Printed copies available from the following:

NASA Center for Aerospace Information

800 Elkridge Landing Road
Linthicum Heights, MD 21090-2934
(301) 621-0390
 FOREWORD

This document is one of nine complementary final technical reports on the development of
advanced composite transport hselage concepts. The work described was performed by
the Boeing Commercial Airplane Group, Seattle, Washington, from May 1989 through
December 1995 under contracts NAS1-18889 and NAS1-20013, Task 2. The contracts
were sponsored by the National Aeronautics and Space Administration, Langley Research
Center (NASA-LaRC) as part of the Advanced Composite Technology (ACT) program.
Direction from NASA-LaRC was provided by M.J. Shuart, J.G. Davis, W.T. Freeman,
and J.B. Nelson.
The nine documents comprising the final documentation for the NASMBoeing ATCAS
program include:
Advanced Technology Composite Fuselage
- Program Overview (CR-4734). Synopsis ofprogram approach, timeline and signifcant
findings. Design synthesis considering manufacturing, materials, processes, structural
performance, maintenance, and cost.
- Manufacturing(CR-4735). Baseline manufacturing and assembly approaches. Process and
tooling developments, and manufacturing demonstration activities to address critical
manufacturing issues.
- Materials and Processes (CR-473 1). Baseline and alternative materials and processes.
Material and process developments. Material performance.
- Structural Performance (CR-4732). Methods usedfor design sizing. Analysis and test
activities supporting assessment of design development methodologiesfor critical performance
issues.
- Repair and Damage Assessment SupportingMaintenance (CR-4733). Maintenance
considerations in design. Detailed repair conceptsfor quadrant design. Fabrication, inspection,
and analytical developments.

Cost Optimization Software for Transport Aircraft Design Evaluation (COSTADE)

- Overview (CR-4736). Synopsis of COSTADE initiative, including integration of cost, weight,
manufacturing, design, structural analysis, load redistribution, optimization, and blending.
- Design Cost Methods (CR-4737). Components of cost analysis and their interactions.
Theoreticalframework for process-time prediction. Methodsfor developing and maintaining cost
equations. Applications to A TCAS quadrant designs.
- User's Manual (CR-4738). COSTADE user instructions, including hardware requirements and
installationprocedures. Program structure, capabilities, and limitations. Basis of cost model and
structural analysis routines. Example problems.
- Process Cost Analysis Database (CR-4739). Rationale for database framework. Database
user's guide, including capabilities and limitations. A TCAS process step equations.

Use of commercial products or names of manufacturers in this report does not constitute
official endorsement of such products or manufacturers, either expressed or implied, by
the Boeing Company or the National Aeronautics and Space Administration.
...
111
At completion of these contracts, Boeing program management included Bjorn Backman
as Program Manager, Peter Smith as Technical Manager, and Larry Ilcewicz as Principal
Investigator. Authors listed for this contractor report prepared portions of the document.
The members (past and present) of the Boeing ACT contract team who contributed to the
work described in this document include:
Program Management: Technical Aide: Developmental Manufacturing:
Phil Whalley Bill Waltari Jose Valdez
Ron Johnson Ponci Puzon
Ray Horton Materials and Processes: Bonnie Luck
Jordan Olson Dodd Grande
Bjom Backman David Scholz Test Laboratories:
Karl Nelson Ron Slaminko
Technical Management: Tony Falcone John Schneider
Peter Smith Brian Perkins Carl Preuss
Joan Dufresne
Principal Investigators: Manufacturing Technology: Tony Phillips
Randy Coggeshall Tom May Dan Moreillon
Larry Ilcewicz Kurtis Willden Bill Hardrath
Val Starkey
Structural Design: Tim Davies Business Management:
George Truslove Mark Gessel Jeff Heineman
Chris Hanson Joe Hafenrichter Marge Apeles
Ken Griess Bob Matetich Kira Goerlich
Mike Schram Ken Goodno
Stephen Metschan Dick Curran QC and NDE Development:
Mike Moms Ken Dull Ken Mackey
TUanLe Rob Biomstad Brian Lempriere
Peter Lohr Bill Fortig
Computing Support: Stan Stawski John Linn
Bob Lundquist Chris Harris
Bill Koch Greg Bell Weights:
Sterling Johnston Jan Koontz Glenn Parkan
Rob Synder
Structural Analysis: Tom Cundiff Repair Development:
Tom Walker Gary Moon Bert Bannink
Ernie Dost Mike Evens
Gary Swanson Cost Estimating: Sherry Marrese
Blake Flynn Kent Venters
Gerald Mabson Will Gaylord Customer Support:
David Carbery Cal Pfahl Dave Berg
Scott Finn David Tervo Jeff Kollgaard
Dan Murphy Len Witonsky
Bernhard Dopker Odo Bormke Materiel:
David Polland Robert Humphrey Maureen Hughes
William Avery Mike Proctor Mark Jones
Jeny Bodine Hans Fredrikson Steve Ruth
Doug Graesser Dennis Stogin Doug Wood
Andre Williams Christal Tyson-Winston
Maik FdiG Fki?'WGdh:iieSS: Hcward Laiie
Peter Grant Jim Peterson Mark McConnell
Adam Sawicki Thomas Murray Tom Hesketh
Pierre Minguet

iv
Jndustrv And Universitv Design-Build-Team Members

University of Washington: Sikorsky Aircraft: United Airlines:

Kuen Y. Lin Christos Kassapoglou Bob Bernicchi
James Seferis John Player
Zelda Zabinslcy Northrop/Grumman:
Mark Tuttle Ravi Deo Cherry Textron:
Steve Russell Howard Gapp
Stanford University: Bob Ley
Fu-KUOChang Ram Vastava Sunstrand:
Ram Ramkumar Glen Smith
Oregon State University: Hossein Saatchi
Tim Kennedy McDonnell Douglas: Bill Durako
Benson Black
M.LT: IC1 Fiberite:
Paul Lagace Lockheed Aero. Systems: Erinann Corrigan
Tim Gutowski Tony Jackson Russ Holthe
David Hoult Ron Barrie
Greg Dillon Bob Chu G.M.I:
Hugh McManus Dan Skolnik Roland Chemana
Jay Shukla
Drexel University: Bharat Shah Intec:
Jonathan Awerbuch Lowell Adams Brian Coxon
Albert Wang Lisa Ott Chris Eastland
Alan Lau Rod Wishart
Frank KO Fiber Innovations: Shreeram Raj
Steve Goodwin Don Stobbe
University of Iowa: Garrett Sharpless
Roderic Lakes Zetec:
Hercules Materials Co.: Chuck Fitch
University of Utah: Doug Cairns Gregg Colvin
William Bascom David Cohen
John Nairn Roger Stirling Draper Laboratory:
Lynn Muir Ed Bernardon
University of Wyoming: Will McCaMll
Donald Adams Yas Tokita Hexcel:
Rhonda Coguill Stacy Biel
Scott Coguill Alliant Techsystems: Julaine Nichols
Carroll Grant Kevin Marshal
U. of Cal. Santa Barbara: George Walker
Keith Kedward Tammy Harris E. I. Du Pont De Nemours:
Todd Brown Jim Pratte
Univ. of British Columbia: Mark Wheeler Hal Loken
Anoush Poursartip Jon Poesch Ginger Gupton
Vern Benson
Brigham Young University: Materials Science Corp.:
Ken Chase American Airlines: Walt Rosen
Jim Epperson Anthony Caiazzo
San Jose State University: Marcus Peter
Robert Anderson Structural Consultant:
Northwest Airlines: John McCarty
DOW-UT Jim Oberg
Rich Andelman Erik Restad EBCO Tooling:
Douglas Hoon Mark Wolf Rich Roberts

V
 TABLE OF CONTENTS

1.0 SUMMARY ............................................................................................. 1-1

2.0 INTRODUCTION................................................................................... 2-1
3.0 DESIGN FOR MAINTAINABILITY .................................................... 3-1
3.1 IN-SERVICE EXPERIENCE ................................................... 3-1
3.2 GENERAL MAINTENANCE APPROACH ............................. 3-3
3.2 SKIN/STRINGER REPAIR CONCEPTS ................................. 3-6
3.2.1 Scope 3-6
3.2.2 Issues 3-7
3.2.3 Concepts 3-8
3.3 SANDWICH REPAIR CONCEPTS........................................ 3- 10
3.3.1 Scope 3-10
3.3.2 Issues 3-1 1
3.3.3 Concepts 3-12
3.3.4 Temporary Repair Options 3-13
4.0 STRUCTURAL ANALYSIS SUPPORTING MAINTENANCE ..........4-1
4.1 EFFECTS OF DEFECTS .......................................................... 4-1
4.2 ANALYSIS AND TEST OFREPAIRED STRUCTURE.......... 4-2
4.2.1 Skidstringer 4-2
4.2.2 Sandwich 4-5
5.0 REPAIR FABRICATION. ...................................................................... 5-1
5.1 SKIN/STRINGER BOLTED REPAIR DEVELOPMENTS AND
DEMONSTRATIONS .............................................................. 5- 1
5.1.1 Large Crown Panel Repair Demonstration 5- 1
5.1.2 Airline Comments 5-2
5.2 SANDWICH BONDED REPAIR DEVELOPMENTS AND
DEMONSTRATIONS .............................................................. 5-3
5.2.1 Subscale Process Trials 5-3
5.2.2 Full-scale Process Trials 5-5
5.2.3 Demonstrations 5-1 1
5.2.4 Temporary Repair Trials 5-14
6.0 INSPECTION METHODS ..................................................................... 6-1
6.1 INSPECTION OF SKIN/STRINGER STRUCTURE................ 6-1
6.1.1 Warpage Due to Damage -A Simulation Approach 6-1
6.1.2 Enhanced Optical Schemes 6-3
6.1.3 Lamb Wave Propagation 6-4

vii
 6.2 INSPECTION OF SANDWICH STRUCTURE ........................ 6-5
6.2.1 Pulse Echo Ultrasonics 6-6
6.2.2 Through-TransmissionUltrasonics (TTU) 6-6
6.2.3 Low Frequency (Dry-Coupled) Bondtesting 6-6
6.2.4 Inspection At Field Bases 6-6
7.0 CONCLUSIONS ..................................................................................... 7-1
8.0 REFERENCES...............................................................e ............ 8-1

APPENDIX A Subcontractor Final Report: "Evaluation of Repair

Concepts for Composite Fuselage Shell Structures,"
Oregon State University

...
Vlll
Under the NASA-sponsored contracts for Advanced Technology Composite Aircraft
Structures (ATCAS) and Materials Development Omnibus Contract (MDOC), Boeing B
studying the technologies associated with the application of composite materials to
commercial transport fuselage structure. Included in the study is the incorporation of
maintainability and repairability requirements of composite primary structure into the
design. Such issues must be addressed to meet regulatory requirements and ensure that
life-cycle costs are competitive with current metallic structure. This contractor report
describes activities performed as part of the ATCAS program to address maintenance
issues in composite fuselage applications.
A key aspect of the study was the development of a maintenance philosophy which
included consideration of maintenance issues early in the design cycle, multiple repair
options, and airline participation in design trades. Furthermore, fuselage design
evaluations considered trade-offs between structural weight, damage resistance/tolerance
(repair frequency), and inspection burdens. To support these design decisions, analysis
methods were developed to assess structural residual strength in the presence of damage,
and to evaluate repair design concepts.
Repair designs were created with a focus on mechanically fastened concepts for
skidstringer structure and bonded concepts for sandwich structure. Repair materials and
processes were established in small-scale trials, then demonstrated on large panels. Both a
large crown (skidstringer) and keel (sandwich) panel were repaired. A compression test
of the keel panel indicated the demonstrated repairs recovered ultimate load capability. In
conjunction with the design and manufacturing developments, inspection methods were
investigated for their potential to evaluate damaged structure and venfy the integrity of
completed repairs.

1-1
 2.0 INTRODUCTION

Boeing's Advanced Technology Composite Aircraft Structures (ATCAS) program

(contract NAS1-18889) was initiated in May 1989 as an integral part of the NASA
sponsored Advanced Composites Technology (ACT) initiative. As an extension of this
work, Task 2 of Materials Development Omnibus Contract (h4DOC, contract NAS1-
20013) was awarded in November 1993. Combined, these two contracts addressed
concept selection and technology development (referred to as Phases A and B). An
additional contract (NAS1-20553) has been initiated to verify this technology at a large
scale (referred to as Phase C). The goal of the ACT initiative is to develop composite
primary structure for commercial transport aircraft with 20-25% less cost and 30-50% less
weight than equivalent metallic structure.
The ATCAS program activities within the ACT framework have focused on fuselage
structure. More specifically, the primary objective of the program is to develop and
demonstrate an integrated technology which enables the cost- and weight-effective use of
composite materials in fuselage structures of future aircraft. The area of study is a
pressurized aft fbselage section of a wide body airplane with a diameter of 244 inches
(Figure 2-1). The structure, located immediately aft of the wing-to-body intersection and
main landing gear wheel well, is designated Section 46 on Boeing aircraft. This section,
highlighted in Figure 2-1, contains most of the structural details and critical manufacturing
issues found throughout the fuselage. It has significant variations in design detail due to
relatively high loads in the forward end which diminish toward the aft end, allowing a
transition to minimum gage structure.

Figure 2-1: Baseline Vehicle and Study Section

2- 1
The fbselage cross-section is divided into four circumferential segments in the baseline
manufacturing approach. These "quadrants" consist of a crown, keel, and left and right
side panels, as illustrated in Figure 2-2. The quadrant approach was adopted to reduce
panel assembly costs (fewer longitudinal splices) and leverage the size-related efficiencies
of the automated fiber placement (AFP) process for laminated skins, while maintaining
design flexibility for regions with differing requirements [l, 21.

- Keel
Figure 2-2: Fuselage Quadrants

Design Build Teams (DBTs), consisting of various disciplines responsible for creating
aircraft structure ( e g , design, manufacturing, cost analysis, materials, structures, quality
control) were formed to develop detailed designs and manufacturing plans for each
quadrant, as well as the associated splices. Design trade studies resulted in the selection
of a skidstringer configuration for the crown, and sandwich construction for the keel and
side quadrants [1, 31. Large crown and keel subcomponent panels were subsequently
fabricated to demonstrate manufacturability and to verify cost and weight efficiency of the
respective designs. In addition to numerous small-scale repair trials, one subcomponent
panel of each design was dedicated to demonstrate repair techniques and serve as a
structural test panel.
Crucial to the success of composite fbselage design is the incorporation of maintainability
and repairability requirements. These issues were addressed by the DBT throughout the
ATCAS program. This report comprises the findings of this effort. Section 3.0 presents
the general approach established for including maintenance considerations into the design,
and describes specific repair designs for crown (skidstringer) and keel (sandwich) panels.
Section 4.0 details analysis and test efforts supporting the design of both maintainable
quadrant panels and individual repairs. Developments and demonstrations of repair
manufacturing processes and inspection methods are discussed in Sections 5.0 and 6.0,
respectively.

2-2
Initial investigations of repair technology under Phases A and B of ATCAS were
necessarily limited due to budget and schedule constraints. For this reason, a relatively
narrow scope of damage scenarios, material options, processing and other design variables
were demonstrated. Nevertheless, due to the building block nature of the approach, some
scaling to other damage states is possible. Subsequent development under Phase C
finding will expand composite repair technology, and will have as its goal the
development of a general repair approach for a variety of damage which is likely to occur
in service, with supporting process development and demonstration.

2-3
 3.0 DESIGN FOR

3.1 In-Service Experience

The first step toward designing reliable and cost-effective design details is to understand
the history of composite structure currently flying in the commercial aircraft fleet.
Composite materials, as we know them today, were introduced into the commercial
aircraft industry during the early 1960's and used mostly glass fiber. Development of more
advanced fibers such as boron, aramid, and carbon offered the possibility of increased
strength, reduced weight, improved corrosion resistance, and greater fatigue resistance
than aluminum. These new material systems, commonly referred to as advanced
composites, were introduced to the industry very cautiously to ensure their capabilities.
The early success of the first simple components such as wing spoilers and fairings led to
the use of advanced composites in more complex components such as ailerons, flaps,
nacelles, and rudders. The increased specific stiffness and strengths of composites over
aluminum, coupled with weight-driven requirements caused by &el shortages, led to the
application of thin-skin sandwich structures. Long-term durability requirements of the
original aluminum parts were not fblly accounted for when these composite parts were
originally designed. To compound the problem further, damage phenomena such as
delamination and microcracking were new and complex in comparison to traditional
aluminum structure.
The original composite parts, particularly thin-gage sandwich panels, experienced
durability problems that could be grouped into three categories: low resistance to impact,
liquid ingression, and erosion, Because the parts were secondary structure, and given the
emphasis placed on weight and performance, the facesheets of honeycomb sandwich parts
were often only three plies thick. This approach was adequate for stiffness and strength,
but never considered the service environment where parts are crawled over, tools
dropped, and where service personnel are often unaware of the fragility of thin-skinned
sandwich parts. Because damage to these components is quite often difficult to detect
with a visual inspection, service personnel did not want to delay aircraft departure or bring
attention to their accidents, which might reflect poorly on their performance record.
Therefore, small damages were allowed to go unchecked, often resulting in growth of the
damage due to liquid ingression into the core. Nondurable design details (e.g., improper
core edge close-outs) also led to liquid ingression.
The repair of parts due to liquid ingression can vary depending upon the liquid, of which
water and Skydrol (hydraulic fluid) are the two most common. Water tends to create
additional damage in repaired parts when cured unless all moisture is removed from the
part. Most repair material systems cure at temperatures above the boiling point of water,
which can cause a disbond at the skin-to-core interface wherever trapped water resides.
For this reason, core drying cycles are typically inciuded prior to pefiorming any repair.
Some airlines will take the extra step of placing a damaged but unrepaired part in the
autoclave to preclude any additional damage from occurring during the cure of the repair.

3-1
This is done to assure they will only need to repair the part once. Skydrol presents a
different problem. Once the core of a sandwich part is saturated, complete removal of
Skydrol is almost impossible. The part continues to weep the liquid even in cure such that
bondlines can become contaminated and full bonding does not occur.
Erosion capabilities of composite materials have been known to be less than that of
aluminum and, as a result, their application in leading edge surfaces has been avoided.
However, composites have been used in areas of highly complex geometry, but generally
with an erosion coating. The durability and maintainability of erosion coatings are less
than ideal. Another problem, not as obvious as the first, is that edges of doors or panels
can erode if they are exposed to the air stream. This erosion can be attributed to improper
design or installatiodfitup.
Assessing airline experience with composite structure is, taken as a whole, an extremely
difficult task. Depending on who is consulted from the airlines, the responses vary fiom
horror stories to outstanding success. Given these limitations, the facts and data that are
currently available are the detailed reports that were received fiom the airlines on parts
involved in the NASA-sponsored Advanced Composites Energy Eflticiency (ACEE)
program, which supported the design and fabrication of composite parts such as those
shown in Figure 3-1 to replace metal parts. Five shipsets of 727 elevators have
accumulated more than 331,000 hrs. and 189,000 cycles; 108 737 spoilers have
accumulated more than 2,888,000 hrs. and 3,781,000 cycles. Also included were five
shipsets of 737 horizontal stabilizers which incorporated laminate torque boxes and
sandwich ribs. These stabilizers have amassed over 133,500 flight hours and 130,000
landings as of May, 1995. The service exposure data collected for these parts has not
indicated any durability or corrosion problems. (Some minor corrosion pitting was found
in fastener holes of 737 stabilizer aluminum trailing edge fittings -this due to an obsolete
sealing practice.) Several repairs have been satisfactorily performed on the 727 elevators
and 737 horizontal stabilizers.
Production graphite-epoxy sandwich parts, such as trailing edge panels, cowls, landing
gear doors, and fairings have demonstrated weight reduction, delamination resistance,
fatigue improvement and corrosion prevention. The poor service records of some parts
can be attributed to fragility or the inclusion of nondurable design details. Many of the
design problems were a result of insufficient technology transfer from development
programs such as NASA-ACEE. Fragility, so much an issue in thin-gage secondary
structure, is expected to be much less important in thicker-gage primary structures such as
the fbselage and wing. The thicker skins of the current 777 composite horizontal
stabilizer, for instance, are much more damage resistant.

3-2
 727 elevator

737 horizontal stabilizer 737 spoilers

Figure 3-1: Prototype Commercial Applications of Composite

Primary and Secondary Structures

3.2 General Maintenance Approach

The maintenance approach for the ATCAS program was based upon input and knowledge
gained fiom a working relationship established between the ATCAS team and airline
maintenance personnel. This was accomplished through repair workshops with airline and
customer support personnel, subcontracts with American Airlines for repair trials and
demonstrations, and involvement with an international composite repair committee. The
time spent within these efforts has provided a broader understanding of the overall
environment in which airlines operate. Boeing's involvement in the Commercial Aircraft
Composite Repair Committee (CACRC) has contributed considerably to addressing the
problems that airlines voice. The CACRC is pioneering standards and recommendations
for the design and maintenance of f h r e composite structure based on current and past
experience.
Figure 3-2 shows maintenance development philosophy established during Phase B of the
ATCAS composite fbselage program. Maintenance procedures such as inspection and
repair, which are applicable to a service environment, must be considered during design
selection. Closed hat-section stringers, for instance, are compatible with inexpensive
manufacturing techniques, but pose difficulties relative to inspection and attachment in
repi-iir CpplicEtitions. Material choices m y also be affected. The designer should avoid the
use of different material systems with different curing temperatures on one part. For
instance, skins and stiffeners are sometimes precured at 350°F and then, for manufacturing

3 -3
ease, secondarily bonded with 250'F adhesive. This can present problems when the skins
or stiffeners are repaired at 35OOF; the integrity of the 250°F adhesive at the bond interface
may be compromised with no indication of degradation.
Design concept developments should include parallel efforts to establish maintenance
procedures. As will be discussed later, the crown was the only ATCAS fbselage q-mdrant
where maintenance procedures were established afrer design features for manufacmring
scale-up were set, resulting in unnecessarily complex repair designs and processes. Skin
layup changes were later found to simplifj7 the crown repair without adding weight.

Ultimate

Allowable Critical Damage

Damage Limit Threshold
(ADU (CDT)
Increasing Damage Size

Figure 3-2: Rulesfor Maintainable Composite Structures

Another important aspect of concept development critical to maintenance is damage

tolerant design practices. The allowable damage limits (ADL) and critical damage
thresholds (CDT) defined in Figure 3-2 must be established to support the structural repair
manual and inspection procedures. The former allows rapid determination of the need for
repair during scheduled inspection, while the latter should be sufficiently large to allow
safe aircraft operation between inspection intervals. Knowledge of residual strength and
inspection capabilities should allow determination of both ADL and CDT as a fbnction of
structural location.
The design of some areas of the structure can be controlled by manufacturing and
durability considerations. Specific examples of these considerations are minimum gage
requirements (to provide a minimum of impact damage resistance and avoid knife-edge at
countersink fasteners), and avoiding rapid ply drops and buildups. Areas of the structure
designed to these considerations will therefore have higher margins for damage tolerance.
Figure 3-3 shows the minimum margins of safety for the side quadrant, illustrating the
"over-designed" regions. These zones have ADLs and CDTs larger than the rest of the

3 -4
fbselage section. Zoned ADL and CDT information should prove usefbl to airlines
desiring minimum maintenance costs.

Margins

Figure 3-3: Strength Margin of Safety Distributionfor the Side

Quadrant Subjected to Ultimate loads

The understanding derived fiom residual strength analyses and tests will also ultimately
lead to DBT cost and weight trades that affect nearly all of the total direct operating costs
(DOC). Small increases in manufacturing cost and structural weight may be traded
against increased damage tolerance to reduce maintenance costs. Decisions may be
required to balance the ADL and CDT. For example, test results for laminate tension
notch sensitivity showed an inverse relationship between small and large notch strength [4,
51. Under such circumstances it may be desirable to have some ADL capability to avoid
having to repair small damages but not at the expense of CDTs that allow sufficiently long
inspection intervals and satisfactory failsafe behavior.
Returning to Figure 3-2, another requirement for maintainable composite structure is the
establishment of nondestructive inspection (NDI) and evaluation (NDE) procedures for
practical damage location and quantitative assessment, respectively, during scheduled
maintenance. The latter, which may require ultrasonic methods, should only be required
to assess the effects of damage found by more easily performed procedures (e.g., visual).
When damage is found, efficient repair procedures are needed which the airlines can
accomplish with available resources (tooling, equipment, etc.) and with a minimum
amount of airplane down time. In order to develop repair concepts for a broad range of
damage scenarios, the repair design philosophy is focusing on more generic repairs which
are not damage-specific. This approach will be beneficial because generic designs and
corresponding repair "kits" can be developed for various levels of damage which are,
within certain limits, independent of specific damages. This is intended to greatly reduce
the need to develop repairs for each damage event as it occurs, providing a higher level of
maintainability. Initially, three damage levels have been defined and are shown in Table
3-1 as they apply to a skidstringer configuration.

3-5
 Table 3-1: Damage Level Definitions
I Designation Damage Description Repair

Level 0 Skin delamination or disbond Fastener restraint or wet

from stiffening elements resin repair
Level 1 Critical damage to a single Mechanically fastened
structural element (skin or patch and/or splice
stiffener)
Level 2 Multiple occurrences of Level 1 Same as Level 1
(and higher) damage

Designs will address repair in a building-block approach in that each bay is looked upon as
a unit, or building block. Restoration of that unit (frame, stringer, and/or skin) will be
designed so that larger multiple-bay damages can be handled with less effort. Structural
units are less easily defined for sandwich structure; however, the same general philosophy
applies. The strategy behind this approach is to address the repair scenarios for a large
range of damage at the beginning of the design process to ease the maintenance burden.
Another aspect of the approach is to provide airlines with multiple options for a given
repair situation. Options might include temporary vs. permanent repair, bonded vs.
bolted, or which specific repair materiaVprocessing combination to employ. An airline's
choice might depend on the severity of the damage, the time available to perform the
repair, the airline's facilities and capabilities, inspectioxdoverhaul schedules, and/or current
field environmental conditions.

3.2 Skin/Stringer Repair Concepts

3.2.1 Scope
The ATCAS tension-load-dominated fbselage crown panel is a stiffened skin design with
cocured hat stringers and cobonded J frames (Figure 3-4). Only mechanically fastened
repair concepts were considered for the crown. Repair design variables were first
investigated on building-block coupons with Level 1 damage. Subsequently, a repair was
designed for a large aft crown panel with Level 3 damage (i.e., a through-penetration,
longitudinally oriented, severing two skin bays and a central circumferential frame). These
efforts focused on an early crown-quadrant design which included a skin laminate which
was much stiffer in the hoop than the axial direction.

3-6
 -A-
w -A-

Figure 3-4: Baseline ATCAS Crown Panel.

3.2.2 Issues
Crown repair concepts are focusing on mechanically fastened external skin patches and
nested frame splice angles. Mechanically fastened repairs require care and accuracy in the
drilling of holes and the alignment of parts during assembly. Fastener hole breakout is a
characteristic problem, commonly solved by using a layer of fabric as the outermost ply
for all laminates. Typically, even though there may be other methods to avoid fastener
hole breakout, there are numerous situations in the real world that challenge a good
mechanic's ability to consistently drill high quality holes. Provisions to locate the position
of the drilled holes in the structure include alignment marks and templates. Also, sealant
must be applied both on faying surfaces and at each fastener location to ensure pressure
containment. Because of these issues, a repair that contains a large number of fasteners,
though sometimesunavoidable, is also less desirable.
Typical issues associated with external patches include weight, surface profile, and load
path eccentricities. The added patch material, in addition to increasing weight, can locally
increase the stiffness so as to create a hard spot, which can affect load distributions in the
area of the repair. Concerns over external patches projecting into the airstream are similar
to those for metal repairs. Eccentric load paths, although contributing to bending stresses,
are less of an issue in the crown quadrant where compression loads are less critical than
tension loads, and cmcerns over stability are therefore low.
Another question is whether patch stiffness should be tailored to a specific damage site, or
be made more generic (e.g., all quasi-isotropic) for more general applicability. An airline's

3-7
inventory of materials can be reduced through standardization of the supplies needed for
repairs, such as fasteners, specialized repair materials, tools, and equipment.
Inspection methods must be sufficient to identif4r the damage area for removal. The
choice of methods for inspecting the repaired structure may depend on the type of repair.
A one-sided repair can be easier to install, but is limited to one-sided NDE techniques and
requires the use of blind fasteners, which are difficult to inspect and have less consistent
strength. A two-sided repair is less convenient to perform, but can make use of more
common fasteners and a wider array of inspection methods.

3.2.3 Concepts
The repair concept chosen for demonstration on the large crown panel is shown in Figure
3-5. A predetermined cutout is used to simulate a failed flame and two bays of skin,
assuming that the damage did not interact with the stringer. If stringer damage had
occurred, it would also have to be repaired. The predetermined cutout does not reflect
any anticipation of damage type, shape, or orientation, but rather is generalized. This
approach lets the design of the repair cover a large range of damage scenarios.
The crown repair design includes a two-layer external skin patch for commonality with the
smaller damage. Based on this design, parts of the skin patch can be used for a Level 1
skin repair. The larger Level 3 damage requires a much thicker patch for ultimate load
capability, hence two layers are used. The Level 3 skin patch design utilizes a stack of
two 0.08"skin patch laminates, with the outer patch being shorter than the inner patch to
allow the first row of fasteners to only interact with one patch, thereby locally reducing
the intensity of the load introduction in the longitudinal direction. The longer skin patch
includes holes at the corners to alleviate the severe load introduction at those locations.
Skin patches are quasi-isotropic layups of fiberglass/epoxy fabric which are precured and
bolted to the base structure. Splice members for the severed frame are precured angles
made of graphite/epoxy fabric, bolted to the frame web and caps.
Analysis efforts described in Section 4.2.1 identified two additional configurations which
provide improved structural performance relative to the above baseline. The second
configuration utilizes E-glasdepoxy tape instead of fabric in the skin patches to achieve
anisotropic properties and better match the stifiesses of the base structure. Similarly, the
flame splice material was changed to graphite/epoxy tape, with a laminate equal to the
skin. Although this concept was shown to enhance structural performance, its generality
to repair of the entire fuselage, where stiffness requirements would vary, is questionable.
The third repair configuration (Figure 3-6) was developed for a more recent version of the
crown design, which was generated to satisfl an increased axial damage tolerance design
goal. The redesign included updated strain allowables and increased axial stiffness
characteristics for the crown skin and stringers. These changes allowed a simpler, one-
layer, mechanically fastened, graphite/epoxy patch design than was possible with the softer
skin and stiffeners of the earlier concept. The modified design also includes a narrower
skin patch, a slightly altered frame splice (also incorporated on the second repair
configuration), and a modified fastener pattern.

3-8
 Q Frame
b
-
234
..' c
I h A A A I A
I I S "I I I

L +.....- L
. . .-. . . . . . . . . -
. .-
.. ++++++ i -7-
A -
19.3 L-
A-A

\ 114" Titanium Hi4.ok

Skin IFx-5~37msk
AS41938 tow
13 plies, t-0.096"
Base Patch [45/-45/90/0/60/-60/~]S
-
AS413501-6 tape
15 plies, t-0.111"
[45/90/-45/0/45/-45/01~]S

E-glasslepoxyfabric
8 ptes, t=0.080 (each)
[~5/0/*4510]s
€ralrLe
GrlEp braided
6 plies, t=0.141"
2-D triaxial braid
Frame Splice
GrIEp fabric
16 plies, 1-0.133
[~5/0/*45/0]2s

Flange
Splice

Figure 3-5: Demonstrated Crown Panel Repair Concept

3-9
 Q Frame Q Frame

T A
4
I 1 I 1
P t
A-A

I I I 1
A A A
I V
114" Xtanium Hi-Lok 1
Protruding Head
Fasteners
- -
AS41938 tow AS41938 GrIEp fabric
16 plies, k0.118 13 plies, kO.096 16 plies, b0.133"
[451-2519010/301010/~0]s [-45/45/0/90/-30/30/Zi]s s
[45/~5/90/0/60/-60/~]
& h e r (Fx-134mSll €ra.IlQ
AS413501-6 tape GrIEp braided
15 plies, t=0.111" 6 plies, t=0.141"
[45/01-4510190/0/15/01-151s 2-D triaxial braid

Figure 3-6: Alternative Crown Repair.

3.3 Sandwich Repair Concepts

3.3.1 Scope
The configuration of the compression-dominated fixelage keel structure is driven by load
concentrations at the forward end due to the large cutouts for the wing center section and
wheel well [1,3]. The resulting baseline keel design features a thick solid laminate at the
forward end which transitions via ply drops and tapered core to an equally thick sandwich
structure with relatively thin (12 ply) facesheets at the aft end (Figure 3-7). The current
study addresses the repair of sandwich structure in the mid keel area where the facesheets
are 30 plies and the honeycomb core thickness is 0.637". The damage scenario under
consideration is a 2" diameter penetration of the outer facesheet, and corresponding core
damage. For the Phase B study, all design, analysis, manufacturing trials, and test
activities were tied to this damage state. The emphasis was on permanent bonded repairs,
while a parallel effort to investigate temporary repair options was also conducted.

3-10
 Skin Material: AS4/8552 gr/ep tow
Core Material: HRP-3/16 (8 to 12 pcf)
glass/phenolichoneycomb

152 Ply Solid Laminate 12 Ply Facesheets .

FaKad A-A

Figure 3- 7: Baseline A TCAS keel panel

3.3.2 Issues
Conventionally, sandwich structure is repaired with in-situ processed bonded scarf
patches. With the thick facesheets in the keel panel, however, scarf repairs with traditional
shallow taper ratios (e.g., 2O:l) result in very large patch sizes and the removal of a large
amount of undamaged material. Also, thick facesheets require thick patches, which may
require special processing to achieve proper consolidation. Patch and bondline porosity
are of particular concern with normal field processing, which is accomplished with vacuum
pressure and heat blankets. Lower temperature cures are generally preferred due to
concerns over causing additional damage via vaporization of water which has infiltrated
the core. Also, the surrounding structure may act as a heat sink, making it difficult to
achieve and control the higher temperatures with heat blankets, and may contribute to
thermal gradients which can result in warpage or degradation of the surrounding structure.
Still, the shorter processing times generally associated with higher temperature cures are
very attractive in terms of minimizing the out-of-service time for a damaged airplane.
Patch materials may have to differ from the base structure to achieve the desired
propeities within the constraints of the processing environment. Some airlines prefer wet
layup material systems for their ability to cure at lower temperatures, and to avoid freezer
storage of prepregs; however, there is concern about the quality of wet layup repairs on

3-1 1
thick primary structural parts. The current study is focused on prepreg materials, which
can be easier to work with, but which require freezer storage. Precured patches are
considered for bolted repair concepts. Core material options include replacing damaged
honeycomb by bonding in a plug of the same material, or by some alternative such as in-
situ foam.
Repair concepts must address the effects of moisture in the core -both by minimizing the
degree of moisture ingression, and by determining what its presence does to the
pedormance of the structure. A drying cycle is typically performed prior to the
accomplishment of any bonded repair. Lastly, the completed repair must be inspectable to
ensure its structural integrity.

3.3.3 Concepts
A DBT was formed to conceptualize and evaluate repair designs for the keel sandwich
structure. Input was solicited from internal Boeing repair specialists and airline personnel.
A number of basic repair concepts were identified, and are shown schematically in Figure
3-8. There are a number of options within each basic concept including various scarf
angles, patch shapes, sizes, materials, core plug vs. foaming core, etc.

Standard Scarf (20:l) Steeper Scarf w/ External Plies

I + + +
External Bonded Patch Partial Scarf w/External Plies

External Bolted Patch

Figure 3-8: Sandwich Repair Concepts

Bonded scarfpatches have the advantages of being lightweight and low profile, but result
in the removal of a large amount of undamaged material, particularly in the thick
facesheets of the mid and forward keel. External bonded patches add weight, have a
higher surface profile, and represent a load eccentricity, but require only the damaged
facesheet material to be removed. As a compromise between the hlly scarfed or hlly
external concepts, partial scarf and steep sea$ bonded patches compensate for reduced-
capability scarfs with the addition of a few external plies. External boltedpatches, like
their bonded counterparts, have the advantage of removing only minimal facesheet

3-12
material, and the disadvantages of high surface profile, weight, and load eccentricities.
Achieving more than one row of fasteners with these concepts would probably require the
use of blind fasteners. Another option is to fasten through the full sandwich thickness, but
this would require the addition of potting or some other means to provide support for
fastener clamp-up. Mechanically fastening precured patches has the potential of
representing a quickly accomplished repair.
Variations on the above concepts were explored which included, for instance, rings of
reinforcement to provide an alternate load path around the damage, or elliptical shaped
patches with variable scarf angles (i.e., steeper scarf angles in the direction of lesser load).
The added complexity of these refinements was found to be too labor intensive,
overshadowing any potential gains in structural efficiency. Later efforts were therefore
concentrated on simple patch geometries.

3.3.4 Temporary Repair Options

In keeping with the goal to provide airlines with multiple options to repair damage,
temporary repairs of sandwich structure were also investigated. Such repairs are
characterized by simple, rapid installation, but potentially reduced durability. They would
be applied to small damages which do not take the structure below ultimate load
capability. The intent is to seal the sandwich core from the environment until a major
aircraft check or overhaul is made, when time can be spent to permanently repair the
structure, or the continued integrity of the repair can be verified.
Typical temporary repair designs have focused on thin precured patches which can be cut
to size and adhered with two-part epoxies or other quick-setting adhesives (Figure 3-9).
The patches themselves could be gr/ep, fiberglass, or even metal. Each
materiaVprocessing combination considered will have its own characteristics in terms of
installation time, performance, and inspection requirements. The goal is to provide the
airlines with a range of options for a given repair scenario where, for instance, when only a
short time is available prior to airplane departure, a very quick but perhaps less durable
repair could be performed. Other times, when an airplane is grounded for a longer period
of time, a less quick but more structurally robust repair could be accomplished which
would last longer and/or require fewer intermediate inspections. Current field
environmental conditions may also play a role in determining which repair option to
pursue. The temporary repairs could even become permanent repairs if the associated
inspection burdens are acceptable to the operator.

3-13
 Precured Patch

Figure 3-9: Temporary Repair Designs

3-14
 4.0 STRUCTURAL ANALYSIS SUPPORTING
MAINTENANCE

Analysis methods are needed to support the design of both the original and repaired
structures. By determining the effects of damage on a structure's residual strength,
informed design decisions can be made when faced with the normally competing goals of
minimum structural weight and improved damage tolerance to reduce maintenance costs.
Furthermore, the analysis can determine repair requirements and ensure repair designs are
sufficient to restore the required structural capability.

4.1 Effects of Defects

Current standard repairs for composite structures are generally inadequate for the range of
damages seen in service, thereby causing airlines to frequently perform special repairs that
require input from the manufacturers. A preferred alternative is to determine at the
beginning of the design process what the flight restrictions of the likely damages are. For
example, when damage is detected, it is often desirable for the airline to perform a quick
temporary repair (such as speed tape) and delay a permanent repair until the next
scheduled maintenance. Since the repair design must satisfy an ultimate load requirement,
it is important to understand the damage tolerance capabilities of the structure in advance
to be able to quickly respond to this type of in-service inquiry. To understand the effect of
that damage on the residual strength of the structure is to have an answer to the question,
"DOwe really need to repair it?"
The most common damage threats for composite structure include impacts caused by
accidental collision or foreign objects (e.g., runway debris, service vehicles, hail, tool
drop), overheated surfaces, lightning strike, and other environmental effects (e.g., U V
exposure, moisture uptake, or thermal cycles) which may degrade existing damage [ 6 ] .
Significant research has been performed to understand the effects of impact on
composites. Unfortunately, most efforts have been focused on understanding the
findamentals of composite material impact resistance and tolerance in controlled
laboratory experiments for ideal impact scenarios. Although providing some support to
the development of impact resistant and damage tolerant structural designs, most studies
have not supported the maintenance of composite structures in service.
Preliminary impact studies have been performed for structural design details in all ATCAS
quadrant concepts to understand critical damage characteristics (i.e., those which are least
visible, while having the strongest effect on residual strength) and suitable NDE
procedures. The most complete study was a designed experiment performed for stiffened
structure with skin gages typical of the ATCAS baseline crown design [7]. Destructive
and nondestructive evaluation of impact damage identified complex combinations of
matrix cracks, fiber failurej and delaminations. The extent of damage and visibility was
found to depend on numerous impact, material, and laminate variables. Matrix damage
size for minimum skin gages was found to be independent of matrix toughness. The least

4-1
visible, yet most serious damage (significant areas of fiber failure) was found to occur due
to high energy and large diameter impactors.
Residual strength as a fbnction of damage size can be determined through a combination
of test and analysis. Nonlinear, progressive damage analysis methods developed under
ATCAS have successfblly scaled laminate coupon test results to predict structural residual
strength [SI. In addition to achieving the desired accuracy for structural analysis, these
methods could potentially be combined with quantitative NDE to create resiaual strength
charts and tables suitable for practical assessment of damage found in service. This
includes the establishment of the ADLs and CDTs discussed earlier.

4.2 Analysis and Test of Repaired Structure

4.2.1 Skinhtringer
Initial analysis and test of skidstringer crown repairs were focused on gaining an
understanding of the critical variables affecting structural performance. A series of
experiments were performed on notched coupons of graphite/epoxy laminate with various
repair patch configurations, materials, and load conditions. Both uniaxial and biaxial
loading tests were conducted. The uniaxial tests indicated a large patch with multiple
rows of titanium bolts was more effective at restoring strength than a small patch with a
single row of bolts around its periphery. Additionally, titanium fasteners were found to be
more effective than thermoplastic rivets at load transfer into the patches.
Finite element models were constructed at Oregon State University (OSU) for each test
case, and the experimental results were compared to the theoretical predictions. In
general it was found that strains in repaired composites could be predicted with reasonable
accuracy, provided that geometric nonlinearities (large deflections) were taken into
account. Likewise large deflection analysis better modeled the nonlinear evolution of the
repair's in-plane stiffness. Generally, strain predictions outside of the repair area were in
excellent agreement with test data; whereas within the repair region less correlation was
achieved, especially for biaxial load conditions. The fact that strains within the repair area
were greater than predicted for both specimen and patch suggests the influence of friction
coupling between patch and specimen. No such part contact nonlinearities were modeled.
Agreement between predicted and measured failure loads was not as good as that obtained
for strains. Generally, the finite element analysis predicted failure loads that were lower
than those measured (Table 4-1). This was apparently due to the fact that the bolt loads in
the fasteners connecting the patch to the coupon were more uniformly distributed than
predicted because of nonlinear material response at the fastener holes. No material
nonlinearities were included in the finite element analysis.

4-2
 Table4-1: Predicted vs. Test Failure Loads for Flat Laminate
Repair Coupons

7 42.3 40.1
9 43.0 47.7
10 43.1 56.0
11 42.8 45.4
12 43.0 41.5
22 42.6 48.7
23 41.3 51.4
24a, 24b 48.4 53.7 (1)
26a, 26b 20.1 26.9 (1)

An additional finite element model was developed by OSU to represent Panel lla, the
large aft crown repair test panel described in Section 3.2.3. Overall panel size was
dictated by test machine constraints, leading to a configuration with three frames (22"
spacing) and four stringers (14"spacing). Four design ultimate load conditions for the aft
crown panel were examined in the repair analysis (Table 4-2). A fifth condition was also
analyzed to represent a flight condition which combines fkelage bending-induced crown
compression with cabin pressure.

The analysis modeled two external patch layers (see Figure 3-5) - a base patch and a
shorter cover patch - effectively stepping the patch thickness, and reducing the peak
fastener shear loads at the axial leading edge of the repair. The closed-cell geometry of
the hat-section stringers made it undesirable to locate fasteners between the stringer
flanges, thereby causing overload of the flange fasteners. The large elongated cutouts in
the comers of the base patch serve to reduce the base patch axial stiffness locally and
redistribute loads away from the flange fasteners. These design details were incorporated
into the repair demonstration on configured crown panel 1la. The repaired panel is
scheduled for testing at NASA-LaRC in early 1996.

4-3
 3 Maximumshear 833 1665 773 13.65
4 Maximum Axial Compression -1690 0 0 0

5* Axial Compression + Pressure 0 1665 0 13.65

The analysis of the Panel 1l a repair shows that ultimate strength is approached but not
hlly recovered; however, load carrying capacity demonstrated by test would likely exceed
the prediction, as indicated by the trend toward underprediction of building block test
results. Table 4-3 is a brief summary of the resulting margins of safety for the most critical
elements of the test panel repair (a). As the table indicates, despite the efforts to reduce
peak fastener loads, bearinflypass is the controlling criteria. Permissible loads on the
fasteners were limited by the allowable bearinflypass strains, a direct result of originally
sizing the panel to have little or no margin of safety. Further evaluation of the repair
design and verification of the corresponding analysis will be performed with the test of
Panel 1l a at NASA-LaRC.
Two additional repair configurations, both described in Section 3.2.3, are included in
Table 4-3. The first alternate configuration (b) was analyzed to assess the potential
benefits of tailoring the patch stiffness to better match that of the cut-out hselage
structure. This configuration produces an increase in strength over the first design, but
still falls just short of ultimate load capability.
The second alternative repair configuration (c) was developed for a more recent crown
design. The new skin and stringer stiffhesses associated with this design, combined with
the aspect ratio of the damage cut-out, better matched what was feasible in a repair patch.
Also, as part of the redesign, repair strength objectives were considered from the outset by
increasing the margin between nominal (no damageho repair) strains and allowable strains
- a margin which was previously too narrow for effective repair. Repair of complex
structure entails some degree of perturbation to the developed strain field and a repairable
fbselage must include a minimal strain margin to account for this. These modifications
allcwed the development of sr si~plifiedrepair for the redesigned crown for which analysis
predicted a full return of ultimate strength (see Table 4-3).

4-4
 Table 4-3: Margin of Safety Summaly for Aft Crown Repair Test Panel
Criterion Test Panel Alternate Test Alternate
Repair Panel Repair Fuselage Repair
(b) (c)
GraphiteEpoxy - 0.01 + 0.10
Stable Stable
+ 0.23 ---
--- ---
(1.15 x Design Limit
Load)

The repair of a fiame alone was demonstrated and tested on Panel 14, a five-stringer/four-
frame crown compression test article [9]. Following initial tests of the panel undamaged
and with impact damage, a portion of the fiame web and flange were removed fiom the
center of the panel as indicated in Figure 4-la. Removing this portion of the frame
effectively extended the frame cutout over two adjacent stringers. The panel was tested
again with this damaged-frame configuration prior to being repaired. To accomplish the
repair, a section was cut from a spare crown panel frame and attached to the panel by
metal fasteners through both flanges and through the web (Figure 4-lb). A layer of
adhesive was also used between the frame flanges and the crown panel skin.
Experimental results for Panel 14 indicate the frame repair was effective in restoring the
test article to its original, undamaged state. The response of the panel with the repaired
frame was nearly identical to that of the undamaged panel, and the failure did not occur
through the repair.

4.2.2 Sandwich
Initial analyses related to the repair of keel sandwich structure were conducted, with the
goal of understanding the effects of repair design variables on load paths and structural
integrity. As with the skidstringer crown repair analysis, a finite element approach was
used to evaluate candidate repair designs. This method permits the inclusion of material
nonlinearities (both in adhesive response and facesheet failure progression), and geometric
nonlinearities (due to load path eccentricities through external patches, peel stresses).
Nonlinear analysis techniques similar to those used in ATCAS damage tolerance studies
[4, 5, 81 were employed. ABAQUS [lo] finite element models were developed for the
single baseline damage scenario described in Section 3.3.1, but for a range of repair
configurations. The models were used to evaluate (1) the influence of a repair on the
overall load carrying capability of the panel, (2) the stress distribution in the vicinity of the
repair, and (3) the effects of design details @e., patch sizes, scarf depths, materials, etc.)
on load paths and the cvera!! panel strength.

4-5
 Frame web and flange
removed here

Second frame from top of panel _c44.

(a) Frame Damage

@) Frame Repair

Figure 4-1: Crown Compression Test Panel -Frame Damage and

Repair

The desire to capture possible interactions between the repair and the overall structural
behavior required a large section of the fbselage to be modeled. Thus a 66" x 88" mid
keel section with four frames was chosen as the reference structure. This size corresponds
to the test panel dimensions and is close to the full keel width between splices. A quarter
finite element model was generated with the repair located at the center of the structure on
the tool (outer) side. To reduce the size of the model, the structure was idealized with
shell elements except in the vicinity of the repair. There the structure was modeled with
solid elements to provide accurate through-the-thickness stress distributions. The quarter
model is shown in Figure 4- 1 and a detail of the repair area is shown in Figure 4-3. The
model included approximately 2500 elements and 10,800 degrees of freedom.
Two basic failure mechanisms were considered - buckling and material failure. In the
detailed area near the repair, the core material was modeled as orthotropic with an elasto-
plastic behavior. The yield stress was assumed to simulate the core crushing strength.
The facesheets here were modeled as shells with smeared orthotropic properties.
Laminate failure was modeled by a strain softening law established from previous testing
and analysis [l 11. The strongly nonlinear behavior evident in the stress-strain relationship
of the repair adhesive was modeled as elastic-plastic. The adhesive material law was
strongly dependent on the temperature and humidity.

4-6
4-7
Boundary conditions and external loads were used to approximate the influence of the
remaining hselage on the component response. Radial movement was restrained at the
loaded ends and simple supports were applied along the sides (near the longitudinal splice
locations). The fiames were allowed to move freely. These conditions also represent a
typical test panel configuration.
Two reference points were analyzed prior to evaluating individual repairs: the undamaged
system and a damaged but unrepaired component . The latter included an upper bound on
the assumed baseline damage scenario comprising a 2" diameter hole in both the outer
facesheet and the core. Both axial tension and compression load cases were considered.
The compression load is dominant in the keel, but tension can determine some key
elements of the repair such as the patch size, thickness, and scarf angle.
Three types of bonded repairs were modeled: s c a ~ e d , externally bonded, and
combinations of the two methods. Tension load strength predictions are presented in
Figure 4-4 for a basic repair of each type. Many variations to the basic design were
modeled, including increased bond areas, reduced patch thicknesses, modified material
properties, etc. The strengths of several of these variations are shown in the figure. The
undamaged, damaged but unrepaired, and applied loads are also plotted for reference.
Note that the applied tension loads for the mid keel area modeled are low enough such
that even the unrepaired damaged panel has sufficient load carrying capability. The design
details (e.g., bond area, patch thickness, etc.) are shown to have greater effect on the
performance than the general type of repair (e.g., scarfvs. external patch).
40 SCARF EXTERNAL BONDED SCARF W/ EX1
Basic Incr. Basic Incr. 114 NO Incr. Basic 1/2
Undam. Adh. Bond Patch Edge Adh. Scarf
Strength Area Thick. Taper Stif.

30
n
.-C
\

s
g 20
-I
a
C
2
10

0 Ultimate loads adjusted to RTD typicals for consistency with analyses

Figure 4-4: Analysis Predictions of TensionStrengthsfor Sandwich

Repair Designs

4-8
Tension loads produce a stress concentration adjacent to the hole in the unrepaired
structure and material failure is initiated at this location. The addition of a patch provides
a second load path around the damage and reduces, but does not eliminate, the stress
concentration at the edge of the hole. In a scarfrepair, the highest laminate stress is at this
location where the thickness of the original structure has been reduced to almost zero.
Nevertheless, in all tension cases modeled, the overall strength was driven by the adhesive,
which entered its plastic regime and could not, therefore, increase its load carrying
capacity.
For compression cases, the failure mechanisms are quite different because of interaction
with the stability failure modes. The adhesive enters its nonlinear range, but the total bond
surface does not fail. Still, it reduces the stiffness sufficiently to force a buckling failure.
Because of this response, there is little variation in the strength predictions for the various
repair schemes, as shown in Figure 4-5. Again, even the damaged but unrepaired
structure is seen to have sufficient capability. This is similar to previous findings from the
analysis of an aft keel compression damage tolerance panel [4],wherein a tool side impact
was found to cause only a minimum strength reduction. The tool side damage was less
critical than the bag side damage due to the superposition of bending stresses from global
panel deformation.
25
SCARF EXTERNAL BONDED SCARF W/EX1
Basic Incr. Basic Incr. 114 No Incr. Basic 112
Adh. Bond Patch Edqe Adh. Scarf
Strength Area Thick. Taper Stif.
20 - Undam.
Applied Dam.
n - Load '
.-C 15.4
15 -

4
x
0

10 -
.-
4
2=

5-

n
U
Ultimate loads adusted to RTD typicals for consistency with analyses

Figure 4-5: Analysis Predictions of Compression Strengthsfor

Sandwich Repair Designs

The h o v e ma!ysis was tied to the same baseline damage scenario as the manufacturing
process trials and repair test activities. The size of the manufacturing and test articles
were limited, and therefore so were the damage and repair sizes. Unfortunately, the small
size of the damage analyzed did not result in a great amount of load being transferred into

4-9
the repair patch. This prevented the full effects of the repair design variables from
becoming evident. Still, some trends were discerned from these analyses. The tension
cases point toward the importance of patch bond area, patch thickness, and adhesive
strength. Also, scarfing through the full thickness may not be necessary, and in fact may
even degrade the overall strength because of the reduced material thickness at the area of
greatest stress concentration. Given the stiff bending section of the mid keel sandwich
structure, the load eccentricities associated with external patch plies do not appear to be
an issue.
These analysis methods were used to predict the response of three permanent repairs
applied to the large mid keel demonstration panel (MKl), described in the next section.
This panel was tested in uniaxial compression in the 1M lb test frame at NASA-LaRC.
Extensive instrumentation was provided for evaluation of load paths around the repairs.
The panel failed away from the repairs at 99% of ultimate load (after adjusting to account
for room temperature dry test conditions) away from the repairs. Test setup, fixturing,
and/or manufacturing anomalies may have contributed to the early failure at the edge of a
frame flange [4]; however, the performance of the repairs appeared to be very good. At
the time of this writing, test strain and displacement data was being assembled for
comparison to predictions and validation of the analysis. In the subsequent Phase C effort,
the remaining panel may be cut into quadrants for additional tests of individual repairs.
These might include additional structural tests or evaluations of moisture ingression.
Subsequent cross-sectioning may provide insight into the quality of the repairs and
verification of previously conducted nondestructive examinations.

4- 10
 5.0 REPAIR FABRICATION

Repair fabrication developments conducted under ATCAS have focused on mechanically

fastened concepts for the crown panel design (skidstringer) and bonded concepts for the
keel panel design (sandwich). Repair materials and processes were evaluated with
consideration given to the capabilities of airline repair facilities and the constraints of a
typical field processing environment. The goal is to achieve a quality repair which restores
the necessary structural capability, with a minimum amount of time and effort required of
the airline. The developments and demonstrations discussed below were coordinated with
the design and analysis activities, and subject to the same limitations (e.g., types of
structures and damage states considered).

5.1 Skin/Stringer Bolted Repair Developments and Demonstrations

The mechanically fastened repair concepts investigated for crown structure offer several
advantages including: compatibility with hat-stiffened structure; repair configuration
flexibility to restore structural integrity to large panel areas; and proven low-cost,
inspectable installation techniques. Repair manufacturing activities included fabrication of
coupon and element test specimens, limited manufacturing trials, and repair by American
Airlines personnel of a klly configured 63" x 72" crown pressure-box test panel,
designated Panel 11a. Extensive manufacturing development of crown repair techniques
was unnecessary due to the fairly straightforward approach of using precured elements
and mechanical fastening. Still, lessons learned fiom coupon fabrication and
manufacturing trials were incorporated into the repair demonstration on Panel 1la.

5.1.1 Large Crown Panel Repair Demonstration

The precured, pretrimmed repair pieces for Panel l l a (Figure 3-5) were fabricated at
Boeing. A readily available fiberglass/epoxy fabric material was chosen for the skin
patches in order to standardize the overall repair design approach. The two patch layers
were stacked on the same 122"-radius cure tool (separated by a slip sheet) and cured
simultaneously. Damage cleanup was also accomplished at Boeing using a hand held
router, followed by deburring of the edges.
Graphite/epoxy fabric material (standard modulus, untoughened resin) was chosen for the
frame splice application based on the material's availability and its compatibility with the
drape forming process. Unidirectional material forms (e.g., tow-placed laminate charges)
are much more difficult to drape-form to shapes with complex curvature. Titanium was
also considered as a viable option, but was not used due to the difficulty of fabricating a
complex curvature with that material within schedule constraints. Also, use of titanium
would likely constrain the frame splice elements to be provided by Boeing, given the cost
and difficulty for the airlines of procuring, storing, and machining titanium parts.
Assembly of the mechanically-fastened repair of Panel 1la was completed by American
Airlines personnel at their Composite Repair Center in Tulsa, OK. The two skin patches

5- 1
were held in place using jigs, and all bolt holes were drilled. All drilling and reaming was
accomplished with a sacrificial backup tool in place to minimize fiber breakout around the
exit of the hole. Carbide drills and reamers were used. After cleaning all bolt holes, the
skin patches were installed with pressure sealant between the skin and the base patch.
Sealant was applied using a roller applicator. Lastly, the frame splice members were
mechanically fastened in place. Titanium lockbolts were installed where access was
adequate for the fastener installation tool; titanium Hi-Loks were used where access was
limited, since the collars for these fasteners can be wrench-torqued.

5.1.2 Airline Comments

Consistent with the objective of involving the airlines in the development of repair
processes and design concepts, American Airlines personnel were requested to provide
their comments on relevant repair design and assembly activities. They were generally
receptive to mechanically attached stock elements used in the crown repair approach.
Their only concerns related to drilling fastener holes in laminates, high costs for small
quantities of composite fasteners, and a desire to produce stock patches themselves for
future applications (as opposed to a commitment for purchasing specialty repair parts from
Boeing).
During the repair assembly of Panel 11a, American Airlines created a logbook and entered
pertinent remarks, difficulties experienced and steps taken to resolve them, design
improvement suggestions, etc. The logbook was defined as a "tell it like it is'' record to be
produced by those doing the assembly work. The following excerpts from this record are
offered with only slight editing:
"?'%estructure itselfl as supplied by Boeing, is fabulous - what a work
of carbon art! All repair materials supplied were very well done,from the
pre-cured skin patches to the doublers for the frames. All parts were
manufactured, cut andprefit quite well.
Not having any engineering background, it is my opinion that the size and
quantity of fasteners is quite sufficient tofasten a Sherman tank to the side
of a Space Shuttle. Bere is no doubt in my mind that this repair will
simply notfall 08
It is also my opinion that a much stronger, aerodynamically smoother
repair can be bonded more quickly by composite personnel than the repair
done by us. First of all, bolt-on scab patches need to be done by
structures guys that know how to lay out a fastener pattern ... fi you
want this krnd of repair to be done quickly. Let's face it, most bonders
can't drill a round hole!
Like the bolt-on scab patch, a bonded repair would also require all repair
materiais to be of the pre-cured nature. 13e outer repair patch could be
made of carbon (not glass), and only a single thickness (not double) would
be sufficient. All plies on the patch would be tapered 08making for a

5-2
 much more aerodynamically smooth repair. Inside, a stacked carbon ply
wouldfill the damage cutout, and a doubler bonded between the stringers
over the repair would complete the skin repair. I would however, do a
'riveted' repair to theframe, as was done in the repair we did

5.2 Sandwich Bonded Repair Developments and Demonstrations

Development of repair manufacturing techniques for mid keel sandwich structure
progressed fiom small-scale trials to hll-scale demonstrations. These efforts were
undertaken concurrently with design and analysis activities to help focus the
manufacturing developments toward the most promising materials and design concepts.
The scope of this effort was consistent with that described in Section 3.3.1; i.e., the focus
was on a specific damage state and a specific area of the keel sandwich structure. The
emphasis was on permanent bonded repairs, although a parallel effort to investigate
temporary repair options was also conducted.

5.2.1 Subscale Process Trials

Initially, an effort was undertaken to understand the cure kinetics and mechanical
properties of candidate patch materials. Laminates representative of patches were cured
under vacuum to evaluate different prepreg materials and determine the effects of
processing on patch quality. As a means of assessing patch quality, cross-sections were
taken and porosity was either measured or estimated. Patch porosity raises concerns
about loss of strength and the potentia1 for moisture ingression. Strength was assessed
with open hole compression (OHC) tests.
Two materials were evaluated: AS4/8552, the baseline system for the keel; and H T M -
20 (formerly known as HTADLS1194). Both of these prepreg systems are typically
cured in an autoclave at 40 to 80 psig. The M-20 is an epoxy normally cured at 250°F,
while the 8552 is normally a 350°F cure epoxy. The laminates were evaluated for OHC
under room temperature dry (RTD) and hot/wet (WW) conditions. Little difference was
found between the OHC strengths of HTA/M-20 and AS4/8552, whether cured at 250°F,
300°F, or 35O0F. An extended 250°F cure cycle produced essentially the same hot/wet
properties as higher temperature processes (Figure 5- 1). Furthermore, laminates cured at
the lower temperatures for longer times produced lower porosity than those cured at
higher temperatures for shorter times (0% vs. 1% for HTARM-20, 1% vs. 3% for
AS418 5 52).

5-3
 60

Tested hot (180°F)

wet

5o ~
z
r+
"

40
2m
3
30

20 =

10 2

0 =- 7- 7-

250°F 300'F 350'F 250'F 300'F 350'F

5 hrs. 4 hrs. 2.5 hrs. 5 hrs. 4 hrs. 2.5 hrs.
350'F Cure System 250'F Cure System
AS418552 HTAIM-PO

Figure 5-1: Open Hole Compression Strengths of Patch Laminates

Patch configuration and thickness were also found to affect laminate porosity. Several
laminates, rather than having squared-off edges, were laid up in an inverted pyramid,
effectively sealing the edges, much as in a scarfed-out repair (Figure 5-2). All such
laminates had at least some porosity, while some of those with squared off edges had
essentially none. Using the inverted pyramid configuration, three patches made from
HTA/M-20 were cured at 250"F, and varied in thickness from 4 to 12 plies. Five patches
made from the standard-grade AS4/8552 prepreg were cured at 350"F, and varied in
thickness from 8 to 30 plies, including some which were staged in 8- or 12-ply
increments. Porosity measurements were taken of the cured laminates; results are
presented in Table 5-1. (Note that maximum porosities of 2% to 3% are generally
considered acceptable.) The lowest porosity was measured in a patch consisting of a 4-
ply stack of HTA/M-20, which had a nominal porosity of 1.4%. Porosity increased with
increasing patch thickness. The 30-ply AS4/8552 patch had 4.55% porosity. Patches
staged at 8 to 12 plies at a time generally had levels of porosity between 2% and 3%.

I Squared-off edges vs. inverted pyramid

f scarf patch

Figure 5-2: Patch Laminate Configurations

5-4
 Table 5-1: Porosity measurements.
Material Total plies Plies per Porosity

4 4 1.40
HTA/M-20 8 8 2.49
12 12 2.10

1E
~ 2.54-
2.90
AS418552 12" 3.15
~ 24 ~ ~ 4.22
I 30 I 4.55
* Intermediate stages at 225F for 1 hour
As in the first set of subscale trials (with squared-off edges), HTA/M-20 patches generally
had lower porosity than comparable AS4/8552 patches. The standard-grade AS418552
prepreg was expected to yield lower porosity than the automated tape layup (ATL) grade
prepreg used in the first set of trials, which has limited flow. The data did not conclusively
show this result; however, the standard-grade 8552 prepreg was old, and the affect of
prepreg aging on porosity is not well characterized.
Development of repair manufacturing techniques for mid keel sandwich structure
progressed from small-scale trials to full-scale demonstrations. These efforts were
undertaken concurrently with design and analysis activities to help focus the
manufacturing developments toward the most promising materials and design concepts.

5.2.2 Full-scale Process Trials

The subscale investigations were expanded in the full-scale trials to represent an array of
design variables including scarf and patch geometry, patch material, adhesive, core
replacement, ply staging, damage cleanup, bagging, and cure cycle. The repairs,
performed on a number of previously fabricated keel sandwich panels, were conducted
both at the Boeing B-2 repair shop and at American Airlines. Some of the attempted
repair concepts reflect the ideas of the technicians performing the work. Each completed
repair underwent a number of inspection procedures, including TTU, pulse-echo,
sectioning, photomicroscopy, low frequency bondtester, and/or degree-of-cure analysis.
Comments from the repair technicians were also solicited to assist in the evaluation of the
various repair concepts. Figures 3a and 3b presents the matrix of variables investigated
through the hll-scale trials, as well as a summary of the results.

5-5
5-6
5-7
The first ten repairs were completed on a large section of aft keel sandwich panel ATS24.
The bolted repair (AK24-7)was successfblly achieved, and with relative ease. This type
of repair is limited in applicability, however, to those areas with sufficient core thickness
to accommodate the blind ends of the fasteners. There are also concerns about the
difficulty of thoroughly removing the core so as to allow proper seating of the bolts.
Additionally, fastener holes present potential moisture paths into the core. Subsequent
damage to a previously repaired area would probably require the difficult step of drilling
out these bolts.
Repairs with complex patch ply geometries, such as the oval scarf (AK24-4) and ring
reinforced external patch (AK24-6), were found to be too labor intensive. The ring patch
also suffers from a poor surface profile. The filler plug (AK24-10) offers no reliable
structural load path. Simple scarf, externally bonded, and scdexternal hybrid designs
were found to offer the best balance of manufacturing ease, surface profile, weight, and
structural performance.
As Figure 5-3 indicates, many of the repairs had a significant amount of porosity in the
patch and adhesive. Furthermore, the repaired core was not adequately bonded to the
back facesheet in several cases. Additional fbll-scale trials were conducted to resolve
these issues prior to repair demonstrations on the large mid keel test panel. All of the
additional trials incorporated simple, circular patches with full or partial scarfs plus
external plies. All used a separate honeycomb core plug bond operation, HTA/M-20
prepreg material, Hysol EA9628 film adhesive, vacuum compaction of each ply, and a
250°F cure.
The core-plug-to-back-facesheet bondline was improved by increasing the adhesive
thickness and raising the temperature of heat application to account for a thermal gradient
through the thickness of 40-60'F. The inclusion of a 15 minute, 150°F debulk cycle every
4 plies did not consistently result in reduced patch porosity as it had in the subscale trials.
There were also persistent problems with gross bondline porosity between the patch and
scarfed facesheet (e.g., Figures 5-4and 5-5).
One potential contributor to the gross bondline porosity was thought to be bridging of the
prepreg during successive debulk cycles, caused by a stiff heating pad, breather and
vacuum bag system. This possibility was investigated in trial AA2 by employing a special
elastic vacuum bag, a stretchable breather (Ainveave), and a heat lamp (instead of a heat
blanket). Debulking in this manner tended to reduce the amount of gross bondline
porosity, although not to an acceptable level.

5-8
 Figure 5-4: Pulse-Echo Ultrasonic Inspection of Repair Trial AAl
(3.5 MHz, WhiteAreas are Regions of High Porosity or
Disbonds)

Figure 5-5: Cross-SectionPhotonticrograph (Shown 34l of Repair

Trial AAl

Another possible contributor was that the escape path for air under the patch could be
getting sealed off, thus preventing sufficient pressure from forming to seat the patch onto
the scarf area. This problem is exacerbated in repairs with thick facesheets and small
exposed core areas, such as were used in this study. Since entrapped air must be vented
fiom the core plug across the patch-to-facesheet bond area, an advantage is given to
thinner facesheets (shorter radial distance for the vent path) and larger core plug areas
(greater volume-to-perimeter ratio, and therefore greater force to open a vent path). The

5-9
gross bondline porosity was solved by inserting two small glass string-breathers under the
adhesive layer to evacuate the core during the vacuum-bag cure. The string breather is
closed off during the cure process by infiltrating adhesive.
This string breather process was successfilly demonstrated on full-scale trial AA3 at
American Airlines, and later confirmed at Boeing with trial B-AK23-1. The patches were
well consolidated and the repairs had no areas of gross bondline porosity. The overall
quality of the repairs appeared to be very good, as illustrated in the ultrasonic inspection
results (Figure 5-6) and photomicrographs (Figure 5-7). The glass string breather was
completely infiltrated with resin and sealed from the environment, as observed in the
photomicrographs. Patch porosity measurements were lower (-2.2%) in repairs using the
string breather also, despite the fact that no debulk cycles were used.

Figure 5-6: Pulse-Echo Ultrasonic Inspection of Repair Trial AA2

(3.5 r n z )

5- 10
 Figure 5-7: a of Repair
Cross-Section Photomicrograph (Shown 3
Trial AA2

5.2.3 Demonstrations
Results from the subscale and full-scale process trials were used to select repair designs,
materials, and processes for demonstration on large mid keel panel MKl . The panel was a
curved composite sandwich structure with 30-ply facesheets and cobonded J frames
(Figure 5-8), originally built to the dimensions 79.5"x 118", but later cut down to 66" x
88" for test. Three permanent bonded repairs were performed by American Airlines at
their repair facility. The repairs included a full scarf, external patch, and partial s c a e and
are shown in Figures 5-9 through 5-11, respectively. As in the most recent full-scale
trials, the 250°F cure Ciba Geigy H T M - 2 0 prepreg and Hysol EA9628 film adhesive
were again used.

Repair location (typ)

Trim line for test panel

Figure 5-8: Mid Keel Repair Panel MKl

5-11
 Figure 5-9: Full Scarf Bonded Repair - Panel MKl

Film
Adhesive
/ / L 2 . 0 0 4 \FilierPly
(as needed)
4I- 0.20

External
Plies

Figure 5-1 0: External Bonded Patch Repair - Panel MKl

film
Adhesive
I
I l l I l l I I I I I ( 1 Ill I I I I I I I I I I cPanelMK1

Figure 5-11: Partial Scarf Bonded Repair - Panel MKl

5-12
Each repair was intended to restore a penetration of the outer (tool side) facesheet and
corresponding core damage. The simulated damage was removed by grinding a 2 inch
diameter area fiom the outer facesheet down through the core, leaving the backside
facesheet intact. The back facesheet adhesive layer was abraded but not completely
removed. Material was ground-away using a hand-held, pneumatic grinding wheel and a
router. Facesheets were scarfed in a similar manner, giving a taper ratio of 20: 1.
Prior to laying up the patch, the damaged core was replaced using a separate cure cycle.
(This can also serve as a drying cycle to remove moisture which may have entered through
the damage). A core plug, made fiom the same type of honeycomb as was removed
(HRP-3/16-12.0), was cut to size and bonded in place with a thick layer of film adhesive
(0.020 inch, 4 plies of grade 5) against the back facesheet. Foaming adhesive (Synspand)
was used to bond the sides of the core plug to the pre-existing honeycomb core. Since the
patch was not yet in place, a thermocouple could be inserted inside the core, touching the
adhesive layer, to control the adhesive temperature. The adhesive was cured at 25OOF for
2 hours with a vacuum bag and a heat blanket. Because the repair was treated as "one-
sided", heat was only applied to the outer (tool-side) facesheet which typically heats 40-
60°F hotter than the adhesive layer because of the thermal gradient through the core. The
temperature gradient has no affect on the foaming adhesive which can be cured between
250-350OF. M e r the cure, the thermocouple was removed and the core plug was ground
down level with the edge of the scarf or, in the case of the external patch, flush with the
facesheet surface.
Patch bonding was always done to an abraded, solvent-wiped surface. The bonding
surface was roughened in all areas, including those intended for external plies where no
scarfing was done. The solvent wipe, accomplished just prior to the patch lay-up,
removed all oils and dirt. String breathers were laid down first. The film adhesive was
then laid down over the string breathers, followed by a vacuum debulk cycle. Debulking
was done with a vinyl vacuum bag and no heat. Next, the pre-cut prepreg was laid-up,
with each patch ply matching the orientation of the scarfed ply it was replacing. External
plies were also added as required. Each patch ply was debulked.
Bagging of the patch is illustrated in Figure 5-12. Liberal amounts of breather were
placed both above and below the heating pad to ensure good compaction and removal of
volatiles. The string breathers were placed in physical contact with the breather system.
Caul plates, such as 0.020" thick aluminum, are sometimes used to smooth the patch
surface. However, in patches with external plies and tapered edges, it is believed that a
caul plate would not be able to sufficiently conform to the patch contour, creating bridging
and leaving areas of high porosity or disbonds. For this reason, caul plates were not used
in the repairs assembled at American Airlines.

5-13
 Air-weave breather

the

Figure 5-12: Bagging Procedure

The patch was cured for 2 hours at 250°F with full vacuum. A slow heating rate
(IL”F/rnin)was chosen to promote removal of volatiles and entrained water before the
resin gels. Four to six thermocouples were placed around the perimeter, but were not
allowed to touch the patch so as to prevent mark-off. Although heating blankets were
placed only on the outer facesheet due to the one-sided nature of the repair, insulation
blankets were used in some cases to provide more even heat distribution.

5.2.4 Temporary Repair Trials

A limited number of temporary repair trials have also been conducted as part of this
activity. As described earlier, the intent of such repairs is to provide a moisture seal for
sandwich structure with small damage, typically through the use of small precured patches
and quick-setting adhesives. Efforts have focused on patch configuration, adhesives,
surface preparation, bond cure, and surface finish.
Without any significant structural requirements, the precured patches can be quite thin.
This allows them to be cured flat but still conform to the hselage curvature. Still, there
must be sufficient plies for the patch to serve as a moisture barrier. As a goal the patches
should be easily sized without the use of power tools or expensive tooling aids. It is also
desirable to produce patch material without excessive warpage or distortion.
Several thin precured patches were fabricated in thickness increments of one to five plies.
Materials included graphite/epoxy tape and fabric, fiberglass fabric, and primed aluminum
(0.016” thick). Ply stacking was such that patches were as close to quasi-isotropic as
possible. The precured patches were visually inspected for warpage and possible moisture
paths prior to being bonded to a panel. It was found that 3-ply fabric patch laminates
exhibited minimal warpage, were conformableto curvature, and could be easily cut to size
with hand shears. The addition of an exterior layer of Tedlar (polyvinyl fluoride) aids in
providing an additional moisture barrier. (Although not included in this study, paint alone
might ais0 provide sufficient additionai moisture resistance.j The fiberglass patch appears
to be the least expensive approach; it can be precured at 250°F and has good adhesion
properties.

5-14
A variety of adhesives are available for temporary repairs. The goal is to offer a large
number of adhesives that vary in cure time, cure temperature, viscosity, pot life, and shelf
life. Naturally, each adhesive will have different material storage and processing
requirements. Furthermore, each may exhibit different mechanical properties and carry
inspection and repair life requirements that directly relate to its expected performance.
Depending on the specific repair situation, an airline would be able to choose an adhesive
with its associated processing time and inspection schedule that best suits its need.
Three different types of epoxy adhesives were used in the trials: a two-part fast-setting
room temperature cure, a two-part higher-strength room temperature cure (which can also
be cured more quickly at elevated temperatures), and a 250'F cure film adhesive. The film
adhesive is easiest to apply but requires freezer storage and a bagged cure. The two part
systems can be stored at room temperature and cured without a vacuum bag, but require
more time to mix and apply. Destructive inspections revealed the film adhesive to provide
a qualitatively better bondline, while the two-part epoxies tended to produce small air
pockets.
Three different methods of bond surface preparation were investigated for use with the
temporary repairs: sanding, peel ply, and grit blasting. Sanding has the most practical
application for in-field use. The basic process steps were to mask off the area, sand with
150 grit paper, and then solvent wipe. Grit blasting uses a similar process, however care
must be exercised in not dwelling too long in. one area, and in preventing grit fiom
penetrating or becoming lodged in the local damage area. Grit blasting also would require
specialized tooling and equipment. Both of these methods were used on the patch and the
repaired structure. The peel ply method can make surface preparation a very quick and
easy process with no tools involved, but can only be used on the patch material (either
sanding or grit blasting would still have to be used for the base material). The peel ply
must be cured into the patch and requires an extra step during lay-up.
Several cure cycles were evaluated for the temporary repairs depending on the resin
system and the ambient conditions. The fast-setting epoxy system was cured at room
temperature both with and without vacuum for 20 minutes. One of these cures was
accomplished at an ambient temperature of 47OF, using a heat gun to apply heat locally.
The remaining fast-setting epoxy patches were cured at 7OOF. The higher strength two-
part epoxies were cured with vacuum and a heat blanket at ZOOOF for one hour, and at
300°F for 15 minutes. The film adhesives were also cured using vacuum and a heat
blanket at 25OOF for one hour. The use of vacuum during cure on the two part resin
systems improved the flow and distribution of the resin around the patch but is not
necessary.
Some of the variables investigated in the temporary repair trials were intended to improve
the surface finish. The inclusion of an outer layer of Tedlar, in addition to improving the
moisture resistance of the patch, provides a visually appealing surface that may not require
paint to match the aircraft color. In some cases, dry peel ply was placed against the
exterior patch surface during the bonding operation to act as a breather and a flash

5-15
breaker. The dry peel ply would impregnate with excess resin and, when removed,
provide a smooth faired-in surface without the need for a post-bond sanding operation.

5-16
 6.0 INSPECTION METHODS

When damage is found in service, there is often little or no detailed information on the
event that caused the damage (e.g., impactor geometry, energy levels, time since
occurrence). Reliable nondestructive inspection (NDI) and evaluation (NDE) methods are
needed to locate damage, quanti@ the extent of the damage, assess its effect on residual
strength, and veri@ the integrity of the completed repair. Such methods can be used to
avoid overly-conservative maintenance procedures, thus gaining airline acceptance of
composite structures.

6.1 Inspection of Skin/Stringer Structure

Evaluation of mechanically fastened repairs of skidstringer crown structure can be
accomplished with standard NDI methods: TTU to evaluate precured patch laminates,
visual inspections and physical measurements to verify sealant and fastener installation.
This study has therefore focused on innovative methods to detect damage and evaluate its
effect on residual strength.

6.1.1 Simulation of Warpage Due to Damage

When analyzing the window belt panels associated with the side quadrants, it was
observed that a significant change in a panel contour was induced by machining the
window cutouts. Similar changes in contour might result from impact damage.
Simulations of this effect were conducted to evaluate its use as a damage detection and
quantification method. Specifically, deformations of a flat five stringer crown panel was
predicted for two specific damage scenarios described in Figure 6-1.

Example: Flat Five Stringer Panel

Stringer: 45/90/-45/0/45/-45/0/90/0/-45/45/~l-45I9~l4
Skin : 45/-45/90/0160/-60/9~/-60/60/0/90/-45/45
Material: AS4/938

Two D m a e S c e n a r i i
/
* The Damage Area is of constant size but the number of damage plies increases
through the thickness. The damage area is between two stringers.

* The amount of unsymmetric damage is being held constant, but the damage size
increases. The damage area grows fromthe center between two stringers
towards one stringer.

Figure 6-1: Warpage Due to Daniage - A Siniulation Approach

6- 1
The first investigation assumed damage of constant size centered between two stringers.
Levels of damage were simulated by degrading the fiber and matrix stiffness by three
orders of magnitude for various numbers of plies. Results given in Figure 6-2 indicate that
the damage induces out-of-plane deformation. For low damage severities, large
deformation changes were observed. However, as damage levels increased, deformations
become less severe; with all plies damaged, deformations differed only slightly from the
undamaged panel. This deformation reversal is likely the result of increasing damage
symmetry. Small changes in the deformation contours shown in Figure 6-2 for the klly
damaged panel are due to overall load redistribution. A more refined mesh surrounding
the damage would help quantify some additional local effects The results indicate that
visual techniques have limitations relative to detecting and quantifying damage. Changes
in panel contour indicate the presence of damage. However, a lack of contour change
does not guarantee that no significant damage exists.

U1 VALui'
-1.905-0:
-1. 89E-01
-1.972-01
-1.86s-0i
-1.55E-01
-1.83s-Oi
-1.82E-01
-1.815-0:
-1.80s-01
-1,762-O?
-1.771'-01
-1.76E-01
-1.753-0:
-1.73E-01

Damage
No Damage 4 Plies Damaged
= Area

7 Plies Damaged All Plies Damaged

Figure 6-2: Increasing Out-Of-Plane Deformation Due to Increased

Damage Through the Thickness

In the second investigation, the amount of through-the-thickness damage was held

constant. The damage was limited to the inner four plies. The damage size was increased
fiom no damage io a 1.5" x 2.5" size damage. Results, shown in Figure 5-3, again
indicate that the damage induces changes in the out-of-plane deformation contours. The
magnitude of out-of-plane deformation initially is very large with the onset of damage, but
appears to converge to a maximum value as damage size increases. This provides firther

6-2
evidence that the magnitude of panel deformations cannot be used to easily assess damage
extent.
u1 VALWE
-1.90E-01
-1.89E-01
-1.88E-01
-1.86E-01
-1.8%-01
-1.84E-01
-1.82C-01
-1.81E-01
-1.80E-01
-1.18E-01
-1.llE-01
-1.16E-01
-1.15E-01
-1.13E-01

Figure 6-3: Increasing Out-Of-PlaneDeformation Due to Increased

Partial Damage Area

Figure 6-4 summarizes the ability of the method to relate out-of-plane deformation
changes to panel residual strength, which in turn relates directly to repair requirements.
Since the relationship between damage size/severity and panel out-of-plane deformation is
not unique, the relationship between residual strength and out-of-plane deformation is also
not unique. This indicates that the out-of-plane deformation cannot be used to determine
residual strength.

6.1.2 Enhanced Optical Schemes

Most airlines use a detailed visual inspection, supplemented with both mechanical (i.e. coin
taps) and electronic NDI methods to locate damage. Advanced NDI methods that have
recently been considered for in-service application include enhanced optical schemes and
thermography. Such procedures have the potential to inspect large surface areas of
structure with minimum costs. The technique shown in Figure 6-5 [12, 131 has the
resolution to locate the local thermal distortions considered in analyses described in [SI.
Any advanced procedures will need to gain airline acceptance as being practical and
reliable, and resulting in lower total inspection costs (including the combined costs of
labor, equipment acquisition, and down time).

6-3
 Out-of-Plane
t Residual
Strength

A Residual
Damage Size
. Out-of-Plane
Deformation

Out-of-plane deformation can be used to

determine damage location
Out-of-plane deforma: ‘1 cannot be used
I z to determine residual si;ength
Damage Size

Figure 6-4: Residual Strength and Out-of-PlaneDeformation

D-siaht Qgticalconfiauration
To enhance the visualization of surface distortions
(depressions or protrusions as small as 2 microns)

Retroreflectivescreen
Camera
______c__)

I Surface of interest

Figure 6-5: Enhanced Optical Schemes to Inspect Large Surface

Areasfor Ksible Indications of Dnmage

6.1.3 Lamb Wave Propagation

As discussed above, visible schemes may provide sufficient resolution for locating damage;
however, they do not provide quantitative data to predict the effects of damage on
structural residual strength. Figure 6-6 shows the most reliable NDE method pursued in
ATCAS h i quaiitifjiiig the effects of impact damage. This techiique utilizes exper;lme;;:al
data fkom Lamb wave propagation (one-sided gated pulsekatch) and analysis based on
long wavelength dispersion relationships for laminated plates [14, 151. Back calculations
of reduced plate bending and transverse shear stifkesses provide an estimate of the effect

6-4
of damage in degrading local load paths [16]. Good correlations have been found
between reduced local stiffnesses experimentally measured by Lamb wave dispersion and
out-of-plane mechanical loading devices in the impact designed experiment discussed
earlier [7]. Other methods evaluated in these experiments, including visible dent depth,
pulse-echo damage area, and other one-sided inspection procedures were found to have
little or no correlation with the mechanical load measurements.

_.-..-I-

i Receiver

Impact-Damaged Laminates
60

50

40
Phase
velocity, 30
in/ms
20

10

OO 20 40 60 80 100
Frequency, kHz

Figure 6-6: Developnient of Reliable Lamb Wave Dispersion NDE

Methods

6.2 Inspection of Sandwich Structure

The full-scale process trials and repair demonstrations discussed in Section 5.3 were
inspected using a number of nondestructive techniques including pulse-echo ultrasonics,
through-transmission ultrasonics (TTU), and low-frequency bondtesting. The capabilities
of different methods are compared and contrasted in light of the special limitations of a
repair environment.

6.2.1 Pulse Echo Ultrasonics

Pulse echo inspection, using a 3.5 MHz focused transducer and a bubbler arrangement,
gave high resolution (characteristically0.010 inch) images of the skin and doubler areas on

6-5
the repaired side. Disbonds and concentrations of porosity could be identified in some
panels, while in others the path of migrating gas could be discerned. Core anomalies can
not be imaged with this technique. Furthermore, porosity and adhesive bondlines can
prevent detection of anomalies at deeper levels of the repair. The technique was chosen to
provide high-definition data of each repair skin.

6.2.2 Through-Transmission Ultrasonics (TTU)

Through-transmission inspection was performed at 1 MHz with tone-burst excitation and
a 100 dB dynamic range. The equipment was Boeing-developed. The TTU inspection
passes sound through the part thickness so that anomalies in the core and both skins are
detected. The technique requires access to both sides of the part, a condition not normally
existing on an in-service airplane. The technique was chosen to provide an accurate
baseline for comparing other methods, as well as to characterize the patch quality. TTU
located the same disbonds and regions of high porosity in the patches as did the pulse-
echo technique. No core-related defects were identified by TTU.

6.2.3 Low Frequency (Dry-Coupled) Bondtesting

A Sondicator S9 and a Staveley Bondmaster, at frequencies of 14 to 40 kHz,were used as
a third inspection method. Both instruments employ a pitch-catch arrangement wherein an
acoustic wavetrain is generated by one probe tip and measured by the receiving probe tip.
The application of this same technique to skidstringer configurations was described in
Section 6.1.3. Sondicator results are interpreted in terms of the wave propagation speeds.
Fast speeds indicate a relatively stiff structure; slow speeds indicate a soft structure
associated with damage or defects. The technique requires access to only one side of the
part.
Sondicator inspections were performed on several of the full scale process trials. The
sondicator was able to detect a disbond in one case, although the indications were not as
strong as desired. In other cases, laminate and bondline porosity was not detected.
However, the method successfblly located a l-inch diameter impacted area (outside of the
repair) on the far side of one repair panel. Images of a frame flange lying on the far
surface were also visible, indicating complete penetration of the part.
Low frequency bondtesting is currently the best choice for an in-service technique that can
detect flaws in the core. Computerized data-acquisition greatly enhances the usefulness of
this method. Boeing is currently using bondtester C-scan techniques in production and is
extending these techniques to field applications.

6.2.4 Inspection At Field Bases

The American Airlines repair shop employs pulse-echo ultrasonics using commercially
available portable C-scan equipment. This inspection equipment represents the leading
edge of conventional technology. Even though several airlines have purchased such
systems, most still rely on interpretation of real-time data from hand held probes, or simple
tap tests conducted with a coin or tapping hammer.

6-6
The large disbond of trial AA1 (Figure 5-4) was first discovered by American Airlines
using a tap test, and later confirmed with pulse-echo. The pulse-echo scans conducted at
the airline's facility had a lower resolution when compared to those conducted at Boeing,
and the color scale had fewer gradations. Although the equipment was capable of
detecting a large area disbond close to the surface, it was incapable of detecting a smaller
disbond deeper in the patch (trial AA2). The tap test could also not detect this smaller
disbond.
Improvements would be needed before the portable pulse-echo method as used by
American Airlines would be acceptable as a field inspection technique. Use of a focused
transducer setup, with distance-amplitude correction and appropriate gating, might
improve the results. Additional C-scanning with a bondtester would still be necessary to
detect defects in the core.

6-7
 7.0 CONCLUDING REMARKS

The design of composite kselage structure must incorporate maintainability and

repairability considerations to satisfjl customer requirements and reduce total direct
operating costs. Maintenance concerns should be addressed early in the design cycle to
ensure realistic designs and easier supportability in the field. This may include providing
operators with multiple repair options for a given level of damage, or "generic" repairs
which can be applied over a broad range of damage scenarios. Airline participation is
encouraged in the design process to ensure the customers' needs are met.
Design decisions are supported by analysis techniques for assessing the "effects of
defects", Le., the residual strength characteristics of the structure in the presence of
damage. In this way, trade-offs can be made between structural weight, damage tolerance
(repair frequency), and inspection burdens. The analysis methods must also be capable of
evaluating specific repair design details, and determining the resulting strength of the
repaired structure.
Repair materials and processes must be consistent with the capabilities of the airline repair
facilities, and practically applied within the constraints of the typical field processing
environment. Given the wide range of airline capabilities and environments, this reiterates
the importance of providing multiple repair options. In all cases, however, the goal is to
minimize the effort needed by the airline to return the damaged aircraft to full service.
Inspection methods must be developed which can both assess the damage state (to help
determine the need for repair) and verifl the integrity of the completed repair.
These maintenance issues were addressed throughout the ATCAS program through the
design of skidstringer crown and sandwich keel structure. Panel design details such as
skin layup were found to have a significant impact on the complexity of mechanically
fastened crown repair designs. Keel repair designs focused on bonded concepts with
simple patch geometries which were found to offer the best balance of manufacturing ease,
surface profile, weight, and structural performance. In the case of both crown and keel,
tradeoffs were possible between highly tailored repair designs with improved structural
performance and simpler, more generic designs with less manufacturing complexity and/or
greater applicability to other areas of the fixelage.
Analysis methods were developed to aid in the design of both the original and the repaired
structure. Nonlinear, progressive damage analysis techniques developed in the ATCAS
program have successhlly scaled laminate coupon test results to predict structural residual
strength. These methods, in conjunction with quantitative NDE, have the potential to
create the tools necessary for practical assessment of damage found in service. Use of the
methods to evaluate repair design details and predict coupon test results revealed the
importance of including both material and geometric nonlinearities in the analysis. Post-
processing of data fioni a ieceriL large ked pane! test a d ar, upcming large crown pane!
test will further validate the analytical techniques.

7- 1
Repair materials and processes were successfully developed for both crown and keel,
initially in small-scale trials, then demonstrated on large panels. Extensive manufacturing
development of crown repair techniques was unnecessary due to the fairly straightforward
approach of using precured elements and mechanical fastening. In-situ processing of
bonded repairs for keel sandwich structure was more difficult. Persistent problems with
gross bondline porosity were eventually solved by the addition of two small glass string-
breathers under the adhesive layer to evacuate the core during the 250'F vacuum-bag
cure. This string breather process also produced well consolidated patches, without the
addition of any incremental heated debulk cycles. Subcontracts were established with
American Airlines to perform the major repair demonstrations and some of the process
trials at their repair facility. Feedback from this interaction was invaluable.
A limited investigation of temporary repair options found 3-ply precured fiberglass patches
to be easily fabricated and applied with fast-setting epoxy adhesives at room temperature.
More robust adhesives can be substituted with the addition of a short heathacuum cycle.
Sanding and peel ply were found to be the most practical methods for surface preparation.
The addition of an outer layer of Tedlar improves the moisture resistance of the patch.
New NDI and NDE techniques have been pursued under ATCAS to find damage and
evaluate its effect on residual strength. Advanced optical surface mapping can rapidly and
efficiently characterize geometric changes due to damage. This, in conjunction with out-
of-plane deformation simulations can be employed to estimate the location of damage.
Besides providing a large-scale, rapid inspection procedure, difficult to inspect areas can
be evaluated without disassembly. Lamb wave propagation techniques were not as
successful as pulse-echo and TTU at detecting disbonds and regions of high porosity in
repair patches; however, this method did provide good estimates of the effects of impact
damage in degrading local load paths.
Further development of maintenancehepair technology will be performed as part of the
Phase C contract: Technology Verification of Composite Primary Fuselage Structures for
Commercial Aircraft (NAS1-20553). Consideration may be given to a wider range of
damage scenarios (different locations, different damage sizes and types) and load
conditions (combined load, environment, and fatigue). Associated detailed design and
analysis would allow the refinement of repair design variables such as patch layups, scarf
angles, scarf depths, bond areas, and edge tapers. Alternative prepregs and adhesives may
be explored, as may a wet layup repair option. Processing parameters would necessarily
evolve with design and material changes (e.g., a different adhesive eliminating the need for
string breathers in sandwich bonded repairs). Low-cost temporary repairs may also be
further investigated.
Developed repair designs and processes would be evaluated through extensive
demonstrations, inspections, and tests. The evaluations would be supported by the
development of inspection methodologies which are more quantitative, useful, and
practical for field application. Inspection intervals could be established for specific repair
types as a fknction of their durability, determined in part through analysis and test.

7-2
 8.0 CE

1. Ilcewicz, L., et al: "Advanced Technology Composite Fuselage - Program

Overview," NASA CR-4734,1997.
2. Ilcewicz, L., et al: "Application of a Design-Build Team Approach to Low Cost and
Weight Composite Fuselage Structure," NASA CR-44 18, December 1991.
3. Hanson, C., et al: "Design Integration of a Composite Aft Fuselage Barrel Section,"
Sixth NASADoD Advanced Composite Technology Conference,NASA CP-3 326,
1995.
4. Walker, T., et al: "Advanced Technology Composite Fuselage - Structural
Performance," NASA CR-4732,1997.
5. Walker, T., et al: "Damage Tolerance of Composite Fuselage Structures," Sixth
NASA Advanced Composite Technology Conference,NASA CP-3326, 1995.
6. Harris, C. : "Assessment of Practices in Supporting Composite Structures in the
Current Fleet," Fourth NASA/DoD Advanced Composites Technology Conference,
NASA CP-3229, 1993, pp. 21-35.
7. Dost, E., et al: "Impact Damage Resistance of Composite Fuselage Structure",
NASA CR-4658,1997.
8. Dopker, B., et al: "Composite Structural Analysis Supporting Affordable
Manufacturing and Maintenance, Sixth NASA Advanced Composite Technology
Conference,NASA CP-3326, 1995.
9. McGowan, D., et al: "Compression Tests and Nonlinear Analyses of a Stringer- and
Frame-Stiffened Graphite-Epoxy Fuselage Crown Panel," Fifth NASA Advanced
Composite Technology Conference,NASA CP-3294, 1994, pp. 321-350.
10. Hibbitt, Karlsson & Sorensen, Inc.
11. Avery, W., et al: "Design and Structural Development of a Composite Fuselage
Keel Panel," Fifth NASA Advanced Composite Technology Conference,NASA CP-
3294, 1994, pp. 463-495.
12. Komorowski, J., et al: "Inspection of Aircraft Structures Using D Sight," 39th
International S M P E , 1994.
13. Reynolds, R., et al: "Theory and Applications of a Surface Inspection Technique
Using Double Pass Retroreflection," Optical Engmeering, 32 (9), 1993, pp. 2122-
2129.
14. Lamb, H.: "On Waves in an Elastic Plate", Proc. of the Royal Society of London,
Series A, 1917.

8-1
15. Tang, B., et al: "Low Frequency Flexural Wave Propagation in Laminated
Composite Plates", Acousto-Ultrasonics: Xheory and Application (edited by J. C.
Duke, Jr.), 1988, pp. 45-65.
16. Dost, E., et al: "Experimental Investigations Into Composite Fuselage Impact
Damage Resistance and Post-Impact Compression Behavior," 37th International
SAA4PE Symposium & Exhibition, SOC.for Adv. of Material and Process Eng.,
1992.

8-2
 APPENDIX A

Subcontractor Final Report:

"Evaluation of Repair Concepts for Composite Fuselage Shell Structures"

Oregon State University
EVALUATION OF REPAIR CONCEPTS FOR COMPOSITE
FUSELAGE SHELL STRUCTURES

bY
Timothy C. Kennedy and Matthew F. Nahan

Department of Mechanical Engineering

Oregon State University
Corvallis, OR 97331
June, 1994

A-i
 SUMMARY

The objective of this project was the development of cost-effective repair techniques for
aircraft fbselage made of composite materials. The primary focus was on a damage
scenario consisting of a 22-inch, through-penetration notch in the aft crown section of the
fbselage of a wide-bodied commercial airplane. The notch ran parallel to the central axis
of the fuselage and included the severing of a circumferential frame member. Only
mechanically fastened repair concepts were considered. These concepts were evaluated
through finite element analysis.
To help gain an understanding of the critical variables that affect the performance of a
repair design, a series of experiments was performed on notched coupons of
graphite/epoxy laminate with various repair patches. Both uniaxial and biaxial loading
tests were conducted. Finite element models were constructed for each case, and the
experimental results were compared to the theoretical predictions. The strain response
predicted by the finite element analysis was generally in good agreement with the strain
gage output provided that geometric nonlinearities (large deflections) were taken into
account. Likewise, large deflection analysis better modeled the nonlinear stiffness
evolution of the repair and its effect on surrounding structure. Generally, strain
predictions outside of the repair area were in excellent agreement with test data; whereas
within the repair region less correlation was achieved, especially for biaxial load
conditions. Agreement between predicted and measured failure loads was not as good as
that obtained for strains. Generally, the finite element analysis predicted failure loads that
were lower than those measured. This was apparently due to the fact that the bolt loads in
the fasteners connecting the patch to the coupon were more uniformly distributed than
predicted because of materially nonlinear response at the fastener holes. No material
nonlinearities were modeled in the finite element analysis.
A final repair design was developed which addressed the 22-inch axial notch in the
fbselage aft-crown panel. Cut-out of the damaged skin and frame took on an hour-glass
shape and reflected removal of damaged structural units rather than being specific to the
22-inch notch. The skin was patched externally in a bi-level configuration using stock
composite laminate of E-glasdepoxy fabric. The frame was spliced using multiple angle
brackets of graphite/epoxy. Analysis of the final repair showed that return of the fuselage
to ultimate strength was not possible. However, true load carrying capacity would likely
be greater than predicted as indicated in the coupon tests described above.
An alternate repair design was developed which differed from that of the final design by
incorporating stiffness anisotropy in the stock patch laminate. This trait was achieved
using a tailored ply layup of E-gladepoxy tape. Analysis demonstrated a strength
increase over the final design. Although this concept was shown to enhance structural
perfr?mance, its generality tc! repair of the entire fise!we
& > where stifi.ess ren11irPnlwc
"
-I-----* I*"

would vary, is questionable.

A-ii
In developing the repair design, it became clear that kselage design development should
include repair strength objectives and costs in determining an optimal configuration. In
the present case the margin between hselage strain and the various allowable strain limits
was allowed to be too narrow for effective repair. Repair of complex structure entails
some degree of perturbation to the developed strain field and a repairable hselage must
include a minimal strain margin for this reason.
Taking repair performance and costs into consideration, an alternate (modified) hselage
design was proposed. A repair design was developed for which analysis predicted a return
of ultimate strength. The repair differed from those presented above in its use of a
significantly reduced patch size, of only one layer, and composed of graphite/epoxy. The
alternate hselage differed from the nominal design in providing a larger margin between
operational strain and allowables. In addition, its skin stifEhess anisotropy and damage
cut-out aspect ratio combined to allow for improved repair stiffness matching.

A-iii
2-
 TABLE OF CONTENTS

A1.O INTRODUCTION A- 1
Al. 1 BACKGROUND AND OBJECTIVE A- 1
Al.2 LITERATURE REVIEW OF REPAIR PROCEDURES FOR
COMPOSITE AIRCRAFT STRUCTURES A- 1
Al.2.1 Flush Patches A-3
Al.2.2 Adhesively Bonded External Patches A-5
AI .2.3 Mechanically Fastened External Patches A-6
A1.2.4 Conclusion A-8
A2.0 REPAIRED COUPON TESTS AND ANALYSIS A-9
~ 2I . W ~ A LLOADING TESTS AND mmysrs A-9
A2.1.1 Repaired Coupon Test Specimens A-9
A2.1.2 Finite element Analysis of Repaired Coupons A-9
A2.1.3 Comparison of Theory and Experiment A-2 1
A2.2 BIAXIAL LOADING TESTS AND ANALYSIS A-3 1
A2.2.1 Test Specimens and Biaxial Tests A-32
A2.2.2 Finite Element Analysis of Specimens under Biaxial
Loading A-32
A2.2.3 Comparison of Theory and Experiment A-3 7
A2.3 APPLICATION OF COUPON EXPERIENCE TO CROWN
REPAIR A-46
A3.0 FUSELAGE PANEL REPAIR DESIGN AND ANALYSIS A-47
A3.1 COMPOSITE CROWN PANEL GEOMETRY AND LOAD
CASES A-47
A3.2 FINITE ELEMENT MODEL OF THE CROWN PANEL A-47
A3.3 FINAL REPAIR DESIGNS A-58
A3.3.1 General Approach A-5 8
3.3.2 Crown Panel Repair Design A-62
A3.3.3 Structural Analysis A-72
A4.0 CONCLUSION A-79
A5.0 REFERENCES A-82

A-v
 A1.O INTRODUCTION

Al.1 Background and Objective

During the past 2.5 years Oregon State University (OSU) has been a participant in the
Advanced Technology Composite Aircraft Structures (ATCAS) program under
subcontract to the Boeing Commercial Airplane Company. The primary objective of this
program is the development of cost- and weight-efficient structural design concepts for
commercial transport fkselages made of composite materials. OSU's involvement in this
program has focused on the development of cost-effective repair techniques for damaged
skidstringer fbselage configurations.
During the service life of an aircraft structure composed of composite materials, damage
of various types may develop. This includes delaminations caused by small impacts,
delaminations combined with broken fibers caused by larger impacts, and large through-
cracks caused by penetration impacts or handling damage. Cost-effective methods for
repair of such damage will be necessary to meet regulatory requirements and insure that
life-cycle costs are competitive with current metallic structure.
The primary effort in this project has focused on a damage scenario consisting of a 22-
inch, through penetration notch in the aft crown section of the fuselage. The penetration
includes the severing of a frame member as shown in Figure Al- 1. Only mechanically
fastened repair concepts were considered in this investigation because of the advantages
that they offer over bonded repairs including: (a) installation is low cost (neither special
equipment nor extensive surface preparation is required), (b) material handling and storage
problems are minimized, (c) sophisticated nondestructive evaluation equipment is not
needed for inspection, (d) they are compatible with hat-stiffened structure, and (e) there is
flexibility to restore structural integrity to large panel areas.
This report documents the work performed at OSU in investigating the structural integrity
of repair concepts for damaged composite fksefage panels. The remainder of this section
gives a review of the relevant literature in the field of composite repair. Section A2
describes the results of tests on damaged coupons with repair patches and a comparison of
the results with finite element analysis predictions. Section A3 presents the development
of repair designs for the 22-inch notch in a crown panel described earlier. Section A4
presents the conclusions drawn from this research project.

A1.2 Literature Review of Repair Procedures for Composite Aircraft

Structures
Although the primary focus of this research is on mechanically fastened repair, this
hmtiturz review will consider bonded repairs as we!!. A number of repair procedures
have been proposed and those relevant to composite repair needed for the ATCAS

A- 1
 5 4 L 4

22” Notch
-7

__J\EL__
122.0

Figure Al-1: Damaged Fuselage Panel

A-2
program will be examined here. The majority of repair procedures involve removing the
damaged material by cutting a circular or oval hole around the damaged area. The hole is
then covered with a patch that is either bonded flush with the parent laminate or attached
externally to it by adhesive bonding or mechanical fasteners. A general discussion of these
procedures can be found in a review article by Baker (1986b) that describes damage
assessment, common repair procedures (including surface preparation, adhesives, and
curing), and some simplified analysis techniques for calculating adhesive stresses.
Trabacco et a1 (1988) also present a review article on composite aircraft repair describing
repair materials, the advantages and disadvantages of bolted repairs vs. bonded repairs, the
effects of moisture on bonded repairs, and several examples of successhl repairs. We will
examine the work that has been done on three different categories of repair procedures:
flush patches, adhesively bonded external patches, and mechanically fastened external
patches.

A1.2.1 Flush Patches

Flush patches are used when a smooth surface is required for aerodynamic reasons, and
the load eccentricity produced by external patches must be avoided. Flush patches are
normally mated to the parent laminate along a scarfjoint. The scarftaper ratio is normally
20: 1 or higher in order to minimize stress in the joint. The patch ply layup usually matches
that of the surrounding laminate. Both precured and cocured patches have been used in
flush repairs. Although the precured patch is easier to install, it is usually more difficult to
match the scarf angle with this type of patch. This problem can be reduced somewhat
with a stepped joint rather than a smooth scarfjoint. The repaired area is usually covered
with several additional plies that extend over the surrounding laminate. Although these
repairs produce the maximum joint efficiency, they tend to be the most difficult and time-
consuming to produce. An analysis of the stress in a scarfjoint was presented by Adkins
and Pipes (1985). They showed that joints with small scarf angles are very sensitive to
stiffness mismatch between adherends and to adherend tip bluntness. Reasonable
agreement was found between experimental data and theoretical predictions.
Flush patch repairs have been the subject of a number of theoreticaf and experimental
investigations. Myhre and Kiger (1980) tested four repaired panels of graphite/epoxy
laminates with oval holes. In each case a scarf joint was used and the original material
was replaced with a cocured patch with a layup similar to the removed material plus
several additional plies. In each case the panel was restored to its original strength. Labor
and Myhre (1979b) presented the results of a testing program to evaluate adhesively
bonded repairs of graphite/epoxy laminates. Both flush repairs employing scarf joints and
external patch repairs were evaluated. Test results generally showed that the repairs
restored over 80 percent of the parent laminate ultimate allowable fiber failure strength.
Labor and Myhre (1979a) also presented detailed step-by-step procedures for making
repairs developed during the testing program just described.
Labor (198 1) studied concepts for repair of monolithic skin panels, hll-depth honeycomb
sandwich structure, and sine-wave substructure. For the monolithic skin panels, two

A-3
concepts were described for flush repairs with outside access only. The first concept
consisted of laying down a thin precured laminate over a scarf joint followed by cocured
replacement plies and cocured surface plies. The second concept was similar but used
precured replacement plies. For honeycomb sandwich structure the repair was similar to
the first one described above with a core plug inserted to replace damaged honeycomb.
For sine-wave substructure two repair concepts were studied. The first replaced the
damaged web with an aluminum honeycomb sandwich element fabricated separately and
bonded to the undamaged web. The second used a precured sine-wave web segment held
in place with precured splice doublers.
Dehm and Wurzel (1989) performed an experimental study of laminates with damage
consisting of a 51-mm hole. The repair consisted of a flush patch bonded to the laminate
on a scarf joint with additional surface plies. The panels were loaded in torsion,
compression, and combined torsion and compression. The failure loads of the patched
material were higher than the undamaged material, apparently because of the extra
overlapping plies. Mahon and Candello (1981) studied the repair of a IS-cm diameter
hole in the skin of a borodepoxy skin with honeycomb core sandwich plate. A stepped
joint was prepared around the hole, and then a flush borodepoxy internal patch was
cocured in place. An external stepped cover was also bonded onto the outside of the
laminate. The repair was tested and failed at 139% of the design ultimate load.
Bair et al. (1991) analyzed a laminate consisting of a borodepoxy laminate with a 15-cm
diameter hole under tension. The repair consisted of internal and external adhesively
bonded stepped patches. Strains calculated from a finite element analysis were found to
be in reasonable agreement with experimental results. Maman (1989) presented a
methodology for evaluating the reliability of a repaired structure using a probabilistic
failure criterion. He performed a sample analysis for the case of a laminated plate under
tension with damage represented by a circular hole. The repair consisted of a flush patch
filling the hole. More development is needed to make this method practical for large,
complex structures.
Lin and Wang (1988) analyzed a novel repair concept developed by Boeing. In their work
they evaluated both theoretically and experimentally three different repairs of a circular
hole in a laminate. The first repair consisted of a laminate patch attached to the parent
material along a scarf joint with overlay patches on the top and bottom surfaces. The
second repair consisted of an aluminum bow-tie splice across the hole and an overlay
patch. The end of the bow-tie had a triangular shape. The third repair was similar to the
second but with the end of the bow-tie having a circular shape. The circular-shaped bow-
tie splice was found to produce a smaller stress concentration near the edge of the bearing
surface than the triangular-shaped bow-tie splice. Analysis of the shear stress in the scarf
joint indicated the importance of the patch material having properties similar to those in
the parent material.
Siener (1992) performed a finite element analysis on single lap scarf+joints.He considered
several patch thicknesses but with the section modulus of the patch material kept constant

A-4
in the direction of loading. This was accomplished by reorienting the plies in the patch.
The purpose of this was to increase the load transfer efficiency of the joint. Experimental
results were found to be in qualitative agreement with the finite element predictions. It
was concluded that more detailed models would be required to accomplish quantitative
agreement.

A1.2.2 Adhesively Bonded External Patches

In an adhesively bonded external patch repair, the repair area is covered with a composite
patch that is bonded to the top or bottom surface of the parent laminate. Baker (1986b)
discussed several options for this type of patch: it may have a ply layup similar to the
parent laminate; it may be a quasi-isotropic layup that is thicker than the parent laminate;
or it may be made of layers of titanium foil adhesively bonded together. The primary
drawback to this type of patch is that it consists of a lap joint with an eccentric load path
that can result in bending in the patch and the development of peel stress in the adhesive.
It can also lead to stability problems in compression. Renton and Vinson (1977) have
presented a stress analysis for this type of joint. In a review article on joining composites,
Baker (1986a) discussed some of the problems associated with the joints. The load
carrying capacity of the joint increases in proportion to the square root of the thickness of
the adhesive. For most practical joints the adhesive layer thickness is in the range 0.13 to
0.26 mm. Peel stresses under tension loading arise at the outer edges of the adherends as
a result of the tendency of the outer adherend to bend away from the inner adherend under
the moment produced by the shear stress. These joints can be improved by tapering the
ends to reduce the peel stresses. Rivets are also occasionally used at the ends of these
joints to reduce these stresses. Unfortunately, the presence of fastener holes allows
moisture to enter into a high stress area and can lead to environmentally induced bond
failure.
Despite these drawbacks, the ease of installation of this type of patch as compared to a
flush patch has made it a commonly used repair concept and the subject of several
investigations. Labor and Myhre (1979a,b) studied it extensively and compared it to flush
patch repair. However, they recommend a flush patch over an external patch because of
the superiority in strength, aerodynamic smoothness, weight, stiffness (ie., avoiding a
"hard spot"), and uniformity of load distribution.
Hunter (1990) considered a graphite/epoxy laminate with a circular hole repaired by an
adhesively bonded external patch. They performed a finite element analysis of the
specimen using spring elements to model the adhesive layer between the panel and the
patch. Experiments were also performed. Experimental and finite element analysis results
were compared and reasonable agreement was found. Cripps (1984) pefiormed a series of
experiments on damaged kevlar/epoxy cloth mini-sandwich panels with cellular foam core.
The damage consisted of a hole, and the repair consisted of a plug and an adhesively
bonded external patch of either kevlar/epoxy or fiberglass. The fiberglass patch was found
to make an effective repair and was easier to apply than the kevlar/epoxy. Paul and Jones
(1989) considered the case of damage consisting of a delaminated region around a circular

A-5
hole (e.g., a bolt hole). They made a 3-D finite element model of the delaminated panel.
The repair consisted of a bonded external patch. They concluded that the resulting
increase in residual strength was proportional to the reduction in net section stress. Myhre
(1981) presented a simplified repair where the damaged material was left in place, a low
viscosity resin was injected into the damaged area, and an external patch was cocured in
place.
For adhesively bonded external patch repair of metal structures, several investigators
considered the situation where the damaged area is not removed but allowed to remain as
a crack through the parent material. Jones and Callinan (1979) considered a crack in a
metal sheet under tension that was repaired by adhesively bonded strips of composite.
Through a finite element analysis, they found that the optimum location of the strip was at
a point centered near the crack tip. They also found that if the patch was made too thick,
the reduction in stress intensity factor was more than offset by a significant increase in the
adhesive shear stress. Jones and Callinan (1981) also analyzed the repair of a centrally-
cracked aluminum plate by adhesively bonded borodepoxy patches on one or both sides
of the plate. Through finite element analysis, they found that the most effective patches
were those on both sides of the plate and having a variable thickness from a maximum in
the center to a minimum on the edges. Chandra and Subramanian (1989) studied the
influence of patch parameters (stiffness, width, and length) on the stress intensity factor in
a plate with a central crack using transmission photoelasticity. They found that there was
a critical value of patch length beyond which there was no fbrther reduction in stress
intensity factor. They also found that a symmetric patch was more effective than an
unsymmetric one.

A1.2.3 Mechanically Fastened External Patches

A mechanically fastened external patch is similar to an adhesively bonded external patch
except for the method of attachment (bolts or rivets rather than adhesive) and for the
greater variety of materials used (aluminum, steel, or titanium as well as composites).
Baker (1986b) recommends bolted patches for thick laminates (8-15 mm) where high
shear stress exceeds the capability of the adhesive. In mechanically fastened joints most of
the shear load is transmitted through individual fasteners, and the shear loads in the
fasteners are transmitted to the joint members as bearing loads on the faces of the fastener
holes. Load transfer between the joint members by interfacial friction does not usually
constitute a large portion of the shear load because friction transfer usually cannot be
maintained at a high level during prolonged service due to the loss of clamping pressure
resulting from vibration and wear. The stress concentration that develops at a fastener
hole in a composite cannot be ignored (as is often done for metals) because general
yielding is not possible. Bearing failures at the holes in the composite must also be
considered as these are usually the result of local buckling and kinking of the fibers and the
subsequent crushing of the matrix.
These types of repairs have been the subject of a number of theoretical and experimental
investigations. Manno (1981) studied two repair concepts for 10-cm diameter holes in

A-6
laminates with skin thicknesses ranging from 4.8 to 12.7 mm. The first repair consisted of
a titanium plate bolted to the laminate. The second consisted of an external cocured patch
formed by stacking graphitelepoxy discs on a disc of fiberglass epoxy prepreg. A quasi-
isotropic patch was formed with a 25 to 1 taper ratio. These two concepts were tested
and were found to exceed design requirements. Bohlman et al. (1981) presented an
analysis of the repair of a circular hole in graphite/epoxy wing skins (4.8 to 12.7 mm thick)
by means of a bolted titanium patch. They have developed a computer program,
BREPAIR, that takes into account bolt clearance and flexibility. Allowables for edge of
hole strains, tensionhearing interaction, and fastener shear were determined from
experiments. Correlations between code predictions and test results were good. Bohlman
et al. (1986) extended the BREPAIR program to include the capability of handling biaxial
and shear loading as well as uniaxial tension. They used this program to analyze a
carbodepoxy laminate with a titanium patch. Hoehn and Ramsey (1986) have also made
enhancements to the BREPAIR program. They performed tests and analysis of
carbodepoxy panels (5.3 mm thick) under uniaxial and biaxial load. Damage consisted of
holes of various sizes and shapes. Repairs consisted of sheets of stainless steel bolted to
the laminate. Good correlation was found between predicted and measured strains.
Shyprykevich (1986) used finite elements to analyze a wing cover featuring compliant skin
(6.6 to 7.8 mm thick) between I-stiffeners. Three damage scenarios were considered. The
first consisted of a 10-cm diameter hole in the skin only. This was repaired by an
aluminum plate bolted to the skin. The second damage consisted of a IO-cm diameter
hole in the skin with part of the hole through the top and bottom flanges on one side of the
stiffener. The skin was patched with an aluminum plate. The flanges on the stiffeners
were patched with titanium straps. The third damage consisted of a 10-cm diameter hole
through the skin and stiffener. The skin was patched with a titanium plate and the stiffener
was patched with titanium angles and straps. Shyprykevich et a1 (1991) conducted a
testing program to evaluate the repair concepts analyzed previously using finite elements.
Good correlation was found between the strains predicted by finite element analysis and
those measured in the tests.
Russel et a1 (1991) presented an analysis procedure applicable to unrepaired and repaired
damaged composites. The unrepaired analyses used either empirical relations or elasticity
solutions. The repaired analysis procedure used the Rayleigh-Ritz method. They applied
the latter technique to the bolted-patch repair of a circular hole and compared their results
to a similar analysis using a boundary element program. Busch and Dompka (1991)
described the evolution of a repair design for damage of a lower wing skin plank and
stringer on the V-22 aircraft. The initial repair design consisted of carbodepoxy plates
bolted to the skin and carbodepoxy C-channels bolted to the stringer. Through a finite
element analysis, they were able to optimize their design and bring strain levels to within
specified allowables. Reisdorfer (1992) also developed a similar design for a wing skin
plank and stringer on the V-22 aircraft. Bolted joints used in the repair design were
investigated through static and fatigm tests on i;oupoiis gad were faund to be adequate.

A-7
Four different repair methods for a 9.5-cm diameter hole were evaluated experimentally by
Deaton (1991). The first consisted of a bolted external aluminum plate with supplemental
adhesive. The hole was filled with the circular piece obtained fiom the hole forming
operation and epoxy. The second method consisted of a precured bonded external
graphite/epoxy patch with the patch having four more plies than the original laminate and
a slightly different layup. The third method consisted of a cured-in-place external
graphite/epoxy patch with the patch having four more plies than the original laminate and
a slightly different layup. The fourth method consisted of a cured-in-place flush
graphite/epoxy patch with a scarf joint matching the original laminate plus additional plies
covering the top and bottom surfaces. After seven years of outdoor exposure plus 1.75
lifetimes of fatigue, only the fourth repair method did not show a loss in residual strength.

A1.2.4 Conclusion

The steps needed for the development of a composite structures repair methodology have
been presented by Hall et a1 (1989). First, mechanical properties of the basic materials as
well as data on mechanical and adhesive joints are determined. Next, tests of repair
concepts are performed on simple coupons and the results compared with theoretical
predictions. The final phase proceeds on to testing and analysis of complex built-up
structures. Dodd and Sandow (1991) recommend that this process be capped off with the
development of an expert system to guide technicians in the selection of a repair
procedure. The system that they described was for battle-damaged aircraft structures. It
used existing software for the analysis of adhesively bonded or mechanically fastened
joints and for the calculation of stresses around the damaged area. The system also
contains rules, developed fiom interviews with experts in the field of repair design, that
assess the adequacy of the repair.

A-8
 A2.0 REPAIRED COUPON TESTS AND ANALYSIS

To help gain an understanding of the critical variables that affect the performance of a
repair design, a series of experiments was performed on notched coupons with various
repair patches. Both uniaxial and biaxial loading tests were conducted. A finite element
analysis was performed on each repair, and the theoretical predictions of response were
compared with the experimental results. Details are given below.

A2.1 Uniaxial Loading Tests and Analysis

Uniaxial loading tests were conducted to determine the effectiveness of repair patches
under simple loading conditions. This also provided a means to evaluate the validity of
finite element models used in the analysis.

A2.1.1 Repaired Coupon Test Specimens

A list of the tests on repaired coupons is given in Table A2-1. The coupons were 10" x
30" laminates with 2.5"-long line notches (representing damage) at the center, as shown in
Figure A2-1. For all but two cases the coupon material was a 13-ply graphite/epoxy
laminate (0.074" ply thickness) with a [45/-45/90/0/60/-60/90/-60/60/0/90/-45/45] layup,
which is representative of fkselage skin material. Three of these coupons also had a 2.5"-
wide tear strap running down the center consisting of a 0.0765"-thick graphite fabric
composite. The remaining two coupons consisted of 15-ply graphite/epoxy laminates with
[-45/45/0/90/-30/30/-75/0/75/3 01-30/90/0/45/-451 layups. These two coupons also had
the notches aligned parallel to the load direction. Three different patch sizes (6.66" x
4.02", 7.60" x 2.77", and 7.60" x 5.27") were examined. The patch materials that were
considered consisted of 13-ply graphite/epoxy laminate (identical to hselage skin), a 24-
ply glass fabric in epoxy laminate (0.0045" ply thickness) with alternating 0" and 45"
layers, a 12-ply graphite/epoxy laminate with a [-45/90/45/90/0/90]s layup, and a 0.1"-
thick titanium plate. The elastic properties for each ply of these materials is given in Table
A2-2. The fasteners that were considered consisted of 3/16"-diameter titanium bolts,
1/4"-diameter titanium bolts, and 1/4"-diameter thermoplastic rivets. Sketches of the
various repairs that were tested are shown in Figures A2-2 through A2-9. For Case 17
there is an additional 2.60" x 5.27" patch on the tear strap side of the coupon consisting of
the graphite fabric composite described earlier. The coupons were instrumented with
strain gages and tested to failure under tension by Intec (Polland and Swanson, 1993).

A2.1.2 Finite element Analysis of Repaired Coupons

A finite element model was constructed for each of the test coupons listed in Table A2-1
using the general purpose finite element program COSMOS/M. Four-node quadrilateral
and thee-node triangle, !ayered shell elements were wed to model the coupons and
patches. Beam elements were used to model the fasteners. The length of the beam

A-9
 j.
1

'

o c
o c
;;
T

2
W

0
0
**
*
00

5 i:;j
cd
GI
 1 Load

2.50

€l
J

3 3

BASEL INE
HOOP LOAD

COtdICIIRATICN:
2 TOT&

0"

1
10.00

Load

Figure A2-1: Test Coupon with No Repair

A-1 1
Table A2-2: Elastic Properties of Patch Materials

A-12
 Q;
m

+++ +-+-+I
3 I
T-
6.02 R3=

+++--+++
--

L + + + -- + + +

TWO ROW
3/16 1NCH FASTENER
HOOP LOAD

PATCH W I G U R A T I O N :
( 2 ) ChASs WlTH TI LocI(Ep3LT
( 2 ) GlUEFI WITH TI LGClU33-T
( 2 ) HAf?D GR/Ep WlTH TI L O N X_T
( 2 ) Ti-6AL-4 W I T H TI LocxBoL r
---
8 TOTAL

Figure A2-2: Repaired Coupon Cases 1 through 4

A-13
 7-60 1

3 I
r
2.77 K€F
+ + + + + +
1
+ + +I+ + +
1

ONE ROW
I/4 INCH FASTENER
HOOP LOAD
PATCH CONFIGtrRAT[ON:
( 1 ) GLASS W I T H T I LCCKB0I-T
( 1 ) GLASS W I T H T/P RIVET
(1) W EP W I T H T I LOCKBOLT
( 1 ) QUEP W I T H T/P RfVET
---
4 TOTAL

10.00

Figure A2-3: Repaired Coupon Cases 5 through 8

A-14
3 5.27 F S

3
.68

TWO ROW
l/U INCH FASTENER
HOOP LOAD

PATCH CGWIGLIRATION:
( 1 ) GLASS W l T H T I LOCKBOLT
( 1 ) U S S WfTH T / P RIVET
( 1 ) GFI/EP WlTH T I LOCKBOLT
( 1 ) (3R/Ep W [ T H T/P RIVET
---
4 TOTAL

10.00

Figure A2-4: Repaired Coupon Cases 9 through 12

A-15
3
~

+ +
+-+ + +
+ + + +
+ r7
t1 - 2 5 TIP

7.60 F15

+ f + +

1
.68

5.27 F€F -

TWO ROW
1/4 [NCH FASTENER
A X I A L LOAD

10.00

Figure A2-5: Repaired Coupon Cnse 13

A-1 6
 Q;
-
+ 1
I
I
+ I
I
1.2s PIP

3
+ +
+ +
+ +
+ 4-
-68 -L

ONE ROW
1/4 [NCH FASTENER
AX[AL LOAD
PATCH COWIGtRATION:
(1) WEP W I T H T I LOCKBOLT
---
1 TOTAL

Figure A2-6: Repaired Coupon Case 14

A-17
 8. 4

IO.00 F 5

Figure A2- 7: Repaired Coupon Case 15

A-18
 r 1

Figure A2-8: Repaired Coupon Case 16

A-19
1.25 M
-L
t
--I-
f
1
30

E + +I
i- 27

1
ci
.68

c 68

Figure A2-9: Repaired Coupon Case 17

A-20
elements was taken as the distance between the middle surface of the coupon and the
middle surface of the patch. The ends of the beams were rigidly connected to the coupon
and patch material. All material response was assumed to be elastic. The mesh for an
unrepaired coupon is shown in Figure A2-10. A very dense mesh is used in the vicinity of
the notch tip. To determine the adequacy of the mesh density, stresses near the crack tip
were calculated and compared with the analytical solution with a finite width correction
factor. Figure A2-11 shows that there is reasonably good agreement between the
analytical and finite element solutions.
When the repaired coupon is loaded, there are two possible failure modes. One is
propagation of the 2.5"-1ong notch. The other is bearing stressbypass strain failure at the
bolt hole in the coupon. The Whitney-Nuismer (1974) point stress criterion with a critical
stress of 130,300 psi at a characteristic distance of 0.0654" was used to predict notch
failure. The bearingbypass failure curves for Cases 13 and 14 are given in Figure A.2-12
and the remaining cases in Figure A2-13. The strains at the locations of the strain gages
were calculated as fknctions of load. As will be discussed in the next subsection, it was
necessary to perform a nonlinear analysis that took into account large deflections in the
finite element calculations. Thus, load versus strain curves are not necessarily linear. The
load required for failure was also calculated.

A2.1.3 Comparison of Theory and Experiment

Plots of the measured and predicted strains as fhctions of load were generated for all of
the coupons. The agreement between theory and experiment ranged from fair to
excellent. In this subsection we will focus on just a few representative cases. The strain
gage locations for Case 7 are shown in Figure A2-14. The measured and predicted strains
for gages 5 and 13, which are closest to the notch tip, are shown in Figure A2-15. Both
linear and nonlinear (large deflection) analyses were performed. It is clear fiom the figure
that large deflection effects are significant, and the large deflection solution is in
considerably better agreement with experimental results than the linear solution. This
apparently occurs because the out-of-plane displacements caused by the eccentric load
path through the patch produce a restoring moment that reduces the total amount of
bending. Because of this significant effect, large deflections were taken into account in all
of the calculations.
The predicted (nonlinear analysis) and measured strains for several gages in Case 7 are
shown in Figures A2-16 and A2-17. The agreement between theoretical and experimental
results is good for these gages. An example of only fair agreement between theory and
experiment is shown in Figure A2-18 which comes from Case 2. Gage 11 is located on
the coupon 1.25" from the crack tip. Gage 5 is in a similar location but on the patch.
From the figure it can be seen that the measured strain on the patch is lower than
predicted. Thus, the load transmitted through the patch is lower than predicted. This
might have resulted from an unusually large gap between the bolt and the hole coupled
with low friction between the patch and the coupon. Neither bolt/hole gap nor
coupodpatch friction are accounted for in the finite element model.

A-2 1
 k
w
Q)
c.

Q)

5
Ccl
0

Id

A-22
 DISTANCE FROM CRACK TIP (in.)

Figure A2-11: Comparison of Analytical and Finite Element Analysis

(FEA) Resultsfor an Unrepaired Coupon

A-23
0 25 50 75 IO0
BEARING STRESS (ksi)

Figure A2-12: Bearing StressBypnss Strain Failure Curvesfor Cases

13 and 14

A-24
>-
a

0 25 50 75 100
BEARING STRESS (ksi)

Figure A2-13: Bearing StressBypass Strain Failure Curvesfor Cases

1-12 and 15-1 7

A-25
 0 0 0 0 0 0
0 ---El--- @
0 0 0 0 0 0

Back Front

Figure A2-14: Strain Gnge Locationsfor Case 7

A-26
 373-07

Figure A2-15: Comparison of Linear (FEA-L) and Nonlinear (FEA-

NL) Theoretical Results with Experimental (EW)
Resultsfor Gages #13 and #5for Case 7

A-27
 373-07

-I ooo 0 1000 2000 3000 4000 5000 6000 7000

STRAIN (microstrain)

Figure A2-16: Coniparison of Nonlinear Finite Elenzent Analysis

Predictions (FEA) and Experimental Results ( E m )
for Gages #I and #9for Case 7

A-28
 373-07

-1 000 0 1000 2000 3000 4000 5000 6000 7000

STRAIN (microstrain)

Figure A2-17: Comparison of Nonlinear Finite Element Analysis

Predictions (FEA) and Experimental Results (EX?)
for Gages #2 and #14for Case 7

A-29
 373-02
50

40

- 30
v)
CL
_-
s.
n
Q
0

-
1'20

-
#11 (FEA)

10
#5 (FEA)
A #5 (EXP)

0
-1 000 0 1000 2000 3000 4000 5000 6000 7000
STRAIN (microstrain)

Figure A2-18: Comparison of Nonlinenr Finite Element Analysis

Predictions (FEA) and Experimental Results ( E m )
for Gages #5 and #I1 for Case 2

A-3 0
The predicted and measured (averaged if more than one test is performed for a case)
failure loads for each coupon are listed in Table A2-1. The origin of failure is also
indicated with either a "B" for bearinghypass failure or an "N" for notch failure. The
predicted and measured failure loads for an unrepaired coupon were 38,200 Ib and
38,900 lb, respectively. All of the patches were effective in increasing the strength of the
coupon. Before comparing the finite element predictions to the experimental results, it
should be pointed out that the coupon used in Case 8 came from a defective panel, and its
results are clearly not consistent with the rest of the data. Also, the part-through nature of
the notch in Case 16 results in a complex three-dimensional phenomenon that the finite
element analysis had no chance of properly modeling with shell elements. If Cases 8 and
16 are discarded, a comparison of predicted and measured results indicates an average
difference of 12.4 percent.
In examining the failure loads, it can be seen that the agreement between theory and
experiment was best for the 7.60" x 2.77" patches which contained only one row of
fasteners above and below the notch. For the other cases there were two rows of
fasteners above and below the notch, and the theory predictions of failure were less than
the measured failure loads in all but one case. For all of these cases the theory predicted
failure to originate at the fastener hole due to bearing stress/ bypass strain limitations. In
actual fact, failure usually initiated at the notch tip. This would indicate that the bolt loads
are actually more uniformly distributed throughout the fasteners than predicted. This
could be the result of a materially nonlinear response of the coupon at the fastener hole.
This type of nonlinearity was not accounted for in the finite element analysis.
An examination of the theoretically predicted failure loads does not indicate any particular
trends regarding the effect of patch size, patch material, or fastener type. However, the
experimental results indicate that large patches are more effective in raising the failure load
than small ones and that titanium bolts are more effective than thermoplastic rivets. The
experimental results do not indicate any particular trend regarding patch stiffness.

A2.2 Biaxial Loading Tests and Analysis

Biaxial loading tests and analysis were conducted to determine the effectiveness of
analysis methods in predicting repair behavior under complex loading conditions. Also,
such tests would provide krther experience with the behavior of the tow-formed laminate
in its role as fbselage skin. Finite element analysis was performed on the single repair and
baseline no-repair specimens. Both linear and nonlinear (large-deflection) analyses were
exercised. Analysis predictions were compared with test results. Details are given below.

A2.2.1 Test Specimens and Biaxial Tests

Both repaired and unrepaired tests were conducted upon a notched flat specimen of
ericifk cmfqgwattiofi (40 in. ?I 40 in.). The repaired version is shown in Figures A2-19
and A2-20. The center circular region of the specimen consists of the 13-ply
graphite/epoxy-tow laminate utilized as fbselage skin. It is also the same as that defined,

A-3 1
A-3 2
 I t t i f f
. I

7 -
l-! a
b-5‘ U

A-33
tested uniaxially, and reported upon in the previous section. Outside of the repair region,
doublers were bonded to the specimen. Within the specimen test region, a 2.5-inch notch
was cut. The notch was aligned parallel to the laminate’saxial fkselage direction.
A two-tiered patch configuration was bolted to one side of the specimen over the
specimen notch area using titanium protruding head fasteners. The patching consisted of
the same glass/epoxy fabric as that incorporated in the uniaxial repair tests but of 8 ply
quasi-isotropic layup of 0.08 inch thicknesses. The axial ends of the patches were
staggered with the cover (top) patch set back from that of the base (bottom) patch. The
bi-level repair configuration was expected to improve the bearinghypass strength of the
critical leading axial row of fasteners.
An unrepaired specimen was tested under biaxial load. As identified in Table A2-3, the
ratio of hoop directed load to axial load was 1.19. Also, the specimen ultimately failed
due to fracture at the notch-tip as desired. The specimen was instrumented with strain
gages at various locations including a series ahead of the notch tip. For brevity, the test
instrumentation data has not been included in this report.
A repaired specimen was tested under a series of four load conditions as indicated in Table
A2-4. The final two test runs were truly biaxial and reflected the two critical load
conditions representative of fbselage crown structure. As indicated the test specimen was
loaded to failure in the fourth test run. Its ratio of hoop to axial loading of 0.52
represented the most critical fbselage crown condition and is representative of an ultimate
high-g maneuver. The repair specimen was instrumented with strain gages. Final load
levels for test runs 1 through 3 were controlled by limits set for certain strain gage
combinations. For the first run, the average axial strain ahead of the patch (back-to-back
gages #24 and #45) was to be limited to 2000 microstrains. For the second and third test
runs, the hoop strain ahead of the notch (gage #0) was to be limited to 3000 microstrain.

A2.2.2 Finite Element Analysis of Specimens under Biaxial Loading

A finite element model was constructed for use in subsequent small and large deflection
(nonlinear) analysis. The model for the repaired specimen is shown in an expanded display
in Figure A2-21. As evident in the figure, only one quarter of the specimen was modeled,
taking loading and geometry symmetry into consideration. Model specifics such as code,
element types and material properties are the same as those used in the uniaxial repair
analysis an discussed in Section A2.1.2. Stifhess properties of the graphite/epoxy tow
were an exception in that a slightly greater ply stiffness was used. The ply properties
were: E,, = 19.8E6 psi, E,, = 1.37E6 psi, G,, = 0.68E6 psi, and v12= 0.321. The
specimen, base patch, and cover patch were modeled independently and joined by beam
elements representing the fasteners. The mesh densities within the patch region and about
the notch-tip are indicated in the figure. The patch area element size (0.125”x0.125”)
enabled a density of 10 elements between fasteners which was judged more than adequate.
The nutch-tip element size (0.02”xO.W) was the same used in the uniaxial repair analysis,
and its rationale was discussed in Section A2.1.2.

A-34
 Table A2-3: Unrepaired Biaxial Test Summary

Test Run No. Loading Phoop/Paxial Result

1 1.19 Notch Fracture

Table A2-4: Reynired Biaxinl Test Summary

~ ~~

Test Run No. Result

1 No damage
Paxialp e a k = 1 5 000 1bS
2 Phoop Only No damage
Phoop p e a k = 1 0 7 0 0 lbs
3 1.19 NO damage
Phoop p e a k = 1 2 r 750 lbs
Failure at specimen
edge round-outf away
from repair
Paxialp e a k = 7 9 500 lbs

A-3 5
’f

A-3 6
Analysis of the unrepaired specimen was based on small deflection analysis. Nonlinear
analysis would not produce results significantly different since no one-sided patch existed
to create unbalanced stiffness and resultant large deflection behavior. Two possible failure
modes were identified for the specimen, i.e. notch fracture and peak strain failure along
the large radius round-out edge between the axial and hoop legs of the specimen. Notch
fracture analysis employed the mtney-Nuismer point stress failure criteria as discussed in
Section A2.1.2. Peak strain analysis attempted to address maximum strain realized away
from the notch and due to any strain concentration and/or flexure. Peak failure strains
were extracted from the ATCAS Side Quadrant Material Property Design Data. A 15
percent B-basis knockdown from coupon failure loads was assumed for the allowables
therein, and thus peak failure strains were factored up a corresponding amount. It is
stressed that the 15 percent value was assumed as reasonable based upon other composite
strength studies but could well be otherwise. Resultant peak failure strains of 8670 and
7640 microstrain were thus derived for the axial and hoop laminate directions respectively.
Nonlinear analysis of the repaired specimen was made necessary because the patching
created an imbalance of in-plane stiffness. When in-plane loads are applied, the combined
repair section of patch and specimen attempts to align itself with that of the applied load
as shown in Figure A2-22. This phenomenon is nonlinear in that the lateral deflection of
the specimen, needed for section alignment, develops at a high rate under initial loading
but abates as the sections approach alignment. The corresponding bending moments
caused by the stiffness offset behaves similarly. Comparison of the lateral deflection
predicted by linear and nonlinear analysis for the specimen center is shown in Figure A2-
23. Note that the nonlinear deflection behavior correctly approaches the dimension of the
axial directed stiffness offset for the axially dominate load condition. Analysis of the
repaired specimens addressed two possible failure modes in addition to those relevant to
the unrepaired specimen: specimen fastener bearinghypass strain failure and patch
bearing failure. These criteria were also applied in the uniaxial repair analysis and reported
upon in Section A2.1.2.

A2.2.3 Comparison of Theory and Experiment

The unrepaired specimen was tested at a load ratio of hoop/axial = 1.19. It failed due to
fracture emanating from the notch at a hoop load level of 98,000 lbs. Failure analysis of
the specimen addressed notch fiacture and peak strain failure. Comparison of predicted
versus actual failure loads is listed in Table A2-5. The point-stress fracture criteria based
upon finite element stress ahead of the notch predicted a higher failure load than the peak
strain criterion. However, both were within 6 percent of the actual failure load. A
possible reason why the fracture criterion predicted too high a strength may be that it was
based upon parameters derived from uniaxially loaded experiments. Biaxial loading may
generate greater damage resulting in reduced strength. A possible reason for why the
peak strain criterion predicted an erroneously low strength may be that the assumed 15
percent B-basis knockdown factor used deriving peak faiiure strain from allowabks dztira
was too high. The slopes of strain versus applied loading were derived from strain gage

A-3 7
 - - -

Offset of repair
Unloaded Neutral axis

t
Loaded
Alligned

Figure A2-22: Repair Section Alignment with Plane of Applied

Loading Responsiblefor Nonlinear Large Deflection
Behavior

0.35 I 1 I I

x Run #4: P-hoop/P-axial = -52

/’
0.3

? Predkted
~+
Offset of Repaired Sec-
.-
S tion neutral axis:
v Failure
0.25 Load Axial - direction.
E:

V
;; -+ Offset:
-ua 0.2 Hoop - direction.
.-f
v)

-
R
Q 0.15
L
V Non-Linear F.E.A
Y

5
8.1

0 .e5

0
0 0.2 0.4 0.6 0 -8 1

Applied Load Ratio : P-hoop/200,000 lbs.

Figure A2-23: Nonlinear Analysis Correctly Predicts Repair Section

Alignment

A-3 8
 Table A2-5: UnrepairedSpecimen, Predicted vs. Test Failure

No Repair
Phoop/Pasial= 1-19
Failure Linear Test
Criteria FEA Result
Peak Strain 95,320 lbs. non-failure mechanism
Notch Fracture 103,600 lbs. 98,000 lbs

A-3 9
data and compared with finite element analysis results. Finite element analysis predicted
half the strain rate at the notch tip as that realized from the gage nearest the notch-tip.
Generally, however, the finite element analysis predicted a response 0- 15 percent more
stiff than realized by the gages. This result is consistent with other ATCAS material and
structural tests.
The repaired specimen was run through three nondestructive tests and a final test to
failure. To ascertain whether the specimen would survive the nondestructive tests,
margin-of-safety checks were made. The ultimate load level for each of the three runs was
arrived at by limiting certain strain readings as discussed in Section A2.2.1. Finite
element analysis prediction of the peak load based upon gage readings at the notch-tip
location are prone to two-fold over-estimation. The reason for the inaccuracy is discussed
below. The analysis nevertheless predicted that the specimen would survive the tests but
with some bearingyield damage likely. Upon reviewing the actual load levels attained for
the three nondestructive tests, analysis would indicate that no damage should have
occurred.
Following the nondestructive tests, the specimen was tested to failure. Linear and
nonlinear FEA were employed in predicting specimen failure. Results of the predictions
and test are listed in Table A2-6. Interestingly, the linear analysis came very close to
predicting the actual failure load for the correct failure mode; that is, the specimen failed
due to strain concentration at the large radius round-out edge between the axial and hoop
legs of the specimen. Nonlinear analysis, however, predicted a significantly higher load
level for this failure mode and a lower load level for the specimen bearinflypass failure
mode. Unfortunately, no instrumentation was placed at the peak strain location.
Nevertheless, indirect instrumentation supports the nonlinear analysis prediction as
follows.
The repair section should align itself with the plane of loading and in doing so exhibit more
in-plane stiffness. This should cause more load to be absorbed by the repair and divert
loading from the outer peak strain location. Linear analysis is unable to reflect this
behavior and also generates erroneously high bending moments. Strain gage data, as
discussed below, agrees very well with the nonlinear analysis outside of the patch area.
Comparison of linear and nonlinear strain prediction in this area showed that greater
concentration of strain developed ahead of the repair for the nonlinear analysis which
could be associated with the increased stiffness of the repair under large deflection
behavior, thereby indirectly indicating that less peak strain should have been realized than
predicted by linear analysis. As for the bearinflypass failure analysis, it could be argued
that it is conservative in multi-fastener joints due to fastener load redistribution. Yet even
if this occurred, the nonlinear peak strain failure analysis still overestimated the actual
failure by 15 percent. Scrutiny of the analysis peak strain failure value would suggest that
a lower value is justified; however, this runs contrary to the results of the unrepaired
specimer, test. The Gn!y remaining iatiofi2k fGr the ~ o i i peak
; Siiaiil faihre is the a ‘rest
load imbalance or specimen flaw, both of which are speculative possibilities.

A-40
Of great interest are the actual loads transferred into the repair via fastener shear.
Unfortunately, instrumentation does not exist to measure fastener shear loads. However,
fastener loads as derived by the nonlinear analysis at the failure load are presented in Table
A2-7 according to the fastener chart in Figure A2-24. It is evident that the peak fastener
load is that in the outermost corner of the repair and principally of axial orientation. The
rationale for such a peak load location is that in addition to being a leading-edge fastener,
the comer fastener must also pull at a section of patch without the beneficial assistance of
an adjacent fastener. Mid-side fasteners on the other hand share their influence.
Measured and predicted (nonlinear analysis) strains as a hnction of load were compared.
The agreement between nonlinear theory and experiment ranged fiom excellent to
marginal, with excellent agreement at gage locations outside of the patch area. In the
outer area little flexure was realized, and a pronounced peaking of in-plane strain in front
of the patch was in contrast with linear analysis predictions. This suggests that alignment
of the repair section with the plane of Ioading generated bending local to the repair area.
Within the repair area the general agreement in the trend of strain development was found
between nonlinear analysis predictions and experimental results. However, only marginal
agreement on the actual values was obtained. Locations of several gages within the repair
are shown in Figure A2-25 and A2-26. All identified gages are of axial orientation,
parallel to the notch. Gages #6 and #34 represent average values of the three symmetric
locations indicated. Plots of the nonlinear prediction versus experiment for the gages are
shown in Figures A2-27 and A2-28 (Test Run No. 4, P,,,JP,,, = 0.52). From
examination of the experimental values, it is evident that strain on the top surface of the
patch increases at greater distances toward its center. This behavior, a product of shear
lag, was expected and was predicted by the nonlinear analysis. Although nonlinear
analysis agreed with trends of the strain development, actual strain values showed poor
correlation. Surprisingly the repair area realized higher strain than predicted. This
suggests a coupling between patch and specimen that is more stiff than that modeled. This
phenomena could in part be explained by fiction between the fayed surfaces. This
behavior was also realized in the uniaxial tests but to a much lesser extent.
Correlation of notch-tip gage results with nonlinear analysis was also marginal. As
experienced with the unrepaired specimen, experimental strain was generally twice as high
as predicted. An exception to this was with the gage nearest the notch-tip which during
the test failure run experienced strain 3.6 times that predicted.

A2.3 Application of Coupon Experience to Crown Repair

From the uniaxial coupon tests it was evident that titanium fasteners were superior to
thermoplastic rivets. Also, patches having multiple fastener rows returned greater strength
to the damaged specimen. These are clear-cut results that should be carried forward. Not
clear is the best patch material choice and the expected influence of the various nonlinear
phenomena. However, a degree of confidence can be assumed with regard to small notch
fracture, bearinghypass failure and peak strain.

A-4 1
 Table A2-6: Repaired Specinzen, Test Failure 19s.Linear and
Nonlinear Analysis

Repaired Specimen 1
Phoop/Paxial= 0.52
Failure Criteria Linear FEA Non-linear FEA Test Result
Specimen Peak Strain 40,230 lbs 48,190 Ibs 41,700 Ibs
Notch Fracture 388,000 lbs --- non-failure mode
Specimen Bearing/By-Pass 41,000 lbs 37,070 lbs non-failure mode
Patch Bearing 50,720 lbs --- I non-failure mode

Table A2- 7: Fastener Loadsfor Repaired Specimen as Derived

from Nonlinear Analysis (Listing Refers to the Failure
Load Level)

Bolt Number == I 2 3 4 5 6 7 8

Fastener loads (lbs) at cover patch level

hoop shear force - -180 -74 -127 -68 -92 -86

axial shear force - - 552 478 277 207 0 0
bolt axial tension 50 38 -19 -14 -5 -6

I I Fastener loads (Ibs) at base patch level I

hoop shear force -51 -64 122 -5 236 154 326 454
axial shear force 1461 1352 827 682 40 1 321 0 0
bolt axial tension 138 122 20 16 -18 -7 -8 2

A-42
 Base Patch

I I I 1 I I I

Figure A2-24: Repair Fastener Identification

A-43
Figure A2-25: Strain Gages on Top of Repair Patches

Figure A2-26: Strain Gages Below Repair on Specimen

A-44
 "
0 2500 5000
STRAIN (microstrain)

Figure A2-2 7: Comparison of Nonlinear Finite Element Analysis

Predictions (FEA) and Experimental Results (EXP)
for Averaged Gages #34 and #6

-500 0 500 1000 1500 2000 2500 3000

STRAIN (microstrain)

Figure A2-28: Comparison of Nonlinear Finite Element Analysis

Predictions (FEA) and Experimental Results (EXP)
for Averaged Gages #32 and #40

A-45
The three patch materials and their respective patch thicknesses represented a range of
patch stiffness which would absorb more or less load depending upon its stiffness and
thickness. For large repair application, stiffness consideration can be more significant and
thus results of tests with patches of equal stiffness would have better extended to crown
repair.
The various nonlinear phenomena can be significant to detail within the repair area. This
detail is of great importance with respect to bearinflypass strain failure. However, the
sometimes marginal predictive capability and limited experimental tools makes
understanding of the repair behavior difficult. However, it has been shown that nonlinear
analysis based upon large deflection can provide good predictive results which are
superior to linear analysis, but additional unrepresented phenomena can be significant
within the repair area. Additionally, nonlinear, large deflection analysis seems to do a
good job of correctly modeling the evolution of the repair area stiffness which has an
effect on regional load path behavior. For stiffness critical structural repair, nonlinear
analysis is recommended. Thus nonlinear, large-deflection analysis should be extended to
the crown repair as the best tool available.

A-46
 A3.0 FUSELAGE PANEL REPAIR DESIGN AND ANALYSIS

As desciibed in the introduction of this report, the primary emphasis of the project has
been to develop an efficient design for the repair of a 22"-long, through-penetration notch
in the aft section of a composite fuselage. This section describes the development of this
design.

A3.1 Composite Crown Panel Geometry and Load Cases

The section of crown panel for which the repair design is being developed is shown in
Figure A3-1. The skin material is a 13-ply graphitelepoxy laminate whose properties were
described in Section A2.1. The stiffeners are composed of a 15-ply graphite/epoxy
laminate with a [45/90/-45/0/45/-45/0/90/0/-45/45/0/-45/90/45] layup. The frame
members are composed of a graphite fabric composite with a thickness of 0.155" and
moduli E,=6.23x106psi, E,=8.89x1O6 psi, G,,=l .66x106 psi, and v,,=O. 144. A 22"-long
line notch, severing both the skin and frame, lies at the center of the panel.
The response of the panel to the four load cases listed in Table A3-1 will be considered.
Load Case 1 consists of a uniform internal pressure of 18.2 psi. This results in the usual
hoop and axial loads present in a cylindrical pressure vessel. Load Case 2 consists of a
uniform internal pressure of 13.65 psi and an axial load of 5,000 lb/in. Load Case 3
consists of an axial compressive load of 1,690 lb/in. Load Case 4 consists of a uniform
internal pressure of 13.65 psi (resulting in the usual hoop and axial loads in a cylindrical
pressure vessel) coupled with an edge shear load of 773 lb/in.

A3.2 Finite Element Model Of the Crown Panel

A finite element model was constructed for the notched crown panel using COSMOSM.
Four-node quadrilateral and three-node triangle, layered shell elements were used
throughout. The development of the finite element model can be simplified by taking
advantage of symmetry in the structure and the loading (except for Load Case 4). We first
note that there is symmetry about a plane along a radial line passing through the notch
parallel to the notch. Thus, only the lower half of the panel in Figure A3-l(b) needs to be
modeled. Symmetry boundary conditions (i.e., zero displacement along an axis normal to
the edge and zero rotations along axes parallel to the edge) were used on all four edges of
the panel. This choice appears to be reasonable for all of the edges except for the lower
edge in Figure A3-1(b). Ideally, the boundary conditions on this edge should be
representative of the conditions on a similar line in the complete cylinder forming the
fuselage. To determine how well the symmetry boundary conditions represent the true
boundary conditions at this edge, an analysis was performed on an enlarged panel with
double the arc length of that in Figure A3-1 for Load Case 1. The hoop stress contours
for the original panel and the enlarged panel are shown in Figure A3-2. It can be seen that
the stress fields are almost identical. In fact, near the notch tip the difference is less than 1

A-47
 fl

ry
c
h
9
C

A-48
 Table A3-1: Load Cases Used in Repair Analysis

Load C a s e Descript- Nx NY NXY Pressure

ion (axial) (hoop) (shear)
lb/in lb/in lb/i n psi
I 1 maximum 1110 2220 0 18.2
pressure
I
acting
alone
2 maximum 5000 1665 0 13.65
axial
load -
2.5g
maneuver
3 maximum 833 1665 773 13.65
shear
load
4 maximum -1690 0 0 0
compress -
ion load

A-4 9
Figure A.3-2: Compwison of Hoop Stress C u ~ ~ ~inuthe
r sOriginal
Panel and the Enlarged Panel

A-50
percent. Thus, we conclude that our choice of boundary conditions should not introduce
significant error in the analysis.
It is also-interestingto note that if a vertical plane parallel to the frames is passed through
the centers of the panels in Figure A3-2, the stress field on the left is virtually a mirror
image of the stress field on the right (near the crack tip the difference is less than 1
percent). Thus, although geometric symmetry does not quite exist about such a plane
(because the frame has a "J" shape), the assumption of such symmetry should still give
reasonably accurate results. Consequently, only one-quarter of the panel was modeled in
the finite element analysis. The mesh for this model is shown in Figure A3-3.
The criterion for failure at the notch tip that was used in the analysis was the Poe-Sova
(1980, 1983) point strain criterion with a critical strain of 0.143 at a characteristic distance
of 0.0585". This criterion was found to work well with this size of notch (Walker et al,
1992). To determine the adequacy of the mesh, a flat, two-dimensional version of the
panel with the stiffeners and frames omitted was modeled under simple tension. The
normalized strain near the crack tip calculated by the finite element analysis was compared
with the analytical solution [Lekhnitskii, 19681. The good agreement between these two
results shown in Figure A3-4 indicates that the density of the mesh is adequate.
In the analysis of repaired coupons (Section A2), it was found that large-deflection
considerations could have a significant effect on the response. To ascertain the
importance of large deflections on the response of the notched crown panel, we performed
both linear and nonlinear (large deflection) analyses on the model shown in Figure A3-3
for Load Case 1. The hoop and axial strains predicted by these two analyses on the inner
and outer surfaces of the panel near the notch tip are shown in Figures A3-5 and A3-6.
The difference between the inner and outer strains is considerably smaller in the nonlinear
analysis than it is in the linear analysis. This indicates that the nonlinear analysis predicts
significantly less bending. Also, for the hoop strain the bending directions are reversed.
The reason for this is illustrated in Figure A3-7 which shows the shape of the deformed
panel predicted by the two analyses. The linear analysis predicts that the panel dishes
inward just ahead of the notch tip. This behavior is absent in the nonlinear analysis.
Figure A3-8 shows the strains at the middle surface of the fuselage skin predicted by the
two analyses. Here, we observe that the difference in response predicted by the two
analyses is considerably smaller than it was for the surface strains. Since the prediction of
failure is based on the middle surface strains, the use of a linear analysis for failure
prediction should result in a small error. Consequently, in arriving at a repair design, a
linear analysis was used to develop the design, and then its adequacy was verified with a
nonlinear analysis.
In developing a repair for the notched crown panel, we will design patches that will carry
load across the severed components. Before developing these patch designs, it is desirable
to reduce the intensity of the stresdstrain riser at the notch tip. This can be done by
blunting the tip by cutting out material to form an elongated hole with rounded ends. For
example, a finite element model for the panel with an elongated hole with a 2"-radius tip is

A-5 1
Figure A3-3: Quarter-Symmetry Model of the Notched Panel

A-52
7.5

ANALYTICAL
m FEM

5.0

2.5

0.0
0.00 0.25 0.50 0.75
DISTANCE FROM CRACK TIP (in.)

Figure A3-4: Comparison of Analytical and Finite Element Analysis

(FEA) Strains Near the Crack Tip in a Flat
UnstiffenedPanel

A-53
0.0300

Q
0.0250 - \
- - - a 7 - - - INNER (L)
t
\

\ - --,I-- - OUTER (L)

0.0200 -
lNNER(NL)
OUTER(NL)

0.0150 -

0.0100 -

0.0050 -
-A,--
-A- - - -A -
0.0000
0.00 0.25 0.50 0.75
DISTANCE FROM CRACK TIP (in.)

Figure A3-5: Hoop Stains on the Inner and Outer Surfaces

Predicted by Linear (L) and Nonlinear (nL) Analyses

A-54
0.0200

..
0.01 50

0.01 00

0.0050

0.0000
-0.00 0.25 0.50 0.75
DISTANCE FROM CRACK TIP (in.)
. .

Figure A3-6: Axial Stains on the Inner and Outer Surfaces

Predicted by Linear (z) and Nonlinear (IVL) Analyses

A-55
 a) linear analysis

b) nonlinear analysis

Figure A3- 7: Defornzed Panel Shapes Predicted by Finite Elentenit

Annlysis

A-56
 0.0250

- - -a-- -
0.0200

0.01 50
- HOOP (NL)
.

--i----AXIAL(NL)

Z
-
Q
u:
i-
CD
0.01 00

0.0050

0.0000
-0.00 0.25 0.50 0.75
DISTANCE FROM CRACK TIP (in.)

Figure A3-8: Hoop and Axial Strains in the Middle Surface of the
Skin Predicted by Linear (L) and Nonlinear (nL)
Analyses

A-57
shown in Figure A3-9. An analysis of this model for Load Case 1 was performed. A
comparison of the strain near the notch tip for the original line notch and for the elongated
hole is shown in Figure A3-10. The strain near the edge of the hole is considerably
smaller than the strain near the tip of the line notch. Specifically, at the characteristic
distance the strain is reduced by 66 percent. Thus, blunting the notch is an important
factor in developing an efficient repair design.

A3.3 Final Repair Designs

A final repair design was amved at which would become the subject of upcoming fbll-
scale crown panel testing. Two additional designs are presented to illustrate the impact of
repair material selection and fbselage design. All three repair designs are closely related.
In pursuit of a practical repair, a general repair approach was developed. The approach
argues that damage cut-out and repair strategies should be incorporated into the original
fbselage design. Also cost-design trade studies are needed to identi& where and when
repair is preferred over factory replacement. Also, damage cut-out shape and repair
building-block strategies are fbselage design dependent.

A3.3.1 General Approach

The general approach recommended below for fuselage repair should be an integral part of
hselage design development to ensure product long-term affordability. It should employ
techniques that are broadly feasible and affordable to the entire family of airline customers.
It should address only that damage for which repair is shown to be cost effective, and the
infinite range of probable damage scenarios must be addressed by a finite hierarchical set
of discrete cut-out repairs. Finally, definition of the discrete cut-out set and corresponding
repair designs should be made part of the delivered aircraft product. A summary of the
approach issues are listed in Table A3-2. The use of pre-cured and stocked laminate
sheets should be employed in single or more layers to transfer load about damage cut-out
areas. Also, fbselage and repair should be joined using mechanical fasteners. These
recommendations are intended to avoid customer material storage and tooling costs as
well as enhance reliability. Patching of fiselage skin should be placed external to the
aircraft, and splicing of the internal substructure (stringers, frames, etc.) should be placed
internally as a practical rule. Repair designs, specific to preconceived damage cut-out
patterns over the various zones of the fuselage, would be developed as part of the aircraft
product. The customer could then associate any potential and critical damage with a
specific predefined cut-out pattern for which a repair design already exists. This method is
obviously intended to minimize customer costs of repair development and aircraft down-
time.
The range of damage scenarios for which cost effective repair is feasible would be
determined by a cost study considering damage probabilities and costs of repair versus
instdling factory mani,ifiactured replacements or aircrai? retirement. This exercise should
define the range of cut-out sizes for the various sections of the hselage for which repair is

A-58
Figure A3-9: Finite Element Model of the Crown Panel with an
Elongated Hole

A-59
 0.0250

0.0200
- CRACK
HOLE

0.O 1.50
Z
Q
a
I-
cn
0.01 00

0.0050

I I
0.0000 I I + I , I

0.00 0.25 0.50 0.75

DISTANCE FROM NOTCH (in.)

Figure A3-10: Comparison of Strain Near the Tip of a 22"-Long

Line Crack and a 22 "-LongElongated Hole

A-60
 Table A3-2: Summary of Approach Issuesfor Fuselage Repair

Damage Repair Approach

9

Incorporate minimum field repair capabilities

mechanical attachments
precured stock elements (patch laminates, composite or metal angles)
e eliminate material shelf life constraints (less waste)
0 reduced refrigeration storage requirements
Predetermine damage cutout pattern
address a range of damage severity
patch size and placement pre-defined relative to the repair cutout
patch size and bolt pattern adjusted to address fuselage location
specific conditions
skin patching external and sub-structure patching internal
Design stock laminate for most severe damage condition
stiffnessconstraints
0 strength requirements
flexibility (single vs. multi-layer patches)
repair vs. factory replacement costs (avoid factory replacement until
very large damage scenarios)

A-6 1
called for. A hrther cost study is needed, however, to determine the best stock laminate
material and respective set of gage thicknesses. It is conceivable that the largest cut-out
repair might employ multiple layers of a stock laminate while the least disruptive cut-out
repair employs only one. Repair and fuselage material stiffness anisotropy should also be
considered at this stage.
The above approach has been referred to as the "Damage Level Approach". As described
above it should be developed from the top-down direction as depicted in Figure A3-11.
However, it would be applied from the opposite perspective; i.e. an incurred damage
would be matched to the smallest pre-defined cut-out pattern. Levels of damage could be
categorized by area of cut-out size, number of structural members involved in cut-out,
and/or number of layers of stock laminate defined in the repair. Specific category
definition was not warranted in the effort described herein. However, a limited example is
described in Table A3-3.

3.3.2 Crown Panel Repair Design

A repair which reflects the above general approach was developed for the fuselage crown
panel and damage defined in Section A3.1. This repair, shown in Figure A3-12, is the
subject an upcoming full-scale pressure box test. The damage (22-inch axial notch) was
cut-out using a general pattern shape. Patching of the skin employed layers of a quasi-
isotropic stock laminate. Splicing of the frame employed yet another stock laminate.
Mechanical fasteners were used to join the repair and fuselage, and no special preparation
or assembly procedures were required.
Two additional repair designs were also developed and are reported herein. An alternative
to the repair described above is distinguished by patch material exhibiting stiffness
anisotropy. The other repair reflects the implications of a repair-friendly fuselage design.
Cut-out Configuration
The same cut-out pattern was employed for all three repair designs mentioned above. As
illustrated in Figure A3-12, the cut-out has an hour-glass shape. This pattern is actually a
union of three general cut-out (block) patterns as follows: two complete skin bay cut-outs
and a joining cut-out of the severed frame and skin assembly. Choice of the block shapes
reflects maximum areas of uniform structure which implies uniform patch design. Block
shapes do not reflect any anticipation of damage type, shape or orientation; thus they are
generalized. Large damage size would be addressed by assembling a union of cut-out
blocks.
The cut-out block of the frame should have included the full breadth of structure between
the stringer flanges as did the skin cut-out block. This would have enabled the arrived-at
repair to address frame damage other than that contained within the central two inches.
The shortened breadth of the version employed in this effort was an artifact of the limited
size of the crown panel employed in testing. The larger frame cut-out would have
necessitated a longer frame splice which would have extended outside of the panel area in

A-62
repair of highest level damage

Figure A3-11: Stock Patch Material Sized for Severe Damage and
Down-Sizedfor Lesser Conditions

A-63
 TableA3-3: Damage Level Approach to Repair Identification

Designation Damage Description I Repair Approach

Level 0 Skin delamination or Fastener restraint
debond from
stiffening elements
Level 1 Critical damage to a Mechanically
single structural fastened patch
element (skin or and/or.splice
stiffener)
I Level 2 (and higher) Multiple occurences
of Level 1 damage
same as Level 1

A-64
 ! I

Figure A3-12: Final Repair Design Configuration

A-65
which flight load conditions could be simulated. The shortened frame cut-out and splice
design were chosen as adequate for demonstration of the repair concept.
External Patches
The two layers of external patching, shown in Figure A3-12, are composed of E-
glass/epoxy fabric lamination. The function of an external patch is to transfer the load that
had been conducted through the removed (cut-out) skin. It also contains kselage
pressure. To avoid complications to surrounding structure, the external patch also serves
to return the damaged area to its original stiffness.
Ideally, the external patch should make a multiaxial stiffness contribution that equals that
of the cut-out hselage structure. In practice this goal is complicated by geometrical
constraints and limited range stiffness modulus available in structural materials. Fuselage
cut-out and repair patch stiffnesses (modulus x area) are listed in Table A3-4. As listed,
the final E-glass fabric patch was more stiff axially and more soft circumferentially than the
cut-out fuselage skin. It should be noted that the listed fabric patch stiffnesses are based
on experimental testing of the actual repair laminate and are 5% less than anticipated.
Depending on how close the original structure was designed to the limits of its allowable
strains, the significance of the patch stiffness can vary. For the ATCAS fuselage crown,
under the ultimate high-g maneuver load condition, little margin existed between nominal
(no damage) axial strain levels and strain allowables that reflected damage tolerance and
fastener bearinghypass strain interaction criteria. Thus, the axial stiffness of the repair
patch was critical. As evident from stiffnesses listed in Table A3-4,the achieved axial
stiffness of the fabric repair patch was significantly higher than that of the cut-out skin.
This was the case even though the employed E-glass/epoxy is the least stiff structural
material available.
The external patching was composed of two layers: a base and a cover patch, as shown in
Figure A3-12. Use of the multi-layered approach was dictated by the need to reduce peak
fastener shear loads at the axial leading edge of the patch. A somewhat low allowable
fastener load was realized due to the limited margin between nominal strains in the
hselage and strain allowables accounting for fastener bearinghypass interaction fracture.
Using the multi-layered patch approach, the base patch could be extended ahead of the
cover patch. In this configuration the leading row of fasteners joins only half the patch
stiffness and thus realizes half the resistance to stretching. Also noted on the sketch in
Figure A3-12 are oblong cut-out holes in each comer of the base patch. Because fasteners
could not be anchored into the closed hat-sectioned stringer cavity, increased duty was
placed on fasteners at adjacent locations. The holes serve to reduce the stiffness of the
combined repair at the leading axial edge over each stringer and thus are an artifact of the
closed nature of the hat-section stringer design.
E-lass/epoxy was chosen as the patch material because its stiffness (lowest modulus
structural material on the market) could best satisfy the stifkess constraints mentioned
above. Patch stiffness is the product of cross-sectional area and modulus of elasticity.

A-66
 Table A3-4: Final Repair Design - Stiffness Comparison of
Damaged Structure Cut-Out vs. Attached Repair

Fuselage cut-out Patch cut-out

Direction Skin E-glass Frame
E x A Fabric WPS)
(MiPS) E x A
W P S1

4Axial 3.30
33.39
8.24
19.94 9.18
-
12.19

A-67
The circumferential cross-sectional width of the patch had to extend to three times that of
the cut-out in order for a second column of fasteners to be anchored into the far stringer
flange either side of the repair. A lesser patch width would be required for stringer
designs other than the hat-section configuration of the present fuselage design. The patch
width constraint combined with the bi-level objective mentioned above would dictate a
low stiffhess (thickness x width x modulus) of each laminate layer. Practical limits exist as
to how thin laminates and sheet metal can be made. Additionally, patch thinning is not
desired from a stability and pressure bulging perspective. E-gladepoxy offered the
lowest modulus of elasticity; and thus, the greatest patch thickness was possible while
satisfying the cut-out stiffness matching criteria. Use of graphite/epoxy or titanium
materials would have dictated a single layered patch approach and thus realized fastener
bearinghypass strain failure at a lower axial load than that required and that achievable
using E-glass/epoxy .
Frame Splice
Design of the frame splice consisted of a four component assembly as depicted in Figure
A3-12. The splice laminate was designed to match the stiffness of the severed woven-
graphite/epoxy frame. Splice members employed a standard graphite/epoxy fabric because
of its general availability, sufficient strengthhtifkess performance, and ability to be draped
and formed to complex curvatures as required of this repair.
Use of multiple splice members was influenced by the benefits of double shear load
transfer from the severed frame via splicing on either side of the frame. Splice member
count could have been reduced using continuous C-sectioned members, but these
encounter practical dimensional tolerance limitations. Not shown in the sketches are shim
plugs placed to occupy the space vacated by the cut-out frame damage.
Alternate Patch Material Design
An alternative to the above final repair design is a repair consisting of external patch
laminate exhibiting stiffness anisotropy. Judicious tailoring of the external patch stiffness
could enable a repair that best matched that of the cut-out fuselage structure and would
thus make possible a repair of greater strength. A repair of this concept was not chosen as
the final design because the chosen fabric patch material does not enable appreciable
stiffness anisotropy. Also, it could be argued that a repair should be of isotropic stiffness
in order to maintain generality to the array of stifkesses realized over the entire fuselage
structure. Although a tape material would have enabled stiffness anisotropy, it could not
be made available within program time constraints, was more costly due to minimum order
conditions and had a lesser track record than that of the fabric material.
The alternative patch design was developed using a ply material of E-glasdepoxy tape.
Resultant patch and cut-out stiffnesses are listed in Table A3-5. The frame design
reflected in this and the following repair eRort did not contain a fai!-safe chord. A!so, the
frame splice utilized a graphite/epoxy tow material laminate equal to that of the skin. As
is evident, the tape patch stiffness is somewhat better than that of the fabric patch which

A-68
 TableA3-5: Optiniized Repair Design - Stvfness Contparison of
Damaged Structure Cut-Out us. Attached Repair

Fuselage Cut-out Patch Cut-out splice

Direction Skin GR/EP Tow Frame Frame
E x A E x A (MiPS) WPS)
WPS) (MiPS)
Axial 3.30 7.77 - -
Hoop 33.39 23.12 6.89 6.21

A-69
was itself 5% better (axially) than predicted from ply property data. Thus, as made
evident in the structural analysis of Section A3.3.3, a tape repair exhibiting stiffness
anisotropy should result in greater strength than an isotropic repair.
The E-gladepoxy tape material, GI913 manufactured by Ciba-Geigy, was chosen for the
alternate design because a good bolted-joint strength database was available (Kretsis and
Matthews, 1985). The reported ultimate bearing strengths (bearing stress at peak load)
were significantly higher than that of the E-glass fabric material used in the final design.
However, bearing yield stresses were lacking from the tape material database. Analysis of
the fabric patch showed bearing yield to be critical. Thus the E-gladepoxy tape material
appears to be a superior material choice but certainty would require additional testing.
Modified Fuselage Repair
A repair design was requested for a modified crown panel which was identical to that of
the other repair efforts except that its skin and stringer laminations were modified to
radically change stiffness characteristics. Impetus for this effort arose from discussion of
fbselage design characteristics that were impediments to repair. The modified fbselage
design was formulated using the ATCAS design-cost trade analysis program COSTADE
and thus was more than a simple example of a repair friendly fbselage design. The
modified fbselage addressed two difficulties encountered in the above repair effort. First,
the margin between nominal (no damageho repair) strains and allowable strains was
increased. Second, the skin was hardened @e., made stiff) axially. This, combined with
the aspect ratio of the damage cut-out, resulted in skin cut-out stiffnesses that better
matched what was feasible in a patch. Cut-out and patch stiffhesses for the modified
fbselage are listed in Table A3-6.
Because of the increased margin between nominal and allowable strains, it was possible to
place a column of fasteners along the perimeter of the cut-out. In the previous designs,
fastener bearinghypass interaction criteria were easily violated for fasteners at these
locations. A second column of fasteners could now be placed into the stringer flange
adjacent to the cut-out. Thus, unlike the previous repair patches, the patch did not have
to be extended to the far stringer flange. The modified fbselage patch was thus much
narrower axially and was thus better able to match the stiffnesses of the cut-out.
The increased margin between nominal versus bypass strain also made possible the use of
a single layer patch design. The greater margin allowed for increased fastener bearing
which made the tiered bi-level patch approach unnecessary. Thickness of the single layer
patch would be twice that of a bi-level patch. Thus, the minimum thickness gage
limitations of graphite/epoxy and titanium were not confronted. Graphite/epoxy tow
laminate, equal to that of the original hselage skin, was chosen for the repair patch due to
its good stiffness matching and an existing database of bolted-joint strengths.

A-70
 Table A3-6: ModiJied Fuselage Repair Design - Stiffness
Comparison of Damaged Structure Cut-Out vs.
Attached Repair
~~

Fuselage cut-out Patch cut-out Splice

Direction Skin GR/EP Tow Frame Frame
E x A E X A (MiPS) WPS)
(MiPS) (MiPS)
Axial 8.31 10.07
Hoop 21.49 23.7 6.89 6.21

A-7 1
A3.3.3 Structural Analysis
All structural analysis was based upon solutions to finite element analysis (FEA). Both
linear and nonlinear (large deflection) FEA were performed. Large deflection E A did not
allow for interactive data retrieval and was thus used only for results found to be of
significant variation from linear FEA. The finite element model employed 114 symmetry
and is shown in exploded form in Figure A3-13. Details of the finite element modeling
technique can be found in Section A3 2.
The designs were analyzed for strength under ultimate load conditions listed in Table A3-
1. Analysis calculations have not been included in this report for brevity. Analysis of the
three repair designs identified margins-of safety for various criteria. The resultant
margins-of-safety are summarized in Table A3-7. Criteria for bearinglbypass strain
interaction were based upon interaction envelopes provided by Boeing. Cut-out strength
criteria was based upon the Poe-Sova point strain fracture analysis method for composites.
Laminate strength criteria addressed peak strain concentration as well as basic strain
levels, for which allowables ensured tolerance to unobtrusive damage. Tolerance to large
damage, a fail-safe load criteria, was factored for inclusion in ultimate load condition
analysis. Stability was confirmed for the ultimate loads, but no effort was made to obtain
final buckling loads.
Bearing/Bypass
The interaction of bearing stress and bypass strain was considered in analysis of the
mechanically fastened composite joints. In particular the joined graphite/epoxy laminates
were analyzed as such; however, the E-glass/epoxy laminates were not since their fracture
toughness far exceeded the hselage strain levels of interest. Thus, E-glasdepoxy
laminates were analyzed for pure bearing failure.
The analysis addressed the interaction of bearing with tensile bypass strain only. The
ultimate compression load condition did not induce strain levels sufficient to warrant
concern. The tensile bypass strain and bearing interaction envelopes were made available
by Boeing. For brevity only the envelope corresponding to bearing and bypass in the axial
direction is shown in Figure A3-14. Based upon limited coupon testing, a Boeing
computer program extrapolated to define a continuous iteration envelope. The program
analysis method was based upon calculating stress concentration about a fastener hole
undergoing bearing and bypass strain. The resultant envelope was then truncated at a
level of bearing strength considered realistic. Boeing provided mean, room temperature,
dry interaction envelopes. OSU applied 0.8 reduction factor to the interaction portion of
the envelope which reflected €3-basis, environmental effects. The 0.8 factor and the B-
basis bearing cut-off stress were derived from previous Boeing, B-basis, interaction
envelope definition for the same material. The effect of fastener diameter versus edge
margin and bolt spacing were taken into account via a knockdown factor.

A-72
 EX

FRCiME

Figure A3-13: Finite Element Modelfor Crown Panel Repair

A-73
 Table A3- 7: Sumntnry of Repair Margins of Safety

Graphite/Epoxy
I

I I
I I I

criteria
I Final Repair Alternate
Repair Fuselage

I1 Bearing/By-
Pass
Basic Strain I
-0.02

-0.14
II -0.01

-0.12
II +o. 10

+0.34
Peak Strain +0.03 +O. 07 f0.25
Damage +O. 04 +O. 06 -
Tolerance
I cutout I +1.03 I +0.92 I +2.3

Buckling None None I None

Glass ‘Epoxy
Final Repair Alternate Alternate
Repair Fuselage
-0.18 +O. 23

+-
-0.30
xDesign
Shear-out +0.21 I -
Pull-Through

A-74
 Bearing. / Bv-Pass Interaction Envelope - Skin

*Graphite/epoxy: AS4/938 Tow

* Layup: 45,'-45/90/0/60/-60/90/-60/60/0/90/-45/45
* Widthhob diameter ratio - w/d = 6
* By-pass strain in 90 degree (axial) direction
* Bearing load orientation: alpha = 0 degrees (axial)

Allowable (B-Basis, sever environment)

R.T.D., Mean

l3gwe AZ-14; Fxanlple BearingLBypass Strain Interaction Envelope

Allowable

A-75
Cut-out
The Poe-Sova point-strain fracture criterion discussed in Section A3.2 was applied in the
analysis of the cut-out of each design. All design cut-outs strengths exceeded the criterion
by at least a factor of 1.92. The point-strain criteria makes the allowance for damage
development in the zone of peak strain. In doing so, it is more representative of fracture
in composites than methods based upon LEFM stress intensity factorization. The
identified characteristic dimension, do = 0.0585", and critical strain of 0.143 were applied
at all locations along the cut-out edge.
It should be recognized that the point strain criterion is an ultimate strength criteria and
does not attempt to address damage tolerance. Therefore, such issues as edge
delamination, impact, etc. were not addressed. Boeing ATCAS has applied a damage
tolerance based criterion to such details as window cut-outs. This criterion was not
utilized by direction of Boeing ATCAS.
Laminate Strength
Strength criteria were applied to the repaired fuselage skin and stringer laminates. The
analysis applied to the stringer identified insufficient strength at ultimate loads. However,
this apparent strength deficiency is also associated with the nominal, no-damageho-repair
fuselage. The criteria addressed tolerance to concentrated strain and hidden damage. No
data for such criteria was available for the frame members. However, in light of their low
strain, their safety was of little concern. Strength analysis calculation for the E-
glasdepoxy patch laminates was also not made. This was primarily due to the lack of
concern considering the high strain capability of the material. "B" basis tensile strain
allowables of 0.01 surpassed all strains developed in the patch under ultimate load
conditions.
A basic strain criterion was applied to the repaired fbselage skin and stringer laminates.
The criterion addressed tolerance to hidden damage, fastener holes and extreme
environments. It was intended not to address concentrated strain areas or flexure. In.
regards to the axially directed strength, a skin laminate allowable of 5575 microstrain was
satisfied for all conditions. However, the stringer allowable of 4660 microstrain was
exceeded under the ultimate high-g maneuver load condition. The basic strain margin-of-
safety for the stringer was M.S.=-0.14. However, the nominal fbselage would experience
an average axial strain of 4960 microstrain for this load condition and thus realized a
margin-of-safety of as least M. S.=-0.06, The allowables were calculated from formulae
provided by Boeing specific to the fbselage material AS4/938 Tow.
A peak strain criterion was also applied to the repair fuselage. The criterion applied "B"
basis unnotched laminate strength to strain concentrations including that associated with
flexure. The criterion was not applied in the vicinity of the cut-out. Allowables in the
critical axial direction of 7680 microstrain for the skin laminate were satisfied uniformly.
Calculations were not applied to the stringer considering that it experienced little variation
of strain and would have realized a significant positive margin-of-safety.

A-76
Damage Tolerance
Damage tolerance was addressed for two levels of damage defined as detectable and
nondetectable. Tolerance to nondetectable damage was incorporated into the ultimate
basic-straidstrength criteria discussed above. Detectable damage definition varied from
that of visible impact indentation to complete structural unit failure. With respect to the
fuselage crown area, the critical structural unit of concern was that of two adjacent skin
bays plus central stringer or fiame member. Tolerance to detectable damage was treated
as a safety-of-flight load criterion. The analysis was conducted by Boeing for the nominal
hselage design and resulted in allowables for far-field strain levels. These allowables were
factored up to reflect ultimate load conditions analyzed in the repair design. Repair
damage tolerance analysis then determined whether the disturbed fuselage strain field
exceeded the damage tolerance allowable.
Damage tolerance analysis of repaired hselage indicated a minimum margin-of-safety of
MS = +0.04 for axially directed strain under the ultimate high-g maneuver load condition.
In general, the repair damage tolerance analysis focused on the fbselage structure just
outside of the repair area. At these areas one can expect the greatest disturbance of the
nominal strain field. Axial and hoop strains were determined over cross-sections in these
areas using finite element analysis results. The cross sections included skin of one bay
length positioned symmetrically over the substructure member of concern. Strains over
the cross-section varied, and an average was arrived at for comparison to the allowable.
Averaging implies a weighting of terms, and in this case strain was weighted according to
the stiffness of the laminate. Thus, strain of the hard stringer laminate was given
proportionately higher weighting than that of the skin.
Stability
Stability of the repaired fbselage was analyzed for the ultimate conditions generating an
axial compressive load of -1686 lbs./in. Stability was confirmed for all repair designs
considered. Further analysis sought to determine the load level at which initial buckling of
the skin or patch laminate would occur. Skin buckling was generally initiated at 113
percent of limit load. The modified fbselage was an exception since no initial buckling
was predicted below the ultimate load mentioned above.
Stability analysis was based upon nonlinear, large deflection finite element analysis. The
incremental load level at which the model stiffness matrix becomes singular was taken as
the instability. Compressive loading of the model was incremented up to the ultimate load
identified above. Initial buckling of the skin would produced nonconvergence of the
analysis. This condition was recognized for what is was, and its load level noted. To load
beyond this level, mode-one skin buckling shape was given a small assistance which
enabled convergence of solution iteration. Loading was then incremented up to the
ultimate load defined above.

A-77
would be identifiable. Repair components and strategy would be defined by design costs
studies to ensure return of ultimate strength for the even the most extreme damage
condition while minimizing costs associated with the more probable small damage
conditions. It was envisioned that multiple layers of stock material could be assembled to
address severe cut-out conditions. A simple reduction of layers would then address less
severe damage. The procedure for assembling patch layering could then be defined in a
building-block/damage level framework.
Repair in general should necessitate only those tools and skills that are widely available
within the customer base. Thus, mechanical attachment would be employed over adhesive
bonding for fkselage/repair joining. Repair material should be available in standardized
stock forms to avoid costs of composite laminate fabrication and perishable material
storage. A final design, reflective of the above approach, was arrived at and analyzed for
safety under ultimate load conditions. The 22" axial notch was removed in block cut-out
fashion. A simple laminate of quasi-isotropic stiffness was employed for external skin
patching in a bi-level manner. Frame patching also used a simple laminate intended as
stock material.
Protruding head fasteners were used throughout the repair. Return of the damaged
fkseIage to ultimate strength was not achieved from a severe environment, "B" basis,
standpoint. A minimum margin-of-safety MS = -0.18 was identified and associated with
bearing failure of the external base patch of E-glass/epoxy. Close behind was a MS =
-0.14 associated with basic strain of the stringer laminate, followed by a MS = -0.02 for
bearinflypass failure of the fuselage. All of these negative margins-of-safety were
associated with the ultimate high-g maneuver load condition. No negative margins were
associated with the remaining load conditions. An additional criteria was imposed on the
bearing strength of the E-glass/epoxy in that it should not yield at load levels of 115
percent of limit. Margins-of-safety associated with this condition under severe
environmental, "B", basis were MS = -0.3.
Inability to derive a repair design which could return the fbselage to ultimate strength was
directly tied to the small margin existing in the nominal fuselage design between
operational strains and allowables. In other words, the nominal fiselage was designed so
close to the allowables for basic strain, damage tolerance, etc., that it could accept little
perturbation of its strain field due to damage cut-out and repair.
Further analysis showed that optimization of the above repair was possible using a patch
laminate exhibiting stiffness anisotropy and a higher performing E-glass/epoxy material.
However, applicability of any specific stiffness anisotropy to fhelage repair in general is
open to question. A minimum margin-of-safety for the optimum design was MS = -0.16
associated with basic ultimate strain of the stringer laminate, followed by MS = -0.01
associated with bearinghypass of the firselage laminate. It should be noted that the basic
u'ltimate strain criterion applied to the nomina! fuselage stringer should alse preduse
negative margins-of-safety.

A-79
The nominal hselage design was arrived at without the benefit of lessons learned in this
repair study. An alternate fuselage design, derived by Boeing design-cost studies, was
identified that seemed better suited to repair. In contrast to the nominal fuselage, its
dominate stiffness was aligned with the dominant axial load and thus reduced peak
operational strains. This allowed fasteners to transfer greater loads without violating
bearingbypass criteria and for them to be positioned where it was previously not possible.
Additionally, the combination of the alternate skin stiffness anisotropy and the aspect ratio
of the bay cut-out shapes made patch stiffness replacement more attainable. The resultant
repair was more simple in that it required only one layer of external patch laminate and
was of significantly reduced size. It also produced all positive margins-of-safety.
Additional fbselage design improvement, fiom a repair perspective, would do away with
hat-sectioned stiffener configurations.

A-80
 A5.0 REFERENCES

Adkins, D.W. and Pipes, R.B., 1985, "End Effects in Scarf Joints," Composites Science
and Technology,Vol. 22, pp. 209-221.
Bair, D.L., Hudson, P.O., and Ghanimati, G.R., 1991, "Analysis and Repair of Damaged
Composite Laminates," 36th International SAMPE Symposium, pp. 2264-2278.
Baker, A.A., 1986a, "JoiningAdvanced Fiber Composites," CompositeMaterialsfor
Aircraft Structures, Hoskin, B.C., and Baker, A.A. (Eds.), AI& pp. 115-139.
Baker, A.A., 1986b, "Repair of GraphiteEpoxy Composites," CompositeMaterialsfor
Aircrafr Structures, Hoskin, B.C., and Baker, A.A. (Eds.), AIAA, pp. 193-216.
Bohlmann, R.E., Renieri, G.D., and Libeskind, M., 1981, "Bolted Field Repair of
GraphiteEpoxy Wing Skin Laminates," Joining of CompositeMaterials, ASTM STP
749, pp. 97-1 16.
Bohlmann, R.E., Renieri, G.D., and Riley, B.L., 1986, "Bolted Field Repairs Subjected to
Biaxial or Shear Loads," CompositeMaterials: Testing and Design, ASTM STP
893, pp. 34-47.
Busch, J.D. and Dompka, R.V., 1991, "V-22 Composite Repair Development Program:
Evolution of a Permanent Repair Design," Third DoD/NASA Composite Repair
Technology Workshop, pp. 24 1-290.
Chandra, R. and Subramanian, A., 1989, "Stress-Intensity Factors in Plates with a Partially
Patched Central Crack," Experimental Mechanics, Vol. 29, pp. 1-5.
Cripps, D.B., 1984, "Developmentof a Field Repair Technique for Mini-Sandwich
KevlarEpoxy Aircraft Skin," M.S . Thesis, Naval Postgraduate School.
Deaton, J.W., 1991, "Evaluation of Repair Durability for GraphiteEpoxy Laminates after
7 Years of Outdoor Exposure, Third DoD/NASA CompositesRepair Technology
I'

Workshop, pp. 393-418.

Dehm, S. and Wurzel, D., 1989, "Fast, In-Situ Repair of Aircraft Pane1 Components,"
Journal of Aircraft, Vol. 26, pp. 476-48 1.
Dodd, S.M. and Sandow, F., 1991, "Optimum Repair Design for Aircraft Battle Damage
Repair," Third DoD/NASA Composites Repair Technology Workshop, pp. 148-167.
Hall, S.R.,Raizenne, M.D., and Simpson, D.L., 1989, "AProposed Composite Repair
Methodology for Primary Structure," Composites, Vol. 20, pp. 479-483.

A-8 1
Hoehn, G. and Ramsey, C., 1986, "Test Verification of Improvements to Bolted Repair
Design Methodology," Second Dod/NASA Composites Repair Technology
Workshop, pp. 341-359.
Hunter, M.W., 1990, "Modeling and Analysis of Orthotropic Plate with Circular Cutout
and Adhesive Bonded Patch," M.S. Thesis, Air Force Institute of Technology.
Jones, R. and Callinan, R.J., 1979, "Finite Element Analysis of Patched Cracks," Journal
of StructuralMechanics, Vol. 7, pp. 107-130.
Jones, R. and Callinan, R.J., 1981, "A Design Study in Crack Patching," Fiber Science
and Technology, Vol. 14, pp. 99-1 11.
Labor, J.D., 1981, "Composite Repair Concepts for Depot Level Use," 26th National
SAMPE Symposium, pp. 705-715.
Labor, J.D. and Myhre, S.H., 1979a, "Repair Guide for Large Area Composite Structure
Repair," AFFDL-TR-79-3 039.
Labor, J.D, and Myhre, S.H., 1979b, "Large Area Composite Structure Repair," AFFDL-
TR-79-3040.
Lin, K.Y. and Wang, F.X., 1988, "Development of Composite Repair Techniques,"
Boeing Advanced Systems Company, Contract No. GU294 1.
Kretsis, G. and Matthews, R.J., 1985, "The Strength of Bolted Joints in Glass FiberEpoxy
Laminates," Composites, pp. 92- 102.
Mahon, J. and Candello, J., 1981, "Development of Large-Area Damage Repair
Procedures for the F- 14A Horizontal Stabilizer,"26th National SAMPE Symposium,
pp. 825-835.
Maman, S., 1989, ''Composite Material Repair and Reliability," M.S. Thesis, Naval
Postgraduate School.
Manno, A., 1981, "Review of Navy Composite Repair Program," 26th National SAMPE
Symposium, pp. 695-704.
Myhre, S.H., 1981, "Advanced Composite Repair-Recent Developments and Some
Problems," 26th National SAMPE Symposium, pp. 716-727.
Myhre, S.H. and Kiger, R.W., 1980, "Problems and Options in Advanced Composite
Repair," Composites in Structural Design, Lenoe, E.M., Oplinger, D.W., and Burke,
J.J. (Eds.), Plenum Press, pp. 354-380.
Paul, J. and Jones, R., 1989, "Analysis and Repair of Impact Damaged Composites,"
Aeronaiilizal Research La3oiatorj;, ARL-STRJC-R-43 5 .

A-82
Poe, C.C., Jr. and Sova, J. A., 1980, "Fracture Toughness of BorodAluminum Laminates
with Various Proportions of 0" and k45" Plies," NASA Technical Paper 1707, 1980.
Poe, C.C., Jr., 1983,"AUnifying Strain Criterion for Fracture of Fibrous Composite
Laminates," Engzneering Fracture Mechanics, Vol. 17, pp. 153- 171.
Polland, D.R. and Swanson, G.D., 1993, Private Communication.
Reisdorfer, D.A., 1992, "Evolution of Permanent Composite Repair Designs," 24th
International SRMPE Technical Conference,pp. "79 l-TSU5.
Renton, W.J. and Vinson, J.R., 1977, "Analysis of Adhesively Bonded Joints Between
Panels of Composite Materials," ASME Journal of AppliedMechanics, Vol. 44, pp.
101-106.
Russell, S.G. and Kan, H.P., 1991, "Rapid Analysis Procedures for Combat-Damaged
Composite Aircraft Structures," Third DoD/NASA Composites Repair Technology
Workshop, pp. 1210-146.
Siener, M.P., 1992, "Stress Field Sensitivity of a Composite Patch Repair as a Result of
Varying Patch Thickness," CompositeMaterials Testing and Design, ASTM STP
1120, pp. 444-464.
Shyprykevich, P., 1986, "Design and Analysis of Field Repairs for High-Strain Composite
Wing," Second DoD/NASA Composite Repair Technology Workshop, pp. 3 15-337.
Shyprykevich, P., Rosenzweig, E., and Corrigan, M., 1991, "Bolted Repairs for High-
Strain Wing: Analysis and Test Validation," Third DoD/NASA Composite Repair
Technology Workshop, pp. 291-316.
Smith, P.J., Bodine, J.B., Preuss, C.H., and Koch, W.J., 1992, "Design, Analysis, and
Fabrication of a Pressure Box Test Fixture for Tension Damage Tolerance Testing of
Curved Fuselage Panels," Third NASA A&anced Technoloo Conference,NASA CP
3154, pp. 789-906.
Trabocco, R.E., Donnellan, T.M., and Williams, J.G., 1988, "Repair of Composite
Aircrafi," Bonded Repair of Aircrap Structures, Baker, A.A. (Ed.), Martinus Nijhott,
pp. 175-211.
Walker, T.H., Ilcewicz, L.B., Polland, D.R., and Poe, C.C., Jr., 1992, "Tension Fracture
of Laminates for Transport Fuselage - Part 11: Large Notches," n i r d NASA
Advanced Technology Conference,NASA CP 3 178, pp. 727-758.
Whitney, J.M. and Nuismer, R.J., 1974, "Stress Fracture Criteria for Laminated
Composites Containing Stress Concentrations, Journal of Composite Materials,
Vol. 8, pp. 253-265.

A-83