Optimisation and Improvement of the Design of Scarf Repairs to Aircraft
Alex Harman BE (Aero. Hons.), GradIEAust
A thesis submitted in fulfilment of the degree of Doctor of Philosophy
School of Mechanical and Manufacturing Engineering The University of New South Wales
August 2006
The work presented in this thesis was conducted as part of a combined research program with the Cooperative Research Centre for Advanced Composite Structures (CRC-ACS) and the Defence Science and Technology Organisation (DSTO). PLEASE TYPE THE UNIVERSITY OF NEW SOUTH WALES Thesis/Dissertation Sheet
Surname or Family name: Harman
First name: Alex Other name/s: Bruce
Abbreviation for degree as given in the University calendar: Phd.
School: Mechanical and Manufacturing Engineering Faculty: Engineering
Title: Optimisation and improvement of the design of scarf repairs to aircraft
Abstract 350 words maximum: (PLEASE TYPE)
Flush repairs to military aircraft are expected to become more prevalent as more thick skin composites are used, particularly on the surface of the fuselage, wings and other external surfaces. The use of these repairs, whilst difficult to manufacture provide an aerodynamic, “stealthy” finish that is also more structurally efficient than overlap repairs.
This research was undertaken to improve the design methodology of scarf repairs with reduced material removal and to investigate the damage tolerance of scarf repair to low velocity impact damage. Scarf repairs involve shallow bevel angles to ensure the shear stress in the adhesive does not exceed allowable strength. This is important when repairing structures that need to withstand hot and humid conditions, when the adhesive properties degrade. Therefore, considerable amounts of parent material must be machined away prior to repair. The tips of the repair patch and the parent laminate are very sharp, thus a scarf repair is susceptible to accidental damage.
The original contributions include: x Developed analytic means of predicting the stresses within optimised scarf joints with dissimilar materials. New equations were developed and solved using numerical algorithms. x Verified using finite element modelling that a scarfed insert with dissimilar modulus subjected to uniaxial loading attracted the same amount of load as an insert without a scarf. As such, the simple analytic formula used to predict load attraction/diversion through a plate with an insert may be used to predict the load attraction/diversion into a scarf repair that contains a dissimilar adherend patch. x Developed a more efficient flush joint with a doubler insert placed near the mid line of the parent structure material. This joint configuration has a lower load eccentricity than external doubler joint. x Investigated the damage tolerance of scarf joints, with and without the external doubler. The results showed that scarf joints without external doublers exhibited a considerable strength reduction following low velocity impact. Based on the observations, the major damage mechanics in the scarf joint region following impact have been identified. These results demonstrated that it is important to incorporate damage tolerance in the design of scarf repairs.
Declaration relating to disposition of project thesis/dissertation
I hereby grant to the University of New South Wales or its agents the right to archive and to make available my thesis or dissertation in whole or in part in the University libraries in all forms of media, now or here after known, subject to the provisions of the Copyright Act 1968. I retain all property rights, such as patent rights. I also retain the right to use in future works (such as articles or books) all or part of this thesis or dissertation.
I also authorise University Microfilms to use the 350 word abstract of my thesis in Dissertation Abstracts International (this is applicable to doctoral theses only).
…………………………………………………………… ……………………………………..……………… ……….……………………...…….… Signature Witness Date
The University recognises that there may be exceptional circumstances requiring restrictions on copying or conditions on use. Requests for restriction for a period of up to 2 years must be made in writing. Requests for a longer period of restriction may be considered in exceptional circumstances and require the approval of the Dean of Graduate Research.
FOR OFFICE USE ONLY Date of completion of requirements for Award:
THIS SHEET IS TO BE GLUED TO THE INSIDE FRONT COVER OF THE THESIS Abstract
Flush repairs to military aircraft are expected to become more prevalent as more thick skin composite structures are used, particularly on the surface of the fuselage, wings and other control surfaces. The use of these repairs, whilst difficult to manufacture provide a more aerodynamic, “stealthy” finish that is also more structurally efficient than standard overlap repairs. A lot of design work has been conducted to improve the state of the art in scarf repair design, but some significant shortcomings were identified that led to this research project.
This research was undertaken to improve the design methodology of scarf repairs with reduced material removal and to investigate the damage tolerance of scarf repair to low velocity impact damage. Scarf repairs often involve very shallow bevel angles to ensure the shear stress in the adhesive does not exceed allowable strength. This is particularly important when repairing structures that need to withstand hot and humid environmental conditions, when the adhesive properties can degrade to 50% of its room temperature values. As such, a large amount of “good” parent material must be machined away prior to the application of flush repairs. With the tips of the bevelled repair patch and the parent laminate being very sharp, a scarf repair is susceptible to accidental damage.
The original contributions of this thesis include x Developed a validated analytic means of predicting the adhesive stresses within optimised scarf joints with dissimilar materials. New governing equations were developed and solved using finite difference algorithms. These equations were validated with finite element methods. x Verified using finite element modelling that a scarfed insert with dissimilar modulus subjected to uniaxial loading attracted the same amount of load as an insert without a scarf. As such, the simple analytic formula used to predict load attraction/diversion through a plate with an insert may be used to predict the load attraction/diversion into a scarf repair that contains a dissimilar adherend patch. x Developed a more efficient flush joint with a doubler insert placed near the mid-line of the parent structure material. This joint configuration has a lower load eccentricity than external doubler joint, thus providing a stronger joint. Experiments and computational analyses have been carried out to assess suitable failure criteria predicting the strength of flush joints. x Investigated the damage tolerance of scarf joints, with and without external doubler. The results showed that scarf joints without external doubler exhibited a considerable strength reduction following low velocity impact. Based on the experimental observations, the major damage mechanics in the scarf joint region following impact have been identified. These results demonstrated that it is important to incorporate damage tolerance in the design of scarf repairs.
i List of papers and Reports
1. Harman, A., (2004), Review of literature pertaining to the design and experimentation of aircraft scarf repairs. CRC-ACS TM-05086. 2. Harman, A., (2004), Testing of thick adherend bonded scarf joint 2D test coupons. CRC-ACS TM 04x25. 3. Harman, A. and C. Wang, (2005), Analytic and finite element stress predictions in two-dimensional scarf joints (Paper 364), Proceedings of 11th Australian International Aerospace Congress. 2005. 4. Harman, A.B., Wang C.H., and D. Kelly, (2006), Improved flush repair design for aerospace structures using modified biscuit joints, Proceedings of Structural Integrity and Failure, Sydney. 5. Wang, C.H. and A.B. Harman, (2006), Optimal Scarf Joint between unmatched adherends, Proceedings of SAMPE 2006, Long Beach, CA, USA. 6. Harman, A.B. and Wang, C.H., (2006), Improved design methods for scarf repairs to highly strained composite aircraft structure. Composite Structures, Vol. 75, 132- 144. 7. Harman, A.B. and Wang, C.H., (2007), Damage tolerance of composite scarf joints, Abstract submitted to ICCM-16.
ii Originality Statement
I hereby declare that this submission is my own work and to the best of my knowledge it contains no materials previously published or written by another person, or substantial proportions of material which have been accepted for the award of any other degree or diploma at UNSW or any other educational institution, except where due acknowledgement is made in the thesis. Any contribution made to the research by others, with whom I have worked at UNSW or elsewhere, is explicitly acknowledged in the thesis. I also declare that the intellectual content of this thesis is the product of my own work, except to the extent that assistance from others in the project’s design and conception or in style, presentation and linguistic expression is acknowledged.
Alex Harman
iii Copyright Statement
I hereby grant the University of New South Wales or its agents the right to archive and to make available my thesis or dissertation in whole or part in the University libraries in all forms of media, now or here after known, subject to the provisions of the Copyright Act 1968. I retain all proprietary rights, such as patent rights. I also retain the right to use in future works (such as articles or books) all or part of this thesis or dissertation. I also authorise University Microfilms to use the 350 word abstract of my thesis in Dissertation Abstract International (this is applicable to doctoral theses only). I have either used no substantial portions of copyright material in my thesis or I have obtained permission to use copyright material; where permission has not been granted I have applied/will apply for a partial restriction of the digital copy of my thesis or dissertation.'
Alex Harman
iv Authenticity Statement
I certify that the Library deposit digital copy is a direct equivalent of the final officially approved version of my thesis. No emendation of content has occurred and if there are any minor variations in formatting, they are the result of the conversion to digital format.’
Alex Harman
v Acknowledgements
The work herein was completed over a three year period with the help and support of many individuals within several organisations. The work was jointly funded and supported by the Defence Science and Technology Organisation (DSTO) and the Cooperative Research Centre for Advanced Composite Structures (CRC-ACS). The author would like to thank my supervisors: Dr. Chun Wang (DSTO) and Prof. Don Kelly (UNSW), who helped the author to establish the key research objectives and to formulate the research strategy/approach.
In addition, the author would like to thank other key contributors to the work program, including many DSTO colleagues: Richard Bartholomeusz, Roger Vodicka, Rowan Geddes, Howard Morton, Noel Goldsmith, David Parslow, Paul Chang, Aaron Charon, Eudora Yeo, John Van Den Berg, Dr. Paul Callus, Dr. John Wang, Dr. Andrew Rider, Peter Chalkley, Ken Houghton Marilyn Anderson, Leo Mirabella, John Nugent, Ivan Stoyanovski and Dr. Richard Chester. The author would also like to thank the following CRC-ACS colleagues: Prof. Israel Hertzberg, Dr. Alan Baker, Daniel Bitton, Andrew Gunnion, Henry Li, and Ken Houghton. Working with these colleagues at Fishermans Bend made the time spent on this project very enjoyable.
The author would like to thank my wife, Nicola, dearly for the support and love she has given and continues to give to both the kids and me. During the course of this work, both of my two children were born, which put considerable burden on my wife, while I had to spend a lot of mental energy away from my family.
Finally, the author would like to dedicate this project to the loving memory of my Grandma, Mrs. Jean Stebbins. She provided much love in her lifetime to many people of which I was a fortunate recipient at critical development times in my life.
vi Table of Contents
1. INTRODUCTION ...... 1 1.1 Summary...... 1 1.2 Aims of the Research...... 4 1.3 Layout of the thesis...... 5
2. OVERVIEW OF SCARF REPAIR STUDIES FOR COMPOSITE AIRCRAFT STRUCTURE...... 6 2.1 General...... 6 2.2 Introduction ...... 6 2.3 Analytic methods ...... 10 2.4 Finite element analyses methods...... 16 2.4.1 Two dimensional scarf joint...... 16 2.4.2 Three dimensional scarf patch...... 17 2.5 Failure mechanics of the scarf joint and repair...... 19 2.5.1 Two dimensional scarf joint...... 19 2.5.2 Three dimensional scarf patch sub component testing...... 21 2.6 Repair shaping to minimise bond-line stress...... 22 2.7 Scarf joint reinforcement methods...... 23 2.7.1 Overlap doublers...... 23 2.7.2 Through thickness pins...... 24 2.7.3 Internal patches...... 24 2.8 Impact resistance and damage tolerance ...... 25
3. SHAPE OPTIMISATION OF SCARF REPAIRS...... 27 3.1 Introduction ...... 27 3.1.1 General...... 27 3.1.2 Analytic solutions...... 28 3.1.3 Finite element solutions...... 29 3.1.4 Experimental Testing ...... 29 3.2 Reduced Modulus Insert Analyses...... 29 3.2.1 General...... 29 3.2.2 Isotropic plate with round inserts...... 30 3.2.2.1 Analyses Description ...... 30 3.2.2.2 Results and Discussion ...... 31 3.3 Scarf Joint Equation Formulation and Solution ...... 35 3.3.1 Formulation...... 35 3.3.2 Solution ...... 39 3.3.3 Optimisation...... 39 3.4 Scarf Joint Finite Element Analyses...... 40 3.4.1 Model descriptions...... 40 3.4.1.1 Isotropic adherend scarf joint ...... 40 3.4.1.2 Orthotropic composite adherend scarf joint...... 41 3.5 Experimental Tensile Testing...... 43
vii 3.5.1 General...... 43 3.5.2 Specimen Descriptions...... 44 3.5.2.1 Phase 1: Linear scarf joint between identical modulus isotropic adherends (Al.2024T3, FM300) ...... 44 3.5.2.2 Phase 1: Linear scarf joint between identical modulus orthotropic adherends (21 ply T300-914c, FM300)...... 46 3.5.2.3 Phase 2: Linear scarf joint between identical modulus isotropic adherends (Al.7075T6, FM300) ...... 47 3.5.2.4 Phase 2: Linear scarf joint between dissimilar modulus isotropic adherends (Al.7075T6/Ti6Al4V, FM73) ...... 48 3.5.2.5 Phase 2: Optimised scarf joint between dissimilar modulus isotropic adherends (Al.7075T6/Ti6Al4V, FM73) ...... 49 3.5.3 Testing Procedure...... 49 3.5.3.1 Test Machines...... 50 3.5.3.2 Data acquisition systems ...... 52 3.5.3.3 Extensometers ...... 53 3.5.3.4 Strain Gauges ...... 56 3.5.3.4.1 Far field adherend strain ...... 56 3.5.3.4.2 Adherend tip strain...... 57 3.5.3.4.3 Strain gauge location summary...... 58 3.6 Results and Discussion...... 58 3.6.1 Scarf Joint Analytic and FE Results...... 58 3.6.1.1 Isotropic adherend scarf joint ...... 58 3.6.1.2 Orthotropic composite adherend scarf joint...... 63 3.6.2 Tensile Testing of Scarf Joint Coupons...... 67 3.6.2.1 Typical Failure Modes ...... 67 3.6.2.1.1 Metallic scarf joint specimens ...... 67 3.6.2.1.2 Carbon-epoxy composite scarf joint specimens ...... 70 3.6.2.2 Stresses and Strains ...... 71 3.6.2.2.1 Phase 1 metallic scarf joint specimen results ...... 73 3.6.2.2.2 Phase 2 metallic scarf joint results...... 75 3.6.2.2.3 Phase 1 composite scarf joint specimen results...... 76 3.7 Conclusions and Recommendations for Future Work...... 78 3.7.1 Conclusions ...... 78 3.7.1.1 Analytic and Finite Element Analyses ...... 78 3.7.1.2 Tensile Testing ...... 79 3.7.2 Recommendations for Future Work ...... 79 3.7.2.1 Analytic and Finite Element Analyses ...... 79 3.7.2.2 Testing...... 80
4. REINFORCEMENT METHODS FOR SCARF JOINT STRENGTH IMPROVEMENT 81 4.1 General...... 81 4.2 Through Thickness Pinning ...... 81 4.2.1 Introduction...... 81 4.2.2 Pin type ...... 81 4.2.3 Pin layout...... 81
viii 4.2.4 Method of pin insertion...... 82 4.2.5 Results and Discussion ...... 83 4.2.5.1 Typical Failure Mode ...... 83 4.2.5.2 Stresses and Strains ...... 84 4.3 Modified Biscuit Joints within Scarf Joints...... 85 4.3.1 Introduction...... 85 4.3.2 Phase 1 and 2 Specimen Descriptions...... 86 4.3.2.1 Phase 1 Pilot Experimental Study ...... 86 4.3.2.2 Phase 2 Experimental Study ...... 89 4.3.3 Phase 3 FEM Details...... 90 4.3.4 Phase 1 and 2 Experimental Results and Discussion ...... 92 4.3.4.1 Phase 1 Pilot experimental study ...... 92 4.3.4.2 Phase 2 Experimental study...... 94 4.3.5 Phase 3 Finite Element Results and Discussion ...... 99 4.4 Conclusions and Recommendations for Future Work...... 105 4.4.1 Conclusions ...... 105 4.4.1.1 Through thickness pinning ...... 105 4.4.1.2 Modified Biscuit Joints within Scarf Joints ...... 105 4.4.1.2.1 Phase 1 Pilot Experimental Study ...... 105 4.4.1.2.2 Phase 2 and 3 Experimental Study and Finite Element Analyses .. 105 4.4.2 Recommendations for Future Work ...... 106 4.4.2.1 Through thickness pinning ...... 106 4.4.2.2 Modified Biscuit Joints within Scarf Joints ...... 106
5. IMPACT RESISTANCE AND DAMAGE TOLERANCE OF SCARF REPAIRS 107 5.1 Introduction ...... 107 5.2 Experimental Investigation and Analyses ...... 107 5.2.1 General...... 107 5.2.2 CAI Specimen Descriptions ...... 108 5.2.3 Tension coupon testing...... 111 5.2.4 Strain surveys...... 112 5.2.5 Test and data reduction procedures ...... 113 5.2.5.1 Impacting...... 113 5.2.5.2 Ultrasonic inspection and damage size measurement...... 115 5.2.5.3 Sectioning ...... 117 5.2.5.4 Compression Testing ...... 118 5.2.5.5 Tension Testing...... 120 5.2.6 Failure prediction methodology...... 121 5.3 Results and Discussion...... 122 5.3.1 Impacting and damage assessment ...... 122 5.3.1.1 Summary...... 122 5.3.1.2 Damage mode ...... 123 5.3.1.3 Stiffness reduction strain survey...... 128 5.3.2 Failure analyses...... 129 5.3.2.1 Parent structure failure analyses...... 129 5.3.2.2 Scarf specimens...... 131 5.3.2.3 Scarf with doubler specimens...... 133
ix 5.3.3 Static test results and discussion ...... 135 5.3.4 CAI Failure predictions ...... 137 5.4 Conclusions and Recommendations for Future Work...... 139 5.4.1 Conclusions ...... 139 5.4.2 Recommendations for Future Work ...... 140
6. CONCLUSIONS AND RECOMMENDATIONS ...... 141 6.1 Shape Optimisation of Scarf Repairs...... 141 6.1.1 Conclusions ...... 141 6.1.1.1 Analytic and Finite Element Analyses ...... 141 6.1.1.2 Tensile Testing ...... 142 6.1.2 Recommendations for Future Work ...... 142 6.1.2.1 Analytic and Finite Element Analyses ...... 142 6.1.2.2 Testing...... 143 6.2 Reinforcement Methods used to Improve Scarf Joint Strength...... 143 6.2.1 Conclusions ...... 143 6.2.1.1 Through thickness pinning ...... 143 6.2.1.2 Modified Biscuit Joints within Scarf Joints ...... 143 6.2.1.2.1 Phase 1 Pilot Experimental Study ...... 143 6.2.1.2.2 Phase 2 and 3 Experimental Study and Finite Element Analyses .. 144 6.2.2 Recommendations for Future Work ...... 144 6.2.2.1 Through thickness pinning ...... 144 6.2.2.2 Modified Biscuit Joints within Scarf Joints ...... 144 6.3 Impact resistance and damage tolerance of scarf repairs ...... 145 6.3.1 Conclusions ...... 145 6.3.2 Recommendations for Future Work ...... 146
7. REFERENCES...... 147
x List of Figures
Figure 1 Manufacturing the hybrid biscuit flush repair joint ...... 3 Figure 2 Schematic diagram of an unloaded and loaded scarf and single lap joint...... 7 Figure 5 Element used by Lubkin to set up scarf joint equilibrium equations [7]...... 11 Figure 6 Scarf joint geometry used to develop governing equations...... 11 Figure 7 Element of adherend between 0 and x ...... 12 Figure 8 Magnified view of the adhesive, exaggerated to show the adhesive deformation...... 12 Figure 9 Finite element mesh used by Soutis and Hu [23] ...... 17 Figure 10 Failure load of a scarf patch repaired laminate versus scarf angle. Characteristic length of 1mm used to correlate test results ...... 18 Figure 11 Configuration of the corner singularity at the interface between two dissimilar materials and traction free edge [35]...... 18 Figure 12 Example of a 2D shallow angle scarf joint specimen to be loaded in tension [16]...... 19 Figure 13 Sandwich specimens used by Baker [19-21] to test two dimensional scarf joints in tension and compression (tension loading shown)...... 20 Figure 14 Failure mode observed within a shallow angle scarf joint loaded in tension [1]...... 20 Figure 15 Scarf joint with through thickness pins [24]...... 24 Figure 16 Scarf joint specimen with an embedded internal patch [24]...... 25 Figure 17 Analytic solution for the stress in a round patch within a uniaxial loaded plate ...... 28 Figure 18 FE mesh used for 2D modelling (left) and 3D modelling (right)...... 31 Figure 19 Stress in the insert with respect to parent-insert Young's modulus ratio...... 32 Figure 20 Peak tangential stress concentration factor with respect to the parent-insert Young's modulus ratio...... 33 Figure 21 Peak radial stress with respect to the parent-insert Young's modulus ratio...... 34 Figure 22 Parent-insert interface shear stress, Applied stress=100 MPa, Units are in MPa, Beginning at top left and moving L to R, (i) Einsert/Eparent=1, (ii) Einsert/Eparent =0.9, (iii) Einsert/Eparent=0.8, (iv) Einsert/Eparent =0.7, (v) Einsert/Eparent =0.6, (vi) Einsert/Eparent =0.5...... 35 Figure 23 Scarf joint geometry used to develop governing equations...... 36 Figure 24 Element of adherend between 0 and x ...... 36 Figure 25 Metallic adherend mesh near to the end of the adhesive bond-line...... 41 Figure 26 Composite adherend mesh near to the end of the adhesive bond-line ...... 42 Figure 27 Geometry for the scarf joint with metallic adherends...... 44 Figure 28 Method of manufacture for the scarf joint specimens...... 45 Figure 29 Baseline CFRP scarf joint geometry ...... 46 Figure 30 Close up of the optimised scarf joint (5q and 10q) profiles...... 49 Figure 31 Phase 1 Test Machine and Apparatus...... 51 Figure 32 Phase 2 Test Machine and Apparatus...... 52 Figure 33 Yokagawa Data Acquisition System ...... 53 Figure 34 Scarf Joint loaded in tension, showing the joint and adhesive CS [1]...... 54 Figure 35 Adhesive displacement (exaggerated) in joint and adhesive CS [1] ...... 54 Figure 36 Extensometer position and mounting method...... 55 Figure 37 Far field strain gauge location ...... 57 Figure 38 Adherend tip strain gauge locations...... 57 Figure 39 Analytic and FEA adhesive shear stress distribution predictions for the linear scarf joint between isotropic adherends ...... 59 Figure 40 Analytic adhesive shear stress distribution predictions, E1=0.9E2, non-linear taper ...... 60 Figure 41 Analytic adhesive shear stress distribution predictions, E1=0.8E2, non-linear taper ...... 60 Figure 42 Analytic adhesive shear stress distribution predictions, E1=0.7E2, non-linear taper ...... 61 Figure 43 Analytic adhesive shear stress distribution predictions, E1=0.6E2, non-linear taper ...... 62 Figure 44 Analytic adhesive shear stress distribution predictions, E1=0.5E2, non-linear taper ...... 62 Figure 45 Optimum Dmax/Dmin with respect to isotropic scarf joint adherend modulus ratio...... 63 Figure 46 Adherend axial stiffness along the tapered section of an orthotropic 5q scarf joint ...... 64 Figure 47 Analytic and FEA adhesive shear stress predictions for the constant angle scarf joint between orthotropic composite adherends with equivalent and reduced modulus plies ...... 65
xi Figure 48 Analytic and FEA adhesive shear stress predictions for the constant angle scarf joint between orthotropic composite adherends with an equivalent laminate and a laminate interleaved with adhesive ...... 65 Figure 49 Analytic adhesive shear stress distribution predictions, E1=0.77E2 with interleaved adhesive, constant and linearly varying taper angle ...... 66 Figure 50 Analytic adhesive shear stress distribution predictions, E1=0.8E2 with equivalent lay-up, constant and linearly varying taper angle...... 67 Figure 51 Phase 1 (top left) and Phase 2 (top right) metallic 5q scarf and phase 1 (bottom) metallic 10q scarf specimens between identical adherends ...... 68 Figure 52 Close up view of a small section of the Phase 2 metallic 5q scarf specimen shown in the top right of the previous figure...... 69 Figure 53 Phase 2 linear 5q (left) and 10q (right) scarf specimens between dissimilar adherends...... 69 Figure 54 Phase 2 optimised 5q (left) and 10q (right) scarf specimens between dissimilar adherends 70 Figure 55 Phase 1 5q and 10q composite scarf joint specimens ...... 71 Figure 56 Adherend far field stress and strain for the phase 1 metallic scarf joint specimens...... 73 Figure 57 Adherend tip strain wrt applied load for the phase 1 metallic scarf joint specimens...... 74 Figure 58 Adhesive shear stress wrt shear strain for the phase 1 metallic scarf joint specimens...... 74 Figure 59 Adherend far field stress and strain for the phase 2 metallic scarf joint specimens between identical material adherends...... 75 Figure 60 Adherend far field stress and strain for the phase 2 metallic scarf joint specimens between dissimilar adherends with a linear taper profile ...... 76 Figure 61 Adherend far field stress and strain for the phase 2 metallic scarf joint specimens between dissimilar adherends with an optimised taper profile...... 76 Figure 62 Adherend far field stress and strain for the phase 1 composite scarf joint specimens ...... 77 Figure 63 Adherend tip strain wrt applied load for the phase 1 composite scarf joint specimens ..... 77 Figure 64 Adhesive shear stress wrt shear strain for the phase 1 composite scarf joint specimens ... 78 Figure 65 Through thickness pin layout ...... 82 Figure 66 Method of pin insertion...... 83 Figure 67 5q metallic scarf joint with through thickness pins...... 83 Figure 68 Adherend far field stress and strain in the phase 1 metallic scarf joint specimens with and without TTP’s...... 84 Figure 69 Adherend tip stress and strain in the phase 1 metallic scarf joint specimens with and without TTP’s...... 85 Figure 70 Pilot study scarf joint specimen with (left) and without (right) an embedded patch...... 86 Figure 71 Pilot study scarf joint specimen with (top) and without (bottom) an embedded patch...... 87 Figure 72 Metallic adherend scarf joint manufacture using paste adhesive...... 88 Figure 73 Metallic adherend scarf joint with composite insert manufacture using paste adhesive ... 89 Figure 74 Hybrid biscuit joint specimen geometry...... 89 Figure 75 Typical mesh near to the scarf and taper tip region...... 92 Figure 76 Load with respect to crosshead displacement for the thick adherend baseline scarf specimen and the embedded patch specimen...... 93 Figure 77 Load with respect to shear strain for the thick adherend scarf joint and the embedded patch specimen ...... 94 Figure 78 Applied stress wrt crosshead displacement for specimens with an aluminium insert...... 95 Figure 79 Applied stress wrt crosshead displacement for specimens with an titanium insert...... 95 Figure 80 Schematic of the failure progression through the joint ...... 96 Figure 81 Failure of joints with Geometry A with an aluminium insert...... 96 Figure 82 Failure of joints with Geometry B with an aluminium insert...... 97 Figure 83 Failure of joints with Geometry A with a titanium insert...... 97 Figure 84 Failure of joints with Geometry B with a titanium insert ...... 97 Figure 85 Initial failure surface in the scarf of specimen 01 with Geometry B with a titanium insert ...... 98 Figure 86 Exaggerated view of the y displacement along the specimen when loaded in tension ...... 99 Figure 87 Shear stress in the tapered bond-lines of joints with an aluminium insert...... 100 Figure 88 Shear stress in the tapered bond-lines of joints with titanium insert...... 100
xii Figure 89 Shear stress in the scarf bond-line of joints with an aluminium insert...... 102 Figure 90 Shear stress in the scarf bond-line of joints with a titanium insert ...... 102 Figure 91 Shear strain in the scarf bond-line at the initial failure stress for each specimen type .... 103 Figure 92 Shear strain in the scarf when the local shear strain exceeds the material allowable ...... 103 Figure 93 Static test results and failure predictions...... 104 Figure 94 Parent adherend (top left), 5q scarf joint (top right) and 5q scarf joint with doubler CAI specimen geometry...... 109 Figure 95 Strain survey reference coordinate system for the compression (left) and tension specimen (right) ...... 112 Figure 96 Impacting apparatus ...... 113 Figure 97 Cross section of AS6/2220-3 showing damage due to a 40J impact (Source: Walsh Paul J. Carbon Fibres. ASM Handbook: Volume 21 Composites. ASM International (2001):pp35-9.). 115 Figure 98 Pulse-echo ultrasonic inspection technique ...... 116 Figure 99 Ultrasonic inspection scanning [51]...... 116 Figure 100 Digitised images using the A-scan (left) and the C-scan (right) technique ...... 117 Figure 101 Cross section of a damaged area within a parent adherend specimen...... 118 Figure 102 SACMA CAI Test fixture ...... 118 Figure 103 Absorbed energy within the specimen with respect to the potential energy of the drop weight for all specimens ultimately tested in compression ...... 122 Figure 104 Damage area with respect to impact energy for all CAI specimens...... 123 Figure 105 Ultrasonic inspection images of the CAI specimens ...... 124 Figure 106 Transverse cross section through the damaged site of the parent adherend specimens. 125 Figure 107 Longitudinal cross section through the damaged site of scarfed specimens...... 126 Figure 108 Longitudinal cross section through the damaged site of scarf with doubler specimens. 126 Figure 109 Longitudinal cross-section through the damaged site of a scarf joint impacted at the tip, x=0 mm, then at x=5 mm, x=10 mm, x=15 mm (mid joint) and x=20 mm...... 127 Figure 110 Longitudinal cross- section through the damaged site of a scarf joint with doubler, impacted at the tip, y=0 mm, then at y=5 mm, y=10 mm, y=15 mm (mid joint) and y=20 mm. ... 128 Figure 111 Typical compression failure modes of the undamaged and impacted parent adherend CAI specimens...... 130 Figure 112 Tensile failure of a parent adherend coupon specimen ...... 131 Figure 113 Typical compression failure mode of the undamaged scarf CAI specimens ...... 132 Figure 114 Typical compression failure mode of the impacted scarf CAI specimens...... 132 Figure 115 Typical tensile failure mode of the undamaged scarf coupons ...... 133 Figure 116 Typical compression failure of the undamaged scarf with doubler CAI specimens...... 134 Figure 117 Typical compression failure of the impacted scarf with doubler CAI specimens ...... 135 Figure 118 Typical tensile failure mode of the undamaged scarf with doubler coupons...... 135 Figure 119 CAI failure strain with respect to damage area calculated using ultrasonic inspection for all specimen types...... 137 Figure 120 CAI failure predictions for the parent adherend, scarf joint and scarf with doubler joint specimens ...... 138
xiii List of Tables
Table 1 Base Metallic adherend and adhesive properties used for FEA...... 41 Table 2 Adherend properties for the base orthotropic composite adherend scarf joint...... 42 Table 3 Metallic baseline specimen geometry parameters ...... 44 Table 4 Typical material properties for the aluminium alloy 2024-T3 obtained from [39] ...... 44 Table 5 Adhesive material properties ...... 45 Table 6 Geometry parameters for the scarf joint specimen with CFRP adherends ...... 46 Table 7 Composite adherend material properties [41]...... 47 Table 8 Material properties for the aluminium alloy 7075-T6 obtained from [39]...... 48 Table 9 Adhesive material properties [38]...... 48 Table 10 Material properties for the titanium alloy Ti6Al4V Grade 5 obtained from [39] ...... 49 Table 11 Test Machine Description...... 50 Table 12 Extensometer parameters...... 55 Table 13 Strain gauge parameters...... 56 Table 14 Strain gauge type and location summary...... 58 Table 15 Specimen failure stress and strain summary ...... 72 Table 16 Through thickness pinning properties ...... 81 Table 17 Through thickness pin layout parameters ...... 82 Table 18 Specimen failure stress and strain summary ...... 84 Table 19 Geometry parameters for hybrid joint test specimens...... 87 Table 20 Metallic adherend material properties...... 87 Table 21 Adhesive material properties ...... 87 Table 22 Composite patch material properties ...... 88 Table 23 Hybrid specimen geometry parameter values ...... 90 Table 24 Average thickness and length of the tapered bond-lines for each specimen type...... 90 Table 25 Hybrid specimen geometry parameter values used in the FEM ...... 91 Table 26 Material properties used within the FEA...... 92 Table 27 Non-linear FEA parameters...... 92 Table 28 Adherend strength and stiffness parameter measurements for the thick adherend baseline scarf specimen and the embedded patch specimen ...... 93 Table 29 Composite adherend material properties...... 110 Table 30 CAI specimen geometry measurements ...... 111 Table 31 Specimen details relating to stiffness reduction due to low velocity impact testing ...... 112 Table 32 Extensometer and strain gauges parameters...... 119 Table 33 Axial strain measurement types and location of strain gauges for the CFRP specimens .. 119 Table 34 Axial strain measurement types and location of strain gauges for the scarf specimens .... 120 Table 35 Strain measurement types and strain gauge loc. for the scarf with doubler specimens...... 120 Table 36 Young’s modulus calculations for the strain survey specimens ...... 129 Table 37 Undamaged parent adherend and scarf joint specimen failure strains...... 136
xiv Nomenclature
Term Definition
DTDscarf Scarf taper angle
Dmin Minimum scarf angle within the optimised scarf joint
Dmin Maximum scarf angle within the optimised scarf joint
Dtaper Angle of the taper at the termination of the biscuit insert
H1x,2x,1z,2z Strain within either the adherend 1 or adherend 2 in either the x or z direction
Htip Strain in the scarf joint adherend near the tip of the scarf
Hx,y,z Strain in either the x, y or z direction within the component
Hu Material ultimate strain
'h Height at the release of the drop weight
'v Change in velocity of the drop weight due to the impact event
'l Extensometer deflection measurement
'u11 Adhesive displacement parallel to the bond-line
'u22 Adhesive displacement perpendicular to the bond-line
'ux Adhesive displacement parallel to the adherend surface
'uy Adhesive displacement perpendicular to the adherend surface
I(x) Force applied to a scarf joint adherend element at location, x
Ipin Through thickness pin diameter
I Diameter
J12 Shear strain in the scarf joint adhesive in the adhesive CS
Je Elastic limit shear strain
Jmax Maximum shear strain
Q Poisson’s ratio of an isotropic material
xv QQ Poisson’s ratio of a CFRP laminae
QxyQxy Poisson’s ratio of a CFRP laminate
V1x, 2x Stress in adherend 1 or adherend 2 in the x direction
V22,n Peel direction stress (2 or n direction in the adhesive CS) normal to the adhesive bond-line, at the adherend interface of the scarf joint
Vadherend Stress in the scarf joint adherend
Vx,y,z Stress in the x, y or z direction
Vyield Yield stress of the material
Vu Material ultimate stress
Vapplied Applied stress
W,p Yield shear stress of the adhesive
W12,W, Shear stress in the scarf joint adhesive
Wpeak, valley Maximum shear stress in the adhesive bond-line between dissimilar isotropic materials (peak) and the minimum shear stress near to the opposite tip (valley)
W,average Average shear stress across the scarf joint bond-line
W,allowable Allowable shear stress in the scarf joint adhesive as given within design documentation
E Elastic moduli of an isotropic material
E1 Elastic moduli of a CFRP laminae in the fibre direction
E2 Elastic moduli of a CFRP laminae in the transverse direction
E3 Elastic modulus of the adhesive in the through thickness direction
Eadherend, insert, parent, A, B, Elastic moduli of the scarf joint adherend, insert, parent structure, adherend A or adherend B.
Eadhesive Elastic moduli of the scarf joint adhesive
Eaverage Average specimen stiffness
Ex Elastic moduli of a CFRP laminate in the x direction
xvi Ey Elastic moduli of a CFRP laminate in the y direction (perpendicular to the x direction)
Eradiusc Young’s modulus measured at the edge of the delamination in the CAI specimen strain survey
Eimpactc Young’s modulus measured in the centre of the delamination in the CAI specimen strain survey
Effcc Young’s modulus measured in the far field, some 50 mm away from the delamination in the CAI specimen strain survey
Eabsorbed Energy absorbed by the specimen during the impact
Epotential Incident energy through the drop weight, controlled by the drop height, 'h and the impacting mass, m
G, G3 Shear modulus of the adhesive
GL Extensometer gauge length
2 g Acceleration due to gravity (9.8ms ) h(x) Adherend height as a function of x h1,2 , tadherend Adherend thickness h3 , ta, tbond Adhesive bond-line thickness
Kt Stress concentration factor
L, Lscarf Scarf joint length
Ldiagonal Through thickness pin spacing diagonally
Ledge_side Through thickness pin spacing from the adherend side edge
Ledge_tip Through thickness pin spacing from the scarf joint tip edge
Lpatch Patch length
Lpin Through thickness pin length
Lscarf Scarf length
Lspacing Through thickness pin spacing horizontally
Lspecimen Specimen length
xvii Linsert Biscuit insert length
Ljoint Length of joint within the biscuit insert specimens m Mass n Number of plies
P,p0 Applied load
Pult Applied load at failure tadherend Adherend thickness tply Ply thickness tpatch Patch thickness vin Drop weight velocity just prior to impact
W Specimen width x Variable x
xviii Abbreviations
Term Definition
2D Two Dimensional 3D Three Dimensional Al Aluminium alloy ASTL Aircraft Structural Testing Laboratory AVD Air Vehicles Division BVID Barely Visible Impact Damage CAI Compression After Impact CFRP Carbon Fibre Reinforced Plastic (epoxy resin in this case) CLT Classical Laminate Theory CRC-ACS Cooperative Research Centre for Advanced Composite Structures CS Coordinate System CTE Coefficient of Thermal Expansion DAS Data Acquisition System DSTO Defence Science and Technology Organisation FEA Finite Element Analyses FEM Finite Element Model HP Hewlett Packard MTS Material Test System MUX Multiplex N/A Not Applicable NL Non-Linear NASA National Aeronautics and Space Administration NATA National Association of Testing Authorities PCL Patran Command Language RAAF Royal Australian Air Force RT Room Temperature SACMA Suppliers of Advanced Composite Materials Association SC Scarf joint specimen
xix SCD Scarf with Doubler joint specimen sps Samples Per Second SRM SACMA Recommended Method (for testing) Ti Titanium alloy TTP Through Thickness Pins UV Ultra Violet Wrt With Respect To
xx 1. Introduction
1.1 Summary
The bonded scarf repair and the scarf joint have been used often within aircraft structures design, particularly during the assembly process to join thick highly loaded structures together, and also in the repair of thick laminated aerospace structures. Their use for these structures was motivated by their efficiency since they can typically restore close to the un-notched tension, compression and shear strength of modern carbon-epoxy composites, provided a high strength adhesive is used during construction. Currently, no other joint is as efficient for joining thick composite laminates together.
The scarf repair itself consists of a shallowly tapered cut-out from the parent adherend, and a mating insert that is bonded into the cut-out with structural adhesive. The cut-out and the mating insert are typically tapered at approximately 3q. By reducing the taper angle, the shear stress and the normal stress in the adhesive is reduced and the joint efficiency is raised. Another advantage of this repair is that the external surface of the laminate remains continuous and smooth following the repair application. As such, these repairs can be applied to the external skin of the aircraft without affecting the aircraft’s aerodynamic performance or raising the radar cross section (RCS).
Several trends in modern fighter design could potentially lead to renewed interest in the use of the scarf repair. These include the use of large unitised structures in airframe construction, the increased use of carbon-epoxy composites in the primary structure and a lowering of the acceptable RCS in the design, not to mention improved aerodynamic performance. The use of large unitised structures has meant that the option of parts replacement when they are damaged is no longer cost effective. The use of composites in the primary structure has effectively raised the thickness and the load capacity of the laminates used to construct the airframe. As such, the bonded joints used for repair will be required to be more efficient if the overall structure is to remain efficient following repair. The lowering of the acceptable RCS of the fighter aircraft has effectively changed the mode of aircraft construction as potential sources of electromagnetic scattering are aggressively removed from the design. These include the removal of any features that may interrupt the geometric continuity of the external skin surface. The scarf repair is the most efficient means of performing a repair that is flush with the parent structure surface.
The efficiency of the scarf repair is typically governed by the shear strength of the adhesive and the taper angle of the joint. The load transfer through the adhesive within the shallow angle scarf joint is almost completely through shear. The maximum shear stress allowable in the adhesive is reduced by temperature and humidity, thus both the design load magnitude and the expected operating environment of the repaired aircraft requires consideration during repair design. Once the design strain in the parent composite adherend is determined, an analysis follows to determine if the maximum shear stress in the scarf joint adhesive exceeds its design allowable at this loading at the maximum operating temperature and relative humidity. If the adhesive is critical at the design strain, the taper angle of the scarf repair is made shallower and the analyses
1 repeated. The adherend tip fracture load may also be considered during design, as the stress typically concentrates in this area. If the taper is made too shallow, the tip can become fragile, thus setting the minimum taper angle that can be used for design. Iteration may follow to determine a taper angle that does not fail or in fact gives the maximum margins of safety in the adhesive and the adherend tip at the design strain. Another limit on the minimum taper angle is the space available to perform the repair. For example, for 50mm diameter damage to a 5mm composite laminate, the repair requires a diameter of 250mm if a 3q scarf angle is used. If the damage occurs close to the edge of the panel, then the repair may not be feasible as there is simply not enough room.
The load transferred by the adhesive within the repair may be reduced if the repair material modulus is less than the parent material modulus. This may divert load away from the repair thus changing the critical element of the design. Whilst this may cause other problems to occur, some of these may be mitigated resulting in a more efficient repair design. The use of a patch material with dissimilar properties to the parent material is normally avoided as this typically raises the stress in the scarf joint adhesive for a given applied load. It was shown by Hart Smith that the peak adhesive shear stress ratio approaches the adherend modulus ratio for straight scarf joints between isotropic materials of dissimilar properties. The first problem to overcome is that the stress within the adhesive concentrates near the stiffer adherend tip. The study herein presents an optimisation analysis such that by varying the scarf angle through the adherend thickness, the adhesive shear stress distribution and magnitude is identical to the constant distribution seen in scarf joints between identical adherends. It was found that for isotropic materials, a scarf profile could be determined for a given Young’s modulus combination that minimised the adhesive shear and peel stresses. These results were confirmed with finite element predictions. Static testing showed that the change in profile did not raise or lower the maximum load that a given two dimensional joint could carry. One possible explanation for this difference is that the adhesive might have reached plastic yield stress along the entire scarf at the point of failure. Time constraints prevented testing of the joints in fatigue, in which the benefit of shape optimisation may be more clearly seen. It was also found that a taper profile could also be determined that minimised the peak adhesive shear stress within the bond-line of composite adherend scarf joints. However, time constraints again prevented experimental testing of these joints in fatigue.
Methods of reinforcing the scarf joint were also proposed to improve the efficiency and fatigue resistance of the flush joint. The new joint utilises a scarf joint in combination with a biscuit insert embedded into the internal region of the adherends. The joint and the method of manufacture are shown in Figure 1.
2 (i) Scarf parent adherend to i
(ii) Perform slot machining
(iii) Machine insert to size, apply adhesive, install then bond to (iv) Apply adhesive to mating patch adherend and bond to insert and parent adherend.
Upper scarf bond-line Biscuit taper Patch adherend Parent adherend Biscuit insert Lower scarf bond-line
Figure 1 Manufacturing the hybrid biscuit flush repair joint
To manufacture a biscuit joint shown in Figure 1, the parent adherend is (i) scarfed at a constant angle near the top and bottom surfaces of the adherend with a vertical step induced in the central region prior to the slotting operation. After scarfing approximately a third of the thickness on each side of the adherend, the central region is (ii) slotted to make room for the biscuit insert. The thickness of the slot comprises the remaining third of the adherend thickness. An insert is machined of identical dimension to the slot with an allowance for the encompassing bond-line around the insert. (iii) Adhesive is applied to the insert and fixed to the parent adherend. The patch is machined with the mating scarf and slot surfaces. (iv) Adhesive is applied to these surfaces and the patch is bonded to the parent adherend and the insert to form the joint.
The new joint was modelled using linear and non linear finite element analyses to determine the adhesive stress and strain distribution when loaded in tension. Representative joints were statically tested to determine the efficiency of the joints with slightly varying geometric parameters and material properties. It was found that by placing an internal biscuit insert within a scarf joint, the strength of the joint could be significantly improved. However, the strength increase was found to depend heavily on the length of insert used. Modest improvements in strength were found over the scarf joint when the length of the insert was limited such that the overall repair length matched the baseline scarf joint repair length. It was also found that conventional failure prediction methods, such as the maximum shear stress and the maximum shear strain approach, were not able to predict the failure of these joints accurately. The maximum shear stress
3 under-predicted the strength by up to 20% while the maximum shear strain approach over-predicted the joint strength.
The vulnerability of the composite scarf joint to tip fracture has been known since they were first applied to aerospace structures in the 1970’s. As discussed earlier, if the taper angle is made too shallow, the stress may concentrate in the tip region causing the adherend tip to become the critical element of the design. When the tip fractures, the efficiency of the two dimensional scarf joint was shown to reduce significantly [1]. Whether this fracture occurs through an overload situation or if accidental impact damage occurs close to the tip region, the effect on the residual strength of the repair may be the same. Little work has been performed to determine if the impact resistance and damage tolerance of the structure is affected by the presence of a scarf repair. As such, specimens were manufactured that allowed Compression after Impact (CAI) testing of the joints using the SACMA standard test method SRM-94 [2].
Firstly, the specimens were impacted, with several parent structure, scarf joint and scarf with doubler joint sectioned to determine the locations of the delamination and the area of damage. The specimens that were not sectioned were tested in compression mainly, but some were tested in tension. It was found that the impact caused a strength reduction in the scarf joint specimens, but was not as significant as the reduction seen in the parent structure specimens. The presence of the doubler further reduced the strength reduction. The applicability of a stress-based failure prediction method in determining the adherend strength of the damaged specimens was assessed. It was found that these methods successfully predicted the strength of the damaged parent structure, but under-predicted the strength of the impact-damaged scarf joints. This indicted that whilst the scarf joint strength was not dramatically reduced by the accidental impact, the damage tolerance of the repair still requires consideration during design.
1.2 Aims of the Research
1. To reduce the amount of “good” parent material that is removed when applying flush repairs to composite structures.
Several approaches were investigated to address this particularly difficult problem, including the application of through thickness reinforcement through the joint, optimising the shape of the joint and embedding a doubler patch in the middle of the adherend to transfer load through the scarf joint. The validity of these methods was assessed by static testing joint coupon specimens in tension, and also by modelling each of the specimens using FE techniques. The joint failure mechanics was investigated, with various failure criteria assessed.
2. To determine if scarf repairs are susceptible to accidental impact damage during operation.
A composite scarf joint, a composite scarf with doubler joint and a parent laminate were tested using standard CAI test methods and apparatus. This was performed
4 to determine the baseline vulnerability of the parent structure to low velocity impact, compared with the scarf and scarf with doubler joint specimens. The effect of increased impact energy and position of impact was assessed, with sectional views used to illustrate the damage mechanics of the impact event, and compression testing used to measure the specimen residual strength. Failure criteria was assessed to determine if existing CAI failure predictions used for the parent structure can be applied to predict the failure load of the damaged joint.
1.3 Layout of the thesis
Chapter 1 gives the introduction to this research project. Chapter 2 presents a literature survey. This survey includes the role of the scarf joint in the design of flush repairs, and also the methods that have been used to improve the design of flush repairs. The survey also included methods that have been used to measure the impact resistance and damage tolerance of scarf joints. Chapter 3 describes the analytic development used to enable the design of optimally shaped scarf repairs, finite element analyses used to validate these methods and a description of the static testing of joint coupons performed. Chapter 4 presents the static testing and finite element modelling of scarf joints with reinforcement added, including the use of through thickness pinning and an embedded “biscuit” patch. Chapter 5 presents the CAI testing and failure criteria assessment used to measure the vulnerability to accidental damage of scarf joints with and without doublers in comparison to the vulnerability of the parent structure. Chapter 6 presents the conclusions of the research and recommendations for future work. Chapter 7 provides a list of references used.
5 2. Overview of Scarf Repair Studies for Composite Aircraft Structure
2.1 General
This chapter reviews studies into methods for designing adhesively bonded scarf patch repairs to aircraft, experimental observations of the failure modes of scarf joints and scarf repairs, design improvements such as geometric shaping of scarf patch repairs to minimise bond-line stress, the addition of reinforcement methods such as doublers and through thickness pins, and the impact resistance and damage tolerance of scarf joints and patches. The common thread between these topics is their application to the improved repair design to thick, highly loaded composite aircraft structures.
The currently accepted design practice for highly loaded composite aircraft structures utilises; x the development of analytic methods to predict the stress in the adhesive line of a two dimensional scarf joint, x the use of finite element methods to model simple and complex joint mechanical behaviour, x knowledge of the failure mechanics of the scarf joint and scarf patch repair, and x methods used to maximise impact resistance and damage tolerance in repair design.
The author has made every effort to acknowledge past studies that contributed to the current state-of-the-art, but some omission is necessary due to the scope of this thesis.
The review also includes reference to methods that have been proven to improve scarf joint design, as well as methods that are currently in various stages of development. These include the: x use of the overlap doubler over the top of the scarf joint and patch, x use of geometric shaping of the scarf adherends to minimise bond-line stress, x insertion of multiple through thickness pins and through thickness stitching of the joint, and the x addition of an internal patch in the middle of the joint, in plane with the adherend.
2.2 Introduction
The use of the scarf joint in aerospace design dates back to the very beginning of powered flight in the early 1900’s. The very first Wright flyer contained a scarf joint to facilitate the use of very long and straight wooden main wing spars. The onset of metallic aircraft saw their use decline as metallic aircraft structural components could be efficiently mechanically fastened, thus there was no requirement for the use of a bonded scarf joint in aircraft design.
6 It was not until the 1960’s that the next materials revolution occurred in the form of advanced fibre and resin composites. High stiffness continuous fibres such as carbon and boron were mixed with an epoxy resin matrix to form composite materials, capable of very high stiffness to weight ratios, showing resistance to corrosion, making them an ideal candidate for aerospace application, particularly as an improved alternative to lightweight metallic alloys.
Adhesively bonded joints have significant advantages over mechanically fastened repairs, particularly in the repair of composite aircraft structure. In fact, the low bearing allowable of composite materials effectively rules out mechanically fastened repairs in most applications. These structures, by their very constituent make up, lend themselves to be bonded together using adhesive, thus making the use of adhesively bonded repairs cost effective provided adequate repair efficiency could be attained.
For thin composite and metallic aircraft structures, the use of a simple overlap patch is typically sufficient to recover a large proportion of the undamaged strength (Myre and Beck in [3]). This type of repair is simple to apply, provided adequate surface preparation is applied to the parent structure prior to bonding. However, both single and double lap repairs suffer from a shear lag phenomena at the tips of the overlap[4, 5]. As seen in Figure 2, at the overlap tip an unloaded patch tip is directly adjacent to a fully loaded parent structure. In this region, a strain incompatibility is introduced, thus raising the local adhesive shear and normal stress. In most repair cases, the tip is bevelled, thus minimising the normal stress, but nonetheless a stress concentration remains. This is exacerbated by the inherent load eccentricity in these repair joints, thus making them only suitable for fully constrained thin composite and metallic parent structures. For more efficient bonded joints, the scarf joint is typically required, which is also illustrated in Figure 2.
A
Unloaded single lap joint Loaded single lap joint B
Unloaded scarf joint Loaded scarf joint Figure 2 Schematic diagram of an unloaded and loaded scarf and single lap joint [1]
The first approaches to design scarf repairs to composite aircraft structures were developed in the early 1970’s [6]. These designers benefited from the legacy left by early pioneers in structural adhesive bonding in the 1940’s 1950’s and 1960’s. The early pioneers [4, 5, 7] observed key behaviour of bonded joints, and managed to develop mathematical equations to determine stress and strain due to applied load.
Governing equations for the shear stress in bonded lap joints were developed and closed form solutions for these equations were found. The governing equations considered the adherends and the adhesive to be isotropic and behave elastically. The adhesive was modelled as a series of tension and shear springs [4, 5]. Similarly, attempts were made to
7 develop models for the scarf joint, although in the initial study by Lubkin [7], equilibrium of the adhesive layer was not considered in the formulation. He showed that for scarf joints between identical adherends, the stress in the adhesive was uniform across the joint. He also showed that for scarf joints between dissimilar modulus adherends, the stress monotonically varies between a peak stress at the stiffer adherend tip and a valley stress at the opposite end of the joint. The scarf angles considered in these analyses were greater than 15q, typically much steeper than joints suitable for aerospace application.
In the early 1970’s, Hart-Smith [6] and Erdogan and Ratwani [8] developed more complete analytic models for the behaviour of scarf joints, typical of those used to repair composite aircraft structures. The adhesive was modelled as a series of tension and shear springs. However, closed form solutions were not available for these equations due to the presence of mathematical singularities which occurred near to the scarf joint tips. Numerical techniques such as finite difference modelling were used to solve these equations [8, 9]. Each of these landmark studies contributed to the field in a unique way.
Hart-Smith [6] showed that the ratio of the peak shear stress concentration at the stiffer adherend tip relative to the average shear stress across the joint was observed to approach the ratio of the adherend Young’s modulus. He observed that other material dissimilarities between adherend materials contributed to non-uniformity in the bond-line stress. Most notably were the effect of differences in the coefficient of thermal expansion (CTE) and the rate of moisture absorption. His model accounted for these effects, as well as proposing a simple method for accounting for non-linearity in the adhesive at high loads. He was also able to suggest that the integrity of the scarf joint tips was important in maintaining the overall integrity of the joint. This effect was modelled some time later by Thamm [10], concurring, and then tested some time after that by Pipes et al [1].
Erdogan and Ratwani [8] developed a model that was slightly more complete than Hart-Smith, in that it took account of both the mechanical behaviour of the adhesive in both shear and normal to the bond-line. Hart-Smith assumed that the normal stress was negligible, due to the inherent tapering of the scarf joint. The model developed by Erdogan and Ratwani lent itself to an easier solution, thus many future researchers [1, 9, 11, 12] were able to build on this work more readily.
Both the study by Hart-Smith and by Erdogan and Ratwani were able to identify that the stress in the scarf joint adhesive between composite adherends was not uniform. The adhesive adjacent to ply drop offs, particularly near to stiffer plies (0q) resulted in stress concentrations, as the majority of load transfer through the joint occurs in these locations. Subsequent studies using both analytic and finite element models attempted to account for this behaviour [1, 9, 13-15]. The emergence of these studies coincided with an increased knowledge of the failure mechanics of the composite scarf joint.
The complexity of the failure mechanisms of composite scarf joints became better understood in the early 1980’s through studies such as that performed by Pipes [1], which were able to shed more light on the intrinsic link between the failure of the joint adhesive and the resin in the matrix. The ply stacking sequence was identified to have a large impact on the site of failure initiation and the progression of failure through the joint. The
8 tip was identified as the site of failure initiation, but it was not clear from these studies whether first failure was in the adherend tip or in the adhesive spew fillet at the tip. To mitigate premature tip failure, reinforcement such as external doublers were proposed, and shown to add significant benefit to the scarf joint design [1, 6, 16].
The use of the doubler was first proposed by Hart-Smith [6] in the early 1970’s. In his study he predicted an infinite strength increase was achievable through longer and longer overlaps. This was disputed by a number of studies [1, 17, 18], although the benefit of a thin overlap doubler is clear. In these studies a critical overlap length was achieved whereby no further benefit was achieved. It was also shown that the addition of the doubler may introduce load eccentricity, thus adding a bending moment that requires consideration in design. Testing with constrained scarf joints with doublers [19-21] showed increased repair efficiency than unconstrained specimens [1, 17]. Coinciding with these studies, tests were carried out on three dimensional scarf patch repair specimens.
The three dimensional specimens aimed to determine if the design techniques developed using the two dimensional models were applicable to the scarf repairs. The significant result from these studies was the identification of a failure locus dependence on scarf angle [16, 22]. For joints with high scarf angles, the stress concentration effect due to the machined hole in the panel was greater than for shallow scarf angles. In other studies [3] it was shown experimentally that the compression allowable for the three dimensional repair was lower than that obtained from testing of two dimensional scarf joints. The behaviour of three dimensional scarf patch repairs was also considered in subsequent finite element activities.
Finite element modelling of the three dimensional scarf patch repair was first reported in the late 1990’s by Soutis and Hu [23]. These studies showed that the existing methodology to design repairs using the two dimensional analyses, gives conservative results. The optimal scarf angle to ensure adherend failure prior to adhesive failure in the joint was reported for the two dimensional composite scarf joint to be around 4q [16]. The same joint was considered in three dimensions by Soutis and Hu [23] who found that the optimal scarf angle was 7q. These studies showed the importance that repair shaping may have on the failure mechanics of the repair.
The use of adherend shaping in bonded joints began with attempts to mitigate the stress concentration in overlap joints due to shear lag. Given the uniform stress that occurs across scarf joints between identical adherends, this was not required for standard scarf joints. In most scarf joint design studies [1, 6], it was recommended to only use repair patches with an identical material and lay-up to the parent structure material. However, some advantage may be possible if a lower modulus patch can be used to divert load away from critical region of the adhesive layer. The surrounding structure may be strengthened as part of the repair to cope with the additional load, and the scarf joint shaped to mitigate the stress concentration seen in the stiffer tip of dissimilar adherend scarf joints.
The use of an external doubler provides the benefit of tip protection to the scarf. However, it contains inherent disadvantages that are difficult to design against. These include the introduction of load eccentricity and a stress concentration in the doubler near the tip
9 region due to local structural non-compliance, as observed in studies by Baker [20]. This is further described in Section 2.7.1. To mitigate these effects, the placement of the doubler in the mid-plane of the scarf joint has been proposed [24]. Other methods such as through thickness pinning [24] of the joint may help to improve impact resistance and damage tolerance, and may also add peel resistance to the tip.
The fragile nature of the scarf joint tip region, combined with the overall repair integrity being heavily reliant on the tip not failing, makes the existing scarf repair designs vulnerable to impact damage. Numerous studies have shown that once damage has occurred in the tip region, damage progression is quite swift, particularly under cyclic loading [25]. To the author’s knowledge, no one has attempted to study the possibility of using design to maximise the impact resistance of scarf joints.
2.3 Analytic methods
Analytic modelling of the load transfer between adherends bonded with a thin adhesive layer began with studies by Volkerson [4] and then in later studies by Goland and Reissner [5]. This work modelled the load transfer by shear within lap joints. They noted the shear lag phenomena that results in stress concentrations near the adherend tips. In this model the adhesive shear stress is determined by the relative displacement of the adherends on either side of the adhesive. The inherent tapering of the scarf joint minimised shear lag considerably.
As seen in Figure 2, when the lap joint is loaded in tension, different parts of the joint are displaced by differing amounts. At location A, for example, the bottom adherend is stretched a lot because it carries nearly the entire load. In contrast, the top adherend at location A is stretched very little because it carries very little load. The result, as the figure below shows, is that the adhesive is stretched a lot in shear at the ends of the bond, but very little in the middle. Damage initiation often occurs at the ends before the middle is fully loaded.
As seen in Figure 2, by scarfing the adherends, the adhesive strain along the bond can be made more uniform. As the joint is loaded in tension, the adherends stretch by the same amount along the entire bond-line. At location B, for example, the top adherend carries little load, however, it stretches by the same amount as the bottom adherend, which carries more load, but is proportionally thicker. As mentioned in the introduction, the uniform shear distribution seen in identical adherend scarf joints is not seen when there are dissimilarities between the adherends Young’s modulus, thermal expansion and moisture absorption properties. These dissimilarities, particularly the difference in the modulus, cause the load to be attracted or diverted, resulting in an uneven load distribution through the thickness of the adherend. As such, load is transferred through the adhesive in varying amounts along the bond-line, particularly near the stiffer adherend tip.
Lubkin was the first to model the scarf joint using analytic methods. His joint model is shown in Figure 3, in which the equilibrium of a small adherend element adjacent to the adhesive is considered. Uniaxial load is transferred through the adherend in the x
10 direction and then picked up by the adhesive through shear and normal tractions. These tractions are given by:
2 xn sin -VV , W V x - cossin - .
T Vx
W Vn
Figure 3 Element used by Lubkin to set up scarf joint equilibrium equations [7]
They represent the average shear and normal stress carried by the adhesive in the scarf joint. They do not account for any local geometric effects, and are independent of adherend modulus, thus do not account for any adherend modulus mismatch.
More accurate solutions for the shear and normal stress were obtained by a model developed by Erdogan and Ratwani [8]. Their contribution included the consideration of the adhesive as a third medium, which act as shear and tension springs in the model. The development of their analytic model is repeated in this section, beginning with a schematic of the scarf joint shown in Figure 4, and progressing through the development of the governing equations describing the mechanical behaviour of the joint (Equations 2.1 to 2.19).
The joint geometry from which the governing equations were derived is shown in Figure 4.
y Adherend 2 Adherend 1 Deformed shape
p0 h3 p0 h1 h2 D x O
Figure 4 Scarf joint geometry used to develop governing equations
It was assumed that the: adherend thicknesses, h1 and h2, were very small compared to the length and width of the plate, the through thickness stress, Vy and Vz, were zero, the contact stress on the adherends in the tapered region acted as body forces, the adhesive acted as a combination shear and tension spring.
11 y 2 x 0
W12(x) I (x)
V22(x) 1
Figure 5 Element of adherend between 0 and x
Considering the equilibrium of an element of adherend 2 (Figure 5) between 0 and x, we find,
x dt x 22 t 12 t DWDVI ))cos()()sin()(()( , (2.1) ³0 D)cos( x dt 22 t 12 t DWDV ))sin()()cos()(( 0 , (0 Since, Equation 2.2 is valid for all x in the region 0 V 22 t D W12 t D)sin()()cos()( . (2.3) Substituting Equation 2.3 into Equation 2.1, becomes, x 2 I x WD 12 )())(tan1()( dtt . (2.4) ³0 'ux 'uy Deformed bondline 'u11 'u22 Figure 6 Magnified view of the adhesive, exaggerated to show the adhesive deformation If we now consider the equilibrium of the adhesive (Figure 6), 12 h3 'u22 V 22 x)( , and (2.5) E3 h3 'u11 W12 x)( . (2.6) G3 where, 'u22 = uu 122222 = relative normal displacement of the adhesive, 'u11 = uu 111211 = relative tangential displacement of the adhesive, h3 = adhesive thickness, E3 = adhesive Young’s modulus, G3 = adhesive shear modulus. From Figure 6, it is possible to transform the relative displacement vector from the local adhesive coordinate system, 1-2, to the joint coordinate system, x-y. ' ' 22 ' uu x D 'uy D)sin()cos( , (2.7) and ' ' 11 ' uu x D 'uy D)cos()sin( . (2.8) It follows from Equation 2.5, 2.7 and 2.8 that, § 2 D][tan1 · hu x ]cos[)( ' x 3 ¨ ¸ 12 DW . (2.9) © 3 EG 3 ¹ Considering the body forces acting on the adherends, I x)( V x)( , and (2.10) 2 x x D]tan[ 0 I xp )( V1x x)( . (2.11) 1 xh D]tan[ The composite adherend was specified to be adherend 2. Hooke’s law with the assumption that H H 21 zz 0 becomes, V 2x H 2x QQ 22 zxxz )1( . (2.12) E2x Similarly, Hooke’s law for the metallic adherend dictated that, V 1x 2 H1x Q1 )1( . (2.13) E1x 13 Substituting Equations 2.10 and 2.11 into 2.12 and 2.13 respectively gives, I x Q Q 22 zxxz )1)(( H 2x , and (2.14) D)tan( Ex 2x 2 0 xp QI 1 )1))((( H1x . (2.15) x 11 xhE D))tan(( Remembering from Equation 2.4, 2 I x WD 12 t)())(tan1()(' . (2.16) Substituting the result from Equation 2.9 into Equation 2.16 gives, 2 D)(tan1 'ux I x)(' 'ux . (2.17) ª 2 D)(tan1 º c h3 D)cos( « » ¬ 3 EG 3 ¼ Differentiating Equation 2.17 gives, 'H I x)('' x (2.18) c Since 'H H H12 xxx , substituting Equations 2.14 and 2.15 into 2.18 gives, I I xgxxfx )()()()('' . (2.19) Where, 2 ª 11 QQ 22 zxxz 1Q1 º xf )( « » , and ¬ 2x D)tan( 11 xhExEc D))tan(( ¼ Q 2 )1( p xg )( 1 0 . 11 xhcE D))tan(( Thus, Equation 2.19 becomes the governing differential equation for the stress distribution along the scarf joint, solvable using numerical techniques. Other models were developed at the same time, but were in a form that was less conducive for further development. A model was developed by Hart Smith [6] at the same time as Erdogan and Ratwani, but did not consider the adhesive to transfer load through normal tractions, thus making it less complete, although equally significant. A computer algorithm was released with this model that was able to account for thermal and moisture effects, as well as non-linearity in 14 the adhesive. This algorithm enabled designers to estimate the overall load capacity of the joint using failure criteria developed specifically for bonded joints. Further information regarding this criteria can be found in [6]. It was identified in both Hart-Smith’s and Erdogan and Ratwani studies that these models were least accurate near the tip area, which may often be a location of high stress concentration. A significant limitation of this model and others similar is that it contained singular points at the scarf joint extremities. This is particularly important, considering that in subsequent experimental studies of scarf joints, failure initiation was observed at one of the adherend tips. This motivated designers to consider detailed finite element analyses to model the joint mechanics, particularly near the tip region. However, there were still notable contributions to the analytic model that considered; x Non-linear deformation of adhesive and adherend [26], x Bending of the scarf joint [1, 17, 26], x Thermal and moisture absorption coefficient dissimilarities [27], x More complete handling of the composite orthotropic properties [1, 26], x Computational routines that were able to compare local stress and strain in the adhesive and adherend with the joint design allowable to predict the overall joint failure load. These models relied on the existing body of evidence surrounding the failure mechanics of the composite scarf joint [18]. The development of these models enabled insight into the extent that material property dissimilarity and geometric variations affected the stress in the scarf joint through the use of sensitivity studies. They are a powerful tool to gain a quick insight into the likely behaviour of the joint between scarfed adherends. Each case may not require the same development time that a finite element model may need, and the solution may be sufficiently accurate for design. An expanded description of these contributions has not been provided since they represent secondary effects when it comes to using these formulae in practical scarf repair to aircraft design, and may well be ignored. Concerning models that account for material non-linearity, in reality the scarf joint is considered to have failed when the adhesive or adherend begins to yield, as it is often indistinguishable the difference between the load at yield and the load that causes catastrophic failure of the joint. Concerning the effect of bending, it is not normally required to consider bending in the scarf joint model since, in design, the joint represents a thin slice through the centre of a scarf repair, and as such, its bending is constrained by the surrounding structure. At present, debate continues as to the relative benefits of analytic modelling of the composite scarf joint. Material heterogeneity is the prime argument for a shift from analytic to finite element methods. Certainly, designers will continue to have to choose the right tool for the right application. The power of quick sensitivity studies achievable with analytic methods must be weighed up with any inaccuracies in the solution. 15 2.4 Finite element analyses methods As commercial finite element codes became available in the 1970’s designers began to use them to predict the stress within adhesively bonded joints. This began with quite rudimentary analyses with very few elements that considered a two dimensional scarf joint, representing a thin slice of the three dimensional scarf patch. Over the years, as the capability of the finite element codes increased, so was the ability to model complex mechanical behaviour such as the mechanics of the composite scarf joint. FEA improved on the analytic solutions by; x Improving stress predictions in the adhesive close to the adherend tip, x Better modelling of the bending that is induced during the uniaxial loading of the scarf joint, x Enabling the modelling of a three dimensional scarf patch, x Better modelling of the effects of thermal and moisture absorption mismatch in the adherends, x Enabling the modelling of extra reinforcement features added to the scarf patch itself such as doublers and through thickness pins. 2.4.1 Two dimensional scarf joint The advent of finite element analyses, even analyses that limited the mesh to be as coarse as one element through the thickness of the adhesive, was very powerful in determining the stress within the scarf joint. The shear lag phenomena described previously was able to be recreated using the finite element method, thus accounting for the major influence on bonded joint mechanics [28, 29]. The limitation on the element size stifled progress somewhat, but certainly today, it is possible to model over four elements through the adhesive thickness at the expense of a very small amount of computing time [30]. Perhaps the major advantage of this technique was the improved accuracy of the stress predictions near to the scarf joint tip. Finite element analyses predicted large normal and shear stress concentrations at the extremities of the scarf joint [28]. As the element size was reduced the published data showed that these normal stress concentrations, within the adhesive spew fillet, were within one to two adhesive thicknesses of the bonded joint extremity. They were largely dependant on the mesh size used, thus indicating the presence of a computational singularity [31]. Certainly, these effects were found to occur very close to the adhesive tip, although it is unclear as to whether failure can be predicted on the basis of peak stress predictions in this region [31]. Other effects such as the effect of ply orientation were easier to model using finite element techniques. Detailed FEM also showed that the effect of material homogeneity assumptions made to enable analytic solutions was significant, particularly when considering adhesive bonds between highly inhomogeneous adherends such as composite materials [13, 20]. When composite plies of mixed orientation were introduced, the load transfer shifted from a monotonic mechanism across the joint length, to a peaky pattern, with the majority of load passing between the plies oriented in the load direction [13, 21]. In these cases, the closed 16 form solutions were considered non-conservative. The use of 0q degree plies near to the scarf joint tip exacerbates the problem. Although the scarf joint contains nominally zero eccentricity between the load plane and the joint midline, a small amount of bending is generated in the joint [1, 12, 17]. This effect is exacerbated by the presence of an external doubler [15] and any dissimilarity between the adherend materials [30]. The use of finite element techniques accounts for this behaviour in a more accurate manner than analytic models. 2.4.2 Three dimensional scarf patch To the authors knowledge, studies by Soutis and Hu [23, 32-34] were the first to include results from a three dimensional model of the composite scarf patch repair. As such, although preliminary, they were able to benefit from the many papers that used two dimensional finite element models of the scarf joint. Their contributions included the following: x Development of a model to predict the stress within a three dimensional scarf patch (Figure 7), x Interpretation of failure modes observed in previous experimental studies within the space shuttle repair development program [16], x The maximum adhesive stress found within the three dimensional patch adhesive was found to be lower that the equivalently loaded two dimensional slice (Figure 8). Figure 7 Finite element mesh used by Soutis and Hu [23] 17 Figure 8 Failure load of a scarf patch repaired laminate versus scarf angle. Characteristic length of 1mm used to correlate test results As seen in Figure 8, the optimal angle for a scarf patch repair was found to be 7q, as compared to 4q predicted for the two dimensional scarf joint considered by Jones and Graves [16]. This is significant as it shows that the 2D analyses models, analytic or finite element, give conservative results. Finite element methods have also been used to quantify the strength of the singularity that exists at the scarf joint tip [35]. In particular the corner singularity where the adherend and adhesive interface intersects with a traction free edge (Figure 9). Results of the analyses showed that the strength of the singularity increased with increasing dissimilarity between materials (adhesive to adherend mismatch) and increasing scarf angle. As yet these methods remain invalidated as a means of predicting failure initiation and subsequent failure progression. Interface Solid 1 Solid 2 T1 T2 Traction Free Edge Figure 9 Configuration of the corner singularity at the interface between two dissimilar materials and traction free edge [35]. 18 2.5 Failure mechanics of the scarf joint and repair The increased application of advanced composite structures to aircraft in the 1970’s inspired a considerable amount of theoretical analysis techniques for predicting stresses and strains in the scarf joint used in bonded repairs. It remained for designers of advanced composite aircraft structures to determine whether in fact these structures could be repaired. The question arose out of a lack of experimental evidence that showed the ultimate load capacity of repairs applied to these structures using techniques that were practical to apply. An extensive United States Air Force funded program was launched in the late 1970’s to address these issues. This was followed by a series of experimental programs that were able to build on this comprehensive study. 2.5.1 Two dimensional scarf joint The two dimensional scarf joint specimen represents a thin slice of a scarf patch repair. It is a relatively simple specimen to manufacture, and can provide insight into the scarf joint failure mechanics, as well as validate failure predictions made using the analytic and finite element models. The specimens used for two dimensional scarf joint testing are shown in Figure 10 below. They include the tension specimen [1, 9, 16, 18, 36, 37] and a combined tension and compression specimen that utilised a four point bend rig [19-22]. Figure 10 Example of a 2D shallow angle scarf joint specimen to be loaded in tension [16] 19 Figure 11 Sandwich specimens used by Baker [19-21] to test two dimensional scarf joints in tension and compression (tension loading shown) In one of these reports [1], which used the simple tension specimen, the failure mechanics of the composite joint was described using post analyses of the failure surface with an electron microscope. The laminate tested was carbon epoxy with the following lay-up: [0,0,r45,90,r45,0,0]s. The key observations from this report, which dealt only with composite joints tested in ambient conditions, were that: x At high scarf angles, 6.2q and 9.2q, the joint failed mainly in the adhesive, but contained regions where the separation was in the adherend. These specimens contained pullout of the 45q and 90q fibres, which was more predominant in the 6.2q specimens. x At low scarf angles, the failure mode was different and more complex. As shown in the figure below, the failure surface begins at one of the scarf joint tips and goes through the adhesive at point A. Inside the joint at B, many 45q and 90q adherend fibres were pulled out. Between the internal 0q plies, the fibres were too strong to pull out, thus forcing the failure surface through the adhesive. Once the failure surface extends beyond the internal 0q plies (point C), the failure surface then breaks through the adherend with considerable delamination (D). Figure 12 Failure mode observed within a shallow angle scarf joint loaded in tension [1] Failure initiation could not be observed in the analyses performed by Pipes [1]. It was clear that failure initiated in the tip region, but it was not clear as to whether failure began in the adherend tip itself or the adjacent adhesive fillet. This region was shown to be highly stressed through analytic and FE models, and is also prone to damage during the manufacture phases. 20 In an analyses by Thamm [10] it was shown that tip damage affects the local stress distribution considerably. In results presented by Pipes [1], rounding the tips in an attempt to reduce the stress singularity, reduced the strength by 30%. In a recent study by Du [31], similar complex failure behaviour was observed for the static failure of shallow angle (5q) aluminium 2024-T3 scarf joint specimens. At higher angles (10q, 20q and 30q) failure was governed by the load capacity of the joint in shear (mode II). At 5q, the failure was observed to be governed by both mode I and mode II fracture contributions. The fractured scarf tip was observed to be bent parallel as well as perpendicular to the length of the joint. It was suggested in these analyses that failure may be governed by tensile overload near the tip along the specimen edges due to excessive local normal stress. Finite element analyses of the region did not support this hypothesis, thus indicating that the failure mechanics of low angle scarf joints are still not completely understood. In other reports, the efficiency of the joints under a variety of environmental conditions with various geometric configurations was reported. Temperature and moisture were found to affect the scarf joint mechanics considerably [16, 18-21]. Studies by Baker [19-21], showed that the failure strain within the scarf region reduced by 50% when the test temperature was raised from ambient to 104qC. In these studies, the failure locus changed from a mixed adhesive-adherend mode at room temperature to an almost purely cohesive adhesive failure at the higher temperature. He noted that the shear compliance of the adhesive was significantly increased under hot/wet conditions making the adhesive stress significantly more uniform than the peakiness observed adjacent to the stiffer ply drop- offs. Tests were conduced by Ahn and Springer [18] using scarf and stepped lap specimens that had been moisturized. At low moisture content (less than 1% by weight), the failure loads were not affected for both room temperature and high temperature dry testing. However, at higher moisture contents, the failure loads were reduced, particularly when tested at high temperatures and high relative humidity. Baker [20] also noted that if laminates contained moisture levels greater than 0.7%, significant voids are introduced into the adhesive during the curing process. The drying process is considered to be essential to reduce the moisture content in the adherends to be bonded to below 0.3%. 2.5.2 Three dimensional scarf patch sub component testing Reports relating to sub component scarf repair testing were released in the 1980’s, which were performed to assist in repair development for the NASA space shuttle [16]. In these studies: x Large flat laminates were tested in tension and compression, and x Large composite laminates were bonded to aluminium honeycomb to form large honeycomb structures that were tested in tension and compression. These studies are considered to be of importance [20] as they represent the true repair case whereby load is diverted away from the repair through the adjacent structure. The 21 modulus of the repair patch may either attract or divert load from the repair adhesive, depending on the design chosen for the particular repair case. It was shown in these studies [16] that higher angle scarf repairs result in lower load carrying capacity than shallow angle repairs. Failures of the specimens were observed to initiate in the scarf region, but then promulgated horizontally, resulting ultimately in net section failure of the wide specimen. As the angle was steepened further, the failure initiation locus moved from the leading scarf tip in the centre of the panel to the two extreme sides of the scarf patch. Thus, for high scarf angle joints, the stress concentrations on the sides of the scarf joint are thought to be more extreme than for low scarf angle repairs. Another extensive experimental study of three dimensional scarf patch repair coupons was conduced by Tomblin [22]. In this study, many of the joints failed in the core region of the scarf repair. As such, a true indication of the repair adhesive or adherend strength could not be obtained. In moving from a large core size to a smaller core size, trends in the strength of the scarf repair with angle showed closer agreement to previous studies. However, even in the small core size specimens, failure was still observed in the core region and not the scarf. Testing by Myre [3] with a four point bend sandwich specimen that contained a circular scarf repair with external doubler plies on one side and an undamaged parent laminate on the opposite face, showed that the compression allowable for the three dimensional repair is lower than the compression allowable for the two dimensional scarf joint. The tension allowable for the three dimensional repair and the two dimensional scarf joint were similar. Failure of the three dimensional repairs in compression initiated in the repair region and extended across the laminate, approximately 30 mm from the scarf patch extremity. Failure of the three dimensional repairs in tension was predominantly through the parent laminate outside the repair region, although damage initiation may still have occurred in the repair. 2.6 Repair shaping to minimise bond-line stress The use of dissimilar adherends in a scarf joint has been recognised to induce a stress concentration in the scarf joint adhesive that varies monotonically from a peak near the stiffer adherend tip to a value considerably less adjacent to the less stiff tip. The stress at the mid-length position is equivalent to the average stress as predicted by Lubkin [7] formulae presented previously. This is shown in recent reporting by the author [30]. The ratio of the peak shear stress to the average shear stress was shown by Hart-Smith [6] to be equivalent to the ratio of dissimilar adherend moduli. Therefore, 1 W peak W valley E max . W average W average Emin 22 Conventional design practice recommends that dissimilar adherend scarf joints be avoided. However, there are some attractive features of using dissimilar adherend joints in design. The use of a scarf repair patch that has a lower modulus than the parent material will divert load away from the patch itself, transferring the load into the surrounding structure. Provided a careful analysis of the surrounding structure is performed, the load that is required to pass into the adhesive may be minimised, and the total repair strengthened. In such an instance, the design load would remain the same, but the scarf angle of the repair may be increased without failing the repair in the adhesive layer. However, if a linearly varying scarf angle is used, the peak stress adjacent to the higher modulus tip would most likely fail the adhesive locally. Therefore, shaping may be used to make the stress distribution more uniform. To the author’s knowledge, the first development of a analytic model to facilitate optimal adherend shaping of dissimilar adherend scarf joints was performed by the author [30]. The work presented in this study is also presented in later sections of this thesis. Equations developed by Erdogan and Ratwani [8] were modified to allow the user to specify a variable scarf angle across the joint. Sensitivity studies showed that for a given mismatch an optimal adherend shape exists to minimise the adhesive stress across the joint. 2.7 Scarf joint reinforcement methods 2.7.1 Overlap doublers The use of external doublers over the top of the scarf joint has been shown to reduce adhesive stress near the scarf joint tip [1, 6, 16, 18]. Testing by Jones and Graves [16] showed that if just three overlap plies were added to their scarf joint compression specimens, improved overall joint efficiencies were attained. Ahn and Springer [18] added one and two external plies, showing that failure loads were increased with increasing number of overlap plies. Baker [20] used five overlap plies with varying ply orientation, resulting in a higher failure load than the equivalent specimens with no external overlap plies. Baker [20] also noted that the use of overlap plies prevents creep, a phenomena where increasing displacement occurs in the adhesive under constant shear over time. It normally occurs in areas where there is a constant shear loading, nominally scarf joints between identical adherend materials. The adhesive in the overlap region under the doubler was shown to follow the shear trough common to bonded overlap joints, thus making the shear stress distribution peakier and less conducive to creep. This may not be the ideal method to prevent creep, as the peakiness introduced into the adhesive distribution may also introduce additional sites of potential failure initiation. Baker noted that a large strain concentration occurred in the doubler just above the scarf tip, which he proposed was the cause for failure initiation of the scarf joint with doubler plies. There is a relatively small displacement between the adherends across the overlap to repair patch transition region, compared to the large displacement that occurs between the 23 scarf adherends near to the tip across the adhesive. This incompatibility is the likely cause of a measured stress concentration in the overlap plies near to the scarf joint tip. As such, the design of the selection of overlap ply orientation and number of plies is of equal importance to scarf repair design as other features such as scarf angle, repair patch lay-up and the type of structural adhesive to use. Ply termination tests were conducted by Myre [3] which showed that a properly designed serrated edge at the external ply extremity improves the peeling resistance. The tests were conducted on the two dimensional scarf patch with a co cured external ply overlap that was shown to exhibit peel failure at the termination. The use of this technique was shown to be practical to apply through the use of pinking shears to cut the final ply prior to lay- up. 2.7.2 Through thickness pins The use of through thickness pins appears to be a powerful tool to minimise peel stress near to the scarf joint tip and improve resistance to tip damage (Figure 13). To the author’s knowledge, the first use of pins to improve the strength and damage resistance of bonded scarf joints was performed in a recent study by the author [24]. In this study the improvement in strength was found to be negligible, however, no reduction was observed, thus making it a potential tool for maximising the damage tolerance of the scarf repair under static and fatigue loading. Pins Pins Scarf joint Top View Side View Figure 13 Scarf joint with through thickness pins [24] 2.7.3 Internal patches The use of an internal patch to reinforce the scarf joint appears to be a powerful tool to improve the load capacity of the scarf joint. Similar to the doubler, an overlap region is introduced to the scarf joint. However, the reinforcing patch is introduced to the 24 mid-plane of the joint (Figure 14). This may improve the strength of the joint through the addition of extra adhesive bond-lines, sites that can transfer additional load across the joint. Compared to the external doubler, the positioning along the mid-plane minimises the load eccentricity, thus reducing the bending introduced as a result of an in-plane load. An elastic trough is introduced to the scarf joint midline, thus reducing the potential for creep [20]. Failure initiation is mitigated via the absence of an unconstrained free edge near to the tip concentration. The constraint at the doubler tip prevents excessive local strain peaks in the adhesive in the doubler tip region. Scarf joint Internal patch Figure 14 Scarf joint specimen with an embedded internal patch [24] To the author’s knowledge, the first use of an embedded patch to improve the strength of bonded scarf joints was performed in a recent study by the author [24]. In this case only one sample was tested as a part of feasibility study. Further work is reported in later sections of this report. The use of reinforcement methods may also be used to gain improvements to the impact resistance and damage tolerance of the scarf repair. This is described in the following section. 2.8 Impact resistance and damage tolerance The fragile nature of the scarf joint tip region, combined with the overall repair integrity being heavily reliant on the tip retaining its structural integrity, makes the existing scarf repair designs extremely vulnerable to impact damage. Numerous studies have shown that once damage has occurred in the tip region, damage progression is quite swift, particularly under cyclic loading [25]. Design according to damage tolerance assumes that a certain amount of damage already present within the scarf joint. Various methods, such as fracture mechanics are then used to estimate the rate at which the damage progresses from its initial value to a predefined critical size. The part is then inspected regularly to ensure that the damage is not moving at a rate that may compromise the part before the next inspection. This is a field of study in its own right, and will not be considered herein. The study herein, focuses on the use of reinforcement methods used to improve the impact resistance of the scarf joint. To the author’s knowledge, no one has attempted to quantify the effect of using design to maximise impact resistance in the scarf joint. Several features have been described already that may be used to improve impact resistance, including doublers and through thickness pins. Other features such as increased doubler to joint bond-line thickness or interleaved 25 structural adhesive with doubler plies, or the use of Kevlar in the doubler plies may also be of use. It is believed that the first study into the effect of impact damage on the efficiency of the scarf joint was performed by Tomblin [22]. In this case, only impact tolerance was considered. The energy required to generate BVID in the undamaged laminate was applied to each of the specimens in the middle of the scarf overlap region. In this study, it was shown that for shallow angle the effect of BVID on the scarf joint efficiency was negligible, but for higher angle repairs, up to a 20% reduction1 in strength was found. To the author’s knowledge, no other study has been performed that considers these issues. 1 Only those results that pertained to specimens that did not fail by core crushing were considered in this review. It is possible that the effect of impact damage was more on weakening the supporting core, rather than the scarf joint itself. 26 3. Shape Optimisation of Scarf Repairs 3.1 Introduction 3.1.1 General In scarf repairs, the use of a patch material with dissimilar properties to the parent material is normally avoided. It was shown by Hart Smith [6] that the peak adhesive shear stress ratio approaches the adherend modulus ratio for straight scarf joints between isotropic materials of dissimilar properties. The results confirmed the notion that the optimal scarf joint profile for an identical adherend scarf joint was of a constant scarf angle. It also showed that if the patch materials differ in properties to the parent material, a large stress concentration would be introduced into the adhesive layer. The study herein presents an optimisation analysis such that by varying the scarf angle through the adherend thickness, the adhesive shear stress is kept constant and of the same magnitude as that induced when the repair patch and the parent structure are made from the same material. The taper angle of a scarf patch repair to a composite structure is typically chosen such that it is shallow enough to reduce the adhesive shear stress below its shear allowable for a given design load. This allowable is typically limited by the adhesive strength, particularly at high operating temperatures. However, if part of the load is deflected away from the repair through the use of a reduced modulus insert, the load carried by the adhesive could be lowered for a given applied load. Provided the adjacent structure retains an adequate margin of safety, higher overall structural loads may be achieved before adhesive failure occurs. This can be demonstrated by considering the stress in a round patch insert within a uniaxial loaded flat plate. As seen in Figure 15, by reducing the modulus of the patch, the stress in the insert is lowered. Clearly, when the patch has an equivalent stiffness to the plate, an equivalent load is carried by the patch and the plate, however if the patch were to have a zero stiffness (i.e. a hole), no load is carried by the patch and all is carried by the surrounding plate ligaments. The patch modulus may be tailored such that an ideal balance may be achieved between load carried by the patch and the plate. A 3D FEA of the scarf repair with a dissimilar modulus to the parent structure was performed with the analyses description and results provided in Section 3.2. 27 1.0 Round repair Uniaxial loading Poisson's ratio=1/3 0.8 0.6 0.4 Stress in patch 0.2 0.0 0.0 0.2 0.4 0.6 0.8 1.0 Modulus ratio (Epatch /Eskin ) Figure 15 Analytic solution for the stress in a round patch within a uniaxial loaded plate The work herein describes the method for adapting the original equations developed by Erdogan and Ratwani [8] to facilitate the shaping of the scarf joint adherends to minimise adhesive shear and peel stress. Sensitivity studies using the analytic equations were performed on isotropic adherend combinations to identify an optimal rate of local scarf angle change for joints between several adherend moduli combinations. For each moduli combination an optimal rate of angle change was determined to achieve as close as possible a constant adhesive stress across the joint. Several cases were modelled using linear elastic finite element techniques to validate the analytical predictions. The adhesive stress within composite materials exhibits a strong dependence on the local ply orientations within the parent and patch adherends. A limited investigation was conducted within this study to determine the effect of using a reduced modulus patch on the adhesive stress within the orthotropic composite adherend scarf joint. Based on the results from this study, the suitability of using optimal shaping to reduce the peak stresses within dissimilar composite adherend scarf joints was evaluated. 3.1.2 Analytic solutions Governing equations have been derived in the literature [1, 8, 9] for the mechanical behaviour of the adhesive in a smoothly tapered joint. There was general agreement in the literature as to the method of formulating these equations. These equations have been adapted to facilitate the use of a variable scarf angle along the joint. Closed form solutions are not available for these equations due to the presence of singularities at the joint extremities. Therefore, a finite difference algorithm was developed to solve for the adhesive shear and normal stress along the bond-line. The formulation and solution of the governing equations is provided in Section 3.3. The results from the analytic modelling 28 and the FE validation are provided in Section 3.6.1, following descriptions of the FEA performed and the static test program undertaken. 3.1.3 Finite element solutions FEA was used herein to verify stress predictions made using analytic tools. MSC Nastran v2000 and MSC Patran v2003 were used to develop and solve all FEM presented in this report. These packages have been certified to use for engineering analyses of military and civilian aerospace structures, thus making them a suitable tool for this validation exercise. An algorithm was written in Patran Command Language (PCL) to facilitate the fast development of the FEM of the optimised scarf profile. Inputs included key geometric parameters that define the optimised profile mathematically and the maximum element length within the finite element mesh. In all cases four elements were employed across the adhesive bond-line and across the composite plies. TRIA3 plane strain elements were used to construct the isotropic adherend mesh and its corresponding adhesive layer. WEDGE6 solid elements were used to construct the composite adherend mesh and its corresponding adhesive layer. A description of the FEA performed is provided in Section 3.4, with results from the analyses provided in Section 3.6.1. 3.1.4 Experimental Testing Scarf joint coupons were tested in tension to failure to characterise scarf joint failure modes and to verify scarf joint failure prediction methodologies. In particular, the failure of scarf joints between dissimilar isotropic (metallic) adherends with and without shape optimisation was of interest. There was also some conjecture in the literature [1] as to the initial failure site and progression of failure through the composite scarf joint, as it was difficult to isolate whether the thin scarf joint adherend tip had failed before or after the adhesive in the same region. The experimental test program is described in Section 3.5. The results from testing, including a description of the observed failure modes is provided in Section 3.6.2. 3.2 Reduced Modulus Insert Analyses 3.2.1 General It is generally well known that throughout redundant structures, load is attracted to regions that are stiffer than the surrounding structure. As such, it is common design practice to reduce the stiffness in the regions of high stress concentration in order to generate a more even stress distribution, thus improving the overall efficiency of the structure. The addition of a reduced modulus insert for a repair application aims to perform a similar task. As the angle of the scarf is increased, the adhesive shear stress increases dramatically, thus for the repair to achieve high efficiencies a very shallow scarf angle is required. As such, the overall length of the repair becomes considerable, often 20-30 times the size of the 29 damage to be repaired. In order to reduce the shear stress within the adhesive, it was proposed to reduce the modulus of the repair insert relative to the parent modulus. In order to measure the effect of a reduced modulus insert on the stress distribution within a plate under uniaxial tensile load, finite element methods were used. Analytic solutions exist for the stress around the boundary of a hole within an infinite plate in tension, and also for the tangential stress in the plate at locations extending radially from the hole boundary, thus these results were used to validate the finite element techniques. Although closed form solutions exist for (i) the stress in a reduced modulus insert and (ii) the stress in the plate adjacent to a reduced modulus insert for an inclusion within an infinite plate subjected to uniaxial loading, no closed form solution is available for the case where the insert is scarfed. As such, FEA allowed the stress to be determined for these cases, with FEM validation possible by modelling cases where a closed form solution does exist. PCL programs were written to allow the fast generation of the FE models, which was particularly useful when ascertaining the effect of changing hole shape on the stress distribution. 3.2.2 Isotropic plate with round inserts 3.2.2.1 Analyses Description Two and three dimensional finite element analyses were conducted to determine the load attraction of a circular scarf insert under uniaxial tension. The case of a tapered hole was effectively an insert with zero modulus. As such, the cases considered were for an infinite plate with a reduced modulus insert that varied in modulus between 0% and 100% of the parent modulus. The two-dimensional analyses using thin shell finite elements (QUAD4) were used to model the case where the insert was butt welded to the plate. The three-dimensional analyses using solid elements were used to model the case where the insert and the plate were scarfed at a shallow angle, then bonded together. The scarf angles considered were 3q, 5q and 10q. For a straight-through inclusion (scarf angle=90q), the stress within a circular insert is given by V insert /3 EE parent insert . (3.1) V applied insert /21 EE parent The closed form solution for the peak stress concentration factor around the edge of the parent structure, just outside the insert, is given by V parent 3 . (3.2) V applied insert /21 EE parent It is yet to be confirmed whether these solutions are applicable to tapered or scarfed insert. To this end, finite element analyses were carried out. To keep the solution as close to that 30 of an infinite plate as possible, the plate width extended four times the transverse radius of the insert and the plate length extended six times the axial radius. The mesh was designed such that it was fine in areas with known stress concentrations and coarse in other areas to minimise computational time. A typical mesh for both the two and three dimensional cases is shown in Figure 16. Applied Stress Applied Stress Figure 16 FE mesh used for 2D modelling (left) and 3D modelling (right) 3.2.2.2 Results and Discussion A plot of the stress in the middle of the insert with respect to the insert-to-parent modulus ratio for the butt weld and scarf bonded cases is shown in Figure 17. It can be concluded from these results that the stress in the insert reduces when the insert modulus is reduced in relation to the parent modulus. It can also be concluded that the effect of scarf angle on the amount of load that bypasses the insert when the insert Young’s modulus is reduced is negligible. 31 Stress within the insert with respect to parent-insert Young's modulus ratio 1 0.8 0.6 0.4 Stress within the insert insert the within Stress 3deg scarf 5deg scarf 10deg scarf 0.2 FE prediction for 90deg butt joint Analytic prediction for 90deg butt joint 0 0 0.2 0.4 0.6 0.8 1 Parent to insert Young's modulus ratio Figure 17 Stress in the insert with respect to parent-insert Young's modulus ratio A plot of the peak tangential (hoop) stress in the parent adherend with respect to Young’s modulus ratio for the butt and scarf bonded cases is shown in Figure 18. The location of the peak hoop stress was adjacent to the insert to parent interface, at the narrowest net section of the parent adherend. It can be concluded from these results that the peak stress in the parent increases as the dissimilarity in the materials increases. The limiting case, when Einsert/Eparent =0, shows a large increase in the peak parent stress when the hole is scarfed, above the stress concentration than is seen in the butt welded case. It is well known that for a straight-through hole the stress concentration, Kt, is equal to 3 under uniaxial loading. The much higher stress concentration factor of a tapered hole was somewhat expected, since the parent tapers to zero thickness in an area of known stress concentration. The other limiting case, Einsert/Eparent =1, shows that the load transfers uniformly across the net section when the insert and the parent are made of identical materials. 32 Peak tangential stress concentration factor in parent adherend 6 5 3deg 5deg 10deg 4 FE prediction for 90deg butt joint Analytic prediction for the 90deg butt joint 3 2 1 Peak tangential stress in the parent adherend parent in the stress tangential Peak 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Parent to insert Young's modulus ratio Figure 18 Peak tangential stress concentration factor with respect to the parent-insert Young's modulus ratio A plot of the peak radial stress in the parent adherend with respect to Young’s modulus ratio for the butt and scarf bonded cases is shown in Figure 19. The location of the peak radial stress occurred 90q around the insert to parent adherend interface from the narrowest net section of the parent adherend. The radial stress at this location is typically used as the design stress for scarf repairs, using conventional two-dimensional analytic methods described in Section 3.3. It can be concluded from these results that by increasing the dissimilarity in the materials the peak radial stress in the adherend close to the tip is increased. This indicates that the surrounding structure is carrying increased load to compensate for the reduced modulus insert. These results were independent of scarf taper angle. The limiting case, Einsert/Eparent =1 showed that the peak radial stress approaches the far field stress. The limiting case, Einsert/Eparent =0 showed that a small compression radial stress is introduced at the hole boundary, possibly due to Poisson’s ratio contraction caused by the tangential stress acting in the area. 33 Peak radial stress in the parent adherend 3deg 5deg 10deg 1.4 1.2 1 0.8 0.6 0.4 0.2 Peak radial stress in the parent adherend parent in the stress radial Peak 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 -0.2 Parent to insert Young's modulus ratio Figure 19 Peak radial stress with respect to the parent-insert Young's modulus ratio Plots of the shear stress distribution at the insert-parent interface of the scarfed insert cases are provided in Figure 20. The magnitude of the uniaxial far field stress, in the direction shown in Figure 16, was 100 MPa. Although 3q, 5q and 10q cases were also modelled, the results from only the 5q scarf joint are shown. The distribution at the other scarf angles was largely the same, although the magnitude of the shear stress increased proportionally with increased scarf angle. It can be concluded from these results that when the insert and parent materials are matched, the shear stress in the scarfed interface is uniform, then as the dissimilarity in material modulus is increased, the shear stress at the interface concentrates, predominantly adjacent to the parent (or stiffer) adherend tip. Interestingly, the area of predominant shear stress concentration is not within the section where the peak radial stress occurs, but in the narrowest net section of the adherend, adjacent to the area of peak adherend tangential stress. 34 Figure 20 Parent-insert interface shear stress, Applied stress=100 MPa, Units are in MPa, Beginning at top left and moving L to R, (i) Einsert/Eparent=1, (ii) Einsert/Eparent =0.9, (iii) Einsert/Eparent=0.8, (iv) Einsert/Eparent =0.7, (v) Einsert/Eparent =0.6, (vi) Einsert/Eparent =0.5. 3.3 Scarf Joint Equation Formulation and Solution 3.3.1 Formulation The joint geometry from which the governing equations were derived is shown in Figure 21. It was assumed that the: adherend thicknesses, h1 and h2, were very small compared to the length and width of the plate, the through thickness stress, Vy and Vz, were zero, the contact stress on the adherends in the tapered region acted as body forces, the adhesive acted as a combination shear and tension spring. 35 y Adherend 2 Deformed shape Adherend 1 D x p0 h3 p0 h1 h2 O x Figure 21 Scarf joint geometry used to develop governing equations Considering the equilibrium of an element of adherend 2 (Figure 22) between 0 and x, we find, x dt x 22 t 12 t DWDVI )cos)(sin)(()( , (3.3) ³0 cosD x dt 22 t 12 t DWDV )sin)(cos)(( 0 , (0 y 2 x 0 W12(x) I (x) h(x) V22(x) 1 Figure 22 Element of adherend between 0 and x From Equation 3.4 it is readily shown that at an arbitrary location, x, along the joint, the condition holds, V 22 D W12 D tttt )(sin)()(cos)( . (3.5) Substituting Equation 3.5 into Equation 3.3 becomes, 36 x 2 I x WD 12 )())(tan1()( dttt . (3.6) ³0 If we now consider the equilibrium of the adhesive (Figure 6), h3 'u22 V 22 x)( , and (3.7) E3 h3 'u11 W12 x)( , (3.8) G3 where, 'u22 = uu 122222 , 'u11 = uu 111211 . From Figure 6, it is possible to transform the relative displacement vector from the local adhesive coordinate system, 1-2, to the joint coordinate system, x-y. ' ' 22 ' x D ' y D xuxuu )(sin)(cos , (3.9) and ' ' 11 ' x D ' y D xuxuu )(cos)(sin . (3.10) It follows from Equation 3.7, 3.9 and 3.10 that, § 2 D x)(tan1 · hu ¨ ¸ xx )(cos)( ' x 3 ¨ ¸ 12 DW . (3.11) © G3 E3 ¹ Considering the body forces acting on the adherends, I x)( V x)( , and (3.12) 2x xh )( 0 I xp )( V 1x x)( . (3.13) 1 xhh )( The composite adherend was specified to be adherend 2. Hooke’s law with the assumption that H H 21 zz 0 becomes, V 2 x H 2 x QQ 22 zxxz )1( . (3.14) E2x Similarly, Hooke’s law for the metallic adherend dictated that, 37 V 1x 2 H1x Q 12 )1( . (3.15) E1x Substituting Equations 3.12 and 3.13 into 3.14 and 3.15 respectively gives, I x Q Q 22 zxxz )1)(( H 2x , and (3.16) )( Exh 2x 2 0 xp QI 12 )1))((( H1x . (3.17) x 11 xhhE ))(( Remembering from Equation 3.6, 2 I x WD 12 t)()tan1()(' . (3.18) Substituting the result from Equation 3.11 into Equation 3.18 gives, 'u I x)(' x . (3.19) xc )( where, ª 2 D x)(tan1 º D xh )(cos 3 «G E » xc )( ¬ 3 3 ¼ 2 D x)(tan1 Differentiating Equation 3.19 gives, ' 'H x § 1 · I x)('' ¨ ¸ I xxc )(')( . (3.20) )( © xcxc )( ¹ Since 'H H H12 xxx , substituting Equations 3.16 and 3.17 into 3.20 gives, I I ' I xgxxfxxpx )()()()()()('' , (3.21) where, 2 1 ª1 QQ 22 zxxz 1Q 12 º xf )( « » , and )( ¬ 2x )( 11 xhhExhExc ))(( ¼ Q 2 )1( p xg )( 1 0 , and 11 xhhExc ))(()( 38 ' § 1 · ' 2 DDD 3)(sec)(tan)( 3 GEGExxx D x)(2cos xp )( ¨ ¸ xc )( 33 33 . ¨ ¸ 2 © xc )( ¹ 2 33 D xGE )(tan Thus, Equation 3.21 becomes the governing differential equation for the stress distribution along the scarf joint. 3.3.2 Solution Several approximate methods of solution have been tried previously, including series and perturbation methods [1, 6]. Given the singularities at the bond-line extremities, solution convergence via these solution methods was often difficult to achieve. A finite difference method was used successfully by Webber [9] and has also been used herein. An algorithm was written to perform the finite difference evaluation. The governing Equation 3.21 was rearranged into the following expression: I I I xgxxfxxpx )()()()(')()('' , (3.22) where, p(x)=0 for the case of straight scarf and is given by Equation 3.21 for the case of varying scarf angle. Equation 3.22 was approximated with the following finite difference expression: 2 xxx xx i1 ii 1 p ii 11 gxf . (3.23) h2 i 2h iii After rearranging, the above equation becomes, § h · 2 § h · 2 ¨ ¸ ii 1 21 xfhxp ii ¨ 1¸ ii 1 ghxp i . (3.24) © 2 ¹ © 2 ¹ Given that singularities of I x)( exist at both ends of the joint, I x)( will not be evaluated at these locations directly. Instead, their values can be obtained from the boundary conditions at the ends as follows, I x1 0)0( and I n )( pLx 0 . Once I x)( was evaluated along the scarf, the derivative I x)(' was determined in order to evaluate W12(x) from Equation 3.18 and the adhesive transverse tension stress, V22(x), from Equation 3.5. 3.3.3 Optimisation Studies by Hart-Smith [6] showed that for the dissimilar adherend scarf joint, the stress concentrated in the stiffer adherend tip, thereby causing a stress concentration in the adhesive stresses. Although the average shear stress across the scarf joint remained 39 unchanged, the adhesive shear stress near to the opposite tips deviated significantly from the average. Therefore, it is desirable to optimise the angle of scarf so that the adhesive stresses become uniform across the joint. From Equation 3.5, it is clear that optimising both the shear and peel stresses simultaneously is not possible due to coupling between the terms, thus only optimisation of the shear stress was considered. This may not introduce significant inaccuracies in the optimisation as the joint strength is typically governed by the shear strength of the adhesive for small scarf angles anyway. Since the coefficients of the governing equation (Equation 3.21) exhibit a complex non- linear dependence on the scarf angle, it was not possible to analytically determine the optimal scarf angle D()x that will result in constant adhesive shear stress. As a first order approximation, a linearly varying scarf angle was proposed as a means of optimising the adherend shape to minimise the adhesive shear stress peak within dissimilar adherend scarf joints. The local scarf angle with respect to x position is given as, x DD()x D. (3.23) minL max A sensitivity study was conducted for several adherend modulus combinations to determine the scarf profile that yielded the minimum shear stress peak. The results of the analyses and the FE validation are presented in Section 3.6.1. 3.4 Scarf Joint Finite Element Analyses To verify the validity and applicability of the analytic solution, finite element analyses were performed to determine the stresses within the adhesive for of scarf joints between isotropic and orthotropic composite adherends. 3.4.1 Model descriptions 3.4.1.1 Isotropic adherend scarf joint Several joints were modelled which contained both a constant scarf angle and also a linearly varying scarf angle to facilitate validation of the analytic results. For comparison purposes, the joints had an equal scarf length corresponding to an average scarf angle of 5q. The adherend thickness considered was 3 mm, and the adhesive thickness was 0.2 mm. The scarf was modelled using two-dimensional triangular plane strain elements (TRIA3), with both the adherend and adhesive treated as isotropic materials. The material properties of the adhesive were based on measured data for CYTEC FM73 film adhesive. The base adherend material properties were for aluminium 7075-T6, with the reduced stiffness properties obtained by simply applying a scaling factor to the aluminium 7075-T6 modulus. The properties specified in the FEM are provided in Table 1. 40 Table 1 Base Metallic adherend and adhesive properties used for FEA Property Adherend Adhesive Young’s modulus (GPa) 72 1.15 Shear modulus (GPa) 27.69 0.503 Poisson’s ratio 0.3 0.3 Care was taken to ensure the aspect ratio of the elements was as close to one as possible. Given the nature of the shallow tapered scarf, the tip was the most difficult to mesh accurately. The mesh tip is shown in Figure 23. Adherend (red) h3 Adhesive at end of bondline (blue) Adherend tip (red) Figure 23 Metallic adherend mesh near to the end of the adhesive bond-line 3.4.1.2 Orthotropic composite adherend scarf joint An average scarf angle of 5q was considered for the analyses of the composite adherend scarf joint. The joint was modelled by firstly creating a mesh of two-dimensional TRIA3 elements, then extruding them normal to the surface plane to create three-dimensional WEDGE6 elements. The extrusion was limited to one element out of plane which was approximately half a bond-line thickness in width. The use of 3D elements rather than 2D elements enabled the composite plies to be oriented in angles other than 0q or 90q. The adherend plies were specified to be three-dimensional orthotropic materials with the adhesive specified to be an isotropic material. The properties of the adherend plies were based on room temperature properties for CYTEC IM7/977-3 carbon epoxy prepreg obtained within certification documents provided by the manufacturer. These properties are provided in Table 2. 41 Table 2 Adherend properties for the base orthotropic composite adherend scarf joint Property IM7/977-3 Young’s modulus (GPa) 162 Transverse modulus (GPa) 16.2 Shear modulus (GPa) 7.2 Longitudinal Poisson’s ratio 0.3 Transverse Poisson’s ratio 0.03 Ply thickness (mm) 0.13 The base laminate chosen for this analyses was representative of the skin on an F/A-18 horizontal stabilator [21], in which the laminate consisted of 21 plies with a lay-up specified as follows: [+45, -45, 90, 0, 0, 0, +45, 0, 0, -45, 90, -45, 0, 0, +45, 0, 0, 0, 90, -45, +45]. The laminate may also be expressed as [48/38/14], with 48% of the laminate consisting of 0q plies, 38% consisting of r45q plies and 14% consisting of 90q plies. A typical quasi-isotropic laminate can normally be expressed as [25/50/25], thus making the skin laminate chosen for this study considerably stiffer than the quasi-isotropic case. The adhesive properties specified in the FEM were based on the measured room temperature properties of FM73 film adhesives [38], and are provided in Table 1. In a similar manner to the modelling of the metallic adherend scarf joint case studies, care was taken to ensure the aspect ratio of the elements was as close to one as possible. Given the nature of the shallow tapered scarf in contact with long thin plies, the scarf tip and the composite adherend plies were the most difficult to mesh accurately. The mesh near the tip of the adhesive bond-line is shown in Figure 24. 0q plies (red) 90q plies (dark blue) h3 45q plies (magenta) Adhesive at end of bondline (light blue) Adherend tip Figure 24 Composite adherend mesh near to the end of the adhesive bond-line 42 3.5 Experimental Tensile Testing 3.5.1 General Scarf joint coupons were tested in tension to failure to characterise scarf joint failure modes and to verify scarf joint failure prediction methodologies. In particular, the failure of scarf joints between dissimilar isotropic adherends with and without shape optimisation was of interest. There was also some conjecture in the literature [1] as to the initial failure site and progression of failure through the composite scarf joint, as it was difficult to isolate whether the thin scarf joint adherend tip had failed before or after the adhesive in the same region. Both elements are stressed above the surrounding area and the far field during uniaxial loading. In total, five sets of specimens were tested throughout the project. Each coupon was 25 mm in width, approximately 3 mm thick, with a scarf angle of either 5q or 10q, except for the optimised specimens, which had a linearly varying scarf angle of which the average scarf angle was either 5q or 10q. These sets were tested in two phases, the initial phase of testing conducted quite early in the project to confirm modes of failure presented in the literature [1]. The second phase of testing was conducted to determine if optimisation of the scarf joint had any bearing on the static strength of scarf joints between dissimilar adherends. During phase 1, the following two sets of specimens were tested in tension to failure; x Linear scarf joints between identical modulus isotropic adherends. The adherend material was aluminium alloy 2024-T3, 3 mm thick, and the adhesive used was FM300. Specimens were tested with 5q and 10q scarf angles, and x Linear scarf joints between identical modulus 21 ply orthotropic composite adherends. The adherend ply material was Hexcel T300-914c, with lay-up [+45, -45, 90, 0, 0, 0, +45, 0, 0, -45, 90, -45, 0, 0, +45, 0, 0, 0, 90, -45, +45] or [48/38/14], and the adhesive used was FM300. Specimens were tested with 5q and 10q scarf angles. During phase 2, the following three sets of specimens were tested in tension to failure; x Linear scarf joints between identical modulus isotropic adherends. The adherend materials were aluminium alloy 7075-T6, 3 mm thick, and the adhesive used was FM73. Specimens were tested with 5q and 10q scarf angles, x Linear scarf joints between dissimilar modulus isotropic adherends. The adherend materials were aluminium alloy 7075-T6 and titanium alloy Ti6Al4V Grade 5, 3 mm thick, and the adhesive used was FM73. Specimens were tested with 5q and 10q scarf angles, and x Optimised scarf joints between dissimilar modulus isotropic adherends. The adherend materials were aluminium alloy 7075-T6 and titanium alloy Ti6Al4V Grade 5, 3 mm thick, and the adhesive used was FM73. Specimens were tested with average scarf angles of 5q and 10q. 43 The specimen descriptions are provided in Section 3.5.2. The test procedure, including descriptions of the test apparatus and data acquisition methods is provided in Section 3.5.3. 3.5.2 Specimen Descriptions 3.5.2.1 Phase 1: Linear scarf joint between identical modulus isotropic adherends (Al.2024T3, FM300) In total, 3 metallic adherend scarf joints were manufactured with a scarf angle of 5q and 3 with a scarf angle of 10q. The specimen geometry for the scarf joint is shown in Figure 25 with the design dimensions for the 5q and 10q scarf provided in Table 3. W Lscarf Lspecimen * Drawing not to scale. tadherend Figure 25 Geometry for the scarf joint with metallic adherends Table 3 Metallic baseline specimen geometry parameters Parameter description 5q scarf joint 10q scarf joint Scarf length, Lscarf (mm) 36 18 Adherend thickness, tadherend (mm) 3 3 Specimen length, Lspecimen (mm) 260 260 Specimen width, W (mm) 25.3 25.3 Typical adherend and adhesive properties are provided in Table 4 and Table 5 respectively. Table 4 Typical material properties for the aluminium alloy 2024-T3 obtained from [39] Parameter description Parameter Material type Al. 2024-T3 Young’s modulus, E (GPa) 72 Poisson’s ratio, Q 0.3 Yield stress, Vyield (MPa) 345 Ultimate stress, Vu (MPa) 485 Yield strain, Hy (PH) 4800 44 Table 5 Adhesive material properties Parameter Value Material type FM300 Cure 1 hour at 180qC under autoclave pressure Shear modulus at RT*, G (MPa) 391** Poisson’s ratio,Q12 0.3 Bond-line thickness, ta (mm) 0.15** Limit shear stress at RT*, Wp (MPa) 34.9** Limit shear strain, Je 0.0895** Ultimate shear strain, Jmax 0.33** * Room temperature (RT) ** Reference: NADEP Engineering Report No. 004-90 page A6.2 Dated 30 August 1990. In order to ensure that all specimens were identical, the specimens were cured as a single plate, and then cut to size. This procedure is shown in Figure 26. Step 1 Step 2 A B Step 5 Step 4 B A A A Step 3 B B B A Step 6 Step 1: Cut adherend plate to the desired dimensions. Step 2: Divide the plate into two equal shapes (A & B). Step 3: Position adherend A on top of adherend B and clamp. Step 4: Machine the scarf to the desire angle. Step 5: Bond adherend A to adherend B, ensuring adherend alignment. Step 6: Following adhesive cure, cut rectangular plate to the specimen dimensions. Figure 26 Method of manufacture for the scarf joint specimens 45 3.5.2.2 Phase 1: Linear scarf joint between identical modulus orthotropic adherends (21 ply T300-914c, FM300) In total, three metallic adherend scarf joints were manufactured with a scarf angle of 5q and three with a scarf angle of 10q. The specimen geometry for the scarf joint is shown in Figure 27 with parameter values for the 5q and 10q scarfed specimens provided in Table 6. W Lscarf Lspecimen tadherend Figure 27 Baseline CFRP scarf joint geometry Table 6 Geometry parameters for the scarf joint specimen with CFRP adherends Parameter 5q scarf 10q scarf Scarf length, Lscarf (mm) 29 15 Adherend thickness, tadherend (mm) 2.54 2.54 Specimen length, Lspecimen (mm) 260 260 Specimen width, W (mm) 25.3 25.3 Typical adherend and adhesive properties are provided in Table 7 below and Table 5 respectively. The laminae properties were obtained from the design standard used for bonded repair certification within the RAAF [40]. The laminate properties were derived from the laminae properties using classical laminate theory (CLT). The method of manufacture was very similar to that described for the metallic scarf specimens in Figure 26. The only significant difference was that the original rectangular adherend plate comprised of a 21 ply CFRP laminate rather than a metallic adherend. The laminate was cured in an autoclave at 177qC for 1 hour at 90 psi under full vacuum. After testing, it was discovered that the adherend had not been laid up in accordance with the design lay-up. Although the difference in ply lay-up was relatively minor, the laminate produced was not symmetrical about its midline. Since the adhesive stresses in the scarf joint are normally sensitive to ply lay-up variation, the error may have influenced the observed failure modes slightly. However, as the difference in ply lay-up did not affect the laminate axial mechanical properties, the impact on test interpretation was expected to be negligible. 46 Table 7 Composite adherend material properties [41] Parameter Value Material type T300-914c Cure 177qC for 60 minutes at 90 psi (full vacuum) Number of plies, n 21 Design lay-up [+45, -45, 90, 0, 0, 0, +45, 0, 0, -45, 90, -45, 0, 0, +45, 0, 0, 0, 90, -45, +45] Actual lay-up [+45, -45, 90, 0, 0, 0, +45, 0,-,45, 90, -45, 0, 0, +45, 0, 0, 0, 0, 90, -45, +45] Laminae Properties Ply thickness (nominal), tply 0.12 mm* Laminae principal direction Young’s 128* modulus, E1 (GPa) Laminae transverse direction Young’s 13* modulus, E2 (GPa) Laminae principal Poisson’s ratio, Q12 0.3* Laminae transverse Poisson’s ratio, Q21 0.03* Ultimate strain in ply, Hu (PH) 7992* Laminate Properties Design Actual Laminate principal direction Young’s 74.8** 74.8** modulus, Ex (GPa) Laminate transverse direction Young’s 38.6** 38.6** modulus, Ey (GPa) Laminae principal Poisson’s ratio, Qxy 0.35** 0.35** Laminate transverse Poisson’s ratio, Qyx 0.18** 0.18** * Reference: Obtained for AS3501 from .[41] ** Reference: Obtained using CLT from the laminae properties. 3.5.2.3 Phase 2: Linear scarf joint between identical modulus isotropic adherends (Al.7075T6, FM300) In total, three metallic adherend scarf joints were manufactured with a scarf angle of 5q and three with a scarf angle of 10q. The specimen geometry for the scarf joint is shown in Figure 25 with the design dimensions for the 5q and 10q scarf provided in Table 3. The adherend and adhesive properties are provided in Table 8 and Table 9 respectively. 47 Table 8 Material properties for the aluminium alloy 7075-T6 obtained from [39] Parameter description Parameter Material type Al. 7075-T6 Young’s modulus, E (GPa) 72 Poisson’s ratio, Q 0.3 Yield stress, Vyield (MPa) 503 Ultimate stress, Vu (MPa) 572 Yield strain, Hy (PH) 6986 Table 9 Adhesive material properties [38] Parameter Value Material type FM73 Cure 1 hour at 120qC with 450 kPa autoclave pressure Shear modulus at RT*, G (MPa) ** Poisson’s ratio,Q12 0.3 Bond-line thickness, ta (mm) ** Limit shear stress at RT*, Wp (MPa) 41** Limit shear strain, Je ** Ultimate shear strain, Jmax ** * Room temperature (RT) ** Reference: [38] In order to ensure that all specimens were identical, the specimens were cured as a single plate, and then cut to size. This procedure is shown in Figure 26. 3.5.2.4 Phase 2: Linear scarf joint between dissimilar modulus isotropic adherends (Al.7075T6/Ti6Al4V, FM73) In total, three metallic adherend scarf joints were manufactured with a scarf angle of 5q and three with a scarf angle of 10q. The specimen geometry for the scarf joint is shown in Figure 25 with the design dimensions for the 5q and 10q scarf provided in Table 3. The aluminium alloy 7075-T6 adherend properties and the FM73 adhesive properties are provided in Table 8 and Table 9 respectively. The titanium alloy material properties are provided in Table 10. 48 Table 10 Material properties for the titanium alloy Ti6Al4V Grade 5 obtained from [39] Parameter description Parameter Material type Ti6Al4V Grade 5 Young’s modulus, E (GPa) 114 Poisson’s ratio, Q 0.3 Yield stress, Vyield (MPa) 1100 Ultimate stress, Vu (MPa) 1170 Yield strain, Hy (PH) 9649 In order to ensure that all specimens were identical, the specimens were cured as a single plate, and then cut to size. This procedure is shown in Figure 26. 3.5.2.5 Phase 2: Optimised scarf joint between dissimilar modulus isotropic adherends (Al.7075T6/Ti6Al4V, FM73) In total, three metallic adherend scarf joints were manufactured with a linearly varying scarf angle, in which the average scarf angles throughout the joint were 5q and three with an average scarf angle of 10q. The overall specimen geometry for the scarf joints is shown in Figure 25 with the design dimensions for the 5q and 10q scarf provided in Table 3. However, in order to determine the optimised scarf profile, the results from the optimisation analyses were used, particularly Equation 3.25 and Figure 43. Through an iterative process, for a given adherend modulus ratio, / EE BA , an optimal D max /D min was obtained. The local angle of the scarf was then varied between these maximum and minimum from one end of the joint to the other. In this case, / EE BA equalled 0.7 and the optimal D max /D min used was 2. A close up of the 5q and 10q scarf joints is shown in Figure 28. Average scarf angle = Average scarf angle = 5q 3mm 3mm Figure 28 Close up of the optimised scarf joint (5q and 10q) profiles The aluminium alloy 7075-T6 and titanium alloy material properties, and the FM73 adhesive properties are provided in Table 8, Table 9 and Table 10 respectively. 3.5.3 Testing Procedure In essence, the testing procedure used for all coupon tensile testing was the same. Variation existed in the particular testing machine used, although they were both Instron test machines with a 100 kN load cell installed. All testing was in displacement control with the rate of crosshead travel the same at 0.2 mm/min. All specimens had an extensometer mounted to measure the displacement across one of the scarf tips, however, the location and number of strain gauges varied between specimens. The data acquisition 49 system used and data sample rate varied, with the minimum rate of data collection at 1 Hz. The load and crosshead displacement was also recorded for each test. 3.5.3.1 Test Machines Two different test machines were used to conduct the phase 1 and 2 testing respectively. The set-up used for phase 1 testing is shown in Figure 29 with phase 2 shown in Figure 30. The test machine specifications are provided in Table 11. Table 11 Test Machine Description Apparatus Phase 1 Specification Phase 2 Specification Test machine Instron 100 kN actuator Instron 100 kN actuator r20 mm stroke r20 mm stroke 250 kN load frame 250 kN load frame Screw end grips Screw end grips End Grips Hydraulic Screw driven Controller MTS 810 MTS 810 50 Test Machine Controller 4 2 7 5 3 Controller Computer 1 8 HP DAS 6 9 DAS Computer 1. Controller to actuator interface. Signal generated by the controller drives the test machine actuator. 2. Controller computer to controller interface. Test operator inputs the desired loading parameters to pass on to the controller to generate an actuation signal. 3. Crosshead displacement signal sent to the controller. 4. Load signal sent to the controller. 5. Extensometer signal sent to the controller. 6. Strain gauge signal sent to the HP DAS. 7. Load, extensometer and crosshead displacement signal output to the controller computer. Computer used to convert load signal into kN, crosshead displacement signal into mm and the extensometer signal into microstrain. 8. Load and crosshead displacement signal sent to the HPDAS. 9. Load, crosshead displacement and strain gauge signals output to DAS computer. Computer used to convert load signal into kN, crosshead displacement signal into mm and strain gauge signal into microstrain. Figure 29 Phase 1 Test Machine and Apparatus 51 Screw Specimen Grips 2,3 1,4 Computer Test Machine with Load cell and Crosshead Load Controller Interface 1: The input parameters for the testing are entered by the user into the PC which interfaces with the load controller. Interface 2: The load controller is used to control the crosshead displacement and rate of loading. Interface 3: The crosshead displacement and the load from the load cell is fed back to the controller from the test machine. Interface 4: The crosshead displacement and the load are fed back to the PC for user output. Figure 30 Phase 2 Test Machine and Apparatus 3.5.3.2 Data acquisition systems Two different data acquisition systems were used for the different phases of testing. The Hewlett Packard Data Acquisition System (HPDAS) (Figure 29) was used in phase 1 and a Yokagawa system was used in Phase 2 (Figure 31). The sample rate within the HPDAS system was 1 Hz, whilst the Yokagawa system sampled at 5 sps (samples/sec.). 52 Figure 31 Yokagawa Data Acquisition System 3.5.3.3 Extensometers Previous studies of the scarf joint [1, 19, 42-44] have also shown that the most common mode of adhesive failure in the scarf joint was by shear strain overload. Therefore, it was of interest to measure the shear strain in the adhesive near the tip during testing. Studies [1] have shown that failure was also precipitated by adherend through thickness overload near the tip due to normal stress concentrations in the adhesive at the joint extremities. Unfortunately, as the transverse tension strain in the adhesive relative to the adhesive shear strain is normally so small, advanced measurement techniques are required to make an acceptably accurate measurement. They were not trialled within this test program, however a description of the method used within this project to measure the shear strain, ignoring the transverse tension strain component, is provided below. The scarf joint loaded in tension is shown in Figure 32. Also defined in this figure are the joint coordinate system (CS) and the adhesive CS. 53 Load, P Load, P y 2 x Joint coordinate system 1 Adhesive coordinate system Figure 32 Scarf Joint loaded in tension, showing the joint and adhesive CS [1] An exaggerated adhesive displacement is shown in Figure 33 Load, P Load, P D 'u2 'uy 'ux 'u1 J12 Figure 33 Adhesive displacement (exaggerated) in joint and adhesive CS [1] The shear strain is then given by: ux cosD ' 'u y sinD J 12 . (3.26) ta The shear strain measurement is normally dominated by the 'ux cosD term in Equation 3.26 for low scarf angles. As such, the 'u y sinD term can normally be omitted as 54 second order, provided the adherend is much stiffer than the adhesive. If the adherend is assumed to be supported against bending, then the 'uy term may be assumed to be zero. The measurement of 'ux was performed using an extensometer placed over the scarf tip on the specimen face (Figure 34). The extensometer was mounted to the specimen via specially tensioned rubber bands that were supplied with the extensometer. The extensometer specifications are provided in Table 12 below. Extensometer positioning 7 mm 12.5 mm Extensometer hooks Rubber band Specimen * Drawing not to scale. Extensometer mounting method (only one blade shown) Figure 34 Extensometer position and mounting method Table 12 Extensometer parameters Parameter Value Type Instron model flex 003 Stroke r2.5 mm Gauge length, GL 12.5 mm The strain reading from the extensometer included components due to the adherend and adhesive deflection. The adherend deflection was subtracted from the extensometer strain reading using Equation 3.27 below to give 'ux . 55 § ta · ' ' x H tip ¨GLlu ' ¸ , (3.27) © sinD ¹ where, 'l =extensometer deflection measurement, H tip =strain in the scarf joint adherend near to the scarf joint tip, GL =gauge length of the extensometer, ta =measured adhesive thickness close to the scarf tip and D =scarf taper angle. 3.5.3.4 Strain Gauges Strain gauges were applied to measure strain in the far field away from the joint (Section 3.5.3.4.1) and also near the scarf joint tips (Section 3.5.3.4.2). Specifications for the types of gauges used are provided in Table 13 and a summary of the strain gauge types used and their location on each of the specimens is provided in Table 14 (Section 3.5.3.4.3). Table 13 Strain gauge parameters Parameter Type A Type B Type Kyowa KFG-3-350-C1- Kyowa KFG-3-350- 23L3M3R axial strain C1-11L3M3R axial gauge with wires strain gauge with attached wires attached Gauge factor 2.10 2.10 Gauge length 3 mm 3 mm Gauge resistance 350 : 350 : Number of wires 3 3 Wire length 3 m 3 m 3.5.3.4.1 Far field adherend strain In order to measure the adherend strain far field of the joint, a strain gauge was attached at locations shown in Figure 35. 56 P 1 1 2 30 mm 12.5 mm 30 mm * Drawing not to scale 3 3 4 P Figure 35 Far field strain gauge location 3.5.3.4.2 Adherend tip strain Previous studies of the adhesively bonded scarf joint have suggested that there was a stress concentration at the adherend tip prior to any local yielding in the adherend. Therefore, it was of interest to measure the strain at the adherend close to the tips of the joint. It may also be used to detect any disbond that may occur during testing. The position of the gauge is shown in Figure 36 below. 7 mm 5 5 6 6 12.5 mm * Drawing not to scale. Figure 36 Adherend tip strain gauge locations 57 3.5.3.4.3 Strain gauge location summary Table 14 Strain gauge type and location summary Phase Test Set Specimen SG1 (ff) SG2 SG3 SG4 SG5 SG6 (ff) (ff) (ff) (tip) (tip) 1 A A A Metallic 5q scarf joint 2 A A A 3 A A A 1 A A A Metallic 10q scarf joint 2 A A A 3 A A A 1 1 B B B Composite 5q scarf joint 2 B B B 3 B B B 1 B B B Composite 10q scarf joint 2 B B B 3 B B B 1 A A A A Metallic identical adherend 5q scarf 2 A A A joint 3 1 A B A B Metallic dissimilar adherend linear 5q 2 scarf joint 3 1 A B A B Metallic dissimilar adherend linear 10q 2 2 scarf joint 3 1 A B A B Metallic dissimilar adherend 2 optimised 5q scarf joint 3 1 A B A B Metallic dissimilar adherend 2 A A optimised 10q scarf joint 3 3.6 Results and Discussion 3.6.1 Scarf Joint Analytic and FE Results 3.6.1.1 Isotropic adherend scarf joint For the linear scarf joint between dissimilar modulus isotropic adherends, the stress typically concentrates near to the stiffer adherend tip. By applying a shallow angle in these areas and a steeper angle in the less loaded areas, the stress may become more evenly distributed along the joint. These studies aimed to determine an optimal curvature to apply for a given adherend moduli combination. Plots of the adhesive shear stress distributions within isotropic adherend linear scarf joints between identical and dissimilar adherends predicted analytically using newly developed formulae and with FEA are provided in Figure 6. These plots confirm the predictions made by Hart-Smith [6] that the peak adhesive shear stress in the scarf joint between dissimilar adherends approaches the adherend modulus ratio. Clearly, for a scarf joint between identical isotropic adherends the optimal scarf shape has a constant taper angle. However, the monotonic variation in the adhesive shear stress between dissimilar adherend scarf joints clearly provides scope for optimisation. 58 Analytic and FEA adhesive shear stress distribution predictions, linear taper (5 deg.), isotropic adherends, bending constrained E1=E2 (analytic) E1=0.9E2 (analytic) E1=0.8E2 (analytic) E1=0.7E2 (analytic) E1=0.6E2 (analytic) E1=0.5E2 (analytic) E1=E2 (FEA) E1=0.8E2 (FEA) E1=0.7E2 (FEA) 2 1.8 1.6 1.4 av 1.2 B W / 1 (x/L) 0.8 W 0.6 0.4 0.2 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 x/L Figure 37 Analytic and FEA adhesive shear stress distribution predictions for the linear scarf joint between isotropic adherends Considering a range of adherend moduli combinations, plots of the shear stress distribution predicted analytically for several curvature severities are provided in Figure 38 to Figure 42. Several of the plots include the results from FEA to validate the result and thus the analytic methodology. These results pertain to a scarf joint with an average scarf angle of 5q. Another set of graphs were prepared for a scarf joint with an average scarf angle of 3q, with only a summary of results provided in Figure 43. 59 Analytic adhesive shear stress distribution predictions, E1=0.9E2, non-linear taper, isotropic adherends, bending constrained amax/amin=1.1 amax/amin=1.2 amax/amin=1.3 amax/amin=1.5 linear taper 1.2 1.15 1.1 1.05 av B W 1 x/L W 0.95 0.9 0.85 0.8 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 x/L Note: Due to limitations in MS Excel, Dmax/Dmin is represented by amax/amin in the legend. Figure 38 Analytic adhesive shear stress distribution predictions, E1=0.9E2, non-linear taper Analytic adhesive shear stress distribution predictions, E1=0.8E2, non-linear taper, isotropic adherends, bending constrained amax/amin=1.2 amax/amin=1.4 amax/amin=1.6 amax/amin=1.7 linear taper amax/amin=1.6 (FEA) 1.2 1 0.8 av B W / 0.6 (x/L) W 0.4 0.2 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 x/L Note: Due to limitations in MS Excel, Dmax/Dmin is represented by amax/amin in the legend. Figure 39 Analytic adhesive shear stress distribution predictions, E1=0.8E2, non-linear taper 60 Analytic adhesive shear stress distribution predictions, E1=0.7E2, non-linear taper, isotropic adherends, bending constrained amax/amin=1.6 amax/amin=1.8 amax/amin=2 amax/amin=2.1 amax/amin=2.2 amax/amin=2.3 linear taper amax/amin=2 (FEA) 1.4 1.2 1 av B 0.8 W / (x/L) W 0.6 0.4 0.2 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 x/L Note: Due to limitations in MS Excel, Dmax/Dmin is represented by amax/amin in the legend. Figure 40 Analytic adhesive shear stress distribution predictions, E1=0.7E2, non-linear taper 61 Analytic adhesive shear stress distribution predictions, E1=0.6E2, non-linear taper, isotropic adherends, bending constrained amax/amin=2 amax/amin=2.2 amax/amin=2.4 amax/amin=2.5 amax/amin=2.6 amax/amin=2.7 amax/amin=2.8 amax/amin=3 amax/amin=3.2 linear taper 1.7 1.5 1.3 av B W / 1.1 (x/L) W 0.9 0.7 0.5 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 x/L Note: Due to limitations in MS Excel, Dmax/Dmin is represented by amax/amin in the legend. Figure 41 Analytic adhesive shear stress distribution predictions, E1=0.6E2, non-linear taper Analytic adhesive shear stress distribution predictions, E1=0.5E2, non-linear taper, isotropic adherends, bending constrained amax/amin=2.8 amax/amin=3 amax/amin=3.2 amax/amin=3.4 amax/amin=3.6 amax/amin=3.8 amax/amin=4 amax/amin=4.2 amax/amin=4.4 linear taper 2 1.8 1.6 1.4 av B W / 1.2 (x/L) W 1 0.8 0.6 0.4 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 x/L Note: Due to limitations in MS Excel, Dmax/Dmin is represented by amax/amin in the legend. Figure 42 Analytic adhesive shear stress distribution predictions, E1=0.5E2, non-linear taper 62 For each of the moduli combinations, the optimal maximum to minimum scarf angle ratio was obtained. That is, the ratio that provided the adhesive shear stress distribution with the most even shear stress distribution across the joint. A plot of the optimal angle ratio with respect to adherend modulus ratio is provided in Figure 43. These results show that the optimal taper angle ratio is independent of the average taper angle or overall scarf length selected and dependant only on the adherend modulus ratio. Optimum Dmax/Dmin vs Adherend modulus ratio (E1/E2) 6.00 5.00 4.00 min D / 3.00 a_average=3deg. max a_average=5deg. D 2.00 1.00 0.00 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00 Adherend modulus ratio (E1/E2) Figure 43 Optimum Dmax/Dmin with respect to isotropic scarf joint adherend modulus ratio 3.6.1.2 Orthotropic composite adherend scarf joint The use of these techniques may also be applied to the repair of composite materials, particularly those that have been designed with a quasi-isotropic lay-up. In the present analysis, we will examine the more difficult problem of an orthotropic skin laminate with more 0q plies along the main loading direction than the quasi-isotropic lay-up. The mitigating factor for the application of gross adherend shaping is that it does not account for the adhesive stress concentrations induced by local ply orientations. Therefore, shaping of dissimilar adherend composite joints may only be applicable for special repair cases. A plot of the adherend axial stiffness within the taper section of an orthotropic 5q scarf joint is shown in Figure 44. As shown, the stiffness of the adherend dramatically increases in areas where the total lay-up is dominated by the 0q plies. This is prevalent at all locations except within the outer two plies near to the adherend tip, which contains off axis plies. 63 Stiffness variation along the tapered section of an orthotropic adherend 5 degree scarf joint 120.0 100.0 80.0 60.0 E(x) (GPa) 40.0 20.0 0.0 -0.175 0.025 0.225 0.425 0.625 0.825 1.025 x/L Figure 44 Adherend axial stiffness along the tapered section of an orthotropic 5q scarf joint Two methods for creating a dissimilar adherend patch were chosen for the trial. The first included the use of a lower modulus ply material with a lay-up identical to the parent laminate. The second included the replacement of several plies from the parent laminate with an interleaved FM73 film adhesive. Plots of the adhesive shear stress between orthotropic constant taper angle scarf joints predicted analytically and with FEA are provided in Figure 45 and Figure 46. Figure 45 shows the stress predicted analytically and with FEA within an orthotropic adherend scarf joint between a skin laminate and a skin laminate with identical lay-up, but with a reduced modulus ply material. Figure 46 shows the stress predicted analytically and with FEA within an identical orthotropic skin laminate scarf joint and between a skin laminate and the skin laminate with two 0q plies replaced by two FM73 adhesive plies. In all constant scarf angle cases, stress concentrations in the adhesive are seen adjacent to areas of predominant adherend load transfer, the 0q plies. The scope for optimising these joints using gross adherend shaping is therefore limited to options that can reduce the peak adhesive stress at these concentration locations. 64 Analytic and FEA adhesive shear stress distribution predictions, constant taper angle (5 deg.), orthotropic composite adherends, equivalent ply layup, equivalent and reduced modulus plies E1=E2 (FEA) E1=E2 (ANALYTIC) E1=0.8E2 (FEA) E1=0.8E2 (ANALYTIC) 2 1.8 1.6 1.4 1.2 12_av W 1 (x/L) 12 W 0.8 0.6 0.4 0.2 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 x/L Figure 45 Analytic and FEA adhesive shear stress predictions for the constant angle scarf joint between orthotropic composite adherends with equivalent and reduced modulus plies Analytic and FEA adhesive shear stress predictions, constant taper angle, orthotropic composite adherends, comparison between equivalent and interleaved adherend E1=E2 (FEA) E1=E2 (ANALYTIC) E1=0.77E2_interleaved (FEA) E1=0.77E2_interleaved (ANALYTIC) 2 1.8 1.6 1.4 1.2 12_av W 1 (x/L) 12 W 0.8 0.6 0.4 0.2 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 x/L Figure 46 Analytic and FEA adhesive shear stress predictions for the constant angle scarf joint between orthotropic composite adherends with an equivalent laminate and a laminate interleaved with adhesive 65 Plots of the adhesive shear stress distributions within orthotropic composite adherend scarf joints with an optimised linearly varying scarf angle predicted analytically and with FEA are provided in Figure 47 and Figure 48. It can be concluded from these results that the use of a reduced modulus ply material with an identical lay-up to the parent material offers better potential for optimisation than the use of an interleaved adhesive material within the repair laminate. In the case where an identical laminate is used for the repair patch symmetrical peaks are observed in the adhesive shear stress adjacent to the location of the plies oriented in the load direction, which transfer the majority of the load. However, when the modulus of the repair adherend is reduced the peaks increase on the half of the bond-line near to the parent adherend tip (the stiffer adherend tip), and reduce on the opposite half. When the scarf angle was linearly varied, the peaks regained the symmetry observed when the repair adherend was identical to the parent substrate. Analytic and FEA adhesive shear stress distribution predictions, E1=0.77E2 with interleaved adhesive, constant and linearly varying taper angle, orthotropic composite adherends constant scarf angle (FEA) constant scarf angle (ANALYTIC) amax/amin=1.7 (FEA) amax/amin=1.7 (ANALYTIC) 2 1.8 1.6 1.4 1.2 12_av W 1 (x/L)/ 12 W 0.8 0.6 0.4 0.2 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 x/L Note: Due to limitations in MS Excel, Dmax/Dmin is represented by amax/amin in the legend. Figure 47 Analytic adhesive shear stress distribution predictions, E1=0.77E2 with interleaved adhesive, constant and linearly varying taper angle 66 Analytic and FEA adhesive shear stress distribution predictions, E1=0.8E2 with equivalent layup, constant and linearly varying taper angle, orthotropic composite adherends constant scarf angle (FEA) constant scarf angle (ANALYTIC) amax/amin=1.6 (FEA) amax/amin=1.6 ANALYTIC) 2 1.8 1.6 1.4 12_av 1.2 W (x/L)/ 12 1 W 0.8 0.6 0.4 0.2 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 x/L Note: Due to limitations in MS Excel, Dmax/Dmin is represented by amax/amin in the legend. Figure 48 Analytic adhesive shear stress distribution predictions, E1=0.8E2 with equivalent lay- up, constant and linearly varying taper angle For the repair with interleaved structural adhesive, significant peaks were introduced adjacent to the adhesive plies. A sharp stress gradient was observed for the case when the scarf taper had a constant scarf angle, adjacent to plies connected to the interleaved adhesive layer. The use of a linearly varying scarf angle served to raise these stress concentrations further, particularly in the locations where the local scarf angle was a maximum. Therefore, optimisation of such a joint may require a complex taper profile whereby the local scarf angle is reduced adjacent to the 0q plies, and then increased in areas adjacent to less stiff plies. This type of joint would be extremely difficult to manufacture, with quite severe stress concentrations likely in the adhesive if a tight manufacturing tolerance was not met. 3.6.2 Tensile Testing of Scarf Joint Coupons 3.6.2.1 Typical Failure Modes 3.6.2.1.1 Metallic scarf joint specimens Plots of typical failure surfaces for the constant scarf angle metallic scarf joint specimens between identical adherends are shown in Figure 49. Plots of typical failure surfaces for the constant scarf angle metallic scarf joint specimens between dissimilar adherends are shown in Figure 51. Plots of typical failure surfaces for the linearly varying scarf angle 67 metallic scarf joint specimens between dissimilar adherends are shown in Figure 52. Each plot is typical of the failure surfaces observed for the remaining specimens in the set. In most cases the adhesive failure was predominantly cohesive, with adhesive remaining on both adherend surfaces. This is indicative of a failure through the adhesive due to overload, and not an interfacial failure, which can be the result of poor surface preparation prior to bonding. The exceptions to this were the optimised profile specimens, in which areas of the bond-line area were not coated with adhesive at all. An explanation of this is provided later in the section. The phase 1 5q and 10q constant scarf angle specimens shown in Figure 49 both contain areas within the bond-line where the adhesive is thinner than in other areas. This is particularly evident near to the lower scarf tips as orientated in the figure. It was concluded that failure was most likely to have initiated in these areas, before rapidly progressing through the joint. The resultant bending moment following load re-distribution was large enough for the 5q scarf specimens to fracture of the opposite scarf joint tip. No evidence of tip fracture was observed in the 10q scarf specimens. Figure 49 Phase 1 (top left) and Phase 2 (top right) metallic 5q scarf and phase 1 (bottom) metallic 10q scarf specimens between identical adherends 68 Figure 50 Close up view of a small section of the Phase 2 metallic 5q scarf specimen shown in the top right of the previous figure The phase 2 5q constant scarf angle specimens shown in the top right hand corner of Figure 49 appeared at first glance to have not failed cohesively, but upon closer inspection, small amounts of adhesive are seen in the areas that appear bare. The failure appears to still travel through the adhesive, but very close to the adherend interface. A close up view of a small section of the Phase 2 metallic 5q scarf specimens shown in the top right of Figure 49 is provided in These specimens also contained evidence of adherend tip failure, which was not unexpected, given the adherend stress at failure was within 5% of the yield stress of the aluminium 7075-T6 material. Failure appears to have begun in areas of adherend yielding, then progressed close to the yielding adherend surface, and then more predominantly through the adhesive. The phase 2 scarf joint specimens between dissimilar modulus adherends shown in Figure 51 and Figure 52 also failed cohesively. However, the specimens with a linearly varying scarf angle did not appear to reach their full load capacity. Upon close inspection of the failure surface it was discovered that areas within the joint bond-line were not wet out with adhesive. This appears to have reduced the potential load capacity of the joint. Figure 51 Phase 2 linear 5q (left) and 10q (right) scarf specimens between dissimilar adherends 69 Figure 52 Phase 2 optimised 5q (left) and 10q (right) scarf specimens between dissimilar adherends This was discovered late in the program following the static testing of the specimens, which prevented additional specimen manufacture. The curvature in the bond-line meant that pressure applied through a “pressure plate” above the joint during manufacture forced an uneven pressure distribution during cure. As such, the method of manufacture used to make all previous linear scarf joint specimens did not work as effectively for the optimised scarf joints. In hindsight, an envelope bag technique may have worked better to ensure an even bond-line throughout the joint. 3.6.2.1.2 Carbon-epoxy composite scarf joint specimens Images of the failure surfaces of the composite adherend scarf joint coupon specimens are provided in Figure 53 below. These images are representative of the entire specimen set, as the failure modes observed were similar. Failure of the specimens was predominantly via adhesive shear overload, as cohesive failure of the adhesive occurred, with adhesive present on both adherends. The scrim pattern can be seen to be observed on both adherends, even if the adhesive itself is obscured. At one end of the 5q scarf specimens, a large amount of fibre separation and fracture can be seen. No evidence of adherend failure can be seen within the 10q scarf specimens. Given the results from the metallic specimens indicated that tip fracture occurred at the opposite end of the scarf joint from which failure initiates, it is possible that the tip fracture was a result of load re-distribution and then bending following failure initiation within the adhesive near to the scarf tip. Also of note was that evidence of failure progression into the off-axis plies in the lower quartile and middle thickness of the joint can be seen in the 5q specimens. This further substantiates the notion that failure initiated near the tip opposite that where fibre fracture occurs. These images indicate that the initial failure of the joint was in the adhesive near to the scarf joint tip, with failure progression into the adherend only in areas where it is weakened by the presence of off axis (45q and 90q) plies. Failure progression through the 10q scarf joint was entirely through the adhesive bond-line. 70 5q linear scarf, plan view (top) and side view (bottom) 10q linear scarf, plan view (top) and side view Figure 53 Phase 1 5q and 10q composite scarf joint specimens 3.6.2.2 Stresses and Strains The adherend strain was measured in a range of locations for each of the specimens with only a select number containing the full set of gauges. The failure loads, maximum adherend stresses and strains and maximum adhesive shear stress for all specimen types are provided in Table 15 below. 71 Table 15 Specimen failure stress and strain summary Phase Test Set Spec. Failure Maximum Max. stress/ Maximum Max. average Max. adhesive No. Load adherend Material adherend adhesive stress/Material (kN) stress (MPa) yield stress strain (PH) stress (MPa) yield stress 1 33.9 426 1.23 N/A 37 1.06 Metallic 5q scarf joint 2 32.9 414 1.20 N/A 36 1.03 3 32.7 412 1.19 N/A 36 1.03 1 17.2 216 0.63 N/A 37 1.06 Metallic 10q scarf joint 2 19.0 239 0.48 N/A 41 1.17 3 16.7 210 0.61 N/A 36 1.03 1 1 25.6 N/A N/A 5936 28 0.80 Composite 5q scarf joint 2 24.8 N/A N/A 5734 27 0.77 3 25.8 N/A N/A 5963 28 0.80 1 14.6 N/A N/A 3377 31 0.89 Composite 10q scarf joint 2 14.6 N/A N/A 3379 31 0.89 3 14.2 N/A N/A 3279 31 0.89 1 36.3 484 0.96 N/A 42 1.02 Metallic identical adherend 5q 2 37.3 498 0.99 N/A 43 1.05 scarf joint 3 35.7 476 0.95 N/A 41 1.01 1 35.8 478 0.95 N/A 41 1.01 Metallic dissimilar adherend 2 35.8 478 0.95 N/A 41 1.01 0.96linear 5q scarf joint 3 35.7 476 0.95 N/A 41 1.01 1 14.4 193 0.38 N/A 33 0.80 Metallic dissimilar adherend 2 2 16.1 214 0.43 N/A 37 0.90 linear 10q scarf joint 3 16.4 219 0.44 N/A 37 0.91 1 23.4 312 0.62 N/A 27 0.66 Metallic dissimilar adherend 2 24.0 319 0.63 N/A 28 0.67 optimised 5q scarf joint 3 29.0 387 0.77 N/A 34 0.82 1 Test machine malfunction – No data. Metallic dissimilar adherend 2 11.3 150 0.30 N/A 26 0.63 optimised 10q scarf joint 3 13.3 177 0.35 N/A 30 0.74 Legend nomenclature for Phase 1 testing The following legend nomenclature applies to Figures 55-57, 61-63 and 67-68. In some cases the identifier “–sg$” was omitted when referring to far field strain gauge data or data from an extensometer used to measure the shear strain near the tip. The identifier used for each specimen test data was as follows: #d&-D-sg$, where, # = scarf angle, 5 or 10, denoting the 5q and 10q scarf joint, & = specimen identification number, 1,2 or 3, D = adherend material identifier, m denoting a metallic adherend or c denoting a composite adherend, $ = strain gauge identification number, 1,2 or 3. In some cases the legend also refers to “Data sheet (E or YM=%%GPa)” for adherend strain gauge results and FM300 data (G=….) for adhesive shear strain results. This refers to an idealised material property curve using established property data. 72 Legend nomenclature for Phase 2 testing The following legend nomenclature applies to Figures 58-60. The identifier used for each specimen test data was as follows: D_#__&-$ where, D adherend material identifier, “al” denotes an aluminium to aluminium joint and “alti” denotes an aluminium to titanium joint, # = scarf angle, 5 or 10, denoting the 5q and 10q scarf joint, & = specimen identification number, 1,2 or 3, $ = adherend material that the strain gauge is applied, “al” denotes that the strain gauge is applied to the aluminium adherend and “ti” denotes that the gauge is applied to the titanium adherend (Note that this identifier has been omitted in the case of the identical adherend joint). The solid lines used in these plots show the idealised material property curves for aluminium and titanium in the linear region. 3.6.2.2.1 Phase 1 metallic scarf joint specimen results Plots of the adherend far field stress and strain for the phase 1 metallic scarf joint specimens are provided in Figure 54. It can be seen in these plots that for the 5q scarf joint specimens, the adherend begins to plastically yield prior to the adhesive failure in the scarf joint. No adherend yielding was observed in the 10q scarf joint specimens, as the adherend was lowly stressed when the specimen failed. Adherend far field stress vs strain Adherend far field stress vs strain 500 300 250 400 ) ) MPa 200 ( MPa ( 300 150 200 av lied Stress 100 lied Stress 5d1-m 10d1-m 5d2-m pp 10d2-m pp 100 5d3-m A 50 A 10d3-m Data sheet (E=72 GPa) Data sheet (E=72 GPa) 0 0 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000 0 500 1000 1500 2000 2500 3000 3500 4000 4500 Adherend strain (PH) Adherend strain (PH) Figure 54 Adherend far field stress and strain for the phase 1 metallic scarf joint specimens Plots of the tip strain in the phase 1 metallic scarf joint specimens are provided in Figure 55. Strain gauges were placed adjacent to both adherend tips. Interestingly, the tip strain in both adherend tips within the 5q scarf joint specimens was similar, but within the 10q specimens, the tip strains seen in the tips were different. This discrepancy was 73 observed for all three specimens tested. As noted in the failure investigation, the bond-line thickness within the 5q specimens was constant along the bond-line, but for the 10q specimens, the bond-line thickness tapered linearly with one end slightly thinner than the other. This may have caused the load to re-distribute in the manner observed in these results. Adherend tip strain vs Load Adherend tip strain vs Load 4000 8000 10d1-m-sg2 5d1-m-sg2 3500 10d2-m-sg2 5d2-m-sg2 7000 10d3-m-sg2 5d3-m-sg2 3000 10d1-m-sg3 5d1-m-sg3 10d2-m-sg3 6000 5d2-m-sg3 5d3-m-sg3 2500 10d3-m-sg3 5000 5d1-m-sg1 (far field) 10d1-m-sg1 (far field) 2000 4000 1500 Strain (microstrain) Strain 3000 Strain (microstrain) Strain 1000 2000 500 1000 0 0 0 2 4 6 8 10 12 14 16 18 20 0 5 10 15 20 25 30 35 40 Load (kN) Load (kN) Figure 55 Adherend tip strain wrt applied load for the phase 1 metallic scarf joint specimens Plots of the adhesive shear stress and shear strain for the phase 1 metallic specimens are provided in Figure 56 below. The data obtained from the metallic 5q scarf joint specimens matches the published data for FM300 fairly well. It shows that a large amount of adhesive plastic yielding occurs prior to specimen failure. The data obtained from the metallic 10q specimens showed that much less yielding of the adhesive occurs prior to specimen failure. This is typical of an adhesive loaded less predominantly in shear, which is the case when the scarf angle is increased. The maximum shear stress at failure was similar for both specimen types. Adhesive shear stress vs shear strain Adhesive shear stress vs shear strain 40 45 35 40 30 35 25 30 25 20 (MPa) (MPa) 12 20 W 5d1-m 12 15 W 5d2-m 15 10d1-m 5d3-m 10 10d2-m FM300 data (G=380 MPa, W = 35MPa) 10 p 10d3-m 5 FM300 data (G=380 MPa, = 35MPa) 5 Wp 0 0 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.00 0.05 0.10 0.15 0.20 0.25 0.30 J 12 J12 Figure 56 Adhesive shear stress wrt shear strain for the phase 1 metallic scarf joint specimens 74 3.6.2.2.2 Phase 2 metallic scarf joint results A plot of the adherend far field stress and strain for the phase 2 metallic scarf joint specimens with identical adherends is provided in Figure 57. It can be concluded from this result that by using an adherend material with a higher yield stress can reduce the amount of yielding prior to failure. The 5q scarf specimens still showed signs of yielding in the adherend, but to a much lesser extent than the phase 1 specimens. No evidence of adherend yielding was observed in the 10q scarf specimens. Adherend far field stress vs strain (5° scarf) 600.00 500.00 400.00 300.00 al_5_01 al_5_02 al_5_03 200.00 Data sheet, YM=72GPa. 100.00 Applied Stress (MPa) 0.00 1000 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 Adherend strain (PH) Figure 57 Adherend far field stress and strain for the phase 2 metallic scarf joint specimens between identical material adherends Plots of the adherend far field stress and strain for the phase 2 metallic specimens with dissimilar material adherends with a constant scarf angle are shown in Figure 58. The adherend far field stress and strain results for the same specimens except with a linearly varying scarf angle are provided in Figure 59. The 5q optimised scarf joint specimens had an equivalent taper length to the linear 5q scarf joint specimens, and similarly, the 10q optimised specimens had an equivalent taper length to the linear 10q scarf joint specimens. It can be concluded from these results that the assumed Young’s modulus for the titanium and aluminium used was fairly accurate. Clearly, the aluminium-to-titanium stiffness ratio was approximately 0.65. No evidence of adherend yielding was observed in this data. It can also be concluded that by using dissimilar adherends to construct the scarf joint, the failure stress of the linear scarf joint does not reduce significantly. 75 Adherend far field stress vs strain (5° scarf) Adherend far field stress vs strain (10° scarf) 600.00 250.00 500.00 200.00 ) ) 400.00 MPa 150.00 ( MPa ( 300.00 100.00 alti_10_01-al alti_5_01-al alti_10_02-al 200.00 alti_5_02-al lied Stress alti_10_03-al lied Stress alti_5_03-al Al. Data sheet, YM=72GPa. Al. Data sheet, YM=72GPa. pp 50.00 alti_10_01-ti pp alti_5_01-ti A alti_10_02_ti A 100.00 alti_5_02_ti alti_10_03-ti alti_5_03-ti Ti. Data sheet, YM=110GPa. Ti. Data sheet, YM=110GPa. 0.00 0.00 500 0 500 1000 1500 2000 2500 3000 3500 4000 -1000 0 1000 2000 3000 4000 5000 6000 7000 8000 Adherend strain (PH) Adherend strain (PH) -50.00 Figure 58 Adherend far field stress and strain for the phase 2 metallic scarf joint specimens between dissimilar adherends with a linear taper profile Adherend far field stress vs strain (5° optimised scarf) Adherend far field stress vs strain (10° optimised scarf) 450.00 250.00 400.00 200.00 350.00 ) ) 300.00 MPa 150.00 ( MPa ( 250.00 200.00 100.00 alti_10_02-al alti_5_01-al 150.00 alti_10_03-al alti_5_02-al lied Stress lied Stress alti_5_03-al Al. Data sheet, YM=72GPa. pp 100.00 Al. Data sheet, YM=72GPa. 50.00 alti_10_02_ti pp alti_5_01-ti A A alti_10_03-ti alti_5_02_ti 50.00 alti_5_03-ti Ti. Data sheet, YM=110GPa. Ti. Data sheet, YM=110GPa. 0.00 0.00 -500 0 500 1000 1500 2000 2500 3000 3500 -1000 0 1000 2000 3000 4000 5000 6000 Adherend strain (PH) Adherend strain (PH) Figure 59 Adherend far field stress and strain for the phase 2 metallic scarf joint specimens between dissimilar adherends with an optimised taper profile 3.6.2.2.3 Phase 1 composite scarf joint specimen results Plots of the adherend far field stress and strain for the phase 1 composite scarf joint specimens are provided in Figure 60. No adherend material non-linearity was observed in most of the results from either the 5q or 10q scarf joint specimens, except for one specimen, 5d3-c, which stopped registering an increase in strain with applied stress, indicating the possibility of an early disbond in the tip region. 76 Adherend far field stress vs strain Adherend far field stress vs strain 250 400 350 200 ) ) 300 MPa ( 150 MPa ( 250 200 av 100 150 lied Stress lied Stress 10d1-c 100 pp 5d1-c 50 10d1-c pp 5d3-c A 10d3-c A 50 Data sheet (E=72 GPa) Data sheet (E=72 GPa) 0 0 0 1000 2000 3000 4000 5000 6000 0 500 1000 1500 2000 2500 3000 3500 Adherend( strain) (PH) Adherend strain( ) (PH) Figure 60 Adherend far field stress and strain for the phase 1 composite scarf joint specimens Plots of the tip strain in the phase 1 composite scarf joint specimens are provided in Figure 61. It can be concluded from these results that the strain in the tips was similar at the lower loads for both the 5q and 10q specimens. However, there is clear evidence of load drop off in the tip region as failure is approached. The strain diverges significantly from the far field stress value. The adherend may have fractured prior to the catastrophic scarf joint adhesive failure, or the tip disbonded prior to complete specimen failure. Adherend tip strain vs Load 8000 Adherend tip strain vs Load 8000 10d1-c-sg2 10d2-c-sg2 6000 10d3-c-sg2 10d1-c-sg3 6000 10d2-c-sg3 10d3-c-sg3 4000 10d1-c-sg1 (far field) 4000 5d1-c-sg2 5d3-c-sg2 Strain (microstrain) 2000 5d1-c-sg3 Strain (microstrain) 5d3-c-sg3 2000 5d1-c-sg1 (far field) 0 0 0 5 10 15 20 25 30 0 2 4 6 8 10121416 Load (kN) Load (kN) Figure 61 Adherend tip strain wrt applied load for the phase 1 composite scarf joint specimens Plots of the adhesive shear stress and shear strain for the phase 1 composite specimens are provided in Figure 62 below. The data obtained from the 5q specimens shows that the maximum shear stress for the material was not attained prior to specimen failure. This could indicate that the adherend failure occurred prior to the catastrophic failure of the adhesive. The data obtained from the 10q specimens showed that the specimens failed when the adhesive maximum shear stress was attained. In a similar fashion to the metallic 10q specimens, very little shear yielding was observed prior to adhesive failure, which is typical of adhesive joints loaded under combined shear and normal direction loading. 77 Adhesive shear stress vs shear strain Adhesive shear stress vs shear strain 40 40 35 35 30 30 25 25 20 20 (MPa) (MPa) 12 12 W 15 W 15 10d1-c 5d1-c 10d2-c 10 5d2-c 10 10d3-c 5d3-c FM300 data (G=380 MPa, W = 35MPa) 5 FM300 data (G=380 MPa, = 35MPa) 5 p Wp 0 0 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.00 0.05 0.10 0.15 0.20 0.25 0.30 J J12 12 Figure 62 Adhesive shear stress wrt shear strain for the phase 1 composite scarf joint specimens 3.7 Conclusions and Recommendations for Future Work 3.7.1 Conclusions 3.7.1.1 Analytic and Finite Element Analyses An analytic technique has been developed to facilitate the optimal design of scarf joints between isotropic adherends with dissimilar material properties. The technique specifies a linearly varying scarf angle that generates a characteristic scarf profile for a given adherend modulus ratio. FEA validated the results obtained from the analytic modelling. Analytic and finite element modelling demonstrated the dependence of the adhesive stress distribution on local ply orientation within the adherend. This showed that the effectiveness of gross adherend shaping in optimising dissimilar adherend scarf joints between orthotropic adherends was limited to cases where the repair laminate could maintain the identical lay-up to the parent adherend. By reducing the modulus of the repair adherend plies, the adhesive shear stress was shown to concentrate predominantly in the bond-line near to the parent (stiffer) adherend tip. The magnitudes of the peaks became equal when an optimal, linearly varying scarf angle was used. In contrast, the linear variation of the scarf angle could not minimise the adhesive shear stress concentration if the repair laminate lay-up was not identical to the parent lay-up. The use of interleaved adhesive layers in the repair laminate served to raise the stress concentration in the adhesive adjacent to the interleaved plies. The use of a linearly varying scarf angle raised the magnitude of the stress peak on the side of the joint where the local scarf angle was a maximum. The use of a more complex taper profile to minimise ply induced stress concentrations in the adhesive may be possible, but may be very difficult to manufacture in practice. 78 3.7.1.2 Tensile Testing It can be concluded from the results of scarf joint coupon testing that the predominant failure mode was shear overload of the adhesive. In many cases, particularly for shallow angle scarf joints, this was complicated by adherend yielding or composite fibre fracture near the scarf tips. It can also be concluded from these results that as the scarf angle is increased, adhesive is loaded less predominately in shear and more in tension. As such, there is less shear deformation prior to specimen failure in these joints. The adhesive within the metallic specimens with shallow angles deformed significantly in shear prior to failure. It can be concluded from these results that when dissimilar materials are bonded with a ductile adhesive, the static strength of the joints do not change significantly with the scarf profile. However, given large stress concentrations are introduced in the bond-line between dissimilar adherends, the benefit of adherend may be better seen with fatigue testing. Unfortunately, in some of the optimised profile specimens areas within the bond-line were not coated with adhesive. This was discovered late in the experimental program, which prevented additional specimen to be manufactured. The curvature in the bond-line meant that pressure applied through a “pressure plate” above the joint during manufacture forced an uneven pressure distribution during cure. As such, the method of manufacture used to make all previous linear scarf joint specimens did not work as effectively for the optimised scarf joints. In retrospect, an envelope bag technique may have worked better to ensure an even bond-line throughout the joint. As such, it was not possible to determine if the optimisation would have enhanced the failure strength of the linear scarf joints between dissimilar adherends in the current program. 3.7.2 Recommendations for Future Work 3.7.2.1 Analytic and Finite Element Analyses It was concluded that peaks in the adhesive stress distribution were observed adjacent to the stiffer regions of the adherend, particularly near to the 0q plies. As such, the macro profiling of the adherend may only be applicable to quasi-isotropic lay-up configurations. To improve applicability, work is needed to develop a closed form solution for the optimal adherend shape that gives a near constant shear stress distribution in the adhesive. An improved shape would have a shallow scarf angle adjacent to the stiffer areas within the adherend and use a steeper angle adjacent to the less stiff regions. As a matter of fact, following the completion of the work program, but before the thesis was completed, the beginnings of this work was presented in [45]. These analyses did not account for any of the thermal and moisture absorption mismatches that may induce further stress concentrations in the adhesive. Given that the 79 hot-wet operational environment is often the most critical for the survivability of the bonded joints, these effects may need inclusion for future analyses and testing. 3.7.2.2 Testing Future work is needed to determine the effect of optimisation on the static strength of the bonded joints in ambient and hot-wet environments. A method of manufacture, possibly using an envelope bag, is required to ensure that the bond-line thickness throughout the optimised scarf joint is uniform prior to testing. It was concluded from the testing that the static strength of the bonded joints between dissimilar materials at room temperature was not significantly below that of the scarf joints between identical adherends. However, the influence of optimisation may show improved fatigue performance, particularly in raising the load in which a disbond initiates. Current design philosophy within the bonded repair manual as used by the RAAF to verify scarf repair design does not consider fatigue performance, but given the overlap repair designs do, it is likely that this may require consideration within future flush repair designs. 80 4. Reinforcement Methods for Scarf Joint Strength Improvement 4.1 General Several different methodologies have been tried to improve the strength of scarf joints and scarfed repair patches using methods of reinforcement or add on features. These include the addition of an overlap patch that fits over the scarf joint once it has been bonded in place, the use of through thickness pinning near to the scarf joint tip to improve peel resistance, and also the use of an embedded patch. The work herein describes the use of through thickness pinning and the development of an embedded patch methodology. 4.2 Through Thickness Pinning 4.2.1 Introduction Through thickness pins (TTP) were inserted to reinforce the structure near the adherend tips to improve peel resistance in this area. In order to test this hypothesis, the phase 1 metallic specimens described in Section 3.5.2.1 were reinforced with TTP’s. The pin layout, pin type and the method of pin insertion are described below. 4.2.2 Pin type Several different pin types were proposed, including composite tubes, metallic tubes, chamfered metal drill bits and solid shanks of metal drill bits. For ease of manufacture and material availability, the solid and chamfered shanks of a small metal drill bit were used. The specifications of the pins used are provided in Table 16. Table 16 Through thickness pinning properties Parameter Value Pin material Solid carbide steel Pin diameter, Ipin (mm) 0.66 mm (#71 drill) Pin length, Lpin 3 mm Adhesive material K-106 Araldite Cure 30qC for 3 hours 4.2.3 Pin layout Two design requirements were considered when deciding on a pin layout. The main requirements were that: the pins be located close to the tip, the pins have sufficient adherend material to bond to. 81 These are competing requirements as one demands the pin as close to the tip as possible and the other demands the pins be moved away from the tip where the adherend is thicker. This indicated that a compromise was required. Standard rivet spacing rules [46] for aircraft design were applied to determine the initial pin layout. The “zig-zag” pattern used herein has also been used to effect in the through thickness stitching of composite lap joints to improve fatigue resistance [47]. This layout has also been shown to be effective in the stitching of composite laminates to improve through thickness strength and damage tolerance. The layout for the specimens is shown in Figure 63 with layout parameter values for the 5q and 10q scarf joint configurations provided in Table 17. Ledge_tip (typ.) L (typ.) diagonal Lspacing (typ.) Lscarf Ledge_side (typ.) Figure 63 Through thickness pin layout Table 17 Through thickness pin layout parameters Parameter 5q scarf 10q scarf Tip edge spacing, Ledge_tip (mm) 2 2 Diagonal spacing, Ldiagonal (mm) 3 3 Horizontal spacing, Lspacing (mm) 3 3 Side edge spacing, Ledge_side (mm) 2 2 Scarf length, Lscarf (mm) 36 18 4.2.4 Method of pin insertion The method of pin insertion is shown conceptually in Figure 64. 82 1. Prepare baseline specimen 3. Cut pin to size 2. Drill holes from a drill bit 4. Apply adhesive to pin and hole * Drawing not to scale. 5. Insert pins and cure Figure 64 Method of pin insertion 4.2.5 Results and Discussion Two sets of scarf joint specimens with through thickness pins were tested to failure in tension, one set for the 5q and 10q scarf specimens respectively. The load, far field strain, tip strain and crosshead displacement were recorded for each specimen at a sample rate of 1 Hz. The loading rate was 0.2 mm/min using displacement control. Observations relating to the failure mode were made following testing. 4.2.5.1 Typical Failure Mode Images of the failure surfaces of the 5q metallic scarf joint specimens with TTP’s are provided in Figure 65. It can be concluded from these images that the failure mode of the specimens was similar to the failure mode of the metallic scarf joint specimens without the pins. These specimens failed by shear overload of the adhesive. The tip fracture observed on the 5q specimens may have occurred after the adhesive had already overloaded. Figure 65 5q metallic scarf joint with through thickness pins 83 4.2.5.2 Stresses and Strains The adherend strain was measured using far field strain gauges located 30-40 mm away from the scarf joint tip and tip gauges located within the taper region just aft of the pins at both scarf joint tips. Plots of the adherend far field and tip stress and strain for both the 5q and 10q specimens are provided in Figure 66 and Figure 67. The failure loads, maximum adherend stresses and maximum adhesive shear stress for all specimen types are provided in Table 18 below. Table 18 Specimen failure stress and strain summary Test Set Spec. Failure Maximum Max. stress/ Max. average Max. adhesive No. Load adherend Material adhesive stress/Material (kN) stress (MPa) yield stress stress (MPa) yield stress 1 33.9 426 1.23 37 1.06 Metallic 5q scarf joint 2 32.9 414 1.20 36 1.03 3 32.7 412 1.19 36 1.03 1 17.2 216 0.63 37 1.06 Metallic 10q scarf joint 2 19.0 239 0.48 41 1.17 3 16.7 210 0.61 36 1.03 1 31.7 399 1.16 35 1.00 Metal 5q scarf joint with TTP 2 30.0 377 1.09 33 0.95 Metal 10q scarf joint with TTP 1 18.4 231 0.67 40 1.15 The adherend stress and strain plots show that the presence of the pins had a minimal effect on the load capacity and the strain field within the scarf joint coupons. No significant difference in the failure load was found between the specimens with pins and the specimens without pins for both the 5q and 10q scarf joint specimens. This is indicative of the adhesive overloading in shear prior to the fracture of the adherend tips. The area of the pin was so small in comparison to the total bond area that their effect on the load capacity of the bond was minimal. The weakening of the adherend tip by drilling holes to allow the installation of the pins did not lower the joint strength significantly. Adherend far field stress vs strain Adherend far field stress vs strain 500 250 400 200 300 150 (MPa) (MPa) av V av 200 V 5d1-m 100 10d1-m 5d2-m 10d2-m 5d3-m 10d3-m 5d4-m (pinned) 100 50 10d4-m (pinned) Data sheet (E=72 GPa) Data sheet (E=72 GPa) 0 0 0 10002000300040005000600070008000900010000 0 500 1000 1500 2000 2500 3000 3500 4000 H (PH) ( ) 1 H1 PH Figure 66 Adherend far field stress and strain in the phase 1 metallic scarf joint specimens with and without TTP’s 84 Adherend tip strain vs Load Adherend tip strain vs Load 8000 4000 7000 3500 6000 3000 5000 2500 4000 2000 5d1-m-sg2 10d1-m-sg2 5d2-m-sg2 10d2-m-sg2 1500 Strain (microstrain) 3000 5d3-m-sg2 10d3-m-sg2 5d1-m-sg3 10d1-m-sg3 5d2-m-sg3 Strain (microstrain) 10d2-m-sg3 2000 1000 5d3-m-sg3 10d3-m-sg3 5d4-m-sg2 (pinned) 10d4-m-sg2 (pinned) 500 1000 5d4-m-sg3 (pinned) 10d4-m-sg3 (pinned) 5d4-m-sg1 (far field) 10d4-m-sg1 (far field) 0 0 0 5 10 15 20 25 30 35 02468101214161820 Load (kN) Load (kN) Figure 67 Adherend tip stress and strain in the phase 1 metallic scarf joint specimens with and without TTP’s The fact that the pins were not able to raise the static load capacity of the scarf joint was not unexpected, as the adhesive within bonded and mechanically fastened joints typically don’t complement each other to provide a stronger joint. The adhesive typically begins load transfer at much lower adherend displacements than mechanically fastened joints, thus the adhesive is close to failure by the time the mechanically fastened joint begins to support a significant portion of the load. The pins in this case, were not designed to carry the large shear loads at the adherend interfaces, and were only in place to improve peel resistance. More significantly, the pins did not lower the load capacity of the joint. As such, their application in arresting crack growth through the adhesive in the scarf joint remains possible. It has been noted previously in [25] that cracks in the scarf joint adhesive propagate uncontrollably once initiating in the tip region. This is because the shear stress distribution along the joint is approximately uniform, thus making the conditions for continued crack growth favourable. The pins may improve the resistance to crack growth. Due to time constraints, and the project’s emphasis on improving the static strength of the scarf joint, no fatigue testing was carried out on scarf joints with and without through thickness pins to test this hypothesis 4.3 Modified Biscuit Joints within Scarf Joints 4.3.1 Introduction The scarf joint with an embedded patch insert, or biscuit joint, may provide the means to improve the load capacity of flush repair joints such as the scarf joint. This approach has been found to be unique, as it has long been thought to be practical impossibility to machine the slot (refer to Figure 1) in the internal region of the adherend to facilitate the repair. As such, there has simply not been any effort using either experimentation or finite element modelling of representative joints to determine the extent, if any, of the load capacity increase that may be possible if the biscuit joint was employed in this application. 85 Advanced CNC machining techniques and optimised patch layout may make this repair method feasible, particularly in locations close to a free edge where the biscuit can be inserted from the side. This part of the project was split into three phases; x Phase 1 was a pilot experimental study to determine if a benefit is possible if a biscuit insert was placed into an internal cavity within the scarf joint, then x Phase 2 entailed a more extensive experimental study examining the effect of insert material and joint geometry on the load capacity of a representative joint between metallic adherends, and then x Phase 3 used extensive linear and non linear modelling of the joint to determine the stress distribution throughout the joint and also to determine if the strength of the new joint could be predicted. The phase 1 and 2 specimen descriptions are provided in Section 4.3.2. The phase 3 FEM description is provided in Section 4.3.3. The phase 1 and 2 experimental results and discussion are provided in Section 4.3.4. The phase 3 FE results and discussion, including the failure load predictions are provided in Section 4.3.5. 4.3.2 Phase 1 and 2 Specimen Descriptions 4.3.2.1 Phase 1 Pilot Experimental Study General A hybrid joint specimen was designed to test the hypothesis that by combining a scarf joint with a double lap joint, the overall strength of the joint may be increased. In order to do this, two specimens were designed and manufactured. These were a: baseline scarf joint made from metallic adherends, and a metallic scarf joint with a composite patch embedded in the adherend structure. Specimen geometry, properties and method for manufacture The specimen geometry for the baseline specimen and the hybrid joint specimen are shown in Figure 68 , with the geometry parameters provided in Table 19 and a photograph of the specimens provided in Figure 69. Lspecimen W Lscarf Lscarf W Lspecimen Lpatch * Drawing not to scale. * Drawing not to scale. tpatch tadherend tadherend Figure 68 Pilot study scarf joint specimen with (left) and without (right) an embedded patch 86 Table 19 Geometry parameters for hybrid joint test specimens Parameter Scarf joint Hybrid joint Scarf length, Lscarf (mm) 54 54 Adherend thickness, tadherend (mm) 9.5 9.5 Scarf angle (q) 10 10 Specimen length, Lspecimen (mm) 200 200 Specimen width, W (mm) 11.0 11.0 Patch length, Lpatch (mm) n/a 111 Patch thickness, tpatch (mm) n/a 0.47 Figure 69 Pilot study scarf joint specimen with (top) and without (bottom) an embedded patch The adherend, adhesive and patch properties are provided in Table 20, Table 21 and Table 22 respectively. Table 20 Metallic adherend material properties Parameter Value Material type Al. 2024-T3 Young’s modulus, E (MPa) 72x103 Poisson’s ratio, Q12 0.3 Yield stress, Vyield (MPa) 345 Ultimate stress, Vu (MPa) 485 Table 21 Adhesive material properties Parameter Value Material type Hysol EA9396C-2 Cure RT with clamp pressure applied for 8 hours then 1 hour at 100qC Maximum shear stress at RT, Wp (MPa)* 20.7 * Obtained from manufacturer’s data sheet [48]. 87 Table 22 Composite patch material properties Parameter Value Material type AS4 Cure 177qC for 60 minutes at 90 psi (full vacuum) Number of plies, n 21 Lay-up [+45,-45,90,0,0,0,+45,0,0,- 45,90,-45,0,0,+45,0,0,0,90,- 45,+45] Ply thickness (nominal), tply 0.12 mm* Laminae principal direction Young’s 128* modulus, E1 (GPa) Laminae transverse direction Young’s 13* modulus, E2 (GPa) Laminae principal Poisson’s ratio, Q12 0.3* Laminae transverse Poisson’s ratio, Q21 0.03* Laminate principal direction Young’s 74.8** modulus, Ex (GPa) Laminate transverse direction Young’s 38.6** modulus, Ey (GPa) Laminae principal Poisson’s ratio, Qxy 0.35** Laminate transverse Poisson’s ratio, Qyx 0.18** Ultimate strain in ply, Hu (PH) 7992* * Reference: Obtained for AS3501 from [41]. ** Reference: Obtained using CLT from the laminae properties [49]. The procedure for manufacturing the baseline specimen and the embedded patch specimen is shown conceptually in Figure 70 and Figure 71 respectively. Step 1 B A Step 2 Step 3 Clamp Force Force B A Step 4 Clamp B A * Drawing not to scale. Step 1: Machine rectangular aluminium plate to size. Step 2: Machine scarf joint adherends from rectangular aluminium plate, then surface treat bonding surfaces. Step 3: Apply paste adhesive with scrim cloth to scarf joint adherend bonding faces. Step 4: Apply clamping pressure to scarf joint specimen until adhesive has cured. Figure 70 Metallic adherend scarf joint manufacture using paste adhesive 88 Step 2 Step 1 B A Step 3 Step 4 Clamp Force B A Force * Drawing not to scale. Clamp Step 5 B A Step 1: Machine rectangular aluminium plate to size. Step 2: Machine scarf joint adherends from rectangular aluminium plate, then surface treat bonding surfaces. Step 3: Cure a composite patch to match adherend slot thickness and surface treat bonding surfaces. Step 4: Apply paste adhesive with scrim cloth to scarf joint adherend and patch bonding faces. Step 5: Apply clamping pressure to scarf joint specimen until adhesive has cured. Machine sides so that patch is flush with adherend. Figure 71 Metallic adherend scarf joint with composite insert manufacture using paste adhesive The test results and discussion is provided in Section 4.3.4. 4.3.2.2 Phase 2 Experimental Study General In order to validate the concept of a hybrid scarf joint with a biscuit insert, four different specimen types were designed. Each comprised aluminium adherends joined with structural film adhesive with either an aluminium or titanium insert. Each specimen was tested to failure using an Instron static test machine with the load and crosshead displacement recorded during tests. Specimen Manufacture The hybrid joint geometric configuration is shown in Figure 72. 12 specimens with a width of 25 mm were manufactured and testing. The geometric parameters for each specimen type are provided in Table 23. The adherends comprised three layers of 3 mm aluminium 7075-T6, bonded together with FM73 adhesive. This simplified the manufacture of the specimens, as there was no need to machine the slots and develop bonding procedures. All bond-lines within the joint were co-cured in a single step using the FM73 manufacturer’s standard cure cycle, i.e., one hour at 120qC. Ljoint Dscarf Linsert Dtaper Lspecime Figure 72 Hybrid biscuit joint specimen geometry 89 Table 23 Hybrid specimen geometry parameter values Geometry description Abbreviation Geometry A Geometry B Specimen length Lspecimen 280 mm 280 mm Scarf angle Dscarf 5q 5q Biscuit taper angle Dtaper 5q 3.5q Insert length without taper Linsert 30 mm 10 mm Adherend thickness tadherend 9.5 mm 9.5 mm Insert thickness tinsert 3 mm 3 mm The thickness of the bond-lines within each specimen type was measured using a travelling microscope at three locations. An average thickness for each of the tapered bond-lines for each specimen type is provided in Table 24. The bond-line thickness for specimens of like type was similar as the specimens were manufactured from the same larger plate then cut to specimen size after cure. The adhesive used contained a scrim cloth embedded within the adhesive film to control the bond-line thickness. Evidence of the scrim effectiveness is seen in the relatively small variance between the average adhesive thicknesses throughout the tapered bond-lines within the joint. Table 24 Average thickness and length of the tapered bond-lines for each specimen type Specimen type Bond-line Average thickness Length (mm) (mm) Geometry A with an Upper scarf 0.12 31 aluminium insert Lower scarf 0.11 31 LH biscuit taper 0.11 33 RH biscuit taper 0.11 33 Geometry B with an Upper scarf 0.10 30 aluminium insert Lower scarf 0.12 30 LH biscuit taper 0.12 38 RH biscuit taper 0.11 38 Geometry A with an Upper scarf 0.12 31 titanium insert Lower scarf 0.10 31 LH biscuit taper 0.10 33 RH biscuit taper 0.10 33 Geometry B with an Upper scarf 0.10 30 titanium insert Lower scarf 0.08 30 LH biscuit taper 0.10 38 RH biscuit taper 0.11 38 The test results and discussion is provided in Section 4.3.4. 4.3.3 Phase 3 FEM Details General FEA was used to verify the use of two failure load prediction theories currently in use for the design of adhesively bonded joints. Linear FEA using MSC.Nastran v2005 was used to verify initially the stress concentrations induced throughout the joint bond-lines under 90 tensile loading. A maximum shear stress criterion was used to predict the failure load. This criterion is currently used for the prediction of failure in the scarf joint [40], thus it was considered reasonable that the failure predictions may have been accurate given the hybrid joint is also a flush repair joint configuration. The failure load of overlap joints is typically predicted using a maximum shear strain criterion [40]. The design allowable for the FM73 adhesive bond-lines within these joints was generated using thick adherend lap joint specimens in a separate study [38]. It was found that the specimens failed at a near constant maximum shear strain. As such, non- linear (NL) FEA using MSC.Marc v2005 was conducted to determine the applied stress that causes the local adhesive shear strain to exceed the design allowable for the adhesive. Model Description A macro was written in PCL to rapidly develop the required models allowing certain parameters within the joint to vary, such as scarf angle, biscuit taper angle and the biscuit insert length, and then automatically creating the geometry and mesh. Loads, boundary conditions, and material properties are applied manually. Two joints with slightly different geometric parameters were created, with the values of the parameters provided in Table 25. Each geometric configuration was subsequently copied so as the effect of different insert material properties could be assessed. In total, four finite element models were created for this exercise. Table 25 Hybrid specimen geometry parameter values used in the FEM Geometry description Abbreviation Geometry A Geometry B Scarf angle Dscarf 5q 5q Biscuit taper angle Dtaper 5q 3.5q Insert length without taper Linsert 30 mm 10 mm Adherend thickness tadherend 9 mm 9 mm Insert thickness tinsert 3 mm 3 mm Bond-line thickness tbond 0.2 mm 0.2 mm In the area of the bond-line, the element length was reduced to half the bond-line thickness, with the element length progressively increased away from the joint. The tip regions were the most difficult to mesh so care needed to be taken to avoid excessive element aspect ratios. Triangular elements were used throughout to also mitigate element distortion. The elements near a typical tip region are shown in Figure 73 Adhesive Adherend tip 91 Figure 73 Typical mesh near to the scarf and taper tip region The aluminium 7075-T6, titanium Ti-6Al-4V and FM73 adhesive material properties used in both the linear and non-linear FEA are provided in Table 26. Plane strain element properties were applied. Table 26 Material properties used within the FEA Property Alum. 7075-T6 Ti-6Al-4V FM73 Young’s modulus (GPa) 72 110 1.15 Shear modulus (GPa) N/A N/A 0.503 Poisson’s ratio 0.3 0.3 0.3 Maximum shear stress (MPa) N/A N/A 41 Maximum shear strain N/A N/A 0.557 Yield stress (MPa) 462 880 N/A Analyses Description Two finite element solutions were obtained for each of the models. Initially, a linearly elastic solution within MSC.Nastran v2005 was used to determine the extent of the stress concentrations that exist in the adhesive prior to any adhesive yielding. The failure stress of the joint was predicted on the basis of the shear stress in the adhesive exceeding the allowable. A nonlinear solution was also obtained using MSC.Marc v2005 to determine the post yielding mechanical behaviour of the adhesive. This facilitated the calculation of the shear strain post yield. Failure of the joint was predicted on the basis of the maximum shear strain exceeding the material allowable. The non-linear FEA parameters used are provided in Table 27. Table 27 Non-linear FEA parameters Parameter Value Solver MSC.Marc v2005 Non linear geometric effects Large displacement/Large strain Maximum time step 0.05 Total time 1 Applied Stress (MPa) 1000 Plastic material behaviour Elastic perfectly plastic The results and discussion following the analyses are provided in Section 4.3.5. 4.3.4 Phase 1 and 2 Experimental Results and Discussion 4.3.4.1 Phase 1 Pilot experimental study The pilot study included the tensile test results from the baseline scarf joint and embedded patch specimens described in Section 4.3.2.1. 92 Failure modes The scarf specimen failed cohesively with the adhesive remaining on both adherends. There was no evidence of tip fracture in the adherend. The hybrid joint failed initially in the scarf joint, but did not result in complete specimen failure. Final failure of the bond-line between the patch and the adherend resulted in final failure of the specimen. Stresses and Strains The maximum load, Pult, adherend stress at failure, Vult, and the average specimen Young’s modulus, Eaverage, taken from the load vs. crosshead displacement curve shown in Figure 74, is provided in Table 28. It was concluded from these results that the embedded patch raised the load capacity of the joint without affecting the overall Young’s modulus of the specimen. Clearly, in this case the overall repair length was increased significantly to achieve this increase in load capacity. However, the testing demonstrated that designers of flush repairs may use this methodology to increase the capacity of the scarf joint, without necessarily making the taper angle shallower. Further testing as a part of the phase 2 experimental test program described in Section 4.3.2.2, aimed to measure the effectiveness of the embedded patch in comparison to a shallower tapered scarf joint. Table 28 Adherend strength and stiffness parameter measurements for the thick adherend baseline scarf specimen and the embedded patch specimen Parameter Baseline Hybrid Pult (kN) 15.57 21.95 Vult (MPa) 156 220 Eaverage (GPa) 65 67 Load vs Crosshead Displacement 20 18 16 14 12 10 Load (kN) 8 Baseline 6 Hybrid 4 2 0 0.0 0.1 0.2 0.3 0.4 0.5 0.6 Crosshead Displacement (mm) Figure 74 Load with respect to crosshead displacement for the thick adherend baseline scarf specimen and the embedded patch specimen An extensometer was mounted to the specimen to measure the adhesive shear strain in a manner described in Section 3.5.3.3, with a plot of the load with respect to shear strain for both specimens provided in Figure 75. It can be concluded from these results that the applied stress in which the adhesive in the scarf begins to yield is similar. However, in the 93 case of the scarf joint with an embedded insert, a further load increase was possible after the adhesive first yielded. The adhesive within the straight scarf joint yielded plastically in a manner similar to the phase 1 metallic 5q scarf joints described in 3.6.2. Load vs Shear strain 20 18 16 14 12 10 Load (kN) 8 Baseline 6 Hybrid 4 2 0 0.0 0.1 0.2 0.3 0.4 0.5 J12 Figure 75 Load with respect to shear strain for the thick adherend scarf joint and the embedded patch specimen 4.3.4.2 Phase 2 Experimental study The experimental study included tensile test results from the hybrid joint specimens described in Section 4.3.2.2. Stresses and Strains Plots of the load with respect to crosshead displacement for each of the specimens are provided in Figure 76 and Figure 77. Also shown is the predicted applied stress at failure of a 5q scarf joint, assuming that the joint fails when the average shear stress in the adhesive reaches its FM73 room temperature allowable (Table 9). W allowable V applied , sin scarf cos DD scarf Where, W allowable =41 MPa and D scarf =5q. Therefore, V applied 472 MPa. 94 Aluminium Insert - Applied Stress vs Crosshead Displacement Geometry A, Specimen 01 Geometry A, Specimen 02 Geometry A, Specimen 03 Geometry B, Specimen 01 Geometry B, Specimen 02 Geometry B, Specimen 03 500 Applied stress=472MPa 400 300 200 Applied Stress (MPa) Applied Stress 100 0 012345678 Crosshead displacement (mm) Figure 76 Applied stress wrt crosshead displacement for specimens with an aluminium insert Titanium Insert - Applied Stress vs Crosshead Displacement Geometry A, Specimen 01 Geometry A, Specimen 02 Geometry A, Specimen 03 Geometry B, Specimen 01 Geometry B, Specimen 02 Geometry B, Specimen 03 500 Applied stress=472MPa 400 300 200 Applied Stress (MPa) Stress Applied 100 0 012345678 Crosshead displacement (mm) Figure 77 Applied stress wrt crosshead displacement for specimens with an titanium insert It can be seen from these results that specimens with Geometry B with a titanium insert failed at the highest load. These specimens failed at a stress approximately 10% greater than specimens with Geometry A with a titanium insert. Specimens with an aluminium insert with Geometry A and B failed at similar stresses, approximately 5% lower than Geometry B specimens with an insert made from titanium. The failure stress of Geometry B specimens with a titanium insert exceeded the straight scarf failure stress by 2-3%. A summary of the failure loads, as well as the failure predictions using FEA are provided with Figure 91 within Section 4.3.5. Failure Progression The joint specimens all failed in a similar mode under displacement control. A schematic of the failure progression is provided in Figure 78. In all cases, failure initiated in the 95 upper or lower scarf bond-line. Failure then progressed almost instantaneously into the adjacent biscuit taper region. The load shed dramatically as the joint separated but did not fail completely. The adherend ligament adjacent to the failed scarf continued to support load through the joint. Significant bending occurred as a result of the eccentricity introduced to the load path as it passed through the outer ligament. Final failure was via tensile overload of the remaining adherend ligament at a load that was approximately half of the maximum. Initial failure in the upper scarf with progression to the biscuit Final failure in the adherend ligament adjacent to the failed f Figure 78 Schematic of the failure progression through the joint The failed specimens are shown in Figure 79 to Figure 82. Only one specimen from each type was loaded to the final failure displacement after the specimen had initially failed in the scarf. It was established that once the initial failure load was reached, the load shed dramatically as significant plastic yielding occurred. As such, it made possible photographs showing both the initial and final failure modes of the specimens. In all cases, the adhesive failed cohesively with adhesive remaining on both of the joining adherends after failure. A typical bond surface after failure is shown in Figure 83. Specimen 03 Initial failure Specimen 02 Final Initial failure failure Specimen 01 Figure 79 Failure of joints with Geometry A with an aluminium insert 96 Specimen 03 Initial failure Specimen 02 Final Initial failure failure Specimen 01 Figure 80 Failure of joints with Geometry B with an aluminium insert Specimen 03 Initial failure Specimen 02 Final Initial failure failure Specimen 01 Figure 81 Failure of joints with Geometry A with a titanium insert Specimen 01 Initial failure Specimen 02 Final Initial failure failure Specimen 03 Figure 82 Failure of joints with Geometry B with a titanium insert 97 Tip of outer adherend Tip adjacent to titanium insert Figure 83 Initial failure surface in the scarf of specimen 01 with Geometry B with a titanium insert It is possible that local bending was induced into the specimen by the insert taper, which may have weakened the specimens. A plot of the y direction deformation in an exaggerated view as observed with the LFEA is provided in Figure 84. This shows that the deformation within the insert taper causes a bending of the specimen which if it were significant, would raise the stress in the upper scarf region, the site of failure. If the taper were reversed, or the insert made longer, the influence of the insert taper on the upper scarf bond-line stress would be minimised. It should be noted, however, that the scarf joint without the insert deforms in bending by a similar amount to the scarf joint with the insert. This indicates that the location of the maximum stress due to bending may be important. This may not be as significant a problem for the 3D scarf repair there is typically much more out of plane constraint provided by the surrounding structure. 98 y Aluminium insert, Maximum y deformation = 0.004mm when 100 MPa li d y Titanium insert, Maximum y deformation = 0.006mm when 100 MPa li d Figure 84 Exaggerated view of the y displacement along the specimen when loaded in tension 4.3.5 Phase 3 Finite Element Results and Discussion Linear Elastic Finite Element Results Linear elastic FEA was conducted for all joint specimen types with an applied load of 100 MPa. The shear stress distributions along the tapered adhesive bond-lines are provided in Figure 85 and Figure 86. The shear stress in the bond-lines parallel to the load direction was negligible. In all cases, shear stresses have been normalised against the maximum shear stress allowable for FM73 adhesive at room temperature (41 MPa), and the position along the bond-line has been normalised with respect to the length of the bond-line under consideration. The bond-line lengths for each specimen type are provided in Table 24. 99 Normalised shear stress in the tapered bondlines of specimens with an aluminium insert (Applied stress =100MPa) Geom. A-Upper scarf Geom. A-LH Biscuit taper Geom. A-Lower scarf Geom. A-RH Biscuit taper Geom. B-Upper scarf Geom. B-LH Biscuit taper Geom. B-Lower scarf Geom. B-RH Biscuit taper 0.35 h 0.3 0.25 0.2 0.15 0.1 the design allowable shear stress (41MPa) stress shear allowable design the 0.05 Maximum shear stress in the tapered bondline adhesive normalised wit normalised adhesive bondline tapered the in stress shear Maximum 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Position along the bondline normalised with the bondline length Figure 85 Shear stress in the tapered bond-lines of joints with an aluminium insert Normalised shear stress in the tapered bondlines of specimens with an titanium insert (Applied stress =100MPa) Geom. A-Upper scarf Geom. A-LH Biscuit taper Geom. A-Lower scarf Geom. A-RH Biscuit taper Geom. B-Upper scarf Geom. B-LH Biscuit taper Geom. B-Lower scarf Geom. B-RH Biscuit taper 0.35 h 0.3 0.25 0.2 0.15 0.1 the design allowable shear stress (41MPa) stress shear allowable design the 0.05 Maximum shear stress in the tapered bondline adhesive normalised wit normalised adhesive bondline tapered the in stress shear Maximum 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Position along the bondline normalised with the bondline length Figure 86 Shear stress in the tapered bond-lines of joints with titanium insert It can be concluded from these results that the location of the maximum shear stress was strongly influenced by the joint geometry. For specimens with a steeper biscuit taper angle (Geometry A), the maximum shear stress occurred in the biscuit taper bond-line. 100 However, for specimens with a shallow biscuit taper (Geometry B), the location of the maximum shear stress occurred in the outer scarf bond-lines. The failure load predictions using the maximum stress theory are provided in Figure 91. Non-Linear Finite Element Results A maximum shear strain failure criterion may also be used to predict joint failure and determine mechanical behaviour once the adhesive in the joint yields. Modelling of the non-linear joint behaviour required an assumption regarding the plastic behaviour of the adhesive once the yield stress is reached. In this case, the adhesive was assumed to retain a constant yield stress until the maximum shear strain was reached, that is elastic-perfectly-plastic [50]. An idealised stress-strain curve was proposed that has the same area under the curve as that under the true shear stress-strain curve. The idealised curve for the FM73 film adhesive was derived within [38] with the maximum shear strain obtained under ambient testing conditions provided in Table 26. Following the analyses, it was shown that for all specimen types the allowable shear strain was exceeded in the tip region of the outer scarf bond-line. Plots of the adhesive shear stress in the outer scarf bond-line when (i) the maximum shear stress was first exceeded locally; (ii) when the applied stress was equivalent to the failure stress determined through static testing and (iii) when the allowable shear strain was exceeded locally for all specimen types are provided in Figure 87 and Figure 88. The scarf tip is on the left hand side of the plots. Plots of the adhesive shear strain in the outer scarf bond-line when (i) the applied stress was equivalent to the failure stress determined through static testing and (ii) when the allowable shear strain was exceeded locally for all specimen types are provided in Figure 89 and Figure 90. 101 Normalised shear stress in the scarf bondline of specimens with an aluminium insert Geom. A-Maximum Shear Strain Geom. A-Maximum Shear Stress Geom. B-Maximum Shear Strain Geom. B-Maximum Shear Stress Geom. A-Failure Stress Geom. B-Failure Stress 1.2 1 0.8 0.6 0.4 design allowableshear stress (41MPa) 0.2 Maximumtapered shear bondline stress in the adhesive normalised with the 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Position along the bondline normalised with the bondline length Figure 87 Shear stress in the scarf bond-line of joints with an aluminium insert Normalised shear stress in the scarf bondline of specimens with a titanium insert Geom. A-Maximum Shear Strain Geom. A-Maximum Shear Stress Geom. B-Maximum Shear Strain Geom. B-Maximum Shear Stress Geom. A-Failure Stress Geom. B-Failure Stress 1.2 1 0.8 0.6 0.4 design allowable shear stress (41MPa)designallowable shear stress 0.2 Maximum stress inshear bondline the tapered adhesive normalised with the 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Position along the bondline normalised with the bondline length Figure 88 Shear stress in the scarf bond-line of joints with a titanium insert 102 Shear strain in the scarf bondline at the specimen failure stress Geom. B, Aluminium insert Geom. A, Aluminium insert Geom. A, Titanium insert Geom. B, Titanium insert 0.09 0.08 0.07 0.06 0.05 0.04 Maximum shear strain 0.03 0.02 0.01 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Position along the bondline normalised with the bondline length Figure 89 Shear strain in the scarf bond-line at the initial failure stress for each specimen type Shear strain in the scarf bondline when the local shear strain exceeds the maximum shear strain allowable (J=Jmax) Geom. B, Aluminium insert Geom. A, Aluminium insert Geom. A, Titanium insert Geom. B, Titanium insert 0.7 0.6 FM73 Shear strain allowable = 0.557 0.5 0.4 0.3 Maximum shear strain 0.2 0.1 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Position along the bondline normalised with the bondline length Figure 90 Shear strain in the scarf when the local shear strain exceeds the material allowable The adhesive shear stress results show that at the failure load for all specimen types a large proportion of the adhesive has plastically deformed. It can also be seen from the same plots that when the predicted deformation exceeds the shear strain allowable, the 103 entire bond-line has deformed plastically. For all specimen types, this occurred at an applied stress that was higher than the failure stress determined through static testing. It can be seen in the adhesive shear strain plots that the shape of the shear strain distribution changes as the load is increased. At the failure stress, the peaks in the shear strain at either end of the bond-line are approximately the same for all specimen types. However, as more of the bond-line yields, the local shear strain near the outer scarf tip becomes the dominant peak. This may indicate that a greater amount of bending with increasing load is induced. The failure predictions using the maximum strain theory are provided in Figure 91. Failure Predictions The failure stresses from the static testing and each of the maximum stress and maximum strain predictions are provided in Figure 91. Static Test Results and Failure Predictions Specimen Failure Stress Failure Prediction using Maximum Shear Stress Criterion Failure Prediction using Maximum Shear Strain Criterion 700 600 500 400 300 Applied Tensile Stress (MPa) Stress Tensile Applied 200 100 0 Geom. A, Geom. A, Geom. A, Geom. B, Geom. B, Geom. B, Geom. A, Geom. A, Geom. A, Geom. B, Geom. B, Geom. B, Al. insert, Al. insert, Al. insert, Al. insert, Al. insert, Al. insert, Ti. insert, Ti. insert, Ti. insert, Ti. insert, Ti. insert, Ti. insert, 01 02 03 01 02 03 01 02 03 01 02 03 Figure 91 Static test results and failure predictions Failure load predictions comparisons between Geometry A and B The maximum stress theory predicts Geometry B with the shallower insert taper to have an 8% higher failure load than Geometry A when the insert material is aluminium. A similar load capacity increase is predicted using the maximum strain theory. However, the test result shows only 2% strength increase. The maximum stress and strain predict that Geometry B will have a 20% increase in strength than Geometry A when a titanium insert is used. The static test results show a 10% increase. 104 Summary It may be concluded that the predictions using a maximum shear strain and a maximum shear stress criterion represent the upper and lower bounds for the failure stress respectively. Certainly, in many cases the use of the maximum stress theory gives very conservative failure stress estimates. However, relying on the maximum shear strain theory is likely to provide non-conservative predictions making it of little use in design. Finally, only one of the specimen types was able to achieve a marginal increase in the failure stress of the linearly tapered equivalent 5q scarf joint (472 MPa) despite the predictions using the maximum shear strain theory indicating a marked improvement. 4.4 Conclusions and Recommendations for Future Work 4.4.1 Conclusions 4.4.1.1 Through thickness pinning Pinning of the metallic specimens to improve peel resistance was found to not significantly affect the mechanism of joint failure or the load capacity of the joint. Further testing is recommended to determine if pinning can reduce the rate of damage progression under fatigue loading. 4.4.1.2 Modified Biscuit Joints within Scarf Joints 4.4.1.2.1 Phase 1 Pilot Experimental Study It was concluded from the pilot experimental study that by embedding a patch insert within a scarf joint, the overall load capacity of the joint may be increased. As such, this methodology may be used to improve the strength of flush bonded joints without necessarily making the taper angle shallower. However, the length of the joint with the embedded patch was much longer than the scarf joint that was used as a comparison. 4.4.1.2.2 Phase 2 and 3 Experimental Study and Finite Element Analyses Static tensile tests of four types of two dimensional scarf joints with biscuit inserts were conducted with the results compared to predicted failure loads using finite element stress and strain calculations. Each of the specimen types failed through the adhesive, thus the failure predictions of two adhesive failure theories were compared. These were a maximum shear stress theory and a maximum shear strain theory, with the material properties generated using thick adherend specimen tests [38]. It may be concluded that the predictions using a maximum shear strain and a maximum shear stress criterion represent the upper and lower bounds for the failure loads respectively. The maximum shear stress criterion gave conservative predictions whilst the maximum shear strain theory predicted joint strengths that were considerably higher than the test results. Two geometric parameters and one material parameter were varied, resulting in four specimen types. Of interest were the effects of the biscuit insert taper angle, un-tapered 105 length and Young’s modulus on the strength of flush joints. The two geometric parameters were linked in that a shallow angle taper had to be used in combination with a short un-tapered length to ensure that the overall joint dimension matched the length of a straight scarf for comparison. Specimens with a shallow insert taper failed at a higher load than when a steeper taper was used with a longer insert length. Specimens with a higher modulus insert (titanium) failed at a higher load than when the insert material matched the parent adherend (aluminium). Results from the static testing showed that one of the specimen types was able to show a 2- 3% strength improvement over the equivalent length 5q scarf joint, with all specimen types achieving a failure stress that was at least 90% of the equivalent 5q scarf joint. Given this joint has never been used before for aerospace application, the joint designs shown herein may be improved through shape and layout optimisation. However, it is possible that bending was induced into the specimen through the method used to taper the insert, which may have weakened the specimens. If the taper were reversed, or the insert made longer, the influence of the insert taper on the upper scarf bond-line stress may be minimised. 4.4.2 Recommendations for Future Work 4.4.2.1 Through thickness pinning The pins may provide a benefit in retarding crack growth under fatigue loading, or even in increasing the energy required to open the crack once it has initiated or the tip damaged. It is recommended to perform fatigue testing of scarf joints with a disbond inserted at the tip of the joint with and without pins to determine if the resistance of the scarf joint to crack growth is improved. 4.4.2.2 Modified Biscuit Joints within Scarf Joints The phase 2 testing showed that the patch with an insert was at least as strong as the baseline scarf joint, but did not demonstrate a significant strength improvement. This may be due to the insert taper introducing a small bending moment to the specimen that then overloaded the scarf joint. It is recommended to taper the insert differently or increase the insert length to minimise bending of the specimen, particularly in the region of the scarf joint. It is also recommended to test in hot-wet conditions, such that the adhesive is sufficiently degraded so that the adherend is not loaded near to the material yield stress. These conditions are often the most critical for bonded joint designs. 106 5. Impact resistance and damage tolerance of scarf repairs 5.1 Introduction It has been known that the presence of a bond-line defect near to the tip region of the scarf joint will cause a local shear and peel stress concentration in the adhesive, thus lowering the joint’s efficiency [1, 51]. As such, during fabrication of scarf joints, care is always taken to ensure that the thin tip is not damaged prior to bonding. Once the scarf joint adherends have been bonded, the thin tip region is supported by the adhesive and the opposite adherend, often with an overlap doubler covering the tip region, therefore protecting the tip from accidental damage. The final repair is considered as part of the aircraft structure, and hence should provide the same level of impact resistance and damage tolerance as the parent structure. Given that the scarf joint efficiency reduces when the tip is damaged, it stands to reason that the scarf joint would be somewhat more vulnerable to impact damage than its surrounding structure despite protection measures described earlier. A knock to the tip may cause a disbond or fracture the tip, which may reduce the repair strength considerably. The work aimed to determine whether scarf repairs are more vulnerable to low velocity impacting events than its parent structure. To this end, a series of compression after impact test specimens, with and without scarf joints, were tested in accordance with SACMA recommended test method SRM-94 [2]. Various failure prediction theories were employed to interpret the test results. The experimental test program description and failure analyses description is provided in Section 5.2. The results from testing and analyses are provided in Section 5.3, and the conclusions and recommendations for future work are provided in Section 5.4. 5.2 Experimental Investigation and Analyses 5.2.1 General To determine the impact resistance and damage tolerance of composite scarf joints, as well as to test the failure prediction methodologies, compression-after-impact (CAI) tests were carried out on scarf specimens. Impact energies varied between 50% and 120% of the energy required to impart barely visible impact damage (BVID) as defined in the SACMA test method. BVID in this case is defined as the damage caused when struck by a low velocity impact with a total energy of 6.7 J per mm of specimen thickness. For example, to induce BVID to 3 mm thick specimens, an impact of 20.1J would be required. Furthermore, undamaged specimens were tested in tension to failure, which were obtained by cutting the coupons to size from spare CAI specimens. This allowed a comparison between the compression baseline results and the tension results. 107 In addition, a strain survey was performed, with strain gauges installed to measure the adherend strain in both the undamaged and damaged part of the specimen. These results would help determine the extent of stiffness degradation by impact damage. These specimens were initially tested within the CAI test fixture, before having their width reduced and then tested in tension. In total, three types of specimens were tested, each with a number of specimens that contained varying amounts of low velocity impact damage, ranging from no damage to damage that was over and above the definition for BVID. These included: x CAI specimens made of 21-ply carbon/epoxy, with lay-up [+45, -45, 90, 0, 0, 0, +45, 0, 0, -45, 90, -45, 0, 0, +45, 0, 0, 0, 90, -45, +45]. The prepreg used to construct the adherend was CYTEC 977-3/IM7, Lot 303706815. x Constant angle (5q) scarf joints between 21-ply carbon/epoxy adherends with identical lay-up and ply material to the parent adherend specimens. The adhesive used was FM73. x Constant angle (5q) scarf joints between 21-ply carbon/epoxy adherends with identical lay-up to the parent adherend specimens, with a 2 ply carbon/epoxy overlap doubler with lay-up [0]2. The adhesive used the bond the scarf joint and the overlap doubler was FM73. The CAI specimen descriptions are provided in Section 5.2.2. The tension specimen descriptions are provided in Section 5.2.3. The strain surveys used to determine the stiffness reduction caused by the impact are provided in Section 5.2.4. The CAI and tension test procedures are provided in Section 5.2.5. The failure prediction methodology description is provided in Section 5.2.6. 5.2.2 CAI Specimen Descriptions In total, 15 of the parent adherends, 15 of the 5q scarf joints and 15 of the 5q scarf joints with doubler specimens were manufactured. Of these, 13 parent adherends, 14 scarf joints, and 12 scarf joints with doubler specimens were tested to failure. Geometry The specimen geometry for each of the specimen types is provided in Figure 92. All dimensions are in millimetres, with the 1-axis denoting the load direction and the principal direction for the laminate, and the 2-axis denoting the transverse direction. The strain gauge locations denoted in the figure are those recommended within the SACMA test method [2]. However, in the present case two methods were used to deduce axial strain data. The first method utilised strain gauges that were located near to the SACMA recommended locations, while the second utilised back-to-back extensometers mounted centrally between the SACMA recommended strain gauge locations. 108 Figure 92 Parent adherend (top left), 5q scarf joint (top right) and 5q scarf joint with doubler CAI specimen geometry Material properties The ply material used to construct the adherends was carbon/epoxy unidirectional tape, CYTEC 977-3/IM7, manufactured in Greenville, Texas, USA. The composite adherend material properties are provided in Table 29. The adhesive used to bond the adherends was FM73, with its properties provided within Table 9 of Section 3.5.2.3. 109 Table 29 Composite adherend material properties Parameter Value Material type CYTEC 977-3/IM7 Cure 177qC for 6 hours at 500 kPa positive pressure (partial vacuum) Number of plies, n 21 Lay-up [+45, -45, 90, 0, 0, 0, +45, 0, 0, -45, 90, -45, 0, 0, +45, 0, 0, 0, 90, -45, +45] Laminae Properties Ply thickness (nominal), tply 0.13 mm* Laminae principal direction Young’s 160* modulus, E1 (GPa) Laminae transverse direction Young’s 16.5* modulus, E2 (GPa) Laminae principal Poisson’s ratio, Q12 0.3* Laminae transverse Poisson’s ratio, Q21 0.03* Ultimate strain in ply, Hu (PH) 10000* Laminate Properties Laminate principal direction Young’s 94.5 modulus, Ex (GPa) Laminate transverse direction Young’s 48.8 modulus, Ey (GPa) Laminae principal Poisson’s ratio, Qxy 0.35 Laminate transverse Poisson’s ratio, Qyx 0.18 Bond-line thickness, adherend thickness and scarf angle measurements The bond-line thickness and the scarf angle were measured using a travelling microscope at three locations along the scarf with the average of these values presented herein. The thickness of the adherends and the doubler were measured using a micrometer at four locations along the specimen. The results are provided in Table 30. 110 Table 30 CAI specimen geometry measurements Specimen ID Average bond-line Scarf angle (q) Adherend thickness (mm) thickness (mm) Parent adherend specimen measurements (CFRP) CFRP-01 N/A N/A 2.75 CFRP-02 N/A N/A 2.79 CFRP-03 N/A N/A 2.75 CFRP-04 N/A N/A 2.78 CFRP-05 N/A N/A 2.75 CFRP-06 N/A N/A 2.70 CFRP-07 N/A N/A 2.80 CFRP-08 N/A N/A 2.82 CFRP-09 N/A N/A 2.73 CFRP-10 N/A N/A 2.78 CFRP-11 N/A N/A 2.76 CFRP-12 N/A N/A 2.80 CFRP-13 N/A N/A 2.75 5q scarf joint specimen measurements (SC) SC-01 0.176 4.72 2.75 SC-02 0.251 5.16 2.71 SC-03 0.385 5.72 2.69 SC-04 0.405 4.72 2.74 SC-05 0.345 No data. 2.75 SC-06 0.287 4.52 2.73 SC-07 0.302 5.69 2.75 SC-08 0.252 4.75 2.74 SC-09 0.193 6.07 2.72 SC-10 0.235 5.58 2.70 SC-11 0.178 5.45 2.77 SC-12 0.233 6.55 2.75 SC-13 0.248 4.73 2.74 SC-14 0.275 5.87 2.72 5q scarf joint with doubler specimen measurements (SCD) Specimen ID Average bond-line Scarf angle (q) Adherend Mid thickness (mm) thickness thickness (mm) (mm) SCD-01 0.226 5.49 2.70 3.21 SCD-02 0.252 6.39 2.58 3.22 SCD-03 0.216 5.27 2.62 3.21 SCD-04 0.307 5.31 2.60 3.11 SCD-05 0.224 5.71 2.69 3.14 SCD-06 0.227 5.70 2.55 3.23 SCD-07 0.162 6.33 2.77 3.20 SCD-08 0.180 4.39 2.68 3.16 SCD-09 0.257 5.74 2.71 3.23 SCD-10 0.276 6.77 2.80 3.21 SCD-11 0.268 4.52 2.68 3.09 SCD-12 0.165 4.81 2.77 3.20 5.2.3 Tension coupon testing As a means of comparing the compression results of undamaged specimens with the tension results, additional coupons were tested to failure in tension. In total, one parent adherend specimen, two scarf specimens and two scarf specimens reinforced with a doubler were tested. The specimens were cut to size from spare compression-after-impact specimens. Both load and crosshead displacement were recorded. Post-failure investigation was performed to examine the failure modes. 111 The specimens had a length equal to the CAI specimen length (150 mm), but each coupon had a reduced width of 25 mm. The parent adherend coupons were machined from specimen CFRP-14, the scarf coupons were machined from specimen SC-15, and the scarf with doubler coupons were machined from specimen SCD-13. 5.2.4 Strain surveys To determine the extent of load redistribution of the impacted scarf joints a more detailed strain survey was conducted. Initially, two spare CAI specimens, one with no scarf and one with a scarf, were fitted with additional strain gauges, and then tested to approximately 40% of the failure load within the CAI test fixture. The specimens were then reduced in width to examine further the local modulus variation throughout the damaged specimen, without altering the location of the strain gauges. The specimen dimensions, impact energy, damage size and the location of the axial strain gauges relative to the impact site for all specimens tested are provided in Table 31. Figure 93 shows the reference coordinate frame used to location of the impact site and the strain gauges. Table 31 Specimen details relating to stiffness reduction due to low velocity impact testing Specimen Potential Damage size Impact site FF SG SG impact SG 10 mm ID energy (j) (mm2) location site (mm) from (mm) ** ** impact (mm) ** CFRP-13 12.90 799 X50,Y75 X65, Y102 X50,Y75 X58, Y80 (CAI) SC-14 12.76 1015 **X50, Y65 X63, Y112 X50, Y63 X53, Y63 (CAI) CFRP-13 12.90 799 X30,Y75 X45, Y102 X30,Y75 X38, Y80 (tension) SC-14 12.76 1015 **X30, Y65 X43, Y112 X30, Y63 X33, Y63 (tension) ** 8 mm aft of the scarf tip within the tapered region of the adherend. 100 mm 60 mm 150 mm 150 mm Y Y X X CAI specimen Tension specimen Figure 93 Strain survey reference coordinate system for the compression (left) and tension specimen (right) 112 5.2.5 Test and data reduction procedures Compression after impact testing by definition requires a low velocity impact, representative of an in-service event such as a tool drop, and then a residual compression strength test. After impact but prior to the compression test, the specimens were inspected with an ultrasonic technique (C-scan) to determine the size of the delamination region. Additional specimens were manufactured to enable sectioning of the impacted region to determine the type of damage that was imparted. Following the compression test, the specimens were sectioned to expose the damaged region to examine the failure modes. The impacting procedures are described in Section 5.2.5.1. The ultrasonic inspection and damage size measurement procedures are provided in Section 5.2.5.2. The methods of sectioning and viewing the delamination regions are described in Section 5.2.5.3. The compression test procedures are provided in Section 5.2.5.4. The tension test procedure is provided in Section 5.2.5.5. 5.2.5.1 Impacting The apparatus used to perform the impacting is shown in Figure 94. Essentially, the specimen was clamped at four locations around the edges to a rigid metallic base with a cut-out underneath the mid-section of the specimen. The cut-out is smaller than the specimen to ensure that there is enough space with which to clamp the specimen, however, it is large enough to ensure that reasonable delamination growth can occur following impact without being resisted by interaction with the support base. Clamping ensures that the specimen does not move during the impact event, thus allowing the drop weight to impart energy into the specimen. Drop weight including tup Digital tape measure Clamps Support base Figure 94 Impacting apparatus 113 The drop weight consists of a tup, support structure for the tup and extra weights that are added to give the desired total drop weight. In the present study, enough weights were added to give a desired total impacting mass of approximately 1.5kg. Spherical tups of 0.5” and 1” inches in diameter were used. A larger sized tup generally results in a larger amount of energy that can be imparted into the specimen before the dent on the impacted surface becomes large enough to become visible. As such, the delamination zone is typically large when the dent becomes visible. Any requirement within a certification document that specifies a 1” tup be used is restrictive, and typically refers to the design of important load carrying structure. Conversely, a smaller tup generally imparts a visible dent at much lower energies, and typically generates a much smaller delamination zone around the damage for a given impact energy. The drop weight slides on a lubricated rail on ball bearings. Prior to impact, it is cranked up to the desired height, which was measured using an electronic tape measure. The potential energy of the drop weight, Epotential, after it has been cranked to the desired height is then given by, E potential 'hmg , (5.1) where, m=impacting mass, g=the acceleration due to gravity and 'h=the height that the drop weight was dropped. A force transducer was fitted to the tup to measure the force on the tup during the impact event with respect to time. This allowed the calculation of the momentum transfer between the tup and the specimen during impact, and then assuming all potential energy that was stored in the drop weight before it was released, converted to kinetic energy just prior to contact with the specimen, the energy absorbed by the specimen could be calculated. The velocity of the drop weight at the impact site, vin, is given by Equation 5.2 below, in 2 ' hgv . (5.2) The area under the tup force with respect to time curve is calculated to determine the impulse, and consequently the momentum transfer from the tup to the specimen. The change in velocity of the tup is then given by the following equation, v A ' m , (5.3) where, A=the area under the tup force with respect to time plot during the impact event. The absorbed energy, Eabsorbed, is then calculated as the difference between the kinetic energy in the drop weight before the impact event and the energy in the drop weight immediately after the impact event. It is given by Equation 5.4 below, 114 2 vvvm 2 E inin ' absorbed 2 . (5.4) These equations were programmed into software that calculated the absorbed energy from the force transducer output and the input parameters, 'h and m for each impact event. The impacting results are provided in Section 5.3.1. 5.2.5.2 Ultrasonic inspection and damage size measurement The impact damage within the internal structure of the composite is typically much greater than that which is seen on the external surface near to the impact site. Depending on the energy of the impact, the thickness of material and the degree of constraint around the impact site, the damage can be extensive. Typical delamination damage caused by a low velocity impact to a composite laminate can be seen in Figure 95. This type of damage can be detected particularly well using ultrasonic techniques. Figure 95 Cross section of AS6/2220-3 showing damage due to a 40J impact (Source: Walsh Paul J. Carbon Fibres. ASM Handbook: Volume 21 Composites. ASM International (2001):pp35-9.) Ultrasonic inspection relies on sound energy passing into the structure, and then reflecting when material of a different media is encountered. The type of inspection used within this project was pulse-echo, which is shown schematically in Figure 96. For example, in the case of an undamaged specimen, signals reflect from the top surface, and then pass into the specimen before reflecting from the back face of the specimen. In the case of a damaged specimen, the signal reflects from the top face, and then passes into the specimen before reflecting from any air gaps introduced via delamination. If there is energy remaining in the signal (that is, if the air gaps are relatively small), it passes further into the specimen before reflecting from further air gaps or the back face. As such, an indication is given as to whether a delamination has occurred and the location of the delamination relative to the specimen top surface. 115 Received signals Generate signal and receive Generated signal Figure 96 Pulse-echo ultrasonic inspection technique The scans used in this report were A-scans and C-scans. The A-scan technique requires an operator to manually position a hand held probe, whilst interpreting the signal received on a viewing platform such as an oscilloscope. For damage detection, a marker was used to identify on the panel where the operator found evidence of damage. The C-scan technique effectively performs an A-scan at every point on the specimen surface, utilising a computer controlled gantry to position the probe whilst scanning and storing the results from each of the scans. This process, a well as a description of a single sweep B-scan is shown in Figure 97. The results are then calibrated with an RGB colour map, to show areas with and without evidence of damage. Figure 97 Ultrasonic inspection scanning [52] For this project, C-scans were performed where possible, and A-scans were used when the C-scanning machine was out of commission. In both cases, a digital image of the specimen, showing areas with and without damage indications was extracted. The damage size was then calculated using image processing software, IMAGEJ [53]. Example digital images obtained using the A-scan and C-scan technique are shown in Figure 98 below. The C-scan image shows “Time of flight” data, indicating the time that the signal takes to enter into 116 the specimen then be reflected and measured with the probe. The red to grey region indicates the baseline back face reflection, with the other colours indicating that the signal had reflected before the back face, most probably due to a delamination in the region. Figure 98 Digitised images using the A-scan (left) and the C-scan (right) technique The damage assessment results using the ultrasonic technique are provided in Section 5.3.1. 5.2.5.3 Sectioning Further understanding of the type of damage that was caused by the low velocity impact event to the parent adherend and scarf joint specimens was obtained by cutting through the damaged region so that a damaged cross section was exposed, and the internal damage within the specimen able to be viewed. In order to make the internal damage to the composite more visible, fluorescent dye was painted on the damaged section, and then wiped clean. The fluorescent dye remaining on the section had flowed into the delamination air gaps. As such, when a UV light illuminated the damaged cross section, the areas where the fluorescent dye flowed, (delamination areas) glowed brightly. As such, it was possible to view the exact locations within the specimen internal structure that had been damaged by the impact event. A damaged cross section of a parent adherend specimen coated in fluorescent dye can be seen in Figure 99 with and without UV illumination. Also shown is a morphed image created using image processing software [53] showing the location of the damage within the specimen internal structure. 117 Normal light UV light Morphed image Figure 99 Cross section of a damaged area within a parent adherend specimen The damage assessment results using the sectioning technique are provided in Section 5.3.1. 5.2.5.4 Compression Testing Compression testing was performed using a fixture designed and manufactured to conform to the SACMA test method recommendations. This fixture allows specimens with length and width dimensions, 150 mm and 100 mm respectively, to be clamped around all edges of the specimen during compression loading. This fixture is shown in Figure 100, but more information regarding the fixture can be found in [2]. The fixture was clamped into the test machine using hydraulic end grips. Figure 100 SACMA CAI Test fixture 118 Compression loading was performed using an Instron test machine with a 250 kN load cell within a 500 kN capacity frame. The load and crosshead displacement were output to the data acquisition system. The loading rate was 0.5 mm/min under displacement control. Two methods were used measure the axial strain and amount of bending induced within the specimen during loading. These included back to back strain gauges installed in far field locations from the damage area, approximately 1” from the specimen top and side edges. For other cases, the strain was measured using back to back extensometers which were located approximately 1” from the top edge of the specimen, and approximately mid-width. The extensometer and strain gauge specifications are provided in Table 32. The type of axial strain method used and the location of the strain measuring device for each of the specimens is summarised in Table 33 to Table 35 below. Table 32 Extensometer and strain gauges parameters Parameter Value Extensometer Type Instron model flex 003 Stroke r2.5 mm Gauge length, GL 12.5 mm Strain gauge parameters Type Kyowa KFG-3-350-C1-23L3M3R axial strain gauge with wires attached Gauge factor 2.10 Gauge length 3 mm Gauge resistance 350 : Number of wires 3 Wire length 3 m Table 33 Axial strain measurement types and location of strain gauges for the CFRP specimens Specimen ID Axial strain Strain gauge locations** measurement type Parent adherend specimen s (CFRP) CFRP-01 Extensometer Back to back, ** X50, Y25 CFRP-02 Strain gauge Back to back, ** X50, Y25 CFRP-03 Extensometer Back to back, ** X50, Y25 CFRP-04 Extensometer Back to back, ** X50, Y25 CFRP-05 Strain gauge Back to back, ** X50, Y25 CFRP-06 Strain gauge Back to back, ** X50, Y25 CFRP-07 Extensometer Back to back, ** X50, Y25 CFRP-08 Extensometer Back to back, ** X50, Y25 CFRP-09 Extensometer Back to back, ** X50, Y25 CFRP-10 Extensometer Back to back, ** X50, Y25 CFRP-11 Extensometer Back to back, ** X50, Y25 CFRP-12 Extensometer Back to back, ** X50, Y25 ** X and Y coordinates refer to the distance in mm from the bottom left corner of the specimen. X is transverse and Y is the longitudinal axis. 119 Table 34 Axial strain measurement types and location of strain gauges for the scarf specimens Specimen ID Axial strain Strain gauge locations** measurement type 5q scarf joint specimens (SC) SC-01 Extensometer Back to back, ** X50, Y25 SC-02 Extensometer Back to back, ** X50, Y25 SC-03 Strain gauge Back to back, ** X50, Y25 SC-04 Extensometer Back to back, ** X50, Y25 SC-05 Strain gauge Back to back, ** X50, Y25 SC-06 Strain gauge Back to back, ** X50, Y25 SC-07 Extensometer Back to back, ** X50, Y25 SC-08 Extensometer Back to back, ** X50, Y25 SC-09 Extensometer Back to back, ** X50, Y25 SC-10 Extensometer Back to back, ** X50, Y25 SC-11 Extensometer Back to back, ** X50, Y25 SC-12 Extensometer Back to back, ** X50, Y25 SC-13 Extensometer Back to back, ** X50, Y25 ** X and Y coordinates refer to the distance in mm from the bottom left corner of the specimen. X is transverse and Y is the longitudinal axis. Table 35 Strain measurement types and strain gauge loc. for the scarf with doubler specimens Specimen ID Axial strain Strain gauge locations** measurement type 5q scarf joint with doubler specimens (SCD) SCD-01 Strain gauge Back to back, ** X50, Y25 SCD-02 Extensometer Back to back, ** X50, Y25 SCD-03 Extensometer Back to back, ** X50, Y25 SCD-04 Strain gauge Back to back, ** X50, Y25 SCD-05 Extensometer Back to back, ** X50, Y25 SCD-06 Strain gauge Back to back, ** X50, Y25 SCD-07 Extensometer Back to back, ** X50, Y25 SCD-08 Extensometer Back to back, ** X50, Y25 SCD-09 Extensometer Back to back, ** X50, Y25 SCD-10 Extensometer Back to back, ** X50, Y25 SCD-11 Extensometer Back to back, ** X50, Y25 SCD-12 Extensometer Back to back, ** X50, Y25 ** X and Y coordinates refer to the distance in mm from the bottom left corner of the specimen. X is transverse and Y is the longitudinal axis. The failure modes are described in Section 5.3.2. The test results are provided in Section 5.3.3. 5.2.5.5 Tension Testing Tension testing was carried out as a part of a strain survey described in Section 5.2.4 and also as a part of a coupon test program described in Section 5.2.3. In all cases, loading was performed using an Instron test machine with a 100 kN load cell within a 200 kN capacity frame. The load and crosshead displacement were output to the data acquisition system. The loading rate was 0.5 mm/min under displacement control. In the case of the strain survey, strain gauge data was also output to the data acquisition system. The strain gauges used, were identical to those used within the CAI testing described in Table 32. The failure modes are described in Section 5.3.2. The test results are provided in Section 5.3.3. 120 5.2.6 Failure prediction methodology Two methods of failure prediction were assessed, to determine their effectiveness in predicting the failure strain of the parent adherend, scarf joint and scarf with doubler joint specimens. Of particular interest, was to determine if the prediction methodology used within the damage tolerance analyses of composite laminates, could be applied to the damage tolerance analyses of scarf joints with and without doublers. The first methodology assumes the parent adherend to behave in a ductile manner, thus any stress concentrations that may have been introduced at the edge of the delaminated region are ignored. The delaminated region is assumed to have zero stiffness, with the failure occurring when the “good” material within the outer ligaments exceeds the design allowable failure stress for the material. That is, the damage area, Ad is assumed to be circular, giving a diameter, I of damage; I Ad /4 S . (5.5) As such, the stress in the ligament, Vadherend is given by; V applied V , (5.6) adherend I /1 W where W is the specimen width (100 mm) and Vapplied is the applied compression stress. Therefore, failure occurs when Vadherend >Vu, where Vu is the ultimate failure stress of the composite adherend, which is determined from the failure strain of the material assuming the material behaves linearly to failure. That is, Vu=Eadherend.Hu. The second methodology assumes the parent adherend to behave in a brittle manner, thus the stress concentration that occurs in the parent adherend at the edge of the delaminated region is taken into account. In this case, we have measured the modulus reduction (Section 5.3.1.3), thus we can use the result from the finite element modelling of the plate with a reduced modulus insert (Section 3.2) to determine the stress concentration factor in the adherend at the edge of the delaminated region. As such, the delaminated region is assumed to be a rigidly connected within the “good” parent adherend. Failure is then predicted to occur when the stress at the edge of the delaminated region exceeds the ultimate failure stress of the material, Vu. As such, V applied V adherend . (5.3) t I WK )/1( Again, failure occurs when Vadherend >Vu. The CAI failure predictions are provided in Section 5.3.4. 121 5.3 Results and Discussion 5.3.1 Impacting and damage assessment 5.3.1.1 Summary A plot of the absorbed energy within each of the specimens with respect to the potential energy of the drop weight for all of the specimens that were tested in compression is provided in Figure 101. These results show that the presence of the adhesive within the laminate allows the specimen to absorb more energy than the parent adherend laminate alone, particularly at higher impact energy levels. At the lower energy levels, the amount of energy absorbed by all specimen types was similar. The scarf joint without the doubler absorbs more energy than the scarf joint specimens with the doubler. This was probably due to an increased rebound provided to the drop weight by the rubber that has been added to the epoxy adhesive, which was present in large quantities in the overlap doubler bond-line. Absorbed energy in the specimen vs potential energy of the impactor cfrp scarf scarf with doubler Linear (scarf) Linear (cfrp) Linear (scarf with doubler) 20 18 16 14 12 10 8 6 Absorbed by energy the specimen (j) 4 2 0 0 5 10 15 20 25 Potential energy of the impactor (j) Figure 101 Absorbed energy within the specimen with respect to the potential energy of the drop weight for all specimens ultimately tested in compression A plot of damage area with respect to impact energy for each of the CAI specimens is provided in Figure 102. 122 Damage Area wrt Impact Energy for CFRP , Scarf and Scarf with Doubler Specimens CFRP Scarf Scarf with doubler Linear (Scarf) Linear (CFRP) Linear (Scarf with doubler) 1800 1600 1400 1200 1000 800 Damage Area (mm) Area Damage 600 400 200 0 012345678910 Energy/Thickness (j/mm) Figure 102 Damage area with respect to impact energy for all CAI specimens As expected, as the energy of the impact was increased, the area of damage measured within the adherend increased. The extent of damage induced by an impact of similar energy was similar for each of the specimens. 5.3.1.2 Damage mode The type of damage that occurred within each of the specimens was assessed using non-destructive and destructive inspection techniques. The damage area was calculated using the ultrasonic inspection results. Dummy specimens were prepared which were of identical geometric and material parameters to the CAI specimens, specifically to determine the internal damage within the parent adherend and scarf specimens using destructive sectioning through the damaged regions. Ultrasonic inspection Typical ultrasonic inspection images for each of the specimen types with increasing impact energy are provided in Figure 103. The plots from the parent adherend specimens show that damage growth with increasing impact energy progresses in the load direction, transforming from a circular delamination pattern into an oval shaped pattern. Also of note was the increased separation of the rear face off axis plies as the impact energy was increased. The delamination growth within the scarf joint specimens showed a similar trend to the parent adherend specimens with respect to increasing impact energy. However, the separation of the off axis plies on the rear face was halted once they interfaced with the joint adhesive. Delamination growth of the off axis plies was much more extensive on the opposite side of the specimen to the impact face. The delamination of the outer ply 123 typically extended to the rear face scarf joint tip then progressed in a line adjacent to the scarf bond-line towards the rear face. The delamination pattern within the scarf joint with doubler was virtually identical to the scarf joint without the doubler. In this instance, the impact energy was greater than the scarf joint specimens, as the thickness of the material at the impact site was greater than the scarf joint specimens by two extra plies (0.26 mm thicker). As such, when normalised with the total thickness of the adherend, it may be concluded that the doubler does not provide increased impact resistance to the adherend. That is, the adhesive between the doubler and the adhesive did not inhibit the growth of delaminations within the scarfed adherends. However, destructive sectioning of dummy specimens was conducted to verify whether the presence of the doubler was in fact, able to alter the delamination pattern through the thickness of the specimens. CFRP Specimens CFRP Specimens 0.6 BVID 0.7 BVID 0.9 BVID BVID (Impact (Rear (Rear (Impact Face) F) Scarf Specimens Scarf Specimens 0.7 BVID 0.9 BVID BVID 0.6 BVID (Rear Face) (Rear Face) (Rear Face) (Impact Face) Scarf with doubler Specimens Specimens 0.5 BVID 0.7 BVID 0.85 BVID BVID (Rear Face) (Rear Face) (Rear Face) (Rear Face) Figure 103 Ultrasonic inspection images of the CAI specimens 124 Sectioning to determine the effect of impact energy on delamination growth A cross section through the damaged region of a parent adherend dummy specimen is shown in Figure 104. The section chosen in this case was through the plane normal to the loading direction. The shiny plies shown in the image represent the 90° or transverse plies. The image was created from two images of the section, one taken in normal incandescent light and the second taken in UV light after the section was coated in fluorescent dye to highlight areas of delamination. The final image is morphed to show the delaminations and the ply orientations simultaneously. The predominant delaminations observed within these images run along the direction of the transverse plies. However, there are delaminations which are shorter that run within the other plies. These results show that the delamination growth in the transverse direction is predominantly through the transverse plies, which implies that delamination growth in the longitudinal direction runs adjacent to the longitudinal plies. The delamination regions near the rear face were more extensive than those near the impact face, which is typical of low velocity impact damage to carbon/epoxy laminates. 0.8 BVID (Top surface is impact face) 1.1 BVID (Top surface is impact face) Figure 104 Transverse cross section through the damaged site of the parent adherend specimens The delamination pattern through the damaged region of a scarfed dummy specimen can be seen in a longitudinal cross section shown in Figure 105. In this case, two impact energies are shown, with the impact occurring approximately 8 mm from the scarf joint tip within joint region. Delamination regions close to the impact face that progress towards the scarf joint tip can be seen. They can also be seen in areas adjacent to the 0q plies on both sides of the scarf joint, predominantly near to the central plies and the rear face plies. The delaminations in these regions are all halted by the presence of the scarf joint. At both energy levels, the plies on the rear face opposite to the impact site have delaminated into a bulge. The added energy of impact did not appear to increase the size of the larger delamination regions, but rather increased the size of the major delaminations. Of concern from a structural integrity point of view was the damage observed to both of the scarf joint adherend tip regions. 125 0.6 BVID (Top surface is impact face) 1.1 BVID (Top surface is impact face) Figure 105 Longitudinal cross section through the damaged site of scarfed specimens The delamination pattern through the damaged region of a scarfed dummy specimen can be seen in a longitudinal cross section shown in Figure 106. The pattern of delamination within the scarfed adherends of these specimens was similar to the specimens without the doubler. Very little damage to the overlap doubler was observed. There was no evidence of a disbond in the doubler adhesive either. These results show that a significant portion of the impact energy was absorbed within the scarfed adherends. Of concern from a structural integrity point of view was the damage observed within both tip regions, particularly at the higher impact energy level. 0.6 BVID (Top surface is impact face) 1.1 BVID (Top surface is impact face) Figure 106 Longitudinal cross section through the damaged site of scarf with doubler specimens Sectioning to determine the effect of impact position on delamination growth within the scarf Several scarfed adherend specimens with and without an overlap doubler were impacted at different locations relative to the scarf tip, then sectioned through the damage site, and assessed. The impact locations considered were (i) a direct strike to the scarf joint tip, x=0 mm, then (ii) x=5 mm from the scarf joint tip within the joint itself, (iii) x=10 mm, (iv) x=15 mm, which was directly over the mid point of the scarf joint and (v) x=20 mm, which was approximately 10 mm from the rear face tip. The cross section images for the scarf 126 joint with and without a doubler are provided in Figure 107 and Figure 108 respectively. The specimens were impacted with 5.36j/mm thickness of the parent structure. x=0 mm x=5 mm x=10 mm x=15 mm x=20 mm Figure 107 Longitudinal cross-section through the damaged site of a scarf joint impacted at the tip, x=0 mm, then at x=5 mm, x=10 mm, x=15 mm (mid joint) and x=20 mm. The pattern of damage within the scarf joint adherends without the doubler was found to depend largely on the impact location. x At x=0mm, delamination damage was confined almost entirely to the adherend on the opposite side of the bond-line to the impacted tip. The exception was a fine delamination adjacent to a 90q ply near the tip. All other delaminations terminated at the bond-line. x At x=5mm, the pattern was similar to x=0mm, except that the delamination near the impacted tip became more pronounced. All other delaminations terminated at the bond-line. x At x=10mm, two additional delaminations were introduced, one through the 45q ply after the first set of 0q plies within the impacted tip, and one through the 45q and 90q on the adjacent side to the impacted tip extending through to the opposite scarf tip. x At x=15mm, delaminations were distributed evenly between adherends. Delaminations were observed near both scarf tips. x At x=20mm, the predominant delaminations were in the impacted adherend, with delaminations observed near both scarf tips. 127 These results show that impacts at x=10mm and x=15mm caused the most amount of damage to the scarf joint adherend tips. This was considered to be tup size dependant, with predominant damage to the plies near the impact face occurring approximately one tup radius from the impact location. It was also observed that as the impact location moved towards the central part of the joint, the delaminations in the plies near to the rear face shortened and the upper ply delaminations extended. It may be concluded from this result that the adhesive provides some energy absorption, thus protecting the plies beneath the joint. x=0 mm x=5 mm x=10 mm x=15 mm x=20 mm Figure 108 Longitudinal cross- section through the damaged site of a scarf joint with doubler, impacted at the tip, y=0 mm, then at y=5 mm, y=10 mm, y=15 mm (mid joint) and y=20 mm. The presence of the doubler did little to affect the delamination pattern within the scarf joint adherend. There was no evidence of overlap doubler damage or any disbonds in the doubler adhesive. The critical location with regards to scarf tip damage was similarly, x=10mm and x=15mm. 5.3.1.3 Stiffness reduction strain survey Testing was carried out on parent adherend and scarf joint specimens to determine the extent of load redistribution and stiffness reduction following low velocity impact. Axial strain gauges were attached in the far field location, on the impact site itself and approximately one tup radius (10 mm) from the impact site. The specimens were impacted with 70% BVID or 12.90J. The specimens were loaded to 50% of the failure load of the scarf 128 joint specimens. Plots of the stress with respect to strain were generated, and the local Young’s modulus calculated. The initial strain survey was conducted on CAI specimens within the SACMA test rig. Following this, the specimens were reduced in width and tested in tension. The local Young’s modulus at each of the required locations was calculated in both test conditions, and provided in Table 36. For the parent adherend specimen loaded in compression, the Young’s modulus was seen to increase by 8% above the far field at the impact site, but reduce by 8% at a distance 10 mm from the impact site. This is probably due to load bypass, around the delaminated area, with the central region simply not attracting the same load as the outer parts of the delaminated region. For the parent adherend specimen loaded in tension, the modulus reduction in the damaged region reduced by 15%. This is a better indication of the true Young’s modulus in the delaminated region as there is less chance of the load bypass. For the scarfed specimen, the Young’s modulus at the impact site did not reduce significantly when loaded in compression or in tension. The load distribution through the adherend is quite complex in this region, as there is evidence of tip strain concentrations. A better indication of the adherend modulus reduction was found in gauge measurements 10 mm from the impact site, with the strain reducing by 24% below the far field value. It may be concluded from these results that the parent material modulus is reduced by a greater within the scarf joint than in impacted regions away from the joint. Table 36 Young’s modulus calculations for the strain survey specimens Loading type Strain gauge Parameter Parent adherend Scarf type (CFRP) Compression Far field Effc 96 GPa 100 GPa (Specimen Impact site Eimpactc 108 GPa 90 GPa width=100 mm) Eimpactc/Effc 1.13 0.90 10 mm from Eradiusc 88 GPa 78 GPa impact Eradiusc/Effc 0.92 0.78 Tension Far field Effc 100 GPa 95 GPa (Specimen Impact site Eimpactc 86 GPa 93 GPa width=32 mm) Eimpactc/Effc 0.86 0.98 10 mm from Eradiusc 85 GPa 76 GPa impact Eradiusc/Effc 0.85 0.80 5.3.2 Failure analyses 5.3.2.1 Parent structure failure analyses Compression Images showing the typical failure mode of the undamaged and impacted CAI specimens are provided in Figure 109. Of particular concern was the failure mode of the undamaged specimen. The compression failure strain of the undamaged parent adherend laminate 129 was quite low in comparison to the data provided by the manufacturer. The failure mode observed was through an upper section of the specimen, corresponding to a small gap in the edge restraint. All of the edges were clamped, except for the small gap (10 mm or 7% of edge length) on the sides of the specimen to allow unimpeded deformation to failure. This indicates that local buckling may have occurred within this region, thus yielding a baseline compression strength that represents a lower bound, with much higher strengths possible if the local buckling could have been constrained. As expected, in all cases, the failed section of the impacted specimens passed through the damaged region. This indicates that the presence of the delamination regions have weakened the parent structure. The off axis failure pattern indicates that failure may have initially occurred within the 45q plies before extending to the other plies. The test results are provided in Section 5.3.3. Impact site BB Undamaged Specimen Failed Impacted Specimen A A VIEW A VIEW B Figure 109 Typical compression failure modes of the undamaged and impacted parent adherend CAI specimens Tension Images of a failed parent structure un-notched tensile coupon are provided in Figure 110. The failure of the specimen was near to the grip region, which was expected given the specimen contained no tapering of the net section. The test results are provided in Section 5.3.3. 130 Figure 110 Tensile failure of a parent adherend coupon specimen 5.3.2.2 Scarf specimens Compression testing Images showing the typical failure mode of the undamaged CAI specimens are provided in Figure 111. The failure mode in compression appears to be in the tapered region near the scarf joint. The results show that failure may have occurred was via interlaminar shear overload between plies that interfaced with the outer groups of 0q plies. In the left side view, there was delamination between the outer 90q ply and the 0q ply, causing the scarf joint tip region to buckle. In the right side view, it shows the adherend fracturing in a line parallel to the taper angle of the scarf joint. Failure may have initiated in the lower tip region then continued parallel to the scarf joint through the adherend. The remaining failure progression is shown in Figure 111. The failure mode in this instance is quite complex, and difficult to interpret. Although fractographic analyses were performed, the initiation location and the failure path shown remain inconclusive. The adhesive does not appear to be the critical element in this particular joint under this type of loading, as there was no evidence of cohesive failure through the adhesive. There was also no evidence of interfacial failure between the adherend and the adhesive, as close inspection of the failures near to the adherend/adhesive interface showed that fibres remained adhered on the adhesive following failure. An interfacial failure caused by poor surface treatment would have resulted in areas where the adhesive was clean, thus free of fractured fibres. However, the fracture adjacent to the joint indicates that the adherend may have been locally weakened near the surface of the tapered region by the grit-blast and abrasion process used during surface treatment. 131 RIGHT SIDE VIEW Failure path TOP FACE Right Left Left Right Local buckle of the outer adherend BOTTOM FACE LEFT SIDE Figure 111 Typical compression failure mode of the undamaged scarf CAI specimens Local buckle of the outer adherend Delamination between groups of 0q Adherend fracture Impact site Local Figure 112 Typical compression failure mode of the impacted scarf CAI specimens Images showing the mode of failure for an impacted scarf joint are shown in Figure 112. The cut section shown was from specimen SC-08, which was impacted with 12.85j, or 70% BVID. The zoomed in shot of the impact site was from specimen SC-05, which was impacted with 18.43j, or 100% BVID. The results show that the failure mode may be governed by the interlaminar failure within the scarfed adherends. As with the un-notched scarf specimens, the results show that local buckling of the outer adherend 132 plies may have occurred causing delamination between the outer 90q ply and the outer group of 0q plies. The degree of outer-ply buckling appeared to reduce at the impact site itself, but was pronounced within the adherend approximately 10 mm from the impact site. The lower part of the cut section shows that failure of the adherend progressed adjacent to the scarf joint, then progressed between the groups of 0q plies. It can be seen within Figure 106 that a delamination was caused in between the groups of 0q plies by the impact event. This indicates that compression failure may have initiated in the delaminated region caused by impact. The test results are provided in Section 5.3.3. Tension Images of the failure surfaces of the undamaged scarf joints loaded in tension are provided in Figure 113. These results show that the tensile failure initiated either in the scarf adhesive or in the adherend tip close to the adhesive. The results show that progression may be through the adherend adjacent to the joint until the fracture progresses past the stronger 0q plies, where the fracture moves into the adhesive. The failure then progresses back into the weaker plies of the adherend before returning to the adhesive as it passes through other 0q plies of the opposite side of the joint. This failure mode is consistent with the failure modes of scarf tension specimens observed within [1] and shown schematically in Figure 12. Given these specimens were cut from one of the spare CAI specimens, it may be concluded from these results that the compression failures observed in the previous section, are valid, in that they were observed were not a results of poor manufacturing quality. The test results are provided in Section 5.3.3. 0q plies Figure 113 Typical tensile failure mode of the undamaged scarf coupons 5.3.2.3 Scarf with doubler specimens Compression Images showing the typical failure mode of the undamaged CAI specimens are provided in Figure 114. In this case, the fracture appears to have initiated in the scarf joint region 133 either in the tip covered by the doubler or more likely in the tip opposite the doubler. Ultimately, the doubler fractured in the region over the joint, most likely after the scarf joint fractured. The results show that the fracture path was similar to the scarf joint without the doubler (Figure 111). However, the doubler appears to have mitigated the local buckling of the outer plies seen in the scarf joint specimen. Doubler failure TOP FACE RIGHT SIDE VIEW Right Right Left side Left side BOTTOM FACE LEFT SIDE VIEW Doubler failure Figure 114 Typical compression failure of the undamaged scarf with doubler CAI specimens The impacted scarf joint specimens with a doubler all failed in the adherend just outside of the doubler region across the section. A side view of the failure mode is shown in Figure 115. It appears that the impact damage has lowered the buckling strength of the adherend, introducing modes that could not be constrained with the test fixture. This instability occurred at loads below the failure load of the un-notched specimen. The test results are provided in Section 5.3.3. Doubler termination 134 Figure 115 Typical compression failure of the impacted scarf with doubler CAI specimens Tension failure Images of the failure surfaces of the undamaged scarf with doubler joints loaded in tension are provided in Figure 116. The failure mode of the scarf with doubler joint in tension is quite complex, with evidence of doubler failure, interlaminar failure of the adherend and also some cohesive fracture of the adhesive. The only evidence of cohesive failure of the adhesive occurred between the groups of 0q plies. The results show that failure initiation may have occurred near either of the scarf joint tips or in the doubler itself. The test results are provided in Section 5.3.3. Scarf tip fracture Doubler failure Figure 116 Typical tensile failure mode of the undamaged scarf with doubler coupons 5.3.3 Static test results and discussion Baseline specimen testing (without impact) The failure strain of the undamaged specimens, tested in compression and tension is provided in Table 37. For the parent adherend, the failure strain in tension far exceeded the failure strain in compression. As indicated in the failure mode discussion in Section 5.3.2.1, this appears to be due to buckling near a small gap in the edge restraint to allow specimen deformation. As such, this represents a lower bound in the parent structure failure strain. For the scarf joint specimens, the failure strains observed in both tension and compression were similar, despite their failure modes exhibiting some differences. This indicates that 135 the buckling observed in the parent adherend specimens did not contribute to the scarf joint specimen failure. For the scarf with doubler joint, the tensile failure strains were around 57% higher than the compression failure strains. The most likely cause for this discrepancy was the load eccentricity introduced by the overlap doubler which may have introduced a local bending moment near to the scarf joint. This may have then caused a local instability to occur, introducing further bending moments, and ultimately specimen failure. As such, the compression failure strains represent a lower bound from which to baseline the impacted specimen results. Table 37 Undamaged parent adherend and scarf joint specimen failure strains Specimen type Loading type Specimen ID Failure strain (PH) Parent adherend Compression CFRP-02 4220 CFRP-04 4000 Tension CFRP-14-T01 14340 Scarf joint Compression SC-02 2920 SC-03 3090 Tension SC-15-T01 3020 SC-15-T02 2980 Scarf with doubler Compression SCD-01 3470 joint SCD-05 3250 Tension SCD-13-T01 5290 SCD-13-T02 5370 Compression after impact Plots of the CAI failure strain in the composite adherend with respect to the damage area for each of the specimen types are provided in Figure 117. The impacted specimens of each of the types failed at strains below the lower bound baseline result. As such, it may be concluded from these results that the presence of low velocity impact has weakened the specimens. For the parent structure specimen, the degree with which the specimen residual strength diminishes following impacting is considerable. This result agrees with the failure strain reduction observed within previous studies [54]. It may be concluded from these results that the current design practice of considering damage tolerance in the design of composite laminates used within aircraft structures is valid, as the failure strain reduction due to impact damage can be significant. For the scarf joint specimens, the failure strain reduction caused by low velocity impact was considerable, but not as severe a reduction as the parent adherend. However, when considering that the baseline scarf joint failed at quite a low strain, a further strain reduction caused by impact damage is significant from a design point of view. As such, it 136 may be concluded from these results that damage tolerance needs to be considered when designing scarf joints for repair, over and above the static strength requirement. The failure strain reduction in the impacted scarf with doubler joint specimens was less severe than in the scarf joint specimens, but a reduction nonetheless. As such, it may be concluded from these results that damage tolerance needs to be considered when designing scarf joints for repair, over and above the static strength requirement, even when an overlap doubler is applied to cover the scarf joint. CAI Failure Strain wrt Damage Area for CFRP , Scarf and Scarf with Doubler Specimens CFRP Scarf Scarf with doubler Linear (Scarf) Linear (CFRP) Linear (Scarf with doubler) 1.1 0.9 y = -0.57x + 4056.1 0.7 y = -0.3263x + 3205 0.5 strain y = -0.5515x + 3060.9 0.3 0.1 Specimen CAI failure strain/Undamaged parent adherend failure 0 200 400 600 800 1000 1200 1400 1600 1800 2000 -0.1 Damage Area (mm) Figure 117 CAI failure strain with respect to damage area calculated using ultrasonic inspection for all specimen types 5.3.4 CAI Failure predictions The predicted failure strains for the CAI specimens, normalised with the failure strain of the un-notched specimen for each of the specimen types is provided in Figure 118. 137 CAI Failure Strain wrt Damage Area for CFRP Specimens CAI Failure Strain wrt Damage Area for Scarf Specimens test data ductile failure prediction brittle failure prediction test data ductile failure prediction brittle failure prediction 1.2 1.2 1 1 0.8 0.8 0.6 0.6 0.4 0.4 0.2 0.2 Specimen CAI failure strain/Undamaged parent adherend failure strain Specimen CAI failure strain/Undamaged parent adherend failure strain failure adherend parent strain/Undamaged failure CAI Specimen 0 0 0 200 400 600 800 1000 1200 1400 1600 0 200 400 600 800 1000 1200 1400 Damage Area (mm) Damage Area (mm) CAI Failure Strain wrt Damage Area for Scarf with Doubler Specimens test data ductile failure prediction brittle failure prediction 1.2 1 0.8 0.6 0.4 0.2 Specimen CAI failure strain/Undamaged parent adherend failure strain failure adherend parent strain/Undamaged failure Specimen CAI 0 0 200 400 600 800 1000 1200 1400 1600 1800 2000 Damage Area (mm) Figure 118 CAI failure predictions for the parent adherend, scarf joint and scarf with doubler joint specimens The parent structure results show that the prediction methodology that takes into account the stress concentration at the edge of the delaminated region best matched the test result. If the stress concentration was not taken into account, the failure strain was over-predicted by around 20%. The scarf joint and scarf with doubler joint results show that both failure predictions under-predict the failure strain of the impacted scarf joint. This indicates that the degree to which the strength of the scarf joint diminishes with increasing impact energy is much less than for the parent adherend. One reason for this may be that the adhesive absorbs part of the impact energy without delaminating, which would otherwise be absorbed into the parent laminate. As such, the amount of damage caused to the parent laminate with an 138 adhesive layer may be less. Another reason may be that the failure strain of the scarf joint is already quite low, so any reduction caused by impacting needs to be quite severe to reduce the failure load further. Although the strength of the scarf joint does not diminish by the same percentage as the parent laminate, the actual failure strains of the scarf joints impacted with BVID energy levels were 20-30% lower than the un-notched scarf joint specimens. As such, damage tolerance of scarf joints will still need to be taken into account when designing scarf joints used for repair. 5.4 Conclusions and Recommendations for Future Work 5.4.1 Conclusions It was concluded that by impacting each of the specimens, significant delaminations were introduced at multiple ply levels throughout the thickness of the specimen. For the parent structure specimens, delaminations were found to progress adjacent to the groups of 0q plies, particularly towards the face opposite the impact face. Damage progression was longer in the load direction than in the transverse direction, probably due to the presence of more 0q plies, thus offering less resistance to delamination growth in this direction. As expected, significant strength reductions were observed in the impacted specimens, relative to the baseline specimens. The impacted scarf joint specimens had a reduced strength than the baseline un-notched scarf joint specimens. Many of the delaminations caused by impact were halted by the presence of the scarf joint adhesive, however, if the scarf joint was hit in a vulnerable location, considerable damage was induced near to the scarf joint adherend tips. As such, the fracture of the adherend within the damaged scarf joint specimens was observed to link up with the delaminations caused by the impacting, thus offering less fracture resistance. Clearly, The impacted scarf with doubler joint specimens had a reduced strength than the un-notched scarf with doubler joint specimens. However, the strength reduction was not as severe as it was for the scarf joint specimens. The presence of the doubler over the top of the scarf joint improved the impact resistance of the scarf joint a little with slightly less delaminations caused. However, its presence was found to introduce significant instabilities within the joint when loaded in compression. The doubler plies create a load eccentricity within the joint, which may be the cause of the compression failure load being less than the tensile failure load. The impacting of the doubler plies served to increase the instability of the joint further to the extent that failure occurred in the adherend outside of the joint itself near to the doubler termination. Two failure prediction methodologies were assessed in their ability to predict the failure strain of the CAI specimens. The methodology that took into account the stress concentrations induced near to the edge of the delaminated region was able to reasonably match the parent adherend test data. The methodology that didn’t take into account these stress concentrations over-predicted the failure strain of the impacted specimens by 139 approximately 20%. Both prediction methodologies under-predicted the failure strain of the impacted scarf joint and scarf with doubler joint specimens, indicating that the percentage strength reduction in the impacted scarf joints was less than the in the parent adherend. 5.4.2 Recommendations for Future Work The use of a thicker parent structure may have avoided the instability problems that affected some of the tests in this program. The use of a thinner or less stiff doubler may have also helped avoid the instability, giving a better indication of whether the overlap plies make the scarf joint less vulnerable. Future work may consider the effect of adherend lay-up on the damage tolerance of the scarf joint, particularly in establishing methodology to reduce damage to the plies oriented in the load direction, and in limiting the size of delamination for a given impact. The effect of tup size and the shape of the tup on the damage that is introduced to the scarf joint area may also be considered. The type of accidental damage that is caused during operation is random in nature, and some assessment of how effective a spherical tup is in inducing representative damage to the scarf is needed. 140 6. Conclusions and Recommendations The aim of the research undertaken within this project was to determine methodologies for reducing the area or footprint of the current flush repair methods, and also to achieve greater understanding of the vulnerability of the scarf repair to low velocity impact damage. The following approaches were taken to address this problem: x Shape optimisation of the scarf joint using analytic and finite element techniques, x Addition of reinforcement such as through thickness pins and embedded doublers to the joint, and x Compression after impact testing of scarf joints with and without doublers to determine their vulnerability to low velocity impact damage. In so doing extensive finite element modelling and experimental test programs were undertaken which have contributed new data and the results were critically examined with respect to the failure mode and damage type that occurred, and then to provide a theoretical framework from which the newly developed joints and also the damaged joints may be designed in the future. A comprehensive literature survey was conducted, covering a wide range of topics that are relevant to the design and testing flush repair joints to aircraft, which is presented in Section 2. This placed the work presented in this thesis in the context of the published literature. Conclusions and recommendations have been made at the end of each section, and have been brought together below. 6.1 Shape Optimisation of Scarf Repairs 6.1.1 Conclusions 6.1.1.1 Analytic and Finite Element Analyses An analytic technique has been developed to facilitate the optimal design of scarf joints between isotropic adherends with dissimilar material properties. The technique specifies a linearly varying scarf angle that generates a characteristic scarf profile for a given adherend modulus ratio. FEA validated the results obtained from the analytic modelling. Analytic and finite element modelling demonstrated the dependence of the adhesive stress distribution on local ply orientation within the adherend. This showed that the effectiveness of gross adherend shaping in optimising dissimilar adherend scarf joints between orthotropic adherends was limited to cases where the repair laminate could maintain the identical lay-up to the parent adherend. By reducing the modulus of the repair adherend plies, the adhesive shear stress was shown to concentrate predominantly in the bond-line near to the parent (stiffer) adherend tip. The magnitudes of the peaks became equal when an optimal, linearly varying scarf angle was used. 141 In contrast, the linear variation of the scarf angle could not minimise the adhesive shear stress concentration if the repair laminate lay-up was not identical to the parent lay-up. The use of interleaved adhesive layers in the repair laminate served to raise the stress concentration in the adhesive adjacent to the interleaved plies. The use of a linearly varying scarf angle raised the magnitude of the stress peak on the side of the joint where the local scarf angle was a maximum. The use of a more complex taper profile to minimise ply induced stress concentrations in the adhesive may be possible, but may be very difficult to manufacture in practice. 6.1.1.2 Tensile Testing It can be concluded from the results of scarf joint coupon testing that the predominant failure mode was shear overload of the adhesive. In many cases, particularly for shallow angle scarf joints, this was complicated by adherend yielding or composite fibre fracture near the scarf tips. It can also be concluded from these results that as the scarf angle is increased, adhesive is loaded less predominately in shear and more in tension. As such, there is less shear deformation prior to specimen failure in these joints. The adhesive within the metallic specimens with shallow angles deformed significantly in shear prior to failure. It can be concluded from these results that when dissimilar materials are bonded with a ductile adhesive, the static strength of the joints do not change significantly with the scarf profile. However, given large stress concentrations are introduced in the bond-line between dissimilar adherends, the benefit of adherend may be better seen with fatigue testing. Unfortunately, in some of the optimised profile specimens areas within the bond-line were not coated with adhesive. This was discovered late in the experimental program, which prevented additional specimen to be manufactured. The curvature in the bond-line meant that pressure applied through a “pressure plate” above the joint during manufacture forced an uneven pressure distribution during cure. As such, the method of manufacture used to make all previous linear scarf joint specimens did not work as effectively for the optimised scarf joints. In retrospect, an envelope bag technique may have worked better to ensure an even bond-line throughout the joint. As such, it was not possible to determine if the optimisation would have enhanced the failure strength of the linear scarf joints between dissimilar adherends in the current program. 6.1.2 Recommendations for Future Work 6.1.2.1 Analytic and Finite Element Analyses It was concluded that peaks in the adhesive stress distribution were observed adjacent to the stiffer regions of the adherend, particularly near to the 0q plies. As such, the macro profiling of the adherend may only be applicable to quasi-isotropic lay-up configurations. 142 To improve applicability, work is needed to develop a closed form solution for the optimal adherend shape that gives a near constant shear stress distribution in the adhesive. An improved shape would have a shallow scarf angle adjacent to the stiffer areas within the adherend and use a steeper angle adjacent to the less stiff regions. As a matter of fact, following the completion of the work program, but before the thesis was completed, the beginnings of this work was presented in [45]. These analyses did not account for any of the thermal and moisture absorption mismatches that may induce further stress concentrations in the adhesive. Given that the hot-wet operational environment is often the most critical for the survivability of the bonded joints, these effects may need inclusion for future analyses and testing. 6.1.2.2 Testing Future work is needed to determine the effect of optimisation on the static strength of the bonded joints in ambient and hot-wet environments. A method of manufacture, possibly using an envelope bag, is required to ensure that the bond-line thickness throughout the optimised scarf joint is uniform prior to testing. It was concluded from the testing that the static strength of the bonded joints between dissimilar materials at room temperature was not significantly below that of the scarf joints between identical adherends. However, the influence of optimisation may show improved fatigue performance, particularly in raising the load in which a disbond initiates. Current design philosophy within the bonded repair manual as used by the RAAF to verify scarf repair design does not consider fatigue performance, but given the overlap repair designs do, it is likely that this may require consideration within future flush repair designs. 6.2 Reinforcement Methods used to Improve Scarf Joint Strength 6.2.1 Conclusions 6.2.1.1 Through thickness pinning Pinning of the metallic specimens to improve peel resistance was found to not significantly affect the mechanism of joint failure or the load capacity of the joint. Further testing is recommended to determine if pinning can reduce the rate of damage progression under fatigue loading. 6.2.1.2 Modified Biscuit Joints within Scarf Joints 6.2.1.2.1 Phase 1 Pilot Experimental Study It was concluded from the pilot experimental study that by embedding a patch insert within a scarf joint, the overall load capacity of the joint may be increased. As such, this methodology may be used to improve the strength of flush bonded joints without necessarily making the taper angle shallower. However, the length of the joint with the embedded patch was much longer than the scarf joint that was used as a comparison. 143 6.2.1.2.2 Phase 2 and 3 Experimental Study and Finite Element Analyses Static tensile tests of four types of two dimensional scarf joints with biscuit inserts were conducted with the results compared to predicted failure loads using finite element stress and strain calculations. Each of the specimen types failed through the adhesive, thus the failure predictions of two adhesive failure theories were compared. These were a maximum shear stress theory and a maximum shear strain theory, with the material properties generated using thick adherend specimen tests [38]. It may be concluded that the predictions using a maximum shear strain and a maximum shear stress criterion represent the upper and lower bounds for the failure loads respectively. The maximum shear stress criterion gave conservative predictions whilst the maximum shear strain theory predicted joint strengths that were considerably higher than the test results. Two geometric parameters and one material parameter were varied, resulting in four specimen types. Of interest were the effects of the biscuit insert taper angle, un-tapered length and Young’s modulus on the strength of flush joints. The two geometric parameters were linked in that a shallow angle taper had to be used in combination with a short un-tapered length to ensure that the overall joint dimension matched the length of a straight scarf for comparison. Specimens with a shallow insert taper failed at a higher load than when a steeper taper was used with a longer insert length. Specimens with a higher modulus insert (titanium) failed at a higher load than when the insert material matched the parent adherend (aluminium). Results from the static testing showed that one of the specimen types was able to show a 2- 3% strength improvement over the equivalent length 5q scarf joint, with all specimen types achieving a failure stress that was at least 90% of the equivalent 5q scarf joint. Given this joint has never been used before for aerospace application, the joint designs shown herein may be improved through shape and layout optimisation. However, it is possible that bending was induced into the specimen through the method used to taper the insert, which may have weakened the specimens. If the taper were reversed, or the insert made longer, the influence of the insert taper on the upper scarf bond-line stress may be minimised. 6.2.2 Recommendations for Future Work 6.2.2.1 Through thickness pinning The pins may provide a benefit in retarding crack growth under fatigue loading, or even in increasing the energy required to open the crack once it has initiated or the tip damaged. It is recommended to perform fatigue testing of scarf joints with a disbond inserted at the tip of the joint with and without pins to determine if the resistance of the scarf joint to crack growth is improved. 6.2.2.2 Modified Biscuit Joints within Scarf Joints The phase 2 testing showed that the patch with an insert was at least as strong as the baseline scarf joint, but did not demonstrate a significant strength improvement. This may be due to the insert taper introducing a small bending moment to the specimen that then overloaded the scarf joint. It is recommended to taper the insert differently or increase the 144 insert length to minimise bending of the specimen, particularly in the region of the scarf joint. It is also recommended to test in hot-wet conditions, such that the adhesive is sufficiently degraded so that the adherend is not loaded near to the material yield stress. These conditions are often the most critical for bonded joint designs. 6.3 Impact resistance and damage tolerance of scarf repairs 6.3.1 Conclusions It was concluded that by impacting each of the specimens, significant delaminations were introduced at multiple ply levels throughout the thickness of the specimen. For the parent structure specimens, delaminations were found to progress adjacent to the groups of 0q plies, particularly towards the face opposite the impact face. Damage progression was longer in the load direction than in the transverse direction, probably due to the presence of more 0q plies, thus offering less resistance to delamination growth in this direction. As expected, significant strength reductions were observed in the impacted specimens, relative to the baseline specimens. The impacted scarf joint specimens had a reduced strength than the baseline un-notched scarf joint specimens. Many of the delaminations caused by impact were halted by the presence of the scarf joint adhesive, however, if the scarf joint was hit in a vulnerable location, considerable damage was induced near to the scarf joint adherend tips. As such, the fracture of the adherend within the damaged scarf joint specimens was observed to link up with the delaminations caused by the impacting, thus offering less fracture resistance. Clearly, The impacted scarf with doubler joint specimens had a reduced strength than the un-notched scarf with doubler joint specimens. However, the strength reduction was not as severe as it was for the scarf joint specimens. The presence of the doubler over the top of the scarf joint improved the impact resistance of the scarf joint a little with slightly less delaminations caused. However, its presence was found to introduce significant instabilities within the joint when loaded in compression. The doubler plies create a load eccentricity within the joint, which may be the cause of the compression failure load being less than the tensile failure load. The impacting of the doubler plies served to increase the instability of the joint further to the extent that failure occurred in the adherend outside of the joint itself near to the doubler termination. Two failure prediction methodologies were assessed in their ability to predict the failure strain of the CAI specimens. The methodology that took into account the stress concentrations induced near to the edge of the delaminated region was able to reasonably match the parent adherend test data. The methodology that didn’t take into account these stress concentrations over-predicted the failure strain of the impacted specimens by approximately 20%. Both prediction methodologies under-predicted the failure strain of the impacted scarf joint and scarf with doubler joint specimens, indicating that the percentage strength reduction in the impacted scarf joints was less than the in the parent adherend. 145 6.3.2 Recommendations for Future Work The use of a thicker parent structure may have avoided the instability problems that affected some of the tests in this program. The use of a thinner or less stiff doubler may have also helped avoid the instability, giving a better indication of whether the overlap plies make the scarf joint less vulnerable. Future work may consider the effect of adherend lay-up on the damage tolerance of the scarf joint, particularly in establishing methodology to reduce damage to the plies oriented in the load direction, and in limiting the size of delamination for a given impact. The effect of tup size and the shape of the tup on the damage that is introduced to the scarf joint area may also be considered. The type of accidental damage that is caused during operation is random in nature, and some assessment of how effective a spherical tup is in inducing representative damage to the scarf is needed. 146 7. References 1. Pipes, R.B. and D.W. Adkins, Strength and mechanics of bonded scarf joints for repair of composite materials [Final Report, Jun. 1981 - May 1982]. 1982. 2. Association, S.o.A.C.M., SACMA Recommended Test Method (SRM 2R-94) for Compression after impact properties of oriented fiber-resin composites. 1994. 3. Myhre, S.H. and C.E. Beck. Repair concepts for advanced composite structures. 1978: Structures, Structural Dynamics and Materials Conference, 19th, Bethesda, Md ; United States; 3-5 Apr. 1978. pp. 117-125. 1978. 4. Volkersen, O., Die Nietraftverteilung in Zugbeanspruchten Nietverbindungen mit Konstanten Laschenquerschnitten. Luftfahrtforschung, 1938. 41. 5. Goland, M. and E. Reissner, Stresses in Cemented Joints. Journal of Applied Mechanics, 1944. A17. 6. Hart Smith, L.J., Adhesive-bonded scarf and stepped-lap joints. 1973, Mcdonnell-Douglas Corp., Long Beach, Calif. 7. Lubkin, J.L., A Theory of Adhesive Scarf Joints. Journal of Applied Mechanics, 1957: 255. 8. Erdogan, F. and M. Ratwani, Stress Distribution in Bonded Joints. Composite Materials, 1971: 379. 9. Webber, J.P.H., Scarf Repair Joints in Carbon Fiber Reinforced Plastic Strips. Journal of Adhesion, 1981. 4: 257. 10. Thamm, F., Stress distribution in lap joints with partially thinned adherands. Journal of adhesion, 1976: 301. 11. Reddy, M.N. and P.K. Sinha, Stresses in adhesive bonded joints for composites. Fibre science and technology, 1975. 1: 33. 12. Harman, A. and C. Wang. Analytic and finite element stress predictions in two-dimensional scarf joints (Paper 364). in Eleventh Australian International Aerospace Congress. 2005. 13. Johnson, C.L., Effect of ply stacking sequence on stress in a scarf joint. AIAA Journal, 1989. 1: 79. 14. Adams, R.D. and N.A. Peppiatt. Stress analysis of lap joints in fibre reinforced composite. in INST.CIVIL ENGRS. FIBRE REINFORCED MATERIALS: DESIGN. 1977. London. 15. Adkins, D.W. and R.B. Pipes, End Effects in Scarf Joints. Composites Science and Technology, 1985. 3: 209. 16. Jones, J.S. and S.R. Graves, Repair techniques for Celion/Larc-160 graphite/polyimide composite structures. 1984. 17. Adkins, D.W. and R.B. Pipes. Tensile Behavior of Bonded Scarf Joints Between Composite Adherends. in Proceedings of the Fourth Japan--U.S. Conference on Composite Materials. 1989. Washington, D.C ; USA. 18. Ahn, S.H. and G.S. Springer, Repair of composite laminates. Federal Aviation Administration, DOT/FAA-00/46, 2000. 19. Baker, A.A., R.J. Chester, and G.R. Hugo, Scarf repairs to graphite/epoxy aircraft components. National Conference Publication - Institution of Engineers, Australia. Publ by IE Aust, Crows Nest, NSW, Aust, 1993. 93 pt 6: 193. 20. Baker, A.A., R.J. Chester, G.R. Hugo, and T.C. Radtke, Scarf repairs to graphite/epoxy components. AGARD, Composite Repair of Military Aircraft Structures 12, 1995. 147 21. Chester, R.J., A.A. Baker, G.R. Hugo, and T.C. Radtke, Scarf repairs to highly strained graphite/epoxy structure. International Journal of Adhesion and Adhesives, 1999. 2-3: 161. 22. Tomblin, J.S., L. Salah, J.M. Welch, and M.D. Borgman, Bonded repair of aircraft composite sandwich structures. 2004. 23. Soutis, C. and F.Z. Hu. Repair Design of Composites and Efficiency of Scarf Patch Repairs. in The 11th International Conference on Composite Materials. 1997. Melbourne, Australia: Australian Composite Structures Society Woodhead Pty. Ltd. 24. Harman, A., Testing of thick adherend bonded scarf joint 2D test coupons. CRC-ACS TM 04x25, 2004. 25. Objois, A., Y. Gilibert, and B. Fargette, Theoretical and experimental analysis of the scarf joint bonded structure: Influence of the adhesive thickness on the micro-mechanical behavior. Journal of Adhesion, 1999. 1: 13. 26. Mortensen, F. and O.T. Thomsen, Simplified linear and non-linear analysis of stepped and scarfed adhesive-bonded lap-joints between composite laminates. Composite Structures, 1997. 1-4 May-Aug: 281. 27. Marcolefas, S., V. Kostopoulos, and S.A. Paipetis, Non-linear analysis of a metal-to- composite scarf joint. International Journal of Mechanical Sciences, 1991. 12: 961. 28. Adams, R.D., N.A. Pepiatt, and J. Coppendale. Prediction of strength of joints between composite materials. 1978: Jointing in fibre reinforced plastics; Proceedings of the Symposium, London, England; 4-5 Sept. 1978. pp. 64-78. 1978. 29. Adkins, D.W. and R.B. Pipes. Planar scarf joints in composite repair. in Second International Conference on Composite Materials. 1978. Toronto, Canada. 30. Harman, A. and C. Wang, Analytic and FEA solutions for adhesive and adherend stresses in two dimensional scarf joints. DSTO, 2004. 31. Du, J., F.T. Salmon, and A.V. Pocius, Modeling of cohesive failure processes in structural adhesive bonded joints. Journal of Adhesion Science & Technology, 2004. 3: 287. 32. Soutis, C. and F.Z. Hu, 3-D failure analysis of scarf patch repaired CFRP plates. Collection of Technical Papers - AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics & Materials Conference, 1998. AIAA-98-1943: 1971. 33. Soutis, C. and F.Z. Hu, Failure analysis of scarf-patch-repaired carbon fiber/epoxy laminates under compression. 2000. 34. Soutis, C. and F.Z. Hu, Strength prediction of patch-repaired CFRP laminates loaded in compression. Composites Science and Technology, 2000. 7: 1103. 35. Liu, D. and N.A. Fleck, Scale effects in the initiation of cracking of a scarf joint. International Journal of Fracture. Vol. 95, 1999. 1-4: 67. 36. Murphy, N.A.D. and M.G. Thorton, Scarf joints in unidirectional carbon fiber reinforced plastics (tests of repairs made using adhesives) [B.S. Thesis]. 1980. 37. Kyogoku, H., T. Sugibayashi, and K. Ikegami, Strength Evaluation of Scarf Joints Bonded With Adhesive Resin. JSME Int. J. Vol. 30, 1987. 265: 1052. 38. Chalkley, P. and J. Van Den Berg, On obtaining design allowables for adhesives used in the bonded composite repair of aircraft. DSTO-TR-0608, 1998. 39. Metallic materials and elements for aerospace vehicle structures, ed. D.o.D.a.F.A. Administration. Vol. MIL-HDBK-5CD-ROM. 1996. 148 40. Davis, M.J. and Australia. Royal Australian Air Force. Logistics Systems Agency, Composite materials and adhesive bonded repairs. RAAF standard engineering. 1995, Laverton, Vic.: Logistics Systems Agency. 41. RAAF, D.G.T.A., Composite materials and adhesive bonded repairs, in DEf AUST 9005 replacing RAAF Std (Eng.) C5033. 1995. 42. Adkins, D.W. and R.B. Pipes, Tensile Behavior of Bonded Scarf Joints Between Composite Adherends. 1989. 43. Baker, A.A., R.J. Chester, G.R. Hugo, and T.C. Radtke, Scarf repairs to graphite/epoxy components. 1995. 44. Chester, R.J., A.A. Baker, G.R. Hugo, and T.C. Radtke, Scarf repairs to highly strained graphite/epoxy structure. 1999. 45. Wang, C.H. and A.B. Harman. Optimal Scarf Joint between unmatched adherends. in SAMPE. 2006. Long Beach, Calafornia. 46. Reithmaier, L.W., Standard aircraft handbook for mechanics and technicians. 6th ed. 1999, New York ; Sydney: McGraw Hill. 47. Jain, L., K. Leong, Y.W. Mai, and L. Tong, Effect of Through-Thickness Stitching on the Fatigue Life of Composite Single-Lap Joints. Applied Composite Materials, 1998: 399– 409. 48. Anon, Hysol EA9396C-2 Material Data Sheet. Last accessed 22/12/04, http://www.loctiteaero.com/Images/Datasheet_PDF/Hysol%20EA%209396C- 2.pdf. 49. Daniel, I.M. and O. Ishai, Engineering mechanics of composite materials. 1993, Oxford, England ; New York: Oxford University Press. 50. Hart-Smith, L.J., Adhesive bonded double lap joints, N. Contractor, Editor. 1973. 51. Ratwani, M.M. and H.P. Kan, Analysis of Composite-to-Metal Joints with Bondline Flaws. Composites Science & Technology, 1985. 1: 53. 52. Anon, Acceptable methods, techniques and practices-Aircraft inspection and repair, AC 43.13- 1B,. 1998, Federal Aviation Administration. 53. Anon, IMAGEJ software, in http://rsb.info.nih.gov/ij/. Last accessed 30/08/06. 54. Eisenmann, J.R. and C.Q. Rousseau, IBOLT: A composite bolted joint static strength prediction tool, in Joining and repair of composite structures, K.T. Kedwood, Editor. 2004, ASTM International: West Conshohocken, PA. p. 161. 149