference
fit.
The
stresses
shown
in
the
plot are always the maximum
increasing of the mean stress. For
principal
stress
at
applied
external
load
(Smax), with no external load
interference values greater than
applied
(S
)
and
the
resulting
amplitude
stress
Sa and mean stress Sm.
min
25-30 m the effect on the
amplitude stress is less relevant
(horizontal asymptote in the plot).
In this case the fit line has shown a very good agreement with the FE results for interference value below 30m.
At this interference value the corresponding stress was still elastic (270 MPa) but increasing the interference the Smin
change following the non-linear trend as expected.
G. Advanced Fastened Repair Solution FE Models
1. Support Doubler Repair Solution
The design principle of this repair solution is described in details in 3. It consists of a single shear joint (3
fastener rows) with a support doubler located at the skin side of the upper and lower doubler run out. Even for this
7
American Institute of Aeronautics and Astronautics

repair solution, a strip of the joint


(half rivet pitch, 11 mm) of coupon
specimen geometry and large/small
flat panel was modeled (coupon
specimen and large/small flat panel
geometry are described in 3). Skin,
doubler and support doubler were
modeled using Aluminum AA2024T351 while the fasteners (HiLoks) were modeled in Titanium
(Ti-10Al-6V). Non-linear analysis
was performed including both linear
and non-linear material behavior.
Figure 13 shows the maximum
stress distribution of this repair
solution after performing non-linear
solution with linear material
behavior. The example reported in
the following figures refers to the
coupon specimen geometry (2.0 mm
skin thickness, 2.5 mm doubler Figure 13. Maximum principal stress distribution (in MPa) for nonthickness and 4.0 mm support linear analysis with linear material behavior of coupon specimen geometry.
doubler thickness). The locations Location corresponding to peak of stresses are also shown.
corresponding the peak stress
(doubler and skin) are also
shown in the figure. To make
easier the location of the
maximum stress the fasteners
are not shown in the figure.
Figure 14 shows a close up
view of the critical locations of
support doubler repair solution.
Stress values at the doubler
location and skin location were
510 MPa and 450 MPa
respectively. The highest stress
resulted at the doubler location Figure 14. Maximum principal stresses at doubler and at skin position of
in the cylindrical surface support doubler repair solution.
contact with the remaining
collar of the fastener (510MPa). This is the region where the countersunk geometrical discontinuity is more effective
as stress concentrator.
In order to establish the
most critical area suitable to
fatigue crack initiation, it is
important to evaluate the size of
the area subjected to the
maximum stress. Fatigue crack
may initiate from material
defects (porosities, inclusion or
mechanical damages) in area
with high stress concentrators.
Hence the bigger the area
subjected to high stress the
more likely a fatigue crack may Figure 15. Von Mises stress distribution for coupon specimen geometry after
initiate. The material used for full non-linear analysis including the non-linear material behavior.
the skin and the doubler in this model was Al2024-T351, which has a yield stress of 335 MPa.
8
American Institute of Aeronautics and Astronautics

A full non linear analysis (including the non linear behavior of the material) was performed and the results (in
terms of Von Mises stresses) are shown in Figure 15.
By limiting the maximum principal stress up to this value, it is possible to have an estimation of the area
subjected to this stress, and therefore the location of the fatigue crack initiation, to be used for the fatigue life
prediction.
At the critical location of the doubler, the region with highest stress (red region in Figure 15) was distributed to
the half thickness of the doubler; in fact, at the critical location of the skin, the region with highest stress was
covering the entire thickness of the skin.
The conclusion of this analysis is that, although the highest stress was found at the doubler position, the most
critical location for fatigue crack initiation
was assumed at the skin position. The fatigue
life prediction was based on the maximum
stress at this location.
2. Parametric Study
In order to optimize the design principle
of this repair solution a parametric study was
done.
The
main
parameter
under
investigation was the support doubler
thickness. The stress distribution, at the
critical locations discussed in the previous
paragraph, is effected by the stiffness of the
joint. An increase of the support doubler
thickness should correspond to an increase of
the general stiffness of the joint, which leads
to a reduced secondary bending effect with
obvious reduction of the maximum stress. In Figure 16. Effect of the support doubler thickness with respect
fact, the increase of the doubler thickness to the maximum principal stress at the skin critical location.
causes a more pronounced eccentricity effect,
which leads to an increase of the secondary
bending. A compromise must be found and the
results of this analysis are shown in the following
paragraphs.
Figure 16 shows the effect of the increasing
support doubler thickness on the maximum
principal stress at the skin critical location of the
support doubler repair solution with reference to
the geometry of small/large flat panels (1.6 mm
skin thickness and 1.8 mm doubler thickness).
An increasing of the support doubler thickness
corresponds to a decreasing of the maximum
stress. Values of the support doubler thickness
greater than 3.2 mm does not lead to a significant
reduction of the maximum stress.
Figure 17 shows the maximum stress distribution
at the critical skin location for the base line repair
solution (figure a) and for the support doubler Figure 17. Maximum principal stresses at increasing
repair solution at increasing thickness of the thickness of the support doubler.
support doubler (figure b,c and d).
Increasing the support doubler thickness had, as main effect, the reduction of the maximum principal stress at the
reference location. The absolute maximum stress location moved through the thickness towards the support doubler
skin surface contact at increasing values of the support doubler thickness (as shown in Figure 17).
H. FE Model Validation
To validate the FE models created by RE-JOINT tool, the results have been compared with strain gauge readings
of a coupon specimen containing a step, chemically milled radius at the doubler run-out area, with increased bending
stress components in this region, where the strain gauges were installed
9
American Institute of Aeronautics and Astronautics

Figure 18 shows the coupon


Kt = 4.7
specimen geometry, details of
Kt = 3.9
the step chemically milled at the
doubler run out are shown on
the right side, including the
maximum principle stresses at
the doubler run out, where the
strain gauges have been Figure 18. Coupon geometry and maximum principle stresses at the
installed.
doubler run out.
The coupon specimen with
three rivet rows shown above has been tested with strain gauge readings applied near the radius (and at the opposite
side of the coupon). Figure 19 shows the test rig and the strain gauge locations.

Figure 19.

Strain gauges positions. Strain gauges have been placed back to back (1-3 and 2-4).

Stress
MPa

Details of the FE model at the rivet connection and at the step radius is shown in Figure 20 (top-left). A very fine
mesh at stress concentration locations (such as rivet connections, chemically milled radius) was used. Figure 20
(bottom-left) shows the position and the length of the strain gauges with respect to the FE model.
The comparison between stresses calculated by FE analysis (FEA) and the ones measured by strain gauges
revealed that the FE
Test and FEM Results - GAP Elements FE Model
models are able to
Stress vs. Nominal Stress
240
predict
stress
distribution
around
200
fastener connection of
160
joints with maximum
error of 7%.
120
Stress
deviation
between strain gauge
80
readings (thick lines)
40
and FEA (thin lines)
mm
demonstrated that a
0
0
10
20
30
40
50
60
good estimation of the
-40
local
stresses
is
possible. The stress
-80
field predicted by FE
-120
models has shown very
Test, SG1
Test, SG2
Sigma Nominal
good agreement with
Test, SG3
Test, SG4
N/mm
mm
the one measured by
FEM, SG1
FEM, SG2
FEM, SG3
FEM, SG4
strain
gauges
in
coupon specimens as Figure 20. Details of the FE model at the rivet connection and position and length of
shown in Figure 20 the SGs in the FE model.
(right).
0,75

1,5

10
American Institute of Aeronautics and Astronautics

III. Curved Stiffened Fuselage Panel FE Model


Two curved stiffened panels, representing a real fuselage structure, in which four repairs were introduced (SRM
repair to be used as reference and three advanced repair solutions including an hot bonded repair), were subjected to
biaxial constant amplitude fatigue loading (Figure 21).
A software tool to study the mechanical behavior of a curved stiffened panel under biaxial loading (internal
pressure and longitudinal load) was developed in this project. This tool consists of several macros written in
MSC.PCL coordinated by a main macro which calls the other subordinated macros to create a Finite Element model
and a MSC.Nastran input
deck.
Figure 22 shows an
example of a curved panel
FE model (deformed shape
after nonlinear analysis)
created by this tool.
This model is a cutout
of about 35 of the fuselage
with a length and width of
about 3200 x 1600 mm,
including
frames
and
stringers.
All
main
measures like radius, frame
sizes, pitches etc. can be
adapted.
Also
a
longitudinal lap joint with
different
configurations Figure 21. Curved stiffened panel with four unlimited repair solutions.
may be placed at the central
stringer.
Figure 23 (right) shows
the mesh size at the central
area of the model which is 34 times finer than the outer
part. Mesh details of the
stiffener elements (frames
and stringers) are also shown
in Figure 23 (left). The outer
part is meshed coarser to
reduce analysis time. A stripe Figure 22. Example of FE model of curved stiffened panel (right: deformed shape).
with
unstructured
mesh
connects the finer and the
coarser mesh. The most
common used element type is
the
two
dimensional
CQUAD4 plate element
(MSC.Nastran notation) with
a rectangular shape. CTRIA3
triangle elements are used as
gradual passage between
coarse and finer mesh. A
MPC connection (multipoint
constraint) is used for Figure 23. Mesh details of stiffener elements (left), refined mesh area at the center
(right).
applying the axial force.

11
American Institute of Aeronautics and Astronautics

The boundary conditions applied to the model are shown in Figure 24. Internal pressure and axial load were the
loading conditions applied. The axial force is applied at a center node, which is connected to the panel with a MPC
element (multipoint constraint of the type RBE2) in z-direction (axial direction).
The
cutting
edge is in between
two stringers and
frames. The FE
model is fixed by
symmetry
boundary
conditions. At the
longitudinal
sections,
the
degree
of
freedoms (DOF)
2,4,6 are fixed in Figure 24. Mesh details of stiffener elements (left), refined mesh area at the center (right).
the
cylindrical
coordinate system CS 3 - that is TT (tangential translation), RR (radial rotation), RZ (axial rotation) (as shown in
Figure 24). At the cross section, the degrees of freedoms are 3,4,5 - that is TZ (axial translation), RR (radial
rotation), RT (tangential rotation). The symmetry boundary conditions are conterminous to mirroring the model.
This simplification ignores small unsymmetrical configurations like the stringer and frame orientation. The wide
range of frames and stringers prevents local influence of the constraints at the model edges on the region under
investigation in the middle. The structural behavior of the panel is the same as of a long, closed cylinder.
A geometric nonlinear static analysis was performed with MSC.Nastran (SOL 106). Within the scope of this
analysis type, large deformations and follower forces are taken into account. Large deformations consideration
means, that the stiffness matrix is updated according to the calculated deformation after every iteration step. This is
important in case of large rotations which are induced by plate offset at repairs and lap joint (secondary bending).
With the follower forces option on, the pressure applied at the plate elements remains perpendicular to them.
The material behavior is assumed as linear and isotropic.
The main macro calls the other subordinated macros. With the !!input commands in the main macro, all
functions stored in the macro files are read and compiled.
Now MSC.Patran can use the functions:
to read additional pcl functions and to open a new
MSC.Patran database
to read the input files (two ASCII files containing all
the geometric figures to create the panel and the
loading conditions)
to create the FE model (geometry, mesh, loads and
constrains)
to write an analysis file for the MSC.Nastran solver
The user can control, which part of the program shall run
by setting switches in the main macro.
Two ASCII input files contain the geometry of the panel
and all other information (lap-joint configurations) are
required to create the FE model. These are the geometric
parameters of the joint such as thickness and rivet pitches.
An example of the parameters required to describe the
stiffeners is shown in Figure 25. The MSC.Patran PCL
macros read the required parameters from this text files and
search for the parameters necessary to create the model.
Figure 26 shows how the stress distribution changes
because of the presence of the four repairs (left). Both panels
were equipped with a large numbers of strain gauges most of
them installed back to back. The FE model of the panel Figure 25. Example of input parameter for a)
containing repairs needs to be edited manually as repairs are Frame/Clip and b) Stringer geometry.
not in the scope of the software tool.
12
American Institute of Aeronautics and Astronautics

All FE models
have been validated
comparing
the
stresses simulated by
the analysis with the
ones read by the
strain gauges during
the fatigue tests.

IV. Bonded Repair


FE 3D Model
Figure 26. Example of FE model of curved stiffened panel containing four repairs (right).
I. Introduction
The FE 3D models to perform parametric studies of two flat bonded repair configurations are described in this
section. The models are created using StressCheck a FE program based on the p-level method (3.2). Two
different bonded joint geometries have been modeled: butt-joints and small flat panels 4. Both joints were modeled
with aluminum for the skin and doubler and combining aluminum for the skin and glare for the doubler. The
adhesive layer was modeled too.
J. StressCheck FE Program
StressCheck (developed by esrd) is based on the p-version of the finite element method. One of the
advantages of using p-method based program is the possibility to reduce the errors of approximation increasing the
polynomial degree of the elements. Moreover, high aspect ratio elements for single ply modeling are feasible
(prerequisite for thin solid layers like adhesive). Models can be created parametrically in such way the geometry
and/or the properties of the model (e.g. element properties) can be changed easy for a full parametric study. One
more important feature of StressCheck is the possibility to handle nonlinear analysis domains (nonlinear material
behavior and geometric nonlinearities).

K. Models Description
Figure 27 and Figure 28 show the technical drawings (left) and the corresponding FE mode made by
StressCheck (right) of coupon test specimen and small flat panel respectively. The skin was always modeled in
aluminum while the doubler was modeled both in aluminum and glare. The models were created parametrically.
Areas where high stresses are expected were meshed with smaller element size. Simple geometry entities (such as
points and curves have been
used in order to minimize
meshing errors caused by remeshing while changing the
parameters. For instance, nodes
and elements created at the
corner of the doubler (small flat
panel) must be located along an
arc to make the element edge
following the curvature. These
nodes cannot be copied with an
offset, because the curvature
information gets lost (the
elements created on base of
these copied nodes will not
have the same curvature). The
possible solution for this Figure 27. Geometry (left) and StressCheck model (right) of coupon
inconvenience is to create arc test specimen.
lines for each curvature in the
model. A FE 3D-model could be also created defining the model in 2D and then copying the plane model with an
offset. The solid elements (eight nodes elements) can be created simply selecting the two planes containing the 2D
models by StressCheck commands: Create Hexahedron Face to Face.
13
American Institute of Aeronautics and Astronautics

L. Boundary Conditions
Only a quarter of the real
specimen
was
modeled
(symmetry constrains at the
symmetry
planes).
The
clamping
area
of
the
specimen was modeled as a
small strip with fixed motion
in the y-direction (direction
out of the specimen plane). A
tensile load corresponded to
100 MPa stress was applied at
the gross section of the skin
as external load. Figure 29
shows
the
boundary Figure 28. Technical drawings (left) and StressCheck model (right) of
conditions of the butt-joint small flat specimen.
coupon test specimen.
Figure 30 shows details of
the model (both coupon
specimen and small flat
panel). In left hand side of the
figure the mesh size at the
adhesive phase length is
shown; the figure refers to the
coupon
specimen
with
aluminum/glare as material
combination (the doubler was
modeled in glare). In the
right hand side of the figure
details of the cutout and Figure 29. Boundary conditions of butt-joint coupon test specimen.
boundary conditions of the
small flat panel model are
shown.
M. Material Modelling
3. Aluminium Alloy AA2024T351
Ramberg Osgood nonlinear properties were used to
model the plates (skin and
doubler) and the outer layers
of Glare in aluminum. A
Ramberg Osgood material is
defined by four parameters:
I) the modulus of
elasticity (E); II) Poissons Figure 30. Model details: mesh size at the adhesive phase length (coupon
ratio (); III) the stress value specimen); details of the cut-out and boundary conditions.
(S70E) corresponding to the secant modulus of 0.7E slope; IV) n shape factor describing the shape of the stress-strain
diagram in the yield region.
The Ramberg Osgood expression is:

3 S70 E

= +
E 7 E S70 E

Typical values of n range between 4 and 90.


14
American Institute of Aeronautics and Astronautics

The material coefficients used in this analysis are shown in Figure 31.

Figure 31.

StressCheck material input and stress-strain Ramberg-Osgood plot for aluminum 2024.

4. Adhesive Modeling
The adhesive was modeled as bilinear material (linear elastic and linear strain hardening). The plastic region was
described by linear stress-strain relationship with lower slope than the elastic region. The stress-strain relationship
are described by four parameters:
I) the elastic modulus (E); II) Poissons ratio (); III) the yield stress (Sy); IV) the tangent modulus to
characterize strain hardening of the material (Et).
The adhesive used in this analysis was FM-73M.06 and the material coefficients are shown in Figure 32.

Figure 32. StressCheck material input and stress-strain Ramberg-Osgood plot for adhesive FM-73M.06.
5. Glare
The Glare material was modeled using a sub-laminate definition available in Stress-Check program. A
sublaminate is a homogenized stack of plies. Using this material definition presents several advantages such as
decreasing the cpu time or allowing parametric models including Glare type as parameter to study (otherwise a
new model must be created if the number of metal and prepreg layers is changed). The aluminum layers used in the
sub-laminate were modeled as linear elastic.
For both models (coupon specimens and small flat panels) the type of Glare 3-4/3-0.3 were used. Three
prepreg layers were modeled, the second and the third layer had the same orientation of fibers and they were only
half of the real prepreg thickness. This way, these two layers act as a single layer, but the possibility to use different
Glare as parameter to be changed is enhanced (for example Glare with three prepreg layers instead of two).

15
American Institute of Aeronautics and Astronautics

Figure 33 shows how the Glare material was modeled. The outer layers were modeled in elastic-plastic
aluminum, after which two prepreg layers were modeled, orienting the fibers at 0 and 90, respectively. The
remaining inner layers were modeled using the sublaminate definition.

Figure 33. Glare model: aluminum layers, prepreg layers and sublaminate layer.
N. Parameters Definition
Previous FE analyses have shown that the length of the adhesive phase and its thickness influenced very much
the local stress. Optimization study was performed and the above mentioned parameters were used as main
parameters for the investigation. In addition to the small flat panel with bonded repair, a larger flat panel with
bonded repair (corresponding to the large flat panel fully described in 3 and 4 was modeled. In the latter, the main
differences in respect to the small flat panel are the panel size (longer and wider panel) and a larger size of the cutout (and doubler).
All models were done in fully parametric way; this means that many parameters were enabled in order to have a
flexible model for further parametric study. For each model (two for the coupon specimen geometry and two for
the small flat panel configuration) 12 analysis were performed: at 0.1 mm adhesive thickness and increasing values
of adhesive phase length; fixed value of adhesive thickness of 0.2 mm and increasing adhesive phase length values.
O. Analysis Parameters
Full non linear analyses (both material and geometric nonlinearity) were performed for all the models. For each
model a linear analysis was performed with a p-level up to 8 (to reduce the cpu time maybe a lower p-level can be
used but the convergence requirement should be satisfied). The energy convergence criterion was used with
Newton-Rapson technique. The tolerance value for the convergence was 0.75%. The linear analysis resulted were
used as reference for the non linear analysis.
P. Coupon Specimen FE-Model
The results of the FE analysis of the coupon specimen geometry are described in this section. Both material
combinations Al/Al for skin and doubler and Al/glare combination (skin in aluminum and doubler in glare)
results are included. The stress analysis was focused in the skin and the doubler at the adhesive line runout. The
fatigue life prediction was based on the stresses obtained at these locations. The influence on the local stress
distribution of both adhesive thickness and adhesive phase length was investigated.
6. Aluminum Skin Aluminum Doubler
Figure 34 shows the non linear analysis results in terms of maximum principal stress distribution for the coupon
model. Both skin and doubler were modeled in aluminum. The adhesive thickness and the adhesive phase length
were 0.1 mm and 0.3 mm respectively.

16
American Institute of Aeronautics and Astronautics

Figure 34. Maximum principal stresses of coupon model with both skin and doubler in aluminum. Non linear
analysis, adhesive thickness of 0.1 mm, adhesive phase length of 0.3 mm.
Figure 35 shows the maximum principal stress distribution at the doubler run out for the coupon model. Both
doubler and skin were modeled in aluminum. The non linear analysis was performed. The adhesive thickness and
the adhesive phase length were 0.2 mm and 0.8 mm respectively.

a) Max principal stress at the doubler (b) Max principal stress at the skin
Figure 35. Maximum principal stresses at the doubler run out of coupon model for both skin and doubler
(figure-a and b respectively).
7. Bondline Optimization
It was found that the adhesive thickness and the adhesive phase length are the main parameter that influence the
stress distribution at the critical location (doubler run out). A detailed study was performed to determine the local
stress at the skin for increasing values of the adhesive phase length. The coupon specimen was the model under
investigation with both skin and doubler made in aluminum. Two values of adhesive thickness were selected for this
analysis (0.1 mm and 0.2 mm).
Based on the experimental evidence that fatigue failure occurred at the skin (ref 4), the analysis was focused on
the stress distribution at the skin. Figure 36 shows the maximum principal distribution at the skin of the bonded
coupon specimen model plotted versus the adhesive phase length. The non linear analysis was performed for two
values of adhesive thickness (0.1 mm and 0.2 mm). Both adhesive thickness models have shown a minimum local
stress above a certain value of the adhesive phase length. The local stress is reducing considerably up a fixed value
of the adhesive phase length and above this value there is not further reduction of the stress. This value resulted
approximately 0.15 mm and 0.4 mm for the adhesive thickness of 0.1 mm and 0.2 mm respectively. The reduction
of stress achievable with optimized adhesive phase length is up to 16% of the maximum stress for the 0.1 mm
adhesive thickness model and up to 20% of the maximum stress for the 0.2 mm adhesive thickness model.

17
American Institute of Aeronautics and Astronautics

Figure 36. Maximum principal stress distribution at the skin of bonded coupon specimen model for
increasing values of adhesive phase length.
8. Aluminum skin Glare doubler
Figure 37 shows the non linear analysis results in terms of maximum principal stress distribution for the coupon
model. Skin was modeled in aluminum and the doubler was modeled in glare. The adhesive thickness and the
adhesive phase length were 0.2 mm and 0.9 mm respectively. Figure 38 shows the maximum principal stress
distribution at the doubler run out for the coupon model. The doubler was modeled in glare and the skin was
modeled in aluminum.

Figure 37. Maximum principal stresses of coupon model with skin made in aluminum and doubler in glare.
Non linear analysis, adhesive thickness of 0.2 mm, adhesive phase length of 0.9 mm.
The non linear solution was performed under a far field stress applied of 100 MPa (based on the skin gross area).
The adhesive thickness and the adhesive phase length were 0.2 mm and 0.3 mm respectively. With this combination
of material the maximum stress at the skin remain as the previous case (both components in aluminum) at
approximately 230 MPa whilst the maximum stress at the doubler resulted 270 MPa. The maximum principal stress
distribution at the doubler and at the skin is shown in Figure 38.

18
American Institute of Aeronautics and Astronautics

Figure 38. Maximum principal stresses at the doubler run out of coupon model for skin made in aluminum
and doubler in glare.
Q. Small Flat Panel FE Model
The results of the small flat panel model are described in this section. For these models, both material
combinations Al/Al for skin and doubler respectively and skin in aluminum and doubler in glare results are
included. The stress analysis was focused to the skin and the doubler at the adhesive line runout.
9. Aluminum Skin Aluminum Doubler
Figure 39 shows the maximum principal stress distribution of small flat panel model with both skin and doubler
made in aluminum. A non linear analysis was performed. The adhesive layer had 0.1 mm thickness and 0.075 mm
phase length. The results shown a maximum stress localized at the skin of 325 MPa approximately. Figure 39 (right)
also shows a fatigue crack occurred at the critical location predicted by the analysis (from ref. 4). Several analysis
were performed at increasing values of the adhesive phase length in order to investigate how the distribution of the
stress at the critical location (doubler run-out) change. The effect of the adhesive thickness was investigated too, two
values of the adhesive thickness were considered: 0.1 mm and 0.2 mm.

Figure 39. Maximum principal stress distribution of small flat panel model with both skin and doubler in
aluminium. Test result (from ref. 4) is also shown with fatigue crack occurred at the stress critical location.
Figure 40 shows the maximum principal stress distribution at the skin of the doubler run-out for two values of
the adhesive phase lengths, 0.05 mm and 0.3 mm. The adhesive thickness chosen for this analysis was 0.1 mm. The
maximum stress in both cases was approximately 325 MPa but increasing the adhesive phase length the area
subjected to higher stresses resulted more localized within a region ahead the doubler run-out. Comparing the
distribution of the stress shown in figure (a) and figure (b), leads to the following conclusion: the longer the
adhesive phase length, the more localized the stress at the doubler run out.

19
American Institute of Aeronautics and Astronautics

(a) adhesive phase length 0.05 mm (b) adhesive phase length 0.3 mm
Figure 40. Maximum principal stress distribution at the skin of the doubler run out in the absence of the
adhesive phase length, figure (a) and with adhesive phase length of 0.3 mm, figure (b).

Acknowledgments
The authors would like to thank a PhD student Goran Ivetic of University of Pisa, Italy for his contribution in
preparation of this paper.

References
1

Armstrong, A., Future structures supportability strategy, Proceedings DSTL Structures Contextual Review, 2003.

Contextual Review, 2003.


2

Furfari, D., Meyer, C., Lafly, A.L., Pramono, A., Advanced Repair Design Principles to Improve Fatigue and Damage
Tolerance Behavior of Fastened Repairs, Proceedings of 24th ICAF Symposium, Naples, Italy, 2007.
3
Furfari, D., Woerden, H.J.M., Benedictus, R., Kwakernaak, A. (2009), Bonded Repair for Fuselage Damage: an Overall
Benefit to Commercial Aviation, Proceedings of 25th ICAF Symposium, to be published, Bos M. (Ed.), Rotterdam, The
Netherlands.
4
Brandecker B., Schmidt H.-J., In: Repair Assessment Program for Airbus A300 Aircraft, Proceedings of 19th ICAF
Symposium, vol. I, p.619-637, Cook, R. and Poole P. (Eds.), Edinburgh, Scotland, 1997.
5
Bakuckas, J. G., Carter, A., In: Destructive Evaluation and Extended Fatigue Testing of Retired Fuselage Structure: Project
Update, Proceedings of the 7th DoD/FAA/NASA Conference on Aging Aircraft, p.1-12, New Orleans, Louisiana, 2003.
6
Bakuckas, J. G., Bigelow, C. A., Carter, A., Steadman, D., In: Destructive Evaluation and Extended Fatigue Testing of
Retired Aircraft Fuselage Structure, Proceedings of the 23rd ICAF Symposium, vol. I, p.229-240, Dalle Donne C. (Ed.),
Hamburg, Germany, 2005.
7
Radhakrishnan R., In: Damage Characterization, Joint FAA/Delta 8th Quarterly Meeting, Destructive Evaluation and
Extended Fatigue Testing of a Retired Aircraft, Atlanta, GA, 2004.
8
Radhakrishnan R., In: Damage Characterization, Joint FAA/Delta 10th Quarterly Meeting, Destructive Evaluation and
Extended Fatigue Testing of a Retired Passenger Aircraft, Atlanta, GA, 2005.
9
Atre A. P., Johnson W. S., In: Analysis of the Effect of Riveting Process Parameters on the Fatigue of Aircraft Fuselage
Lap Joints, Proceedings of the 9th Joint FAA/DoD/NASA Aging Aircraft Conference, p.1-13, Atlanta, Georgia, 2006.
10
Mller R. P. G., An experimental and analytical investigation on the fatigue behaviour of fuselage riveted joints, Ph.D.
Thesis, Delft University of Technology, The Netherlands, 1995.

20
American Institute of Aeronautics and Astronautics

Вам также может понравиться