 
....
p'
DESIGN METHODOLOGY FOR BONDEDBOLTED COMPOSITE, JOINTS Vol. 1. Analysis Derivations and Illustrative Solutions
L. J. HartSmith
Doyiglas Aircraft Company McDonnell Douglas Corporation Long Beach, California 90846
S_
DTjC
L
February 1982
TECHNICAL REPORT AFWALTR813154 Final Report for Period August 1979  June 1981
0IL
SJUL221982d
U4
I
FUGH7 DYNAMICS LABORATORY
'
::.
Qg, 07 212
U 5 .5
NOTICE When Covernment drawings, specifications, or other data are used for any purpose other than in connection with a definitely related Government procurement operation, the United States Government thereby incurs no responsibility nor any obligation wnatsoever; and the fact that the government may have formulated, furnished, or in any way supplied the said drawings, specifications, or other data, is not to be regarded bit implication or otherwise as in any manner licensing Lhe holder or any other person or corporation, or conveying aiy rights or permission to manufacture use, or sell any patented invention that may in any way be related t~hereto. Thi's report has been reviewed by the Office of Public Affairs (ASD/PA) and is releasable to the National Technical Information Service (NTIS). At NTIS, it will be available to the general public, including foreign nations.
/I'
C,
USAF
"If your address has changed, if you wish to be 'emoved from our mailing list, or if the addressee is no longer employed by your organization plea.se notify AWALFERA, WPAFB, OH 45433 to help us maintain a current mailing list". Copies of this report should not be returned unless return is required by security considerations, contractual obligations, or notice on a specific document.
Unclassified
S9CURITY CLASSIFICATION or THIS5 PA040 (l~ia" bealmEntered)
AFFOF
UMSR
4.
TTLE
andSubttle
DESIGN METHODOLOGY FOR BONDEDBOLTrED COMPOSITE JOINTS Volume I. Analysis Derivations and Illustrative Solutions
7. %UVHOR(s)
j.Auqust
L. J. HartSmith, Ph.D.
2. PERFORMING ORGANIZATION NAME AND ADDRESS

F3361579C3212
1. PROGRAM ELEMEN4T, PROJECT, TASK
Project_____ 2401__02_28
12. REPORT DATE
Proet
62401F0
CONTROLLIN
Flight Dynamics Laboratory (AFWAL/FIBRA) AF Wright Aeronauti cal Laboratori e WrightPatterson AFB, OH 45433
IT7MOITORING AGENCY NAME A ADDRIESS(If dillorent froem CmoItoallif Office)
99
Unclassified
ISO.DCL ASSI FICATI ON/ DOWN GRADING
SCiULt
IV.
KEY WORDS (Conrfnue on, twvora. aide It necessary and identify by' blocl,rnmnba)
Computer Programs
_nalsi
Lvanced
*This
report contains recent developments in three aspects of joints in adfibrous composite structures: (1) nonlinear analysis of adhesivelybonded stepped lap joints and doublers, (2) niultlrow mechanically fastened
anotdjoints.i aeosac strutures, deenpd in3)d nonlinear~e alyi cobneeddbonde metal structures as well as those for composites. The analyses are based on continuum mechanics techniques and have beer. coded into three Fortran IV
DD 'J A1
1473
1 OBSOLETE OITIOP14 OF' I NOVS IS SECUJRITY CLASSIFICATION OF THIS5 PAGE flea~n Data Entered)
Unclassified
SIMUNITY CLASSIFICATION OF THIS PAG@U(;;
Dm. Kniew.40
0. ABSTRACT (Continued) *digital computer programs A4EI, A4EJ, and A4EK respectively. The report contains explanations of the derivations of the solutions as well as sample worked problems to illustrate both the capabilities of the programs and the behavior characteristics of the real structures.
Unclassified
SECURn' V CLASSIFICATION
PREFACE This oloqy report for presents the by results Composite the California of the investigation Contract Company,
Joints, during
F3361579C3212.
Douglas
Aircraft
Corporation,
the period
Investigator. Wright
This work was sponsored by the Air Force Flight Dynamics Laboratory, Patterson Air Force Base, Ohio. the Project Engineers. Dr. V. B. Venkayya and Lt. P. J.
Conrad were
The
computer
coding
for
the
programs
A4EI,
A4EJ
and
A4EK will
be
made
tiii
NONLINEAR ANALYSIS OF ADHESIVEBONDED STEPPEDLAP JOINTS ......... .......................... AND DOUBLERS ........ 2.1 INTRODUCTION ....... ....... ........................ ....... .............. ....... 2.2 SYMBOLS........ ....... 2.3 2.4 2.5 2.6 2.7 2.8 NONLINEAR ANALYSIS OF ADHESIVEBONDED STEPPEDLAP JOINTS . ... ...................... SAMPLE SOLUTIONS ......... ....... ....... ... EXPERIMENTAL EVIDENCE ...... ....... LOAD REDISTRIBUTION DUE TO DISBONDS IN ADHESIVE ....... ....... .... IN STEPPEDLAP JOINTS . ..... ....... ........... CHECKS ON ACCURACY CF THE SOLUTIONS ........ CONCLUSIIONS ...... ......... ........................
5 5 6 7 19 27 30 35 35
NONLINEAR ANALYSIS OF MULTIROW BOLTED JOINTS IN FIBROUS COMPOSITE ... 37 ...................... AND METAL STRUCTURES ...... ... ... 37 3.1 INTRODUCTION ....... ... ........................ ... 38 3.2 SYMBOLS ...... ....... .......................... ;39 3.3 LOADDEFLECTION CHARACTERISTICS FOR A SINGLE FASTENER "
.
3.4
3.5
3.6 3.7 3.8
"44 LOAD SHARING BETWEEN MULTIRUW FASTENERS.."."...."." FAILURE CRITERIA AT FASTENER HOLES . . . .. EXPERIMENTAL DETERMINATION OF InPUT DATA. FOR COMPUTER.',..' . ANALYSIS . . . . .. . ... . . . . . . ... . . . .. I. . . 55 ... ............ SAMPLE SOLUTIONS............ ......... ..... '64 CONCLUSIONS ...... ....... ................... 67 67 69
S4.1
. . ......... NONLINEAR ANALYSIS OF BONDED/BOLTED JOINTS . .. ....... .. ....... INTRODUCTIOM . . ..... ....... 4.2 REPAIR OF DEFECTIVE BONDED JOINTS BY MECHANICAL ATTACHMENTS
4.3 4.4
REPAIR OF DAMAGED STRUCTURE BY BONDING AND BOLTING,........ 5 79 COMBINATION OF BONDING AND BOLTING IN FAILSAFE STRUCTURES .
TABLE OF CONTENTS (Concluded) Section mage 4.5 4.6 ANALYSIS OF FAILSAFETY OF BONDED/BOLTED STRUCTURES CONCLUSIONS ..... ..... . . . . . . . . . . . . . . . .. .............................. ...............................
.2
. . . . . ..
. ..
85 87 89
...
 .
INSpECTEED
vi
II
LIST OF ILLUSTRATIONS Figure 1 2 3 4 5 6 7 8 9 10 Page
HI
NOTATION AND GEOMETRY FOR ADHESIVEBONDED STEPPEDLAP JOINT ANALYSIS ............... ........................... REPRESENTATIONS OF ADHESIVE NONLINEAR SHEAR BEHAVIOR ..... .. EIGHTEEN TYPES OF ADHESIVE BEHAVIOR IN BONDED JOINTS . . . ... ADHESIVEBONDED JOINT LOADED BY INPLANE SHEAR ............. .... STEPPEDLAP ADHESIVEBONDED JOINT ..... .............. ADHESIVE SHEAR STRESSES AND STRAINS IN STEPPEDLAP BONDED JOINT .................................. ADHEREND STRESSES IN STEPPEDLAP BONDED JOINT ..... ........ IMPROVEMENTS DUE TO FIRST REDESIGN OF STEPPEDLAP BONDED JOINT IMPROVEMENTS DUE TO SECOND REDESIGN OF STEPPEDLAP BONDED JOINT ....... ..... ............................ ... RFSIDUAL THERMAL. STRESSES FROM BONDING TITANIUM TO GRAPHITE.
7 9 14 18 19 21 21 23 23
25
. . ..................... .?.
F
.
II 1? 13 14 15
16
TENSILE LOADS ON STEPPEDLAP BONDED JOINTS WITH THERKAL MISMATCH .................................... .... COMPRESSIVE LOADS ON STEPPEDLAP BONDED JOINTS WITH THEPMAL MISMATCH ....... ............................ ... FAILURE OF STEPPEDLAP ADHESIVEIONDED JOINT ..... ......... PREMATURE FAILURE OF STEPPEDLAP BONDED JOINT BY DELAMINATICN STRENGTH LOSS AND LOAD REDISTRIBUTION DUE TO DISBONDS IN .................... .......... STEPPEDLAP JOINTS . ....
STRENGTH LOSS
AD
26 26 28 29 31 32 33 34 34 40 41 42 42
LOAD ,~
ETDBoIiTI
nutD Tn
nTcRnNnK
TN
....... ....... ....... STRENGTH LOSS AND LOAD REDISTRIBUTION DUE TO DISBONDS IN STEPPEDLAP JOINTS ........................ ....... . COMPRESSIVE LOAD ON SMALL STEPPEDLAP JOINT WITH DUCTILE .... ADHESIVE .................................... PREMATURE FAILURE OF ADHERENDS DUE TO DISBOND IN ADHESIVE . FASTENER LOADDEFLECTION CHARACTERISTICS .............. .... DEFORMATIONS IN MECHANICALLYFASTENED JOINT ..... ......... IDEALIZED FASTENER LOADDEFLECTION CHARACIERISTICS..... ..... STRESS TRAJECTORIES AROUND BOLTS FOR TENSILE AND COMPRESSIVE LAP SHEAR ....... ..........................
vii
!.
Page LOADS AND DEFORMATIONS ON ELEMENTS OF BOLTED JOINT .... ...... RIVETED FUSELAGE SKIN SPLICES ..... ................ ... WING PANEL JOINT AT SIDE OF FUSELAGE .... ............. ... RAMBERGOSGOOD NONLINEAR CHARACTERIZATION OF STRESSSTRAIN BEHAVIOR .......... ........................... .... BEARING/BYPASS LOAD INTERACTION FOR LOADED BOLTS IN ADVANCED COMPOSITES ............ .......................... ... EXTREMES OF BEARINGBYPASS LOAD INTERACTIONS ........... ....
IDENTIFICATION CODE FOR BEARINGBYPASS LOAD INTERACTIONS
45 47 47 49 52 52
53
31 32
33
OUTER ENVELOPE OF BEARINGBYPASS LOAD INTERACTIONS .... ...... BOLTED COMPOSITE JOINT ............ ....................
INFLUENCE OF HOLE CLEARANCE ON STRENGTH OF BOLTED JOINTS
. .
54 57
58
34 35
36
BOLTED METAL JOINT .............................. ... COMPARISON BETWEEN BASIC AND REFINED FUSELAGE SKIN SPLICES .
BOLT LOAD DISTRIBUTION iN POORLYDESIGfED MULTIRW* f OLT ED
59 61 63 63 69 70 71 73 73 74 75 76 77 79 77:! 81
JOINT ........... 37 38 39 40
41
.......
............................
BOLT LOAD DISTRIBUTION IN IMPROVED DESIGN FOR MULTIROW BOLTED JOINT. .......... ....... ............................ STEPPEDLAP BONDED/BOLTED JOINT ........... ....... ..... LOAD TRANSFER THROUGH BONDED/BOLTED STEPPEDLAP JOINT WITH NO FLAWS ............ ........................... ... LOAD TRANSFER THROUGH ADHESIVEBONDED STEPPEDLAP JOINT WITH NO FASTENERS ........................................
LU M r naNU iRn O ULTLUEuV I
nIIlnIIIuVI
nn
IIl J
.s
LOAD TRANSFER THROUGH FLAWED BONDED JOINT REINFORCED BY BOLTS LOAD TRANSFER THPOUGH FLAWED BONDED JOINT REINFORCED BY BOLTS LOAD TRANSFER THROUGH FLAWED BONDED JOINT REINFORCED BY BOLTS POTENTIAL MANUFACTURING PROBLEMS WITH STEPPEDLAP BONDED JOINTS ............ ........................... .... ADHESIVEBONDED REPAIR OF DAMAGE TO FIBROUS COMPOSITE STRUCTURES ............... ..................... .... SIMPLIFIED ANALYSIS OF LOAD SHARING WITH BONDED REPAIRS ... LOAD TRANSFER THROUGH BONDED/BOLTED FIBROUS COMPOSITE PATCH EDSTIMPLI IEDONAYI &FELOA SHRNGWTHBNEDFREPAIRS.. ... . ... . INDEPENDENT ACTION OF FASTENERS AND ADHESIVE IN LOAD REDISTRIBUTION DUE TO BOND FLAW.. . . . ...... . ...... . .... viii
.,, USE OF ADHESIVE BONDS TO PROVIDE FAILSAFETY FOR RIVETED JOINTS IN THIN STRUCTURE ..... .................. ..... NEED FOR FAILSAFE FASTENERS IN THICK BONDED STRUCTURES USE OF BOLTS AS FAILSAFE LOAD PATHS IN BONDED STRUCTURES .
81 83 83 84
53
tj
Ix
~~i.
ix
TABLE Number 1 STRENGTH OF VARIOUS FLAWED BONDED STEPPEDI.AP JOINTS . ..... Page 33
,1
Ii
SECTION I SUMMAR Y
i
feature of aerospace where for fatigue advanced structural damage can They often represent reductions in sites demanding compositc
It
has lon, been recognized that the critical the joints between the elements. and stressconcentration are even more and These situations they strength,
because mask
are brittle in
redistribute of
(and
imperfections
analysis)
characteristic
conventional metal alloys used widely throughout the industry. This report examines three aspects of the analysis and design of joints in advanced fihrous composite structures: (1) nonlinear analysis of adhesivelybonded steppedlap joints and doublers, and (3) (2) multirow mechanically fastened for IV work and were joints in aerospace structures, and bolt.d jr";nts. metal st.,'ctures computer continuum digital the USAF m~rhanics as well
nonlinear analysis of combined bonded iociude nuriiieaiiIties been and coded into three needed Fortran This The analyses are based on respectively.
programs
research for the NASA Langley Research Centel Laboratory in which elasticplastic analyses
develooed for adhesivelybonded the A4EG and A4EH programs. The miteri a in Section 2
on
iionl i near
analysis
of
adhesivebonded for elastic, the load A have been coded nany of t.yne of
steppedlap joints anJ doublers conta.ns elasticplastic, and biinear adhesive models. into the computer The program samrle A4EI which is used characteristics conposites. applicatior,
demonstrate of the
of thick steppedlap bondeu .loints beLween metal solutions cover the or inplane effect shear; compression, the
tension,
rceidual
stresses due to curing the adhesive at elevated temnperature or to cocure and bonding of the composite to the metal; the improvement of joint strengtth
particularly
importance of having such an analysis method available is unsuitably proportioned steppedlap joints. The nonlinearities included in the new in analysis method
emphasized in terms
in
Section
for the
multirow
mechan;cally
fastened
joints
aerospace
structures
include
characteristics,
elastomechanical interaction
deformation between
and
establishing the failure loads at each station. into a Fortran IV digital computer program
continuum mechanics rather than finite elements, and the computer run times are therefore extremely short. This greater definition of the internal load transfer within older fibrous simpler required such a joint requires analyses and the Sample and report thin considerably indicates are metal greater just input data than data will to illustrate be the cover thick what to
solutions
presented,
strucLure,
in Section 4 of this report needs 2 the and 3. The is is illustrative a nominally typically so no better than adhesive
no
new
derivations show
beyond
those
in is
examples perfect
how the
combination Joint
adhesivebonded benefits
alone,
However,
the combination
in the context of repair of improperly bonded structure of it the is for is inservice fibrous perfect widely repair of damaged composites structures, realworld A4EK or that structure, laminated in thick bonded are and
structures. difficult
recognized
joints
justify warrant
thus
having the
capability to design
allowance single
made for
separate analyses A4EI for bonded joints and A4EJ for bolted joints have been into the computer program combined bonded/bolted
joints.
The
sample
solutions
presented
pertain
to
fibrous
composite
structures, If one
structures.
singled
out
from.the of
in this fibrous
investigation, composite
to be the use
fasteners
structures to provide
by prevent
inq the widespread delaminations that could otherwise be initiated by quite local damage. While the fasteners carry virtually no load as long as the bond and "iaminate are intact, to work in the more efficiently structure, not they enable any remaining adhesive after damage as accept load themselves to relieve splices. the The This condition could arise anywhere as well
examples shown for this prohlem encourage the belief that, with the analyses generated by the A4EK program, it should be possible to thus design sufficient strength of failsafety the damaged into thick composite could be structures. so high as to The residual safe structure
permit
the damaged area, with no notch effect to reduce the load carrying capability
ILIn
3I
2.1
The elasticplastic
steppedlap adhesivebonded joints Is documented in References 1 and 2. Those analyses are modified here to account for the features added since then. The orlginal solution in Reference 1 produced the digital computer program A4EG for bonded joints. 'This was expanded in Reference 3 to produce the program A4EH for bonded douhlers as well as bonded joi,ts. The difference between the two is that, for bonded joints, the entire load is transferred throuqh the adhesive while, for bonded doublers, only a fraction of the load is so transferred, with the remainder staying in the continuous member. As part of the Primary Adhesively Bonded Structure Technology (PABST) program, these earlier programs were expanded to include variable adhesive properties along with the earlier variable adherend properties. This new analysis, coded as the Fortran IV program A4EI is reported in Reference 2. (The earlier proqrams should now be considered superseded.) For the first time it became possible to analyze such effects as load redistribution around bond flaws, the strength loss due to pinch off of the adhesive at the edges of the overlap, and the consequences of porosity. Also, the logic was improved to reduce the already short run times and to minimize the amount of data input. The nonlinear analysis presented here covers all of the material in Reference 2, as coded in the program A4EI, as well as new developments not yet coded. These include the bilinear adhesive model used for doublelap joints An Reference 4. This is not necessary for improved accuracy, but facilitates the generation of a set of solutions for a range of applied loads. The solution here also includes a variable bond width along the length of the joint, as with finger doublers, which is incorporated in the combined program
(S
ODfO
Integration constants Effective disbonding (zero to unity range) Young's moduli of adherends Number of ddhesively bonded surfaces (one or two) Adhesive shear moduli overlap (length of bond) Length of individual step in joint Running loads introduced along length of joint Direct stress resultants in adherends Temperature change (ToTu)
step
P1, P2
T1,
T2
AT
ti,
operating cure
t:
Thicknesses of adherends w2 Widths of adherends Axial (longitudinal) coordinate parallel to load direction of plastic portion of step Coefficients of thermal expansion of adherends Adhesive shear strain Elastic and plastic adhesive shear strains Axial (longitudinal) displacements of adherends
Thickness of adhesive layer
ws, w, x
I
Tel, Tp Subscripts
1, 2 e, el, p, p1
Adhesive shear stre~s Elastic and plastic peak adhesive shear stresses
Different adherends at each end of joint Elastic and plastic adhesive behavior Value of quantity at start of step (x 0)
ref
1
I
(I
(I
iI
*
______
2.3
A representative idealized steppedlap joint is the sign convention of this joint and is nomenclature conveniently analysis
necessary subdivided
G. 1,7.
P
2T.,
S
6x (REFERENCE)
LEic,O,.01 t t /2
JOINT GEOMETRY
t~ E,\
T+

* PLASTIC.TO.ELASTIC TRANSITION
S
ep
ELASTICTOLASTIC TRANSITION
FIGURE 1. NOTATION AND GEOMETRY FOR ADHESIVEBONDED STEPPEDLAP JOINT ANALYSIS Equiiibrium And Compatibility Equations The analysis begins with the equilibrium equations for a differential element of one of the steps.
dx dT2 P2 dx in which w is the bond width. pl + F(ID)w = ,
(1)
:1+
 F(iD)T'w
= 0,
(2)
W is
(3)
The factor F is used to distinguish hetween single and double adhesive bond
I7
surfaces.
F = 1 for one layer of adhesive (single shear) F = 2 for two layers of adhesive (double shear) 4 (4)
The term D represents the effective disbond or porosity at that step, and the shear stress T characterizes the adhesive on the reduced area that is fully stressed. The factor 1D is needed to relate the adhesive shear stress T to the dirferential motion between the adherends in terms of adhesive stressstrain characteristics measured adherend forces T, and T2 refer to on unflawed adhesive bonds. The the total load in the adherends,
rather than the loads per unit width modelled in Reference 2. Those loads may actually be in one or two physical members, depending on the number of adhesi vebonded surfaces. The thermoelastic relations for the adherends are: d6l
=

Ti 11 t2w 1
+ a,(AT),
d6 2 dx
T2
= + a2(AT),
(5)
dx
E2 t 2w 2
in which
AT = Toperating  Tstressfree = Toperating  Tcure (6)
/ n
and is assumed to be constant across the thickness of the adhesive. This approximation violates the stressfree edges of the adhesive and results in a significant overestivate of the elastic adhesive shear stresses at the ends of the overlap; however, the error becomes progressively smaller as the adhesive is strained more into the nonlinear or plastic state. Differentiati)n of Equation (7) and the *elimination of the displacements via Equations (5) leads to the result
dy 1 /d6I  = 1 2 dxf
T
d61 dx
Il(

Ta
T1
Eitiwi
+ (__  a)AT).
(8)
n \E2tav
f.t
of Equation (8),
dy F(1..D)w( 2dx T)
+
1 fl\~it Eiw
II
:
E2t2ww
+ E.
ttW2W1(p E2 t 2 w2
(9) (9)
or near the middle of the bonded overlap, the shear strain by the equation . = Gely (for y  ye
(10)
tel + G 1 (Yye)
(Ge
)y
+ G Y.
These various
any more complicated adhesive models. ELASTICPLASTIC MOOEL FOR INTERMEDIATE LOADS 1CnM MODEL I r. ~ELA1LST
1
S\
.
x,

M LBILINEAR CHARACTERISTIC
Elastic Solution The combination of Equations (9) and (10) yields, tial equation for the elastic behavior,
21yG
2
+y 1
T)
P1
F(ID)
(Etltwi
)
flE 1t1Wi
P2 Ezt~w2
(4
X=
F(1D ).
+(14))
(15)
The elastic adhesive shear strains are then expressed as y = A cosh(Xx) + B sinh(Xx) + C
in which
1(P1 P2
,
(16)
C= X
+ Ezt 2 w2 Eitwj
/
in turn, making use of Equation
(17)
(10),
and
T2
Tr
ref
P2X + F(lD)l
elPX
sinh(Xx) + ""
osh(Xx)1
Cxl.
(19)
depend on the origin The values of the integration constant TIe and T2 In the coding of the computer program A4EI, it is convenient adopted for x. "to adopt the start of each step as the origin of x for that step. Integratino Equations (18) and (19) again and substituting the integrals into Equations (5) yields the following
10
<1
1 11
IIX
refi
+
pix
cqi(AT)x +
Ti
Br
+ 
F(
De
A
Cosh(
inh(xs)
.x x
(20)
and
62= 62
ref
+ c2(AT)x +
E2 t 2w 2
T2
ref BC
x 

F(ID)Gew
Xx )
(21)
computer program A4EI used to solve those in Reference 2 without the The
more
for the elastic adhesive behavior employs the followto determine the integration constants A and B.
notation ref is used to identify the start of any step or bonded zone within the overlap. The constant A in Equation (16) follows from evaluation of Equation (16) at x O. Thus
y
ref
C,(2
(22)
with the constant C given in Equation (17). The other constant B can be deduced by differentiating Equation (16) once and equating the result to that given by Equation (8). Thus
+ (ca.al)AT
(23)
I T'rf
BXn

ref
+
(a'i)tATJ
E 2 t 2w
E1 t 1 w 1
6, and
(18), (19),
11
]I
special
case
of
the
it
is
necessary However,
to the
or prescribe, So it is
a complete necessary
conditions.
actual conditions which must be satisfied are shared between both ends of the adjust of the initial conditions iteratively until the other houndary condition is the bonded overlap. Plastic Solution The inclusion of plastic nonlinear behavior in the analysis is similar to that above for the elastic solution, with one of the major complications being the unknown location of any transitions between elastic and plastic behavior. The plastic adhesive shear stesss is constant throughout that : satisfied at the far end of
portion (or all) of the step, as given in Equation (11). (9) becomes, for the plastic zor,es,
Therefore,
Equation
d'y
xz
Ge
2 x 
n \ Eitw,
P1
E~tqW9
(25)
Thus,
1,
P,
P2
w2 /
constant,
(26)
K ref and,
(28)
x x= J =
retf
TrI
E~tJwJ
+ (a2 aflT(9
\Elt2w2
The adherend loads in the plastic zone follow from integration of Equations (1) and (2), by means of Equation (11), as
12
Ti = Tiref + pix
F(ID)
Px
(30)
and T2 T2
ref
(31)
Equations (5) 61
6
/I
T 1 refX +

2
2P  F(1D)WT
2 X

(32)
T2r f
2
P2X x
+ F( D~w
2
(33)
steps of steppedlap bonded joints or doublers have fully Any plastic (or other nonlinear) steps, behavior are is rebecause in structures order to
throughout. fatigue
nowned, it is desireahle not to operate the adhesive beyond its proportional limit for frequently recurring loads. (11he piastic strength is best reserved for overloads and for local load redistribution dround flaws.) in performing it is elasticplastic aualyses of adhesivelybonded joints, necessary to compute the extent of the step a transition to elastic behavior. are shown in Figure 3. Starting from Therefore, steppedlap
under considera
The
various
behaviors
Equations (27)
7 lX2 . + Jx + yref'
in
the lesser value of x for which Y = Y Equation (34) ir first rearranged to read 0, given by (36) (35)
13
0ZLL1
(OR FULLY NEGATIVE
EOQJIVALENT BEHAVIOR
A FULLY P.A.
IOR
(ORINVERTED SEQUENCE.TO.)
(OR INVERTED

0
Tat
OR INVERTEO SEQUECE TO )
IE, C
2(Yf
(37)
The minus sign in front of the radical holds whenever d~y/dx is less than zero, as at the near end of the joint. Now, if the loading is such that y is negative and dy/dx positive at the start of the joint, the appropriate answer is the lesser of
 +
re
(38)
In evaluating this dimension, it is important to note that the expression for H in Equation (26) is modified by reversing the sign of the term containing Likewise, the expression for J in Equation (29) the plastic shear stress T After x has is altered because the adherend loads T, and T2 are negative. been computed from Equation (37) or (38), it is compared with the actual step
eng
step
If
step'
that
particular
step
is
fully
plastic
14
throughout, t
and the
values of the various quantities at the far Pnd of the (30), (31), and (33).
the difference is examined elastically to see Should x be less than X step' p whether the step remains elastic thoughout the remainder or becomes plastic again at the far end. or It is necessary that dy/dx be in maintained order to at any plastictoelastic equilibrium, elastictoplastic trnsitions one must ensure
In solving for the maximum prov,,de both for the at each end of stress. The
possible length of the elastic adhesive zone, the step and for the case of a reversal
in the shear
latter can arise physically in the presence of adherend thermal mismatch (as between fibrous composites and metals) but occurs most fiequently as a result of an J zone is imbalance between the known and assumed boundary conditions at the The procedure to determine the length of the elastic of the distance required for
the adhesive strain to become Y If Y or y e e the elastic zone it is does not extend to beyond the far the load In doing so, zone and (30), and (21) (31), end of the it is
analyzed,
necessary
compute
transferred
adherends throughout the elastic trough. the length computed (10), for (18), the elastic (20), (11), Equations I (16), (19), (27),
Should the elastic trough not extend to the far and (33)
Bilinear Solution The linear portion of the bilinear adhesive model bY the elastic analysis above, is necessary to discuss only shown in F igure 2 is given So it case
just as for the elasticplastic model. the nonlinear portion here. In this
15
S(
l
Ejtjwj
l1
E2t 2W,!
dx dx2
T r2
/
F(lD)

1

1
+

1/ri

E2taZV2
/1
(G elG 1

Eit1w
)
E2t2w2
(40)
P aN
Ye +iP
(41)
dx
eGl
It is
El
It can be seen that this has the same basic form as Equation (15)
linear elastic solution. G
X, .
G el
()2
(42) is
+
C'
(43)
CI
2
P1
(X,') T El twl
P2 E2 t 2W2
G1
(4 (44
.G
(45)
D'GPis
(46)
I
D ('t
PD
P2
.
C. (47)
The loads in the adherends then follow from the integration of Equations (1) and (2) as
Ti
f+ pix
F(ID)G 1w
,sinh(V'x) +
cOsh(X'x))
DIX
(48)
16
and
T ref T2=TT2.
P 2 X + F(ID)Gplw
Xp sinh(X'x) +

cosh(X'x)Cx1
+ D'x
(49)
in the form
6=
c(LTT)x + ref + a
Eitiwi
T1
X2
F(ID)Gpw i )).x)
+
.
rA'
B'
(x)
and
(ii(' (oh('))+
(50)
1
6 62e + az(AT)x + E2t2w2
~ pz2
T2efX  2
)PI + F(ID)G
A'BI
I()2
D IX
. L(inha(Xlx)Xlx)
I'
=
(os(Xx)1)+
(w))I
t.(51)
J)
These
equations
in
much the
same way
as
th se above
elastic and
plastic solutions.
evaluated at x
A'
= yref 
_Ye C'.
(52)
constant Thus 
B'
is
determined
in
exactly
the
same
form
as
in
/T
2 t 2w,
(av)AT)
(53)
the
analysis
above,
it
is
necessary
to
compare overe
the bethe
extent
of each
entire step length and determine whether or not the strain at the far end of The step can be subdivided the step lies in the linear or nonlinear regime. into nonlinear and linear zones as necessary.
17
InPlane Shear Loading The derivations above have been presented for inplane tensile or compressive loading across the bonded joint. 1. Only two changes are needed. The governing equations for inplane shear as explained in Reference
E.
loading (Figure 4) have precisely the same form, must be replaced by the shear moduli G, and G2 . the distribution parameters A and X' which, adhesive tension weaker shear under strains inplane the than if the than still same under usually shear shear or compression. However, shear is since the in load
moduli
and E2 in in or the
This causes an increase results in were materials inplane the weak intensity applied tension link those the in
compression,
adhesive
inclusion since in
induce
preloadin2
That effect must be ignored until some more complex analysis is formulated to account for local effects which peak in two corners of the bond a;rea rather than all along two edges. With ductile adhesives, the resultant effect of a small preload strain orthogonal to those induced by the basic applied load is often insignificant anyway.
tAD
7"
/A
I.
18
2.4 The
SAMPLE SOLUTIONS sample solutions for to adhesivebonded graphiteepoxy bond stepppedlap joint. joints and doublers
which are given below are mostly concerned with. variations on one basic highloadintensity elastic, stresses titanium and The capabilities residual of the computer program A4EI are illustrated in terms of the different solutions for ultimate, potential strengths, of the thermal induced during cooldown after cure, of the different strenqths in
tensile and compressive load application, and of the load redistribution due to various bond flaws. Part 4 of this report uses this same basic joint as a means of showing how effective mechanical fastening is on its own, in con
disbonded
augment the information about the features and capabilities of this program.
iTAI UWTS
TS'RAMITE.IPOXY.\___
STATION WU#msNERs
2 SV " STlps
(TYPICAL)
52.
5 1 2 SCALE (cm) 3 A. SPECIMEN GEOMETRY 6 SCALE IIW.) 1
A
7 'T",  10UPSI
ADENN I
AOAEINEID!
Fu (/
"7,  4JI/?l
1. 4 SHEAR $TRAIN
"
SI'
'1
"yf 62
CL. a C D EEDPOETE
B. ADHESIVE PROPERTIES
f19
Basic Joint Description This basic joint is ately not optimized, shown in Figure 5. It is a hypothetical joint, deliber
are not specific to any particular adhesive hut representative of an adhesive which has a higher service temperature capability than the ductile adhesives used on the PABST program, in conjunction with somewhat more ductility tnan exhibited by the first generation of hightemperature brittle adhesives. The following illustrative example of solutions the analysis perform two functions. and, in They the and
demonstrate process,
the
program
A4EI in
explain
important
considerations
the
design
joints.
itself is
listed in Volume II of this report, along with the user instructions. Solutions In Absence Of Residual Thermal Stresses Figure 6 shows the adhesive shear stresses and strains Figure 5, associated with the
stresses.
for this case, the adhesive properties used in the analysis should those which are apply to curves at the stressfree temperature, about 3000F, he compared later with equivalent solutions stress effects and it would be confusing
thermal
to cover two perturbations simultaneously.) Three curves are shown in Figure 6; for the adhesive elastic capability, for the failure of the joint in the adherend, and for the potential is sought. bond strength of the adhesive had the adherends been stronger. when the joint right, there is on the left. strength The peaks in This information is typical of the answers obtained curves or, the
Note how,
in the strain
none of the flattening out associated with the stress curves of the stress and strain curves properties at each occur at the steps It can he seen adherend station.
(discontinuities)
on the right,
fraction of the adhesive is loaded beyond the elastic adhesive strain, so the accumulation of creep under sustained load is improbable. Indeed, over half of the adhesive is strained to less than 15 percent of its failure strain
20
I.
II
I'
I

L,1/I01.
0.20(AT
'
'6
9. ULTIMATE STRENGTH * 3015 L1MN.11 STATION 1 CRITiCAL AT SADHEREND SONSSTRENGTH $1,511 LIAN. C.POTENTIAL ADHESIVECRITICAL AT STATION'l
A. 0*15c
0.071
oN
0.1"
0
.200
w4
ADHESIVE
$HEAR STRESS WSHl 5 ADHESIVE SHEAR 0.10 STRAIN U
SCALE:
Of AO.IE9ENDS2 STRIIENGTH
LOAD S
LOAD
NI
CRITCAL
E CRITICAL
STATION
$
LOADT RISA
',I
to
AOIIER14OS2 SIRENIITI OF
I ADONIRNO OF STRENGTH
LOlAD KIAL OVERLOAD KSIN Al TWO STA004 IAIFN~, III TITANIUM 3
CRITICAL..~.,w 4PAILUREOF
*I
N I LOADINAD IENO
Is
LOADIts AOlIIREhI(
fi
ISCALE
I
S~~TATIO
STATIO SAII
111CHIlS I
This design is
in Figure 5 was
deliber'Ately
not optimized,
to highlight some
characteristics of steppedlap adhesivebonded joints, limited not by the adhesive, but by the thin, the titanium plate, further information, as shown in Figure 7. beyond that in
fatigue if the joint proportions are not sized carefully. Figure analysis of such joints with the program A4EI. Improvement Of Joint Strength By Optimizing Proportions Armed with information of the type shown in about modifying the that, equal, design to improve the good.
since the adhesive shear strains at each end of the joint are roughly the overall stiffness balance is Any further impovement in that
regard would probably be thwarted by the constraint that the fibrous composite laminate must consist of an integral number of plies. The need for stiffness balance is why there is such a pronounced thickness buildup at the right side of Figure 5. thin, it too long, or both, lest it too thick, However, it joggle the is since the titanium end step is It fibers in either too weaken
should be modified.
the other member more than it should shorten that step, decrease the minimum
strengthened itself.
adherend is be shortened.
as shown in Figure 7,
the ultimate strength was increased from 34075 lbs/in. to 38033 lbs/in. This is a rather small increase, and was sufficient to reduce the potential bond strength from 51519 lbs/in. to 38033 lbs/in., since the adhesive is now critical Therefore, one should perform a further redesign of the joint with the only variable left  an increase in the number of steps. The results of such a further design iteration are shown in Figure 9. The ultimate (and potential bond) strength was increased to 46363 lbs/in. initial This represents a substantial design the in Figure step 6, increase over the 34075 lbs/in, how important it it niu st is to be of the showing rather than the adherends.
properly recorded
proportion
details.
However,
22
IKIPSJAMI
F4gum7
VERY SMALL MARGIN IN TITAURISTEP LA hHHHN
4*SIIEAq
*
ADHESIVE R.ig
STRAIN
STATION
STRENGTHOF AOWIIRNDS I
Ij
916
ILOAD

114
,EiP/E 1011.

.
LOADIN AHIHISRSO9,2
1E
STATIONSTATION SCALE
isTREIIGTH
I OFADHRERING
6.2S.
F
E.IN[ US POR
**
LOAD
RIPSflN 46
Ii113I.
STATIUMN ADHESIVE
0325
LOAD
*ICN
2 2
11 I2 13 14
'
STATION
SCALE IINCHETI a 1 2
STATION
also
that, is
for
this
relatively
thick
steppedlap the
bonded
joint,
the
bond of
strength
still
substantially
less than
lesser
adherend
strength
further refinements
straightforward for thinner adherends, which are subjected to lighter loads, For a real design, one would need to analyze the joint throughout the entire operational condition. environment, A comparison and of iterate Figures 6 the (and design 7), 8, for the most severe the and 9 shows that
So it critical location in every instance is the thin end of the titanium. is here that further design refinements would be applied. These would take the form of a greater number of shorter steps, up below that for the strength build up. to keep the rate of load build However, a word of caution is and
necessary. step
increments be it
limited or cloth.
finite
composite design
material, iteration
would
be easy to continue
this
turing techniques would be beyond any reasonable tolerances. the program A4EI increase only one more design iteration, its strength,
and the next step likewise to 0.080 inch and then to tolerances in chemmilling the titanium and laying of such a final analysis, with all
1J
use the computer program to assess the strength iosses to be expected due to reasonable manufacturing
The results
were an elastic strength of 21032 lbs/,n,, are, in fact, significant strength decreases that it is
and an ultimi~te strength of 39957 lbs/inch, with the adhesive critical at the
I,
with respect to the configuration analyzed in Figure 9, confirmin. going to be difficult to improve upon that design. One very important characteristi: of the design of steppedlap is not apparent such or in Figures 5 through 9 is have The 00 fibers is on adjacent that the the best
900
that experience has shown that the to the adhesive, fibers to the a
joints ply.
t:450 a
reason
transverse available
roll This in
under the shear transfer loads and the resin splits between the fibers. imposes severe constraint options
designer
24
particularly if
that the joint area should be considered more in setting the fiber pattern. The fiber pattern must not deviate much from a uniformly dispersed quasiisotropic pattern if the laminate is to be free from the weak cleavage fibers are bunched planes associated with laminates in which parallel together rather than interspersed.
temperature attack,
to
cure
the
adhesive
and
give
it
resistance
to
environmental
"residual
(half an shrinkage
the metal tends tc, shrink during the cooldown. That induces stresses and strains in the adhesive for all but joints so short inch also or so) that they are alleviated in by adhesive creep. The introduces tensile stresses and compression in in Figure 10 for the basic steppedthe metal
These effects are shown the composite. lap bonded joint shown in Figure 5. Figures 11 and 12 show what happens when
NOT! THATINTEGRAL OF
SHEAR STRESSES S ZERO
[
ADHEISVE SHEAR 2,
~LOADh
;KIFS/IN.!
STAT'A1N
STRESS
(Kill
42
,8I
'
AHE LOAD
SCALE
IINCHESI
25
RESIDUAL THERA:AL STRESSES A. ELASTIC STRENGTH  I1.20 LIAN LN!IN. 8. ULTIMATE STREGI4  24 Sl.S ADHERER') I CRITI(CAL AT STATIOIN C POTENTIAL BONO STRENGTH  4,164 ADHESIVE CRITICAL AT STATION I ASOCIATED WITH IT
L
LAM. 011 TENSILE LOADS IN ADNEREND I C 4,S
LOAD (KIPS/IN.)
a A
ADHESIVE
4I 3 2
5 4 STATION
0 1
(KIPSAiN.I
[
1
nJ
2 3
I_
I
4 SCALE
INCGHES a
I
7
I
1 2
1 1 2 2 3 4 S 4 7 Il
4 5 STATION
STATION
RESIDUAL THERMAL STRESSES ASOCIATED WITHAT 3SF A. EtASIlCSTRENGTH  lELS/IN. 0. ULTIMATE STRENGTH 235.I LRIN. ATSTATION 2 AOHEROS 2 CRITICAL 0 22 LOAD (IIPS/1N.
r.
[
1

STATION 2 3 4 s
'
~~
AlHIESIVc
ST"EFss .4
STATION
2 UIPSJlN.)
@VYERY CLOSE TOc PURELY COMPRESSIVE LOADING IN ADHERENDS 2 SCALE: (INCHES)
LOAD
4a
1
FIGURE 12.
.*
26
tensile and compressive shear loads are superimposed on the residual shown in Figure 10. It is significant that 'he tensile shear
stresses stresses
combine with the tensile residual stresses at station 8, on the right hand side of the joint while station 1, on the left., is critical for compressive loading. tensile different, It is very clear that it is necessary to analyze the for both the are modes
separately,
because
strengths
are different
under compressive
loading,
selfevident
adhesive to help resist the accumulation of adhesive creep. 2.5 EXPERIMENTAL EVIDENCE
The
importance at
of
having
available
computer
program
like in
A4EI,
or the R&D
can be gaged
metaltocomposite
steppedlap
Some of these problems programs over the years at various establishments. are discussed here, to explain them and to encourage their avoidance in the future. Since their history is irrelevant to the present it discussion, no
identification
Nevertheless,
should be noted
that these are real problems and have occurred far toi frequently. happens ".n th end step on the metal is made too Iona
3
Figure 13 and
4 too thin.
a problem
ws
ab
It breaks off!
permitting a secondary failure elsewhere with no indication of this cause of The great bulk of evidence that the long thin end step load redistribution. is has occurred in fatigue the Figur, testing. bond, 10, there Quite is apart also the from the tensile
mechanical
load
transferred
through
as shown in
due to the steps, which are more frequently chemmilled rather than machined. The test specimen shown in Figure 13 had a thickness of 0.030 inch on the end tab and a length of over an tab of that thickness, inch. The author's experience if 27 it were in double suggests that a shear. The A4EI length of only a quarter of an inch would have been more appropriate for a particularly
S!
program permits a systematic parametric study of such joints to be made for a small fraction of the cost of even one such test specimen. This can greatly reduce the need for test specimens and maximize the usefulness of those
actually tested.
SE LOAIN
END METAL TAB STILL GLUED ON, BROKEN OFF METAL ADHEREND
steppedlap titanium to graphiteepoxy bonded joints is given in Reference 6. This was a much easier task than similar work on the F15 tail and F18 wing and tail because, for the A4 graphiteepoxy stabilizer program it was ,, 4.u a L,,,,. ductile adhesive, Epon 951, and the titanium was only 0.?5 inch thick. During several tests, at progressively more complexity, the failures were consistently in the fibrous composite outside the Joint area. Tests on that same program, described in References 1 and 7, highliqhted the potential problems that can occur due to wrinkles in the fibers at the tip of an embedded titanium plate. That is why the joint shown in Figure 5 includes a tapered fiberglassepoxy wedge beyond the end step, cocured with the basic graphiteepoxy laminate to eliminate both the wrinkling and the ahrupt termination of such fibers if they had simply been butted up against the end step. Quite apart from reducing the structural efficiency by carrying around excess
28
material, such abrupt termination of the fibers just beyond the tip of the titanium splice plate can actually weaken the strength to below that which would be attained had those plies been replaced by lowmodulus filler material 14, to prevent them from picking up load. a simplification of a which is This is explained in Figure somewhat more complicated condition in
which the delamination was actually initiated by the combination of shear and peel stresses. The example on which Figure 14 is based had a tip thickness in excess of 0.1 inch, with ove. 50 percent of the central plies terminated The consequent stress concentration was far worse there being longitudinal. The number of plies which can be than normal design practice permits. simultaneously terminated safely in a graphiteepoxy laminate, without fear of inducing delaminations, is closer to four rather than more than twenty, whether there is a joint in that vicinity or not. Likewise, it is important to thoroughly intersperse the cross plies rather than lumping them together, because of the substantial residual tensile stresses in the resin, which has typically two to three times the coefficient of thermal expansion of aluminum alloys. Just because a laminate is not warped, it does not mean that it is ................. +........^............... i actually due to relief of Such internal Any a, a ,tsrss. n stresses which would otherwise be worse.
"TEST AREA
' ..... FRACTURE
FRA TUR
FIB R O U SC O M P O S ITE
DELM.INTIU
~LOAD
INTRODUCTION AREA
METAL
While on this subject, a 900 ply within at all to the resin is than a behave case, rather lateral like a
it
is
and explain that 250F In does not latter such t450 the
laminate
transvers
test
free to contract laterally (perpendicularly develop internal stresses, while in the former is resisted layer If as by adjacent a fibers in the is 8, the 900 occur, within typical in laminate
directions.
and usually also precracked. delaminations tensiontension fatigue loading. The purpose of mentioning
if
certain good design principles are not followed in designing bonded steppedlap joints in fibrous composites, the analysis given here would become quite inadequate and would and need to be replaced by a far more within the complicated for model which checked not only for the adhesive and gross adherend behavior but also checked each every resin interface laminate possible
instead.
when titanium to fibrous composite steppedlap joints are cocured and bonded with an adhesive film, as is done on the FI1 and F15 by McDonnell and was done on the A4 R&D horizontal tail by Douglas, it is virtually impossible to manufacture a bad joint unless there is something like a bag failure which would ruin the laminate easy as well as the joint. However, caused if one were to it precure all the details separately and then try to secondarily bond them. would be relatively details, Indeed, to have major disbonds the joint is the A4EI (PABST) by poor particularly if introduced
fit of the
thick enough to warrant several steps. The variable adhesive Primary Adhesively in Reference 2, steppedlap again bonded program program, during the as
Structure
Technology
described below,
an assessment
of the'effects
of such misfits in
using the
.1
30
I!
Figure 15 shows the consequences left end 'of the joint, Figure 16 of the two end steps being disbunded at the shows the load redistribution due to the
arid Figure 17 shows what happens when tne It is immediate'ly obvious that the
P!
S
LOAD
XIOAD.) 4 IP M I
S ~7
7.o a2
ADHESIVE CRITICAL AT StATiON1 3 ULTIMATE UN1iAAT LOAD OF 8.h11LIINII.
,NADHERCtiO, LOA DA
I
1I
ADHESIVE SHEAR
STFISS
tOISBOND ! ".'~~~
i6
aI
I I ILOAD
VSOlstO"D
LOAD IN ADHERENDS 2
,2
7 2 1 2 3 e 5 STATION 11 7
S
FIGURE 15. STRENGTH LOSS AND LOAD REOISTRIBUTION DUE I'O DISSONDS IN STEPPEDLAP JOINTS load redistribution due to the disbonds is Figures 8 to hond flaws is could central 11 of Reference 2 for far more substantial thick bonded then shown in The
uniformly
adherends.
reason for this much qreater sensitivity of the larger more complex joint to that there are no large areas so lightly loaded that the flaw Figure 16 is are no be missed or to which the load transfer could be moved. flaw is so intense that the critical adhesive
particularly significant in the sense that the load redistribution around the conditions longer located at the ends of the overlap, Figures but at the edge of the flaw.
15 to 17 have been prepared from the thermalstressfree solutions an6 the predicted strengths are well down from the 34067 lbs/inch in Figure 6 for nominally perfect bonds in the same specimen. Table I enumerites some of the predictions when thermal stress terms are included. The strength losses
31
ERo
LOAD
ACRO13
4 STATION
LOADIN AOFIERENDS 2
LOAD
I
5 SCLtE.L N 1 6 a
FIGURE 16. STRENGTH LOSS AND LOAD REDISTRIlBUTION DUE TOO ISBONDS IN STEPPEDLAP JOINTS
I
I
TENSILE LOADING
so
L
LOADINIADMRENEN I
so
SHEARI STRESS
4
sb
.
s.~ INO
LOADIN ADIIERLNOS 2
I
I 2
2I
~~
'
4J
STATION
FIGURE 17. STRENGTH LOSS AND LOAD REDISTRIBUTION DUE TO DISBONDS IN STEPPEDLAP JOINTS 32
THERMAL STRESSES ASSOCIAl ED WITH .AT  300F ULTIMATE STRENGTH SOLUTIONS GIVEN
TENSION
21,814
0.110 AT S
15,911
0.200 AT 6
8430
0.200 AT 6
0.106
1 AT 7
Further analyses,
,reo areas.
fld
^ involve
have been made to predict how much of that strength loss could be
bolting through the disbonded Those solutions
the use of the combined bondedboltea joint analysis program A4EK. Figures 15 to 17 fail to show one other important effect of disbonds in steppedlap bonded joints only because the adherends are so thick and strong that the adhesive is the weak link. This the end steps are disbonded, the entire section of the adherends, just as if they side a sound joint. This effect is shown
a more ductile adhesive.
other important effect is that, if load is now carried in a reduced had been notched or cut down outin Figures 18 and 19 which is not
for the thick
for the same joint as used in Figures 15 to 17, but for a thinner joint with
This effect could not be demonstrated
joint in Figure 5. This situation of a disbond weakening the adherends more than the adhesive would be prevalent for welldesigned joints between thinner adherends, in which the adhesive had a considerable margin with which tolerate some reasonable level of disbonding. to
33
STATION
ASSOCIATED WITH.1T 3 F ULTIh;ATE ST ENGTH  ;IM1 Ll/IN. ADHERIENDS CRITICAl.ATSTATION 4 STATION (KI0150M
2
.
ZO.D
lf
LOADIN AD1ER1 N, I
I3
. MAXIMUM EFFECTIVE THICKNESS OF FINDOUS COMPOS4TE SECTION (ATFAILURE LOCATION)  !.INICH AS OPPOSE TO CAI INCH IF THERE WERE NO DISSONGS
2 AOI~rIIVE
SHEAR STRESS
3
,iI
I.
N 1 2 3 4
i
1
i "
a
"
7I

I
FRACTURE
CONSTANT LOAD
..
I
4
KISIJLOAD zo
"II.'
6 6.o1 9 ' Y.s A2
Nl
I
IN AOIIREISOS2 LOAD
hIN
IES ) CH AL E: (IN SC
FIGURE 18, COMPRESSIVE LOAD ON SMALL STEPPEDLAP JOINT WITH DUCTILE ADHESIVE
STATWXN ,,
RESIDUAL THERMAL STISS5 WITH,IT  3110F. ASSOCIATED ULTIMATE STRENGTH  64N LU/ IN. AIRhtfNOD I CRIIICAL ATSTATIONZ
l
ADHESIVE
STRESS
2
20
*.12 1V,
INADHEREND I 5 N 7S
IRSI)
If..
_1 ;.*..LGAD
(KIPILNI
LOAD
INAOHFRENOSZ
4

$10o
ADHESIVE PROPERTIES
I i
HTS GRAPHITE.
. J
"
""JOINT GEOMETRY
EPOXy
31
2.7
CHECKS ON THE ACCURACY OF THE SOLUTIONS the computer program A4EI is A4EG version, that it is
in comparison sample
to devise
solutions to exercise all options within the coding to verify its accuracy. The last hugs were not found until after over fifty complex joints had been analyzed with several different conditions for each, However, in there are some maintaining the
techniques which have been developed which can be useful program. The most obvious give a precisely can be ones are mathematically trivial, of uniform adherends solution. residual with in a Likewise, thermal
important. joint
The should
thermalstressfree solutions
case devised
doublielap
symmetric
antisymmetric Also, a
problems.
problem run with a tensile load and negative AT, for example, can be compared with what should be a precisely opposite answer for a compressive load and the positive AT. Such checks cannot verify that all parts of the program are
operational, hut they are an easy way of exposing a problem due to misreading a card, or to cards out of sequence, which could not be detected from a very complex solution because there is grossly wrong. 2.8 CONCLUSIONS joints and account for variations is a valuable in no basis for questioning it, unless it is
or geometrically,
asset.
This is particularly true in regard to the load redistribution around flaws. * The computer tions of graphiteepoxy that the program A4EI laminates. of is particularly joints, as usaful between in optimizing titanium edge the propormembers and
steppedlap
bonded
Both analysis and test over many years have shown joints can be particularly in particular, sensitive as well to poor as to any
strength
such
detailing of the end step of the titanium, gross mismatch in adherend stiffnesses.
.35
elements, the computer run times are extremely short. However, the actual run time depends on just where in a joint it becomes critical. Run times are typically fivc times as long if the joint is critical at the far end rather than at the near end where the analysis began. The sample solutions described here show that the computer program A4EI will be a useful tool in analyzing adhesivebonded steppedlap joints and doublers, particularly because of its nonlinear capabilities, without which failure prediction is virtually impossible.
36
?I
SECTION 3
NONLINEAR ANALYSIS OF MULTIROW BOLTED JOINTS FIBROUS COMPOSITE AND METAL STRUCTURES
T 3.1
22IN
INTRODUCTION
The analysis of the load sharing in multirow mechanically fastened joints in aerospace structure tions metals and empirical are which has, very in the past, and been limited have to tend to linear elastic for typical without solutechniques. forgiving These sufficed deform ductile
catastrophic
Some of these existing analyses have been used very effectively to failure predictions are lacking. The manner in which the end
obtain great insight into the behavior of multirow bolted joints, even though fasteners tend to pick up a disproportionate share of the load has long been understood. at some And any sculpturing load to redistribute which to analyses the art ago the have of load been transfer run. more is These evenly will be just as effective at some indeterminate failure load as it lesser tools which specified have is had for existing splice contributed quite evident been designed much a designing to
mechanical established
However,
fibrous composite materanalysis methods for is that fibrous comanalysis is or could are needed even be
ials has necessitated thc development of mechanically fastened splices. One reason posites are so brittle that a more precise
loadsharing
no material yielding to mask approximations because there is inaccuracies in the predictions. The analysis for that aspect linear. feature However, in the same brittleness causes Since a need around the analysis. the clearances fasteners
times about the same as the maximum possible stretching of fibers between the fastener rows, it is obviously necessary to account for any initial
37
Even if it should transpire that clearances or preloads at the fasteners. hole clearances in bolted fibrous composite structures are eventually shown to he unacceptable for much primary structure, this new analysis might well There is another need for he the key to demonstrating such a conclusion. including nonlinear behavior in the analysis of mechanically fastened joints This need is due to the noncatastrophic bearing in fibrous composites. damage that occurs in the iwnediate vicinity of fastener holes whenever the holes are not close enough together to fail by tension through the hole. Thus, even though fibrous composites are customarily regarded as linearly elastic to failure, they still need a nonlinear analysis for bolted or riveted joints While the justification for improving the existing analysis capability is
greater for fibrouscomposites than for ductile metals, it makes sense to include in the analysis developed any extra features necessary to permit a A better characterization ,ett.er analysis of bolted metal structures also. of the load transfer distribution at typical fatigue loads can then be For this reason, the analysis developed here includes nonlinear obtained. behavior of the members between the fastener rows. 3.2 SYMBOLS
d E F
Kell Kp1 k
Fastener diameter, can vary along length Young's modulus Stress level in RambergGsgood model deflection characteristic Stiffness of element of bolt load versus
2.Step P
P1, P2
T 1 , T2 T AT
t
1
, t
WI,
w2
38
n
U2 6L1
II
Coefficients of thermal expansion
0 Suhscripts 0, +, 1, 2 k e, p
Ulzt

Stress in members
Sign convention for displacements Identification of each member Station number identification Elastic and plastic values, respectively
Ultimate value, at failure
fr
3.3 LOADDEFLECTION CHARACTERISTICS FOR A SINGLE FASTENER The transfer of load through mechanical fasteners (bolts and rivets) within a joint is characterized in terms of the relative displacement between the Figure 20 depicts the simplest possible members at each fastener station. The takeup of any initial mathematical model for such load transfer. clearance is shown to be friction free but, in principle, it would be quite force throughout that straightforward to include a constant (Coulomb) Any further load applied to the extremities of the displacement increment. joint induces a shear load in that fastener as the relative motion is resisted by the shear strength of the fastener. hut there linear, as shown, contribution prior to failure. is also The initial load transfer is usually a significant nonlinear more complex possible to (horizontal) to adopt a or to
No in Figure 20 is the simplest such representation possible. In some instances it would be representations are necessary. have an ambiguous (nonunique) solution if a perfectly plastic nonlinear mathematical model were used, so It is preferable positive slope, however small, interferencefit fastener, of in the model. course, there In is
contend with while, for multirow joints, it is possible that the "other"' fasteners may induce a positive or negative preload on the fastener under consideration.
39
.i'RE
"WCOMLINEAR
GREATvy O"A
TRANSiER
"
aT
SCTERFER)NC
FITS)
RELATIVE DISPLACEMENT
FIGURE 20. FASTENER LOADDEFLECTION CHARACTERISTICS in uri.;r to ... ipleent such a necessary to c eti ie also ?odf precisely ect.on what curve as in Figure 20 it is meant by the is
relative
c to its equivalent for doubleshear fasteners. apart under the load P. Those points are
sufficiently repiote from the fastener, at station 1, that the displacenients at A and C can be considered to be uniform across the section. The relative motion between stations A and C can be considered as the sum. of three 1 b.tween .components; the gross section stretchi.nt (or co.;ressior) ,n member A and B, the gross section stretchiio
combination of th.,, local disturti.ons
and the
f the
fastener and the shear deformatiGn (arid/or ;otatlo'KO of the fastener itself. By isolating out the first two such conpunents, one is left with what is
It Iz speflfc to a given fastener referred to as the fastener flgxibillty. in particular materials, but is independent of the location chosen for stations A and C. Such charActeristics rre customa!ily deduced by measuring the relative motion between stations A and C and removing mathematically the amounts that the members I nd 2 wold and BC, respectively,
13ye
%M7A.UREV
P
OR COMPUTED)
(MEASURED EXPERIMENTALLY)
REF R
~LENGTHS
*Il
A REF B C FINAL FASTENER LOCATION DEFORMED POSITIONS SSHOWN BY DOTTED LINES
OTAL EFFECTIVE FASTENER RELATIVE DISPLACEMENT AND INCLUDES DISTORTION OF CROSS SECTION AT FASTENER STATION
FIGURE 21. DEFORMATIONS IN MECHANICALLYFASTENED JOINT include the local distortion of the members around the fastener holes within
the characterization of the fastener flexibility to simplify the analysis of such joints for the load sharing between fasteners, as is explained in the next section. The loaddeflection curve shown in Figure .20 applies direction only. The more general case, in Figure "?, characteristics for load in the opposite direction. It for load in one includes also the is customary to
associate positive loads and deflections with tensile lap shear, as depicted in Figure 21. The elastic hehavior for both tensile and compressive behavior so the initial elastic lines have the same slope. However, since the stress trajectories in each case are so different, as shown in Figure 23, one must provide for different proportional limits and The mathematical model shown in Figure 22 is nonlinear behaviors. appropriate for the fastening of both fibrous composite and metal members, with the use of suitable (different) coefficients to characterize the is usually identical, lobddeflection curves.
41
0 P = U
'
UDi
PUTP PV=POSITIVE
KV NEGATIDLT = DISPMLENT
M ELOIW
IUTN(GEL)
Wu.rSTWN
OFASTENER FL1EPY
Oi.JTi
AIK
RECTrl (KYGEL)
DLM
BNOE TINIALIiA
S~~~MIGHTf
All ORl O.T411 IMTIVE. HUMqDR, ULQP N luTIV WME DLTOIW. AS ASGREAT LEAST BE ALG5E3ICALIY AT DLTOY MWUST
MIGH MAEMWYNGTIE0TLTN
USlY
HWVR
*1
,
1 II '
;I
tI
, i
/,,I
:
*
. ,,
I t]
FIGURE 23. STRESS TRAJECTORIES AROUND BOLTS FOR TENSILE AND COMPRESSIVE LAP SHEAR
42
is as follows:
0
Kel (6  6+)
for 60
for 60+
6
<* 60+1
(54)
(55)
  6el+'
for 6 = for
6
e+
el+ _ 6 56 p+I
6
for 6 =
pl+,
66
Eel(0_ 6) P =  Pe (ve)
Pel e:K
pl (el(
)fo 6)
.6el, 6
(61)
ueIt)
The negative relations cover thermal the possibility mismatch
for 6
(62)
are needed both for compressive external loads and to of internal dissimilar (selfequilibrating) members fastened loads induced at by one together
between
temperature and operated at a different temperature. the iterative analysis, prior to convergence. described below,
Load transfer due to friction associated with tightly clamped assemblies can easily be accommodated in such a model. However, particularly in the case of fibrous composites with a resin matrix, the reliance upon frictional The load
structures is
countersunk clampup
fasteners
to achieve a in the
there
the modifications
Iq
to (62)
would be for 0 _6 6+
(63)
and for 60+ _ 6 < 6el+ Likewise, Equation for6 0 (63) S (64)
would he complemented by
(65)
6
(66)
The other equations would remain unchanged. The preceding discussion concerns the loadtransfer characteristics of either a singlefastener joint or of one fastener isolated out of a multirow joint. The next section of this paper adresses the subjet or calculating just how the load paths. is shared between multiple it is fasteners when there are redundant load
After that,
station
by station, to identify just where any failure would occur, and at what load. 3.4 LOAD SHARING BETWEEN MULTIROW FASTENERS A variety of analysis techniques is this task because of a restriction here is chosen because it is adhesivebonded analysis, Section 4 of this report. The Reference available for the solution of the problem Most are rendered unsuitable for The method adopted equivalent nonlinear in later integrated, with the is to linear behavior.
1, with which it
known
boundary
are not it
at
both
ends to
joint,
while
the in
unknown
conditions
confined
Therefore,
is
more critically
44
I
of the joint) induced by a specified load. along the joint, one can interpret By then calculating progressively requirements, all boundary satisfying both equilibrium and compatibility the consequent predictions
dt
The initial
conditions are satisfied at both ends of the joint. iterative technique is initial assumption is the progressive
Actually, unless the explained in References 1 and 2. very close, or has been refined by prior iterations,
calculations will diverge and the appropriate change to the previous estimate is recognized in terms of the nature of the divergence. the typ4cal equations for each step of the joint is
of
in Figure 24.
for member 1
"TX(k+l)
T(k) P(k)
+ P1X(k)
(67)
(k21
T(k3)
P2I/I
p_ _J ,
,I
STATION NO.
k2
h2l
k FASTENER ON
LSTATI
T Zilk+l)
Pkl
T20lkl
44Pik)
*
REFERENCE INITIAt STATION Wk
B. FREEBODY DIAGRAMS
DEIrERENCE NATION (kW1)
POSITION
II'
i
*
(2(k'l1k
FIGURE 24.
load,
The remaining term pi9 is usually absent from test coupons shear, as shown. but represents the running shear load that may be present in such real structures as the splice at thV root of a wing under bending loads. The sign convention for pl. and when it acts the joint area. Tz(k+l)
p2
likewise,
taken to be positive
T2(k) + P(k)
P2L(k)"
(68)
In a doubleshear joint,
Figure 24 illustrates
a singleshear application.
the two portions which would make up members 1 or 2 would be combined and the fastener load transfer P would be changed from single to double shear values. Similarly, The form of Equations (67) and (68) would not be altered. in what follows, the subscripts summation of properties as well as of loads. To ensure that the analysis (and complies possibly It the I and 2 refer to any appropriate
use the mechanicil of uniqueness sometimes what is to use any ideal could
Because
characterization, terms
fasteners,
Effectively,
the stiffness
adJacent pair of fastener stations, making due allowance for variations in width (as with lollipopping to reduce the load picked up by an end fastener) as well as in thickness. Figure 25 shows typical sculptured skin splices for the fuselages and of it pressurized is to transport the is aircraft. fatigue Because than such tn skins taper are the relatively thin, splices more effective improve It to rout the edges of uniformly strength thick
doublers
thicknesses of sich members. cyclic variztion in are turn, in might be initiated clistomarily It
in the sense that the ioint geometry along the splice permits any cracks V:hich  om the hples to grow slowly into visible areas. That, repairs. tapered the Thicker members, members, capacity as to shown account such as wing skins, in Figure 26. The
analysis
techniques.
for both of these appropriate to also include in the analysis provision for 46
has
thermally
Induced
strains
and
for
the
running
shear
load
discussed above.
SKINN
DOUBLER
II
p'
TAC
U ATC
~J~hOARD STRINGEWi
composite materials, but, limit. can be for ductile behavior nonlinear The
aluminum of
some
loading segment
beyond
material can
stiffness by
to any level
of accuracy desired but any improvement the publicised (bilinear) model the is
in accuracy model
achieved
RambergOsgood
(Reference 9)
fibrous
composites
represented
fiber patterns,
some 00 fibers
load direction,
can be treated as linearly elastic to failure. The extension of the members between stations k and k+l is 6'(kI)  61 (k) = o,6T.(k) and
6 2(k+l)

given by (69)
': 2 .(k)
(k)
(k) + stepped
(k) members, for the the strains strain local (54) derive can be considered to
(70)
be the
stations, is the
except
around
fasteners, and
accounted
for in
Equations
In Equations temperature
thermal lyinduced
(71)
c, and C2 are deduced on the basis of as depicted in Figure 27.
length of
in calculating
the stretching
is taken to be that which would he associated with the average member load in each segment. Since the fastener loads are treated as step discontinuities at each station, it follows from Eq.ations (67) and (68) that the average loads causing the stretching of each member between stations k and k+l are
""( (k)
(72) 48
and
TZ(k)
= T2(k) + '(k)

P2(k)/
(73)
T1(k) / [LW(k)t1(k)]
(74)
and
0
2(k) =
(k)
/[W2
For linearly
elastic materials,
(k) /
(k)(76)
E(k)" E2
GIII
/LRCYAN (0.7E)
I
(77)
utt
F47
F.i..E
1+
STRESS, 0
PROPORTHM&A
LIII?
E=l
r~AORCAN (E)
49
J.4~v
**Mi
In
terms
of
the
RambergOsgood
material
characterization
for
ductile
materials,
C.= +
I Ei
(E F o . n  )i
(78)
E iF.7
Care must be exercised in using the formulations in Equations (78) because some of the published mathematical models do not agree precisely with the The discrepancies are usually small material properties given in MILHDBK5. is necessary to input into the solution strength and strain failure properties at each station which are precisely consistent with Equation (78). but it Having determined the member strains, per Equations the (76) to at (78) the as next
appropriate, the relative displacement between station, follows for Equations (69) and (70) as
members,
6(k+l)
as shown in Figure 24. This, in turn, permits a new increment of fastener load transfer to be evaluated on the basic of Equations (54) to (62). Actually, for the example shown in Figure 24, there would be no transfer of
load at station (k+l). Therefore, between stations (k*l) and (k+2), Equations (67) and (68) would be modified to read
(0 (80)
(81)
Equations (69) through (79) would be applied in precisely the same manner for the next increment of the joint. In analyzing such a multirow joint, one can specify the load(s) in members 1 and 2 at the start (left end) of the joint and seek such a value of the first fastener load that the boundary conditions are satisfied at the far (right) end of the joint. Alternatively, in seeking the elastic and ultimate joint strengths, one can specify the displacement differential 50 between the members
However,
at the first row of fasteners and iterate on the assumed applied load until satisfaction of the boundary conditions at the other end of the joint indicates the convergence has been attained. The application situations requiring relations of the methods described above, for tensile is loading, to
straightforward,
only some sign changes and modifications to the RambergOsgood (78). The same method can be adapted also to inplane shear
loading, just as was done in Reference 2 for bonded steppedlap joints, by replacing the various Young's moduli E by the shear moduli G = E/2(1+v). While the analysis can thus be made to compute the load sharing for a specified applied load, the lack of suitable test data make it difficult to predict the failure strength for inplane shear loading. 3.5 FAILURE CRITERIA AT FASTENER HOLES 'Having computed the bearing loads and bypass loads at each station throughout the joint, it is still necessary to assess whether or not that combination is capable of causing the joint to fail. In the case of conventional metal members, this customarily is a simple problem since only interaction composites, interaction
illustrated
ductile minimal
is
considered
between
the
With fibrous
however, that is not so. The use of a linear or kinked for composites is explained in References 10 and 11, and
in Figure
23
narrow
strips
is
the
loading, The linear interaction for I.t.ns.le consequence of a tensile (throughthehole) failure for
regardless of the ratio of bearing to bypass load. or greater bolt spacing, is seen to permit
distinct bearing
failure modes, so a twostraightline interaction is necessary. of just what constitutes the optimum widthtodiameter ratio, in the sense of maximizing the joint strength, cases it is discussed fully in Reference 11. However,
is found that having the bolts just a little too close together to sometimes a weaker
joint is preferred, on the basis that bearing failures are more forgiving. That aspect of joint design in composites is beyond the scope of this paper. For the present it suffices to say that any analysis of multirow structural joinits in fibrous ctinposities should cover both possibilities. 51
t,
tBEAR~ING CUTOFF
~TOTAL LWAD
TTLLA
0 IGLOAD
SEARING LOAD
S,
YPASS LOAD
TENSIONTHROUGHTHEHOLE FAILURES
'BYPASS LOAD FIGURE 28. BEARING/BYPASS LOAD INTERACTION FOR LOADED BOLTS IN ADVANCED COMPOSITES Figure 28 has been prepared for the imost usua! test case, However, the general case must include compression as well, tensile loading. as in Figure 29.
THEHOLE FAILURES
[mm!P
52
NOTE: ONLY TWO BASIC CONDITIONS POSSIBLE. OTHERS ARE MIRROR iMAGES OF THESE TVIO.
It is important to note that, as shown in the bottom of Figure 29, all possible seemingly different combinations of bearing and bypass load can in fact be reduced to one of only two cases, by identifying as the bypass load the numerically smaller load on one side of each fastener station. For fibrous composite materials, these bearingbypass interactions are strongly dependent upon the fiber pattern as well as the material constituents and, in the case of tensionbearing well. It is evident, interactions, are dependent on the geometry as however, tlat each of the interactions can be defined
BEARING
LOAD .F F C NOTE HOW TENSILE AND COMPRESSIVE BEARN STENGTHS
BEARING
LOAD D
OOSE FIT
MAY DIFFER
FITI
\
BYPASS LOAD
BYPASS LOAD
FIGURE 30. IDENTIFICATION CODE FOR BEARINGBYPASS LOAD INTERACTIONS The physical explanation of the various line segments in Figure 30 is as pius
Along AC, the criterion is that the total load (bearing follows. bypass) is sufficient to fall the net section while, along BC, there would be The a local hearing failure, which would transfer load to other fasteners. line EF represents a grosssection compressive failure through the filled fastener hole while, along E'F', the inability to transmit direct bearing through a loose fastener reduces the compressive load capacity because of the Along FD, the combination of bearing and bypass reduced effective section. The heights of the points B and D above the origin 0 are strongly dependent on whether the load is stress is sufficient to induce a bearing failure.
53
through
simple
shear
pin, be
a countersunk in the
Likewise,
one difference
apparent
Figure 30
for fibrous
outer
envelopes
in
Figure data.
29
are
only
mathematical
curiosities which,
nevertheless,
must be input
included
be able to recognize improperly shown load is the difference must be some explained there
Since the tensile bypass load load and the bearing materials.
between the total net section interaction for even The physical
the most
ductile
This is
in Figure 31.
in Figure 31 are that the net section strength is and is not reduced by any applied bearing constant and is stresses, total bearing allowable is bypass (or carrythrough)
load while,
for compression,
Ti]PTI:I
o.,,o
"0 0

"
FAILURE ENVELOPE
CONSTANT TOTAL LOAD
COMPRESSION
TENSION
P~q
bq
 Frb
_.F_'
P wivl (fU. klhI
"""
+, b ''' y"
SF~bypni
or lgeas
FIGURE 31.
54
incorrect If possible, any violation of this situation. modified and noted. If not, the solution is aborted.
input
data
is
Obviously, fasteners
the likelihood is
remote because of the lack of any bearing stress there, sculpturing criteria to soften be the load on the end fastener, each location.
computer coding should cover that possibility for such a case as improperly proportioned the failure along the fibrous The it is lines of Figure 25. composites, are Since the joint strength varies from station to station, must determined for For in References and 13. It 10 and 11 can be used. should be noted that
customary to express the failure criteria differently  gross section strains for fibrous composites and net section stresses for ductile metals.
3.6 EXPERIMENTAL DETERMINATION OF INPUT DATA FOR COMPUTER ANALYSIS The analysis above has been coded as the Fortran A4EJ which instructions the amount is described are provided. of data needed in Volume II of this IV digital computer program report where complete user for
In comparison with the earlier program A4EI there is the a considerable for to define problem bolted
are specific to individual fasteners in particular members and must be repeated for different thicknesses, diameters, fiber patterns, and so on. Further, many of the past tests on this subject, less ambitious it recording as in References 10 and 11,
of the data It is
hoped that,
will be possible to plan a new test program fibrous composite structures. situation and structures. for the present, data 14
for bolted and riveted joints in Despite this seemingly body of for
is
a the an
empirical
available of
attachment formula
Reference
loaddetlection
charateristics
individual
55
.5
5t
E
(82)
where E is Young's modulus of the members being joined, diameter, and t, and t 2 are the member thicknesses. Established values of the coefficients A and B are
A = 5.0
, ,
d is
the fastener
B = 0.8
(83) (84)
A = 5.0/3
o.86 B =
5.0/1.6
0.82
(85)
would seem to
be appropriate to use the lower modulus as the primary value of E and to ddjust the thickness of tj or t Unti 2 to compensate tor the higher modulus. there is a significant data base established to endorse or refute the use of Equation (82) for fibrous composites, it should be '.%ed with caution. But, in the absence of other verified information, there would seem to be little alternative to the use of that or similar formulae. Now, having established the slope of the linear portion of the loaddeflection characteristic in Figure 22, it remains to establish the proportional limit and the nonlinear behavior. The end of the elastic curve can be estimated as the yield bearing strength of the weaker member at each fastener station. This strength Pe will obviously vary along the length of a stepped joint as the member thicknesses progress from the unbalanced ends of the joint to the balanced middle. The ultimate strengths p can be established the same way. are not as well defined. Obviously, there is much testing remaining to be done to make full use of this analysis, but the considerably increased definition available in the solutions demands a conunensurate increase in the input data. However, the slopes K of the nonlinear portions
56
3.7 SAMPLE SOLUTIONS Three examples illustrate the capabilities of this new analysis program. concerns a design for fibrous of composite structure splice while another is is afterthefact analysis a thick metal and the third One an
a design
study in thin metal splices. GraphiteEpoxy Wing Spar Cap Splice Figure program, large 32 summarizes the results which of an be analysis, built from by the A4EJ computer
of a design for a spar cap splice at the side of a fuselage for a aircraft might advanced composites. much more the load intensity is
transport
Because of the high aspect ratio of the wing, severe than on any existing F18 wing skin root splice length bolts joint. shown fittings, application, which is is
with the possible exception of the presumably much more heavily loaded F16, a,,4i, it or F18. wi't ,, thrae The total rows of inches
in Figure 32
is quite a massive
ITS GRAPHITEEPOXY FIBROUS COMPOSITE 50% 45. 12.5% W0 FIBER PATTERN: 374% Ow, STRIP WIDTH = 3D, 216 an (1,125 IN.)
T _Z1miOS
302 an (1.188 IN.) 3.1 alm (1 IN.)(125 4D PITCH
PREDICTED FAILURE LOAD=225.5 kN (50703 LB) PREDICTED GROSS SECTION FAILURE STRAIN= OOM PREDICTED NET SECTION FAILURE STRESS=349A MPS (50.7 kW) PREDICTED FAILURE MODE: TENSION THROUnIH FIRST FASTENER HOLE RATIO 1.5 PEAK BOLT LOAD (AT ENDS)= 22.7 kN (S093 I) MINIMUM BOLT LOAD (IN MIDOLE)= 14.8 kN (3317 LB)
FIGURE 32.
One very
significant
in Figure 32 is roughly
gross
strain is
level
for unloaded
holes of 0.0042.
expect any significant improvement in strength to be possible of even the most thorough redesign. The other significant
finding is that, because the members cannot possible be tapered down to a point, and the first fastener could not be located there anyway, the load distribution is not uniform. The end bolts pick up 1.5 times as much load at that ratio would have much higher still The siqnificant point to be recognized to design a linearly tapered had here those in the middle. is that it is not Actually, possible
multirow
mechanically fastened joint which approaches the uniformity in load transfer of a bonded scarf joint.
CLEARANCE RADIAL 10.0I0.NGHI /NET FIT
( Z I(si Pbr
73.111
A 70
O4b2 2 (PSI)
V.411
57,446
ULTIMATi
JOINT STRENGTHS
I]
A. 00,713LS
S 0. 7.S7 LB
C. 45.60A L.
amt 1 (PAU I)
0r
II ll2
(r )
48
FIGURE 33.
INFLUENCE OF HOLE CLEARANCE ON STRENGTH OF BOLTED JOINTS on the basic solution the end fasteners. because it in Figure 32 dur The 0.001 inch
around
clearance
actually
increases
the strength
softens
the end
I!58
..
I
0.010 inch around the end fasteners, on the, other hand, unloads them so much that more of the basic load is carried through to the next fastener station, It Is reduree and the laminate strength also. clear that, in tapered splices like this, it is important that the outermost be a Otherwise there will loose. not are in particular fasteners In metal structures, however, there would disproportionate loss of strength. where the thickness has been be enough yielding to redistribute the loads and failure. increase strength prior to
Metal Ving Skin Splice The splice shown in one of many splices actually tested in the development of the wing skin splice at the side of the fuselage for a large The fasteners near the tip of the splice plates are of transport aircraft. Figure 34 is
(IJON.I
iJ~li.)
0.751 .1N
ET.(
PLC&
483ZI
T8 IN
2 13
(Q 84 IN I
 E.,VtTSWNNLiMEAMANALYSIGAT4FIPLIEDOROG 11.6 1,.605 05.4 37.0 8.323 146. 42 7.64 193.5 C7iONSSSQF273,tMPS 39.3 4,947 2%.2 544 12.232 "1.9 276.1 (4KOM)
NOr BAL.ANCE NEIWM END FASTUNER LOAVS. DUE TO THM6WQ OF V!CE STmA7S
40.675 40.041 37 14S 2.o0, 21.2b4 12.379 SF.CTM4 STWMMS (OWS) 151.6 16.6 1.4.7 169.6 (uPa) 179.5 StJC. STRAP 21.97; 24.012 23.U8S (2.2 IS)PN4 24."Z PKS)26.027 NET SECTIO KS))lS, SPCE NP.(92.3 6109,66.5 I MI PASTURER: 360O.0 ThES.ESONIP4*TEPW IMANG $ESS .APEM ST
*
*
.EST FALUME IN SKIN THRIR, RGH.T DiW FASTEN1R. WWHS .iOTH krT TENSON AND iKRM STWSE ARE PRUICTED TO UV MMM
ANALYYSISPTEDIC1TATEFAL,. AT MM6 UECTI 11(
TROECOFP$W*)50.6K$u)
FIGURE 34. BOLTED FETAL JOINT The success of lesser diameter than the others to equalize the load sharing. this design, using older method, established at least ten years prior to this analyis, can be gaged from the predicated net section failure stress of 50.8
59
,i
1
ir
ksi,
in
comparison
with
the
design
target
of
52
ksi.
The
seemingly
disproportionate thickness of the splice platEs wheee the skins butt togethe*r is needed to try and equalize the load transfer at each end of the overlap. The first and last bolts shown do transfer close to the same load. indicates some of the quite extensive internal definitions of Figure 34 the load
transfer and stress states within the joint that this analysis can provide. Metal Fuselaq& Skin Splice
i'
The thick tapered splice shown in Figure 34 is thinner fuselage edges structure, on so it is more and unsuitable for use in the much usual to employ there, scalloped as shown or in lovipopped splice plates doublers Figure 25. This is true for bonding such structures also. Apart from the difficulty of handling thin tapered strips without damaging them, there is a further reason for preferring a wavy edge to the splices. The variation avony the ed..e ,mTkes earlier crack detection possible, minimizing both the likelihood crack of catastrophic failure and the size of if it repair needed. were concealed exposed for A skin under a could grow far longer before detection tapered doubler rather than
uniform
periodically
visual
Figure designs,
35
presents
comparison uniform
between with
fuselage number of
skin
splice for
with a
ba.ic
splice
rivets
reference. Since it has been shown in Reference 1i that the load on a fastener induces a hoop stress around the hole of the same order of magnitude as the average bearing stress, the fatigue life of the lollipopped designs shown by test. The skin is The it most reasons for increases
through
the static stength of the joint, because of the reduced net area loss, and it means that any cracks which may initiate have fur t her t o grow before they These advantages outweigh the could join up and rip the'skin apart. consequent increase of bearing stress on those end fasteners, with respect to
60
3/IDIN1.ADRIVETS.,
I~. 714.IC
=o
3163 IN.o '
Y24
112
224
213
118
103
215
1 6.3
1i 11.5
II 17.1
i3.3
SKIN1EARING STRESS is (KIl) SPLICE NEARING STRESS 12 (KSIl SKINSTRESS (KSl)I 5. SPLICE STRESS (KSIh 11.1
is 11 1.6 14.1
is 1 11.8 6.0
LI
1. ULTIMATE STRENGTHS ARE2121 LI/IN. FOR BASIC JOINT AND2411 WiIN.FOR LOLLCPOPPED SPLICE. CORRESPONDINGI WITH REMOTE SKINSTRESSES OF 48 AND64 KSI.REPECTIVELY
2. EACH SPLICE IHASTHE NAMENUMNBER OF RIVETS.
3. FURTHER OTIMIZATION OF REFINED SPLICE COULD ENHANCE FATIGUE LIFE9Y REDUCING ENDFASTENER LOAD. NUTULTIMATE STRENGTH WOULD NOT INIPROVED.
"E
FIGURE 35. COMPARISON BETWEEN BASIC AND REFINED FUSELAGE SKIN SPLICES
The softening by scalloping which, is shown in Figure 35, is also applied to slightly thicKer structure, where strirgers are spl 4 ced, for example. The computer program A4EJ now makes it possible to perform parametric studies to optimize the proportions of all such design details. It might even be possible now to design by analysis rather than test, for the first failure to occur in the splice plate, which is easier to replace than the skin. In thin sheetmetal structure, it is common to have multiple splice elements which do not all start at the same fastener location. There are therefore two or more rows of fasteners which are subjected to peak loads. Once the use of this program A4EJ has been mastered, it is possible to obtain a reasonable representation of the internal loads in such splices, as shown in Figure 25, by selectively lumping the skin and doubler together to cover one end of the joint, while the 6oubler and splice would be lumped together to characterize the other end. This takes two runs and obviously will not cope with all such situations. There is a need, therefore, to try and modify the existing coding to account for multiple splice elements some day. The limit to such work is anticipated to be in the computer run times. For example, if 61
it takes 100 iteration cycles to solve a problem with a single (or symmetric) splice, it would take 10,000 cycles if there were two splice plates of different widths. That must be weighed against the cost of inputting two models to approximate the problem with two runs on the present program. The author believes that there is a case for a threemember splice analysis, but not for any more than that, even though the recoding effort would be the same. Eventually designers must face up to standardized design concepts within the capability of then current analyses for any future project or incur the ever increasing expense in money and time of specific detailed finiteelement analysis or physical testing, for gains that are often imperceptible with respect to simpler concepts if dore well. Improvement Of Bolted Joint Designs To Enhance Fatigue Lives None of the examples discussed above would qualify as a poorly designed joint, so they provide no standards against which other joints can be assessed., It is appropriate, therefore, to include a comparison between a multirow bolted joint in which no design finesse at all was employed and one in which a reasonable degree of expertise had been applied. This comparison is evident in the different designs and analyses shown in Figures 36 and 37. These joints are idealized in the sense that there would have to he support structure to react the eccentricity in load path and that the bolt failure in Figure 36 could easily be avoided by a doublestrap splice design with the bolts in double shear. Nevertheless, there are several important conclusions to be drawn from these examples. While the uniformity of load transfer in Figure 37 is far superior to that
shown in Figure 36 (withing 4 percent for normal operating loads), the inefficiency in Figure 36 is nowhere near as severe as similar analyses in adhesively bonded structures would suggest. The forty percent inefficiency shown would become much higher for thinner more extensible members and for more rows of bolts, however, but would probably never approach the 10 to 1 ratio needed between the peak.and minimum adhesive shear stresses in durable bonded joints. The ultimate strengths of the two joints shown in Figures 36 and 37 differ by
62
3IN.
BOLTS
0IN.
TOTAL LOAD (LB/ROW) 10.000 20.000 30.000 40.000 03,042 (ULTIMATE) 2.337 4,675 7,032 9,533 10,490"
INDIVIDUAL FASTENER LOADS 1928 3656 5472 7162 7597 1669 3338 4992 6500 667 1828 3656 5472 7162 7597
. .
'SHEAR FAILURE (HIGHEST ASSOCIATED NET SECTION ALUMINUM STRESS  68.9 KS4I NOTE: COMPATIBILITY OF DEFORMATIONS AGGRAVATES NONUNIFORM LOAD DISTRIBUTION AS LOAD IS INCREASED, BECAUSE MEMBERS YIELD ONLY AT HIGHLY LOADED ENDS, NOT
EVERYWHERE.
.1.01
OF FASTENERS
) 2030
INDIVIDUAL FASTENER LOADS 1995 1951 1995 3989 5973 3902 5844 7722 8881 3989 5973 7582 9044
RATIO ,.,/P.m 2030 4060 6105 8257 1 .0 1.04 1.04 1.07 1.09
46,387 (ULTIMATE)
_
9709
ii
7832
j9044
_iilii 9709
ULTIMATE FAILURF IN NET SECTION OF ALUMINUM (AT AVERAGE STRESS OF 74.2 KSI) AT OUTEvhAOsT ROV . OF BOLTS. NO BOLT FAILURES.
FIGURE 37.
63
only
percent.
While
that
represents
significant
improvement,
the
inevitable criticality of the net section at the outermost fasteners makes it very difficult to make major improvements to the ultimate strength of joints by techniques which yield substantial benefits in the fatigue resistance of the joint. fasteners The fifteen percent reduction in bearing stress on the outermost in Figure 37 could easily double the number of load cycles neeeded
to fail the joint in Figure 36. Figure 37 joint in transfer. represents Figure 36. only a Yet it single came estimate of proportions within 4 percent of a to improve the perfect load (which
are in practice not fully obtainable because of the finite thickness at the outermost bolts) can be approached very closely with relatively far less effort. In other wuds, tne benefits fron; a little bit of design finesse are substantial and should always be sought. to be gained from precise optimization Conversely, the residual benefits are disproportionately small in
comparison with the effort expended to improve on the first refinements. In the example shown, it is not obvious enhance that reducing the diameter because of the
outermost
fasteners
would
the jesnt
stren~gth
the bolt
strength would decrease faster than the stiffness. IT a similar design for a doubleshear joint, however, the computer program A4EJ w3uld provide a rapid estimate of any benefits to he gained by such a modification. most unlikely that a joint with such an abrupt load transfer as shown in Figure 36 would be designed today. However, that situation could arise in service due to a fatigue crack grown through a continuous member like a wing spar cap, for example. Thus, tolerance the new analysis can be used as also to investigate the damage of bolted structures as well the load It is
transfer in the virgin structure. 3.8 CONCLUSIONS The nonlinear behavior and joint analysis represent increased internal definition of this new bolted beyond the prior state of
considerable
extensions
the art.
64
This
increase
in
analysis
capability
in
turn
demands
more
extensive
Some such information may be provided also loaded fasteners in specific stress
precise
data,
however,
the
program
can
be very effective
for
parametric studies using only estimated input values. Since this is elements, needed is a continuum mechanics analysis, rather than one based on finite short, The amount of input
significantly greater than for the earlier steppedlap bonded joint number of parameters needed
analysis proqram A4EI because of the increase to define the behavior at cach station.
The sample solutions oa\crr1ed :,t useful tool in analyzing and ietal structures.
a M
multirow boltNd
6S
SECTION 4 NONLINEAR ANALYSIS OF BONDED/BOLTED JOINTS 4.1 INTRODUCTION Separate nonlinear analyses for the load transfer in adhesivebonded joints
and multirow bolted joints are given in Sections 2 and 3 of this report. When adhesive bonding is used in conjunction with mechanical fastening, one cannot sum the individual of each load path differ. of the combined regarded bonding, thorough such a analysis load method transfer, joint strengths because the individual So also do the strains to failure. reported here for is to be able accounting the compatibility stiffnesses The function the
to characterize
of deformations. may be
as exposing once and for all the futility of combining bolting and to increase joint strength, by means of a nonlinear analysis so that there combination. can no longer be any doubts However, that would of not there is installing about the inefficiency of attitude a to be 4 a reasonable
adopt because,
in a broader context,
joint not to augment the shear transfer but to tie in to some other structure Also, of course, the introduction of load in test coupons or fitting. throuqh a bolt via bondedon reinforcing doublers is a widespread standard practice. By far the biggest joint justification analysis for the development is to be of such in a combined imperfect
bonded/bolted structures.
capability
found
of bonding and bolting offers unique advantages, and these provide the real The sample solutions presented in this payoff from this new capability. report have been selected specifically to explain such applications as well as to illustrate the capabilities of the analysis program. of these applications is the repair of what was originally purely adhesively bonded The first group intended to be
67
which
occurred was
during even
manufacture employed as
or a
from
damage
in The
a a of
combination
production
problem on those
particularly
understanding of failsafety due to the combination of rivets and bonding has been far less than adequate, so the opportunity is taken here to present a thorough discourse on this subject in the process of illustrating the capabilities of the analysis program. Briefly, for lightloaded structure,
adhesive bonding provides a failsafe load path to overcome the weakness caused by tearing along a line of fasteners. For heavilyloaded structure, on the other hand, mechanical fasteners can be very effective in preventing widespread unzipping triggered by what was initially quite localized load redistribution around a damaged or defective bond. There is
,
been covered fully in Sections 2 and 3 of this report. is can do and computer difference the reader in analysis is referred to the user manual for the full is that
capable of much but not all of what the separate programs A4EI and A4EJ codes (Volume II of this report) details.
capability
absence of adhesive
permitted the inclusion of nonlinear adherend behavior as well as nonlinear loaddeflection curves for the fasteners in the bolted joint program A4EJ, while remain for the bonded/bolted program A4EK, the adherend deformations of is must the not
adhesive bond. Since prior work has shown that gross yielding adherends triggers progressive failure of adhesive bonds, this considered a significant limitation. analysis program A4ET has permitted
The shorter length of the purely bonded the inclusion of an extra subroutine to
massage the input data and improve the computer run times. Thus, these three new program3 each have unique capabilities. However, the earlier steppedlap joint analysis programs A4EF, MEG and A4EH have no capabilities not found in the A4EI program. Yet their more limited capabilites are associated with a far shorter computer code so that they may be more desirable for small computers or for subroutines in bigger programs on larger computers.
68
4.2 REPAIR OF DEFECTIVE BONDED JOINTS BY MECHANICAL ATTACHt ENTS Section transfer joint. 2 of this report both contains the several to sample solutions of the and load three
through a These
substantial
titanium basic
adLesive join,
bonded
include fasteners,
nominally
to show how the addition can modify The shown in Figure 38.
solutions.
composite laminate is
and the overlap is 5.0 inches. It was shown in Section 2 of this report how, by progressive redesign, the joint strength could be increased significantly by optimizing the proportions. However, these refinements are not compatible with the use of bolts to provide an alternative load path with which to repair local disbonds due to misfit during manufacture or damage in service. So the original equallystepped geometry is used in these new examples, with only a shortened end step on the titanium.
S0.312S IN.
BOLTS
I '
1
I
11 17
19
"I
1 Ib14
7
B~I
0.077u , 0.2
ADHERE,NOS 2 / MATERIAL: HTSORAPHITE.EPOXV 11 45ESACtWT 00oE0 PLIES) C 10.0.R10SP'  to KS1 C. AOHERENS PROPERTIES
FIGURE 38.
69
joint and is
directly comparable with Figure 25 in Section 2 for a purely bonded joint. The addition of seven rows of fasteners to this joint is actually predicted to decrease by the joint comparing strenqth, Figure 39 from with 34,322 the lb/in. to 33,096 lb/in., here as in revealed equivalent solution
Fiqure 40 for precisely the same joint without any fasteners. this decrease is might given a change In failure mode and location. increase it in strength, clear shown that in the the assessment addition increase 39, the of the of
39 makes In
fasteners joint
unflawed
adhesivebonded
joint
does
not
significantly. in
Figure alone,
adhesive any
since it adhesive,
have transferred 28,380 lb/in., or 83 percent as much asthe adhesive alone. The reason why the combination a does not work argument well together is the gross these of dissimilarity in stiffness hetweeio the two load paths. three solutions seem to make strong Taken together,
NO RESIDUAL THERMAl. I1MESSESACCOUNTED FOR (ATG) 0 .i ADHESIVE NOT CRITICAL TENSILE LOADING 4 
LOAD IN AOIEREND I
1KIPS)
LOAD
ADHESIVE (KIII .: .{ j i I 1 3 4 5 4 7 $ 1 10 11 12 13 14 IS
,,STATION
"
IV..",'
LOAD INADHEMERD 2 FIOROUS COMPOSITEADHEREIIO 3 CRITICAI. AT FIRST EItI _._ . . _.... .r .,., ..
11 __
13D14 1
LOA (KIPSII.)
IOIELOADS
17
44
If STATION
II
TOTAL LOAD TRANSFER

33.@N Ll
3 .T71 LB A ISE to
IARELY I PERCENT (B SCALE.
OF TOTALI
(INCHES)
FIGURE 39. LOAD TRANSFER THROUGH BONDED/BOLTED STEPPEDLAP JOINT WITH NO FLAW 70
NO RESIDUAL THERMAL STRESSID ACCOUNTED FOR (,%T01 I f.. )i "0lle ADHES1IV NOT CRITICAL TENSILE LOADiND 4IL
LOAD IN ADIEREIfO I
2l .
TITANIUM CRITICAL AT
(KIPS) LOA
,ig,.'.,
ADHESIVE
''
I"
STRESS 44511
1 F 2 3 4 N 7 1 STATION I I411
SHiAFI
"I.
12 131415
4LOAD IN ADHEREND 2
_____________________________
FIBROUS COMPOSITE
7 1 10 0 It 11 131411 STATION
LOA5
NOTIRiTICA"
LOAOTEIANSFER 34,32l L1 IONI.0 1N.WIDTfI1 SEVEN J01N1.WITH Of SAME IRANSFER 0.3215 IN.FASTENERS AUOE9  33.096 LB
"LOAD
N IPATiON 7 B
I 10 11 02 131410
SLIGHT REDUCTION INSTRINGTHI DUE TOBOLT HOLE IN THICK FNDOF FIBROUS COMPOSITE
SCALE iNCHES)
FIGURE 40. LOAD TRANSFER THROUGH ADHESIVEBONDED STEPPEDLAP JOINT WITH NO FASTENERS
NORESIDUAL THERMAL STRESSES ACCOUNTED FOR (AT0) TEIKSILE LOADINO 4 3fTSLOAD LOAD IWADHEREND I TITANIUM NOTCRITICAL
777
2
11IKIEII
BOLT13
3 4
1MN
Ib7 1
I 10 i
12
13041
.1li
j
Ol
1 0 1 STATION
10II 1
1213 14 B
STATION
18) 11 12
13141B"
I.EIN WIDTH)
(NHSALE0
MS
FIGURE 41. LOAD TRANSFER THROUGH BOLTED JOINT WITHOUT ANY ADHESIVE 71
Such a pessimistic prediction has adhesive bonding and mechanical fasteners. been found to be true for several joint geometries investigated during the course of this work. A well manufactured, well designed bonded joint is very
difficult to improve upon, particularly when one considers the design refinements described in Figures 8 and 9 of Section 2 of this report to increase the bonded joint strength. additions of rivets or bolts. Before proceeding to the more fruitful bolting, Figure 41. middle. joints, below, The This where is the it end is appropriate bolts pick up to applications of combined comment less the the with at on a surprising than The do of bonding and feature in of the Such techniques as the special shortened
end step on the titantium plate are not quite compatible with the subsequent
much
load ends.
those
in marked load
contrast peaks
behavior
transfer
alleviation in Figure 41 is that the end steps are quite thin and, therefore, " those fasteners have a much lower effective stiffness than those nearer the If the members being bolted together were uniformly middle of the joint. thick, the end f steners would incur a voJyuPuIortflodte share of the load, just as for bonded joints.
or damaged bonds,
the
pessimistic picture above changes dramatically. This is shown in Figures 42, 43, and 44, which are directly equivalent to Figures 15, 16 and 17 of Section 2 of this report for flawed bonded joints without fasteners. The
strength increases due to such bolted repairs are seen to be substantial even though they remain much weaker than the baseline joint What is of special interest in Figures 42, 43 and 44 is in Figure 39 or 40. that the fasteners
still transfer so little of the load. Their effectiveness follows from relieving the stress and strain concentrations in the adhesive immediately adjacent to the flaw, other words, rather than in the load they transfer themselves. from the presence of the bolts is more effectively. are shown in the (While In the benefit
SFigures
42,
Figure 38 here and Figure 5 of Section 2 is slightly different hecause of the shortened and titanium step here, the effects are insignificant, as can be
t72
Ik
I
NOIC RESIDUAL THERMAL STRESSES CCOUNTED FOR (ATR)
a
0a
TENSILE LOADING
3 n LOAD
[
I 1Ki)2 1 4 S 6
DSA
7 7
STATION
1511
12t13141k
.LOADINEAOHEREN
E2
FIBROUS COMPOSITE AOHfREIIG CRITICAL
"
.. :"
2 1213
3 4 1IM1
E1 64
11 12 131411 63 I 12
LOAD qKIP$) 2
OLT LOAD
TRANSFER
a 6
9 A
I10
11
141 13 1 z 16 34
STATION
LDMPRABLE STRENGTH WITH ADHESIVE ALONE 1.11 LI STRENGTH OF UNFLAWIED JOIN'  33,869 LE
FIGURE 42. LOAD TRANSFER THROUGH FLAWED BONDED JOINT REINFORCED BY BOLTS
NO RESIDUAL THERMAL STRESSES ACCOUNTED FOR IAT0)
"B
20
AfIHESIVE
LOAD
STATION
... f.. 1 2 3i
3TAT i
10 11 1tttS
'1
LOAD
T
BOLTLOADS ILEI
I,)3
1 16 20 '
12'62 I I'I I
B 7 S I STATION
10 II
12 13 1406
SCALE [INCHES)
FIGURE 43. LOAD TRANSFER THROUGH FLAWED BONDED JOINT REINFORCED BY BOLTS 73
4f
F ..
"I "'. ".. 11 4 2
I
V1.0153
1 MSP
44.45$I
2t~~~OAD

tL
4S 6 0 313 4 4. 1642 14 I
2 IgLO IIIIAMP4EREU
IT3TA30
A.. 2 OA$ I 1 1 0 7 PTATIOS 61
L 111.1,1. M
4K
4OAI4SALI3AOT113'tfl
S23,t1% 1 2 3
is 11
1,)14 1i,
S'1M114AE *OT.k
TIO
C(m&Alaa
SRMOTH.TN A
VALSS
14,P93 L1
,4)31
FIGURE 44, LOAD TRANSFER THROUGH FLAWED BONDED JOINT REINFORCED BY BOLTS seen by comparing the pure.ly bonded joint wi~alyses here and ?4,'075 lb. in Niguc 6 of Section 2.) W'iMle it is is true that i'e predicted qr:ater
load

strangth
for, the
fasteners
alone,
in in
repaired
bonded/bolted joints
condition,
The bonded/bolted repaired joints would last much longer same Joint with no adhesive.
weak link with fibrous
reoaired ,jointI would default to the greater ultimate strength with fasteners alone composite
delamination of the adhe.rends well trigger quite a different fin.il failure. adlesive  it
so the initial
failure
This use of tha combination ofbonding and bolting for repairs is thus seen to expose very real benefits which the assessment of the nominally perfect '. hond in Figure 39; could not revEal.
~74
/TITANIUM
ADHIIV1
LAYI!
/r
STEE
FI,
If HALF TIIARIUM PLATE CAW3OTSE SUtOD SIDEWAYS. STEPS IS PLATE AqO SREAWILL INTERFERE. CAflING LARGE.WIDESPREAD VOIDS IN 5010
~~..
COWCrITE FISRtO&S
FIGURE 45.
The likeClihoiod of needing such repairs due to manufacturing difficulties be minimized hv the selection of the design and fabrication concept. explained in Figure 45, conjunction ensures an with excellent in which the central and bonding any of of fit without titanium the the fibrous cocure composite
can
This is virtually to be
difficulties
encountered with the bonding together of precured details. the prepreq between split external iiJ... quality bond
uuF I i ....
because
....
there
is
no escape
the
cur , t..
if
4.3 REPAIR OF DAMAGED STRUCTURE BY BONDING AND BOLTING The new analysis metnods can be applied also to the repair of general damage to structure, part icuarly for fibrous composite structures, regairdless of whether such damage occured at tne site of an existing joint or not. any such patch can are redundant between tht he recarded as two joints backtoback, However, sense in tbintact that the and the same damage, is member and analysis methods apply. load paths surrou(irin for the repair of local Since theory there shared itself.
structure 75
Actually,
than be
if the repair is too stiff, it could even attract more load locally
structure encountered. the most basic This Is not the place for a full be explained to make the
discourse on the topic of the details of bonded repairs and the pitfalls to principles must present discussion effective.
A, M
ES MIMIAL 3
AGE
___
_TAMEDi
NIcTItN o U PATCH
A HA.V E THICKENIE AtN MOLE IN SRO A ER FO f LIEEVE CRITICAL I& &,ADIlhaivi 0O01 CO*OITI 0 COINSNITIE
REPAIR
ArNE$1VE90NHDED REPAIR OF DAM4AGE TO FIBROUS COMPOSITE STRUCTURES proc in in the redair of local FIRo area is first trimmed to a to
struct~ure
is
shown
transmit so*,w (or all) of the load which had been Interrupted by the damaged area. There are two potential sites at which the bord may become critical. One is at the outer pnrimeter of the patch and the otner is at the edge &f the hole. overlooked. The fact tPat care must be taken at both locations is often here have been premature failures, with excellently detailed patch, which were due to the abrupt thickness
I
! _
76
NM
discontinuity at the edge of the hole. been wh.n complicated scaled and up. by the While fact that the hond The understanding of this problem has seems had to have been up, in the the very same
what area in a
repair concept which had been tested satisfactorily in small scale had failed been scaled change adhesive thickness had not been and that omission changed the relative strength of the adhesive Therefore, adherends, resulting expect pronounced the stress failure mode. at both to develop one should to relieve concentrations
the maximum strength from such a bonded repair. Having attended to such design details for bonded repairs, to analyze the next step is severely in strip loaded can be line with the
Iithe
is a load
repair
the of
most that
location principal
of the hole,
performed, however, it is necessary to establish how much of the load remains in the skins, to be diverted around the hole, and how much passes through the patch, over the hole. establish Strictly, oe would need a finiteelement analysis to but the method described in Firare 47 provides a
that prpcisely.
.
E'. tI
STRIP
DiU
___,
Ep'
'_
STIFFFdeS OF PATCH FROM A TO a LOAD IS SHAFRED ROUGHLY IN RATIO OF RL LATIVE STIrFNESSES. S0 LOAD PASSING THR.UGH UNIT STRIP OF BOND SHOWN IS
*" 3 E Pt Of TOTAL
FIIUJRE 47.
* I
77
reasonably pocket
with nc to be
more
than
calculator.
strip
analyzed
can
then be established by load sharing in proportion to the relative stiffnesses of the load paths, A simple but is as shown in part B of Figure 47, powerful in an rule isotropic to remember plate under in this kind of
assessment
uniaxial
loading
increases in diameter or radius precisely three times as much as the same length in an area remote from the hole, as shown in part A of Figure 47. Thus, third the stiffness of the load path diverted around an open hole is as great as for the original structure. This softening only one is quite
heneficial, around an
of the factor
relation
between
load
the
stress
established,
A sample solution is shown in Figure 48 to bonded or bonded/bolted repairs. illustrate the capabilities of the A4EK program in this context. On theirj own, the bolts of would lb. be predicted inch it tr, be instad should capable of be 9535 noted of an even per the greater inch for is load the an
transfer,
9581 Again,
per
lb. that
combination.
however,
9581 lb.
K"
ultimate value,
while the 9535 lb. strength would be achieved with no such damage.
the repair shown in Figure 48 would often be the most practical the patch That and clamping the parts together, particularly of heat if to would leave only the need The reliable inservice is for application of as fibrous it seems
adhesive.
repair
however. is that
which
recognized
of moisture strengths of
bond
or cocured
unless
removed
drying
impracticality
particularly
for larne thick composite structures, means that highstrength bonded repairs cannot always be relied on. Therefore, the repair of large composite structures may he forced to depend on mechanical fastening and an
78
No
9.1~
11 ~~~
WSp
i
n .. ams
too.
rr
Il.
/.
PATCH
i0
'r Da l4
PATCHGCOWTAY
r' 0 l
(THINNED) AT SOFTEPED (I[F PATCH LOU TOTALLOADTRANSFER ADHESIVE LOAI? I1ANSF LOADTIIANFER
ATAILIPJT OF EDGE
PATCH ..
NOLL DEO
STIFFNESS OF LOAD PATH ACROSS HRU6 . HOLE ,. .., A."ix1 TI ROUGH PATCM STIFIFHlItZ OF LOAD PATH Q IN WJ AOIJUONOLL 'CEskip Mn  3. Igo ACCEPMS $I11.3 OF LOAD 20 PATCH
"
"A"HEiff
STRESS IKSfl
LT
(I PNCE31'T OF TOTALI
4 I
'I
Sa~%LT LOADI
IOLT LOAM WNITHOUT
ADHEIVE
3
1146
1'aO~IN7
r
~ ~
14"
1184
ZS%'4
Mal)
FIGURE 48. LOAD TRANS'ER THROUGH HON!I EU/5'OLTED FIBROUS .1.. adhesive bond of o,0y about half its normal strength. (In
ru
that case,
the
In the case predicted strength ahove would be reduced to (777 lb. per inch.) of thin composite laminates, on the other, hand, purely bonded repairs are quite practical, because the bond starts oiut much ttrcnger than thin adherends, th6 dryine out prior to bonding is both quicker and easier.
4.4 COMBINATION OI BONDIVG AND BOLTING IN FAILSAFE SiRUCTURES a quite wies.pread misconception that mechanical ftsteners can be In considered to always provide failsafety for adhesivelybonded joints. actual fact, they can do so only in relatively fe.4 instarnces in welldesigned bonded structure and not nmany more in poorlydesigned structure. The key to There is the differences In behavior lies in the relative strength of the adhesive and Briefly, whenever Th;s is exolained fully in Reference 16. the adherends. the bond is stronger than the miembers being joined, no failsafe load path is
79
.t tU.*Lf.* U b & a. a*,*.
would
be
located)
before
it
could
propagate. because,
In
such
case,
rivets adjacent
to to
he superfluous
even in
the area
In the case of thick members bonded together with such a simple joint concept that the adhesive is the weak link, and catastrophic hord would structure. a local disbond could cause a widespread fasteners through such a represent be used the an improved such being would without members geometry unzipping of the bond. Mechanical
provide a degree of failsafety and could Certainly, no adhesive bond should ever whenever some such the bon, would use be of weaker a
reinforcement joined. In
than
cases,
the
complex
joint
provide the better solution, by increasing the joint strength and removing the weaklink fuse from the structure. That is not always possible, however,
particularly section, so
structures,
as
is
explained the
in
structures
for which
strongest
d damaqetolerant structure is obtained by bonding and bolting to gther. It should be understood that, even so, the ultimate strength of the intact structure remain is effectively unloaded defined until by the some adhesive alone, and the to fasteners a load virtually
damage
has occurred
cause
redistribution,
WWhereas
rivets
as failsafe
can be appreciated for Figure 50 by starting with the riveted, bandfree case The fasteners cause a weakness being joined, such skin like a seam of perforations along the joint. And the nature of strucoture is that the more fasteners are broken for the longer is the rrack), the more severe is the load in the adjacent skin and
fasteners.
These conditio.ns permit a catastroDhic tearing whenever the damaged area is la'ge enough or the load sufficiently high. Those unsatisfactory characteristics can he avoided by including adhesive bonding as weil, both as a stronger alterrmative load path and as a means of reducing both the skin stress and the bearing stress along the fastener then, the that rivets! the adhesive bond is Ir actual fact, the rivets or seam. It are is then fair to say, Icad path for su .erfluous then acting as the failsafe bolts
Sto
00__ _/0
0
iFASTENURS
........
0 0 0 001_1__
AHESIV
ooo
.. ,,V
LOAD DUETO
AROUND FLAW
"REDISTRIBUTION
SHEAR STR
1 1
LOADTRANSFERHED THROUGH
FASEiNERS IN AREA OF D( ECTIVELOND 
I i
N0 LOADON FASTENERS WHERE B9O SINTACT
"
FtrR
49. INDEPENDENT ACTIv," oF Z.,i ll U. ~ ul ;DUE  r%.Ii .Vv ~ LOAD ~ TO Lis BOND FLAW REDISTRIBUTION IN
...
PASTEINEI ALGA OS
.L I'N'AlT S "
0 c; 0 0
Iilli0
A.
I.
AFTERINITIAL DAMAGE STRESSR BOFAILUE WEAr.UE FUSE .DUCEO .kEDW STRONGER 303%1T PANTLS THAN MOVEtGUT OFSPLICE INTO PANEL
V
ADEOSIV ODOUTSIDE
f
V~
001 0
K0?o
aI
INITIAL DAA/
ADHESIVE
5k
THANAROMERENDS 5050 It STAONGER IN WHICH AGTE IF SPLICE A.IA RAIS'CRCEEl, NY ADOESIVELY SOOEI[D DUUqLLR$.WEIIK LINK1$ tOVED OUTOF . IN 0 F"sft, AS SPLICE INTOBAXLL
FIGURE 50. USE OF ADHESIVE BONDS TO PROVIDE FAILSAFETY FOR RIVETED JOINTS
IN THlIN STRUCIURRE 81
ll
whenever the bond is stronger than the adhesives and should be eliminated to save weight and cost and to remove possible sites for damage initiation. If socalled the chicken rivets are of the added to a design only because of doubts preparation or quality control for
about
reliability
surface
that cost would be better spent on improving tne bonding techniques There are sufficient good service records of honded to remove the doubts. structures to have full confidence In bonding alone, proved that the design. bonding, and manufacture are done properly. Advanced composite structures have done much to expand the use of adhesive bonding in highlyloaded structure because they are so brittle that the structural efficiency of purely bolted or riveted composite structures is often unacceptably low. That is one reason
why so much such structure is purely bonded. * 4.5 ANALYSIS OF FAIL.SAFETY OF BONDED/BOLTED STRUCTURES an explanat...on of why the cominatior,. of bonding and bolting can be the most appropriate for thicker structures, particularly The purpose of this section is to when made from fibrous composites. illustrate the use of the new analysis program A4EK to characterize such a Thi section above contains situation. Figure 51 shows the area of concern representative of a bonded/ool"ted spar Even under cap and skin combination in fibrous composite construction. normal pounds. operating Now, if loads, either the total the load would be some tens of thousands of skin or the spar were to be broken, but not the other, a resin interface adjacent to the bond would be overloaded and tend to The only restraint would be the fasteners whch have much more shear unzip. The adhesive itself strength than the resin or possibly even the adhesive. is not the weak link because it is so much tougher properties ind stronger than the in in Therefore, the resin resin matrix. place of those for the adhesive. are used in the analysis The analyses of this problem shown
'
Figures 52 and 53 show that the structure would indeed start to unzip and that the fasteners are capable of arresting such delamlr, ations before they spread very far. This behavior is directly analogous: with that repnrted in Reference
tI
17
for
purely
bonded
stiffened
metal
structures.
In
fict,
82
BONDIG STRUCTURE I. GROSS DELAMINATION OF THICK ARESTRONIIR THANADHESIVE WHEN ADHERENOS FASTENERS,,.._ /DELAMINATION ARREITED AS
!]
FAFTEN(SF LOAD
TRANSFER
C. SELFARREST OF DELAMINATION WHEN FASTENERS AREUSED IN COMSINATION WITHADHESIVE FOR HIGHLY LOAOED SYRUCIURE RENODELAUEMNATIONS. THE &ELOND MEMBER WOULD NOTE: IF ADHESIVE WERE STRONGER THAN ADHERENOE.THERE WOULD
IV
i
SIMPLY FAILAT THESAMELOCATION ASTTHE FIRST. IN THATSITUATION. FASTENERS WOULD NEVERFEELANYLOAD AND WOULD DESUPERFLUOUS
C Cn,rl
I J I IIL~r
C
III
,iL.v Onln,,,
IIIA, U/IIL LJVb U Jl
I I"
rUU
r eIrlL.
$11RME T5
SIII..
I ZERO
t/
'IT
!AtURE AT1
Ii_,I iLI+
IB
l
3 1 l_
i
. ! l_
"
TSIN
I.I
j
!N
.104 IN.TP
.;_
'
.
@I
"40"
SSTRUCTURAL
SAADJACENT
GEOMETRY
FRCUE
3'
II
I 0 1
II
11
1'
I1:1
1:
' '1
10
A01o15,,..,,I.,,,, II
"A,
"
PARTMELDAUD ILBI
A 40Il ANIJIATION, tAILOB! AT 12.0% I L'
112111 11"
III ST
H
I "i
SI
ia
SIIi
l
0
:
IB 1 1 i
, a
I:1 I
low,
I1
I
7 i1
1, 4., i
4121
Ii
2181,
11
~
lFAVliLN
I
ai2 is
I:
tlus
113
~ji
'1. 04
BY ADHERINDSTRNOMH
FIGURE 52.
".
83
S[
I).!2LIMITED
I.
IELFARSITOF
S;
SIAKFASTINER
INITIALDELAMINATION.
0.04
42 ADHESIVE SHEAR STRAIN CONTINUES TO0IIdNU1H. SHOWING R1 TENOENCY FUR 2OLAMINATION TOGROW INITIALARREST
II1.42
2AFTAR
Is
FIGURE 53.
stiffnesses calculated by the A4EK program for various joint geometries and
assumed or known disbonds could be fed into the pocketcalculator solutions in Reference 17 to expand the capabilities of each individual analysis. Several problem significant observations analyzed in Figures in thick, result structure can be drawn 53 is a from Figures 52 and 53. good example of the need Local The for
52 and
failsafe to one
fasteners if the
highlyloaded bonded structures. in catastrophic were only failure due to bonded, without
member
would
unrestrained
delamination
fasteners.
to quite a small
making repairs quite practical. While the bond alone is not as strong as the adherends, the fasteners are and they can reduce the loss of strength due to the initial damage to only a loss of effective area without any notch as
factor.
The contrihution of the adhesive in this case is of fasteners in with a good the bond. particularly since the delamination
combination laminate
rather than
sealant
84
Iof
'
the
structure.
It
is
important
that
the
bonding
medium
should
flow
Sover
as not to wet the surfaces nor so much as to flow out all neither so little So the ideal material for such an application will have the the place. compatihle with tle available heatup rate for a given most cure cycle
application.
4.6 CONCLUSIONS The new nonlinear analysis of bonded/bolted joints is particularly useful in The method can be used to the context of damaqed or imperfect structures. analyze the residual strength of such structures. i These achieve analyses any have confirmed that bonding over and bolting together do not in
significant
advantage
adhesive
bonding
alone
welldesigned Intact structures. !11The program makes A oEK ~damaged or defective bonded structures by matching the various deformations through each load path. The question of using failsafe rivets in bonded structures is more complex generally recognized and a thorough explanation of the entire story than is Rivets are usually superfluous in lightly loaded structure, but is provided. bolts can be very valuable for heavily loaded bonded structures. possible the analysis of the bonded/bolted repair of

"i
11
85
REFERENCES
1. L.
J.
Hart.Smith,
"AdhesiveBonded
Scarf
and
SteppedLap
Joints",
of
Minnesota, April 1980; published in ASTM STP 749. L. J. HartSmith and C. G. Dietz, Company, "Interactive Composite USAF Technical Report
Desi qn",
Aircraft
HartSmith,
"AdhesiveBonded
DoubleLap Joints",
Douglas Aircraft
Company, , _ 5. L. J.
NASA Langley Research Center Report CR112235, January 1973. HartSmith, Joints Paper in 6224, "Advances Composite presented California, in the Analysis to SAMPE 19th and Design of AdhesiveDouglas Aircraft and Symposium
Bonded Company
Aerospace
Structures",
National
"AdhesiveBonded Douglas
Phenomenological
Considerations",
Aircraft Company
6707,
presented to Techonology Conference Associates Conference Composite Technology, El Segundo, California, March 17978. 7. C. E. Thompson and L. J. HartSmith, "Composite Material
on Advanced
Structures
Joints", Douqlas Aircraft Company IRAD Report MDCJ0638, July 1971, 8. T. K. O'Brien, Composite January 1981. "Characterization NASA of Delamination Onset and Growth in a Technical Memorandum 81940,
Laminate",
Langley
87
9.
W. Ramberg
amd
W. R. Osgood,
"Description of
StressStrain
Curves
by
NACA TN 902, 1943. "Bolted Joints in GraphiteEpoxy Composities", NASA Langley Contract Report NASA CR Douglas 144899,
Aircraft
Composites  Pnenomenoloqical Considerations and Simple Analyses", Douglas Aircraft Company Paper 6748, presented to Fourth DOD Conference on Fibrous Composites in Structural Design, November 1978. 12. F. P. Cozzone, Pins Made M. A. of Melcon Aluminum and F. and M. Hoblit, Steel Alloys, "Analysis Product of Lugs and Engineering, San Diego, California 1407
Shear
Val. 21, Pp. 113117, May iqn. 13. M. A. Melcon and F. Pins", 14. M. Hoblit, "Developments in the Analyses of Lugs and 24, pp. 160170, June 1953.
T. Swift, the
"The Effects of Fastener Flexibility and Stiffener Geometry on Stress Intensity in Stiffened Cracked Sheet", Douglas Aircraft Report MDCJ6502, Mechanics", pp. 419436, February Noordhoff 1974. 1974. Also published in of Fracture
international
Publishing
Netherlands,
S. Timoshenko and J. N. Goodier, "Theory New York, Second Edition, 1951, pp. 7880. L. J. HartSmith,
of Elasticity",
McGrawHill,
16.
and Structural
Joints",
Paper 7066,
I
2!
"Adhesive Bond Stresses and Strains at Discontinuities ASME, Jnl Eng Matls & Tech, Vol. 100,
88
APPENDIX This appendix (AFWALTR813154, Vol II) has been intentionally omitted from this report because the computer software contained therein is limited to DOD agencies only. Other requests for the software should be submitted in accordance with AFSC Sup 1 to AFR 3006 (DOD Dir 4160.19 dtd 2 Apr 73). must be submitted to AFWAL/FIBRA, WrightPatterson AFB, OH 45433. Requests
89