DOT/FAA/AR-07/24 Evaluation of Composite Joints

Air Traffic Organization

Operations Planning
in General Aviation Structures
Office of Aviation Research
and Development
Washington, DC 20591

January 2009

Final Report

This document is available to the U.S. public through

the National Technical Information Service (NTIS),
Springfield, Virginia 22161.

U.S. Department of Transportation

Federal Aviation Administration
 NOTICE

This document is disseminated under the sponsorship of the U.S.

Department of Transportation in the interest of information exchange. The
United States Government assumes no liability for the contents or use
thereof. The United States Government does not endorse products or
manufacturers. Trade or manufacturer's names appear herein solely
because they are considered essential to the objective of this report. This
document does not constitute FAA certification policy. Consult your local
FAA aircraft certification office as to its use.

This report is available at the Federal Aviation Administration William J.

Hughes Technical Center’s Full-Text Technical Reports page:
actlibrary.tc.faa.gov in Adobe Acrobat portable document format (PDF).
 Technical Report Documentation Page
1. Report No. 2. Government Accession No. 3. Recipient's Catalog No.
DOT/FAA/AR-07/24
4. Title and Subtitle 5. Report Date

EVALUATION OF COMPOSITE JOINTS IN GENERAL AVIATION January 2009

STRUCTURES
6. Performing Organization Code

7. Author(s) 8. Performing Organization Report No.

Gerald V. Flanagan MSC TFR 6002/GE02

9. Performing Organization Name and Address 10. Work Unit No. (TRAIS)

Materials Sciences Corporation

11. Contract or Grant No.
181 Gibraltar Road
Horsham, PA 19044 DTRS57-03-C-10013
12. Sponsoring Agency Name and Address 13. Type of Report and Period Covered

U.S. Department of Transportation Final Report

Federal Aviation Administration 1/2/03 to 1/2/06
Air Traffic Organization Operations Planning
Office of Aviation Research and Development
Washington, DC 20591
14. Sponsoring Agency Code
ACE-120
15. Supplementary Notes
The Federal Aviation Administration William J. Hughes Technical Center Technical Monitor was Peter Shyprykevich.
16. Abstract

The report describes a Small Business Innovation Research program to develop an advanced analysis tool that addresses failure
prediction of bonded joints. The joining issues associated with general aviation aircraft were of particular concern. Many of
these structures are of thin-gage construction, which makes bending of the joint an important component of the response. The
resulting computer code accurately handles the high-stress gradients typically present in bonded joints. It allows for the bending
of the joint and will calculate interlaminar tensile stresses as well as interlaminar shear stress. The code allows for material
nonlinearity in the adhesive and for geometric nonlinearity of the overall joint. General boundary conditions can be applied at the
ends of the joint.

This report covers the software implementation of this solution. Efforts to validate the code using alternate solution methods and
available experimental data are also presented. Fundamental mechanics issues related to the response of an adhesive as bondline
thickness varies are also addressed.

17. Key Words 18. Distribution Statement

This document is available to the U.S. public through the
Bond, Adhesive, Joint, Composite National Technical Information Service (NTIS) Springfield,
Virginia 22161.
19. Security Classif. (of this report) 20. Security Classif. (of this page) 21. No. of Pages 22. Price
Unclassified Unclassified 137

Form DOT F 1700.7 (8-72) Reproduction of completed page authorized

 TABLE OF CONTENTS

Page

EXECUTIVE SUMMARY xv

1. INTRODUCTION 1

2. WORK PLAN 3

2.1 Task 1 Adhesive Inelasticity 3

2.2 Task 2 Adhesive Thickness Effects 3
2.3 Task 3 Composite Adherend Failure 3
2.4 Task 4 Inelastic Mixed Mode Fracture 4
2.5 Task 5 Core Code Enhancement 4
2.6 Task 6 Front-End Development 4
2.7 Task 7 Documentation 5
2.8 Task 8 Beta Test Program 5
2.9 Task 9 Typical Joint Elements 5

3. MATERIAL NONLINEARITY 6

3.1 Curve Fitting of Adhesive Stress-Strain Data 6

3.2 Nonlinear Material Models 9
3.3 Direct Iteration Process 10

3.3.1 Transverse Shear Model 12

3.3.2 von Mises Model 12

3.4 Discussion on Solution Convergence 15

4. EXISTING MODELING APPROACHES FOR THE BONDLINE THICKNESS

EFFECTS 16

4.1 Introduction 16
4.2 Conclusions 17
4.3 Experimental Observation 17
4.4 Modeling Approaches 18

4.4.1 Strain Gradient Plasticity Theories 19

4.4.2 Stress Singularity Model 20
4.4.3 Global Yield Model 21
4.4.4 Cohesive Zone Model 21
4.4.5 Statistical Model for Scaling Effect 22

iii
5. GEOMETRICAL NONLINEARITY 23

6. FAILURE OF BONDED COMPOSITE JOINTS 24

6.1 Factor of Safety Calculation 25

6.2 Failure Load for Nonlinear Adhesive—Maximum Strain Criterion 28
6.3 Stress Singularity 29

7. FRONT-END DESCRIPTION FOR Bond Master 29

8. Bond Master ANALYTICAL VERIFICATION PROBLEMS 48

8.1 Introduction 48
8.2 Description of ABAQUS Models 49
8.3 Bond Master Versus ABAQUS Linear Adhesive 50
8.4 Results of Verification of Linear Models Using ABAQUS 52

8.4.1 Single Lap Joint, Linear Adhesive, Tensile Load 52

8.4.2 Single Joggle Joint, Linear Adhesive, Tensile Load 55
8.4.3 Scarf Joint, Linear Adhesive, Tensile Load 58
8.4.4 Double Lap Joint, Linear Adhesive, Tensile Load 62
8.4.5 Double Joggle Joint, Linear Adhesive, Tensile Load 64

8.5 Alternate Bond Master Nonlinear Adhesive Formulations 66

8.5.1 Introduction 66
8.5.2 Problem Description 67
8.5.3 Bond Master Versus ABAQUS, Nonlinear Adhesive 72

8.6 Bond Master Versus ABAQUS, Elastic-Plastic Adhesive, Increasing Load 82

8.7 Bond Master Versus ABAQUS, Linear Adhesive, Out-of-Plane Load 88
8.8 Bond Master Versus ABAQUS, Nonlinear Adhesive, Aluminum Adherend 91

8.8.1 Introduction 92
8.8.2 Conclusions 92
8.8.3 Double Lap Joint Problem Description 92
8.8.4 Double Lap Joint Results and Discussion 94

8.9 Bond Master Versus ABAQUS, Single Lap Joint, Ply-by-Ply Failure Mode 97

8.9.1 Approach 98
8.9.2 Flush Lap Joint, Tension Loading 99
8.9.3 Flush Lap Joint, In-Plane Shear Loading 102
8.9.4 Scarf Joint 105

iv
 8.10 Bond Master Versus Test Data, Linear, and Nonlinear Adhesive 110
8.11 Bond Master Versus Test Data, Different Adhesive Thicknesses 114

8.11.1 Discussion of Results 117

9. REFERENCES 119

v
 LIST OF FIGURES

Figure Page

1 Typical Joggle Lap Joint Used in GA Application 1

2 SUBLAM-Generated Displacement Plots for Joggled and Straight Overlap Joints

Typical of GA Construction 2

3 Piecewise Linear Curve Fit to Scattered Data With (a) Four Segments and
(b) Eight Segments 7

4 Elastic-Perfectly Plastic Curve Fit to Scattered Data 8

5 Linear Elastic Approximation for Scattered Data 8

6 Example of Elastic-Perfectly Plastic Model 9

7 Examples of Ramberg-Osgood Model 9

8 Data-Based, Stress-Strain Curve With Linear Fitting and Spline Fitting 10

9 Approximated Elastic-Perfectly Plastic Model With 10-Term Legendre Polynomials 11

10 Schematic of the Iterative Process 11

11 A Typical Uniaxial Stress-Strain Curve 12

12 Shear Stress-Strain Curve and von Mises Equivalent Stress-Strain Curve for
AF126 (RT) Adhesive, With Shear Test Data From Reference 1 15

13 Shear Stress Distribution at the Lower Interface of the Bondline for a Single
Lap Joint With Converged Solution Corresponding to the Formation of a Total
Plastic Region Within the Adhesive Layer 16

14 Experimental Result From Single Lap Adhesive Joints (ASTM 3165—Thin

Adherend and ASTM 5655—Thick Adherend): Apparent Shear Strength vs
Bondline Thickness for Various Types of Adhesive 17

15 Apparent Stress-Strain Curves with Different Bondline Thickness 18

16 A Scarf Joint—Failure Load vs Bondline Thickness 18

17 Average Stresses in the Adhesive Layer of a Single Lap Joint (a) Shear Stress
Distribution Along the Bondline With t = 0.005″, 0.02″, 0.05″, and 0.1″, and (b) Peel
Stress Distribution Along the Bondline With t = 0.005″, 0.02″, 0.05″, and 0.1″ 19

18 Corner Points on the Interface of Adhesive Layer and Adherends in a Single

Lap Joint 20

vi
19 Schematic of Two Cohesive Zone Models for a Peel Test of Adhesive Joint
(a) A Single Cohesive Zone and (b) Two Elastic-Plastic Adhesive Layers With an
Embedded Cohesive Zone 22

20 Defects (Voids) in an Adhesive Material 23

21 Maximum Deflection of a Cantilever Beam Under Load vs the Length-Thickness

Ratio 24

22 Typical Failure Modes of the Bonded Joints 25

23 Schematic of an Element in a Bonded Joint Model 26

24 Stresses Spike at the Corners of the Lower and Upper Interfaces of the
Adhesive Layer 29

25 The Four Adhesive Material Models Supported by Bond Master 31

26 The Opening Screen Provides Access to the Main Menus 33

27 The “?” Symbols on Many Forms Link to Context-Sensitive Help Files 33

28 There is a General Help Utility Built Into the Program 34

29 Definition of Layer Properties for Incorporation in the Adherend Plate Laminates 34

30 The Option is Available to Define Strength Properties for the Plate Laminate Layers 35

31 General Laminates can be Defined Using the Existing Database of Layer Materials 35

32 The Program Comes With a Library of Adhesive Materials 36

33 Adhesives may be Defined in Terms of Model Parameters or as the Point-Wise Data

for a General Stress-Strain Curve 36

34 The Plotting Option Allows the User to View Different Adhesive Models for the
Stress-Strain Curves 37

35 Typical Starting Form for a Joint Model 37

36 Geometry Data Entry for Single Lap Joint 38

37 The Ends of the Plates may be Tapered 38

38 Geometric Inputs for Single Lap Joint With Flush Exterior 39

39 The Joggles Require Additional Data to Fully Define the Geometry 39

40 Geometric Data Needed for Symmetric Lap Splice 40

vii
41 Geometric Data Needed for Symmetric Lap Joint With Flush Exterior 40

42 Geometric Data Needed for a Scarf Joint 41

43 General Plate Boundary Conditions may be Applied at the Ends of the Adherends 41

44 Additional Solution Parameters are Available for User Control 42

45 There is an Extensive Array of Available Solution Plots 42

46 Example Report File Generated by Bond Master 43

47 Geometry for Single Lap Joint 52

48 Mesh of Single Lap Joint Showing Yellow Adherends and Blue Adhesive 52

49 Single Lap Joint, Linear Adhesive, Tensile Load 53

50 Geometry for Single Joggle Joint 56

51 Mesh of Single Joggle Joint Showing Yellow Adherends and Blue Adhesive 56

52 Single Joggle Joint, Linear Adhesive, Tensile Load 57

53 Geometry for Scarf Joint 59

54 Mesh of Scarf Joint Showing Yellow Adherends and Blue Adhesive 59

55 Close-Up of Mesh for Scarf Joint Showing Yellow Adherends and Blue Adhesive 59

56 Scarf Joint, Linear Adhesive, Tensile Load 60

57 Geometry for Double Lap Joint 62

58 Mesh of Double Lap Joint Showing Yellow Adherends and Blue Adhesive 62

59 Close-Up of Mesh for Double Lap Joint Showing Yellow Adherends and
Blue Adhesive 62

60 Double Lap Joint, Linear Adhesive, Tensile Load 63

61 Geometry for Double Joggle Joint 65

62 Mesh of Double Joggle Joint Showing Yellow Adherends and Blue Adhesive 65

63 Close-Up of Mesh for Double Joggle Joint Showing Yellow Adherends and
Blue Adhesive 65

64 Double Joggle Joint, Linear Adhesive, Tensile Load 66

viii
65 Schematic of the Single Lap Joint Configuration 67

66 Shear Stress-Strain Curve and von Mises Equivalent Stress-Strain Curve for
EA 9359 Adhesive 68

67 Shear Stress Distribution at the Lower Interface of the Adhesive Layer, for the Single
Lap Joint Under Axial Tensile Load 69

68 Peel Stress Distribution at the Lower Interface of the Adhesive Layer, for the
Single Lap Joint Under Axial Tensile Load 69

69 Strain Distribution (γyz, εy, εz) at the Lower Interface of the Adhesive Layer, for the
Single Lap Joint Under Axial Tensile Load 70

70 Shear Stress Distribution at the Lower Interface of the Adhesive Layer, for the
Single Lap Joint Under Bending Load 70

71 Peel Stress Distribution at the Lower Interface of the Adhesive Layer, for the
Single Lap Joint Under Bending Load 71

72 Strain Distribution (γyz, εy, εz) at the Lower Interface of the Adhesive Layer, for the
Single Lap Joint Under Bending Load 71

73 Single Lap Joint, Nonlinear Adhesive, Tensile Load 73

74 Single Joggle Joint, Nonlinear Adhesive, Tensile Load 75

75 Scarf Joint, Nonlinear Adhesive, Tensile Load 77

76 Double Lap Joint, Nonlinear Adhesive, Tensile Load 79

77 Double Joggle Joint, Nonlinear Adhesive, Tensile Load 81

78 Stress-Strain Curve for Adhesive 20 From the Bond Master Database 83

79 The Length of Plastic Deformation Along the Bond Increases as the Applied
Load Increases 84

80 The Single Lap Joint is Loaded With 1010 lb (This plot compares the shear stress
along the bottom of the bond as derived by Bond Master and ABAQUS.) 84

81 The Single Lap Joint is Loaded With 1010 lb (This contour plot from ABAQUS
shows the shear stress in the bond and adjacent adherends.) 85

82 The Single Lap Joint is Loaded With 1010 lb (This contour plot from ABAQUS
shows the plastic strain in the adhesive. Only the very end of the bond shows
plastic deformation.) 85

ix
83 Load on the Single Lap Joint was Increased to 1400 lb (The plastic zone increased
in length as evidenced by the higher length where the shear stress equaled the
yield stress.) 86

84 Load on the Single Lap Joint was Increased to 1400 lb (The length along the bond
where the shear stress equaled the yield stress has increased.) 86

85 Load on the Single Lap Joint was Increased to 1400 lb (The plastic zone has
increased in length.) 87

86 The Load on the Single Lap Joint was Increased to 1800 lb (The length of the
plastically deformed adhesive has increased as well.) 87

87 The Load on the Single Lap Joint was Increased to 1800 lb (The length increased
where the shear stress equaled the yield stress.) 88

88 The Load on the Single Lap Joint was Increased to 1800 lb (The length of the
plastically deformed adhesive has increased as well.) 88

89 ABAQUS Model of the Single Lap Joint 89

90 Single Lap Joint, Linear Adhesive, Out-of-Plane Load 90

91 Configuration of the Double Lap Joint 92

92 Shear Stress-Strain Curve for an Elastic-Perfectly Plastic Adhesive 93

93 Shear Stress-Strain Curve of AF126 Film Adhesive at RT 93

94 Elastic-Perfectly Plastic Adhesive: Shear Stress Distribution of the Top Interface

Along the Adhesive Bond 95

95 Elastic-Perfectly Plastic Adhesive: Peel Stress Distribution of the Top Interface

Along the Adhesive Bond 95

96 AF126 Film Adhesive: Shear Stress Distribution on the Top Interface Along
the Adhesive Bond 96

97 AF126 Film Adhesive: Peel Stress Distribution on the Top Interface Along the
Adhesive Bond 96

98 Flush Lap Joint 97

99 Close-Up of FE Mesh, Lap Joint 98

100 Flush Lap Joint Tension Loading Linear Elastic Adhesive First Ply Failure
Along Plate 1 99

x
101 Flush Lap Joint Tension Loading Linear Elastic Adhesive First Ply Failure
Along Plate 2 100

102 Tension Loading Nonlinear Ramberg-Osgood Adhesive Shear Stress Along Bond 101

103 Tension Loading Nonlinear Ramberg-Osgood Adhesive First Ply Failure

Along Plate 1 101

104 Flush Lap Joint Tension Loading Linear Elastic Adhesive First Ply Failure
Along Plate 2 102

105 In-Plane Shear Loading 102

106 In-Plane Shear Loading Linear Elastic Adhesive First Ply Failure Along Plate 1 103

107 In-Plane Shear Loading Linear Elastic Adhesive First Ply Failure Along Plate 2 103

108 In-Plane Shear Loading Nonlinear Ramberg-Osgood Adhesive Shear Stress

Along Bond 104

109 In-Plane Shear Loading Nonlinear Ramberg-Osgood Adhesive First Ply Failure
Along Plate 1 104

110 In-Plane Shear Loading Nonlinear Ramberg-Osgood Adhesive First Ply Failure
Along Plate 2 105

111 Scarf Joint 106

112 Close-Up of FE Mesh, Scarf Joint 106

113 Scarf Joint Linear Elastic Adhesive First Ply Failure Along Plate 1 108

114 Scarf Joint Linear Elastic Adhesive First Ply Failure Along Plate 2 108

115 Scarf Joint Elastic-Perfectly Plastic Adhesive Shear Stress Along Bond 109

116 Scarf Joint Elastic-Perfectly Plastic Adhesive First Ply Failure Along Plate 1 109

117 Scarf Joint Elastic-Perfectly Plastic Adhesive First Ply Failure Along Plate 2 110

118 Bond Master Matches the Failure Load of the Single Lap Joint but Overpredicts
the Stiffness 111

119 Bond Master Accurately Predicts Failure Load and Stiffness of the Joggle Joint 112

120 Bond Master Results Agree Well With Measured Data for the Single Lap Joint
With an L-Section 112

xi
121 Bond Master Slightly Over Predicts the Failure Load and Stiffness for the
Double Strap Joint 113

122 Bond Master Shows Good Agreement With the Measured Data for the L-Section
Joint With 0.127-mm (0.005-in.) Thick Adhesive 115

123 Bond Master Over Predicts the Failure Load of the L-Section Joint With 0.635-mm
(0.025-in.) Thick Adhesive 115

124 Bond Master Over Predicts the Failure Load of the L-Section Joint With
2.54-mm (0.100-in.) Thick Adhesive 116

125 Two Single Lap Joint Configurations 117

126 Deformation of Two Single Lap Joint Configurations 118

xii
 LIST OF TABLES

Table Page

1 Beta Test Sites 5

2 Material Property Inputs for Linear Studies 51
3 Geometry Inputs for Single Lap Joint 52
4 Geometry Inputs for Single Joggle Joint 55
5 Geometry Inputs for Scarf Joint 59
6 Geometry Inputs for Double Lap Joint 62
7 Geometry Inputs for Double Joggle Joint 65
8 Dimensions of the Single Lap Joint 67
9 Mechanical Properties of the Adherends and the Adhesive 67
10 Ply Properties 97
11 Adhesive Properties 98
12 Ply Properties 106
13 Adhesive Properties 107
14 Data From Figures 118 Through 121 in Tabular Form 113
15 Failure Modes 114
16 Data From Figures 122 Through 124 in Tabular Form 116
17 Failure Modes 117

xiii
 LIST OF ACRONYMS

ADO Active Data Objects

FAA Federal Aviation Administration
FE Finite element
F.O.S. Factor of safety
GA General aviation
MSC Materials Sciences Corporation
RT Room temperature
SBIR Small Business Innovation Research
2-D Two-dimensional
3-D Three-dimensional

xiv
 EXECUTIVE SUMMARY

This report addresses the development of a novel bonded joint analysis methodology, and the
inclusion of the methodology into a software product.

A major objective of the program was to release a software product for joint analysis along with
extensive documentation. This objective was met with the completion of the Bond Master
program. Bond Master is a sophisticated analysis tool that goes well beyond any of the existing
bonded joint codes in terms of including all the mechanics features relevant to a bonded joint. In
addition, the code includes an intuitive, easy-to-use front-end. The front-end guides the user
through the development of a model. The front-end uses a database-centric software model that
stores all the material data, laminate construction information, and model details in a series of
relational databases. The result is that a company can maintain all of its material data and
models in a central file that assures the analyst will have access to the correct and current data for
a project. These features help improve the reliability of the final structural assessment. The key
to making the model creation easy is to parameterize a set of common bonded joint
configurations. The user can simply enter the case-specific data to perform an analysis. The
front-end takes care of setting up an analysis model, using heuristic rules for ensuring the
accuracy of the model and results.

The Bond Master front-end also performs the failure analysis for both the adhesive and the
adherends. This approach isolates the failure theories from the core analysis software. Bond
Master is unique in its ability to check for interlaminar failure in composite adherends.
Examination of failed test specimens indicates that this is an important failure mode to consider.

Bond Master is built on top of an existing Materials Sciences Corporation (MSC) analysis code
called SUBLAM. Several major enhancements to the SUBLAM analysis code were
accomplished under the Small Business Innovation Research program. These included:

• The addition of material nonlinearity. This included the ability to handle several common
models to represent the adhesive nonlinear characteristics.

• The improvement of the nonlinear solvers to increase efficiency and accuracy.

• The addition of geometric nonlinearity. The geometric nonlinear feature required

extensive restructuring of the code.

• An iterative algorithm for determining failure loads for nonlinear adhesives was devised.

MSC performed research in several areas related to the detailed mechanics of bonded joints.
Specifically:

• A new model was developed for treating the multiaxial stress-state in the bondline. This
model is programmed into Bond Master.

xv
• Several approaches for handling thickness effects in adhesives were examined. A theory
based on a damage mechanics approach was developed and tested.

Extensive verification and validation testing was performed on the Bond Master code. The
majority of the verification work consisted of building and solving detailed finite element models
of all the joint types Bond Master addresses. All the bondline stresses and the overall joint
displacements were compared to the finite element results. Linear and nonlinear problems were
considered. In addition to the analytic verification, experimental validation was performed using
data that had been generated by general aviation for typical general aviation joint configurations.

xvi
1. INTRODUCTION.

Evaluating bonded joints for composite and metallic structures presents a number of challenges.
Some of the challenges are analytic in nature. Stresses change rapidly in the thin bondline. This
makes creating accurate numerical models of the joint difficult. In addition, the adhesives are
typically nonlinear, adding to the numerical difficulties. Other challenges are related to the basic
mechanics. For example, how should one predict failure when stress gradients are very large, or
when elastic analysis predicts a singular stress? Finally, there are challenges related to quality
assurance and flaws.

There are additional challenges unique to the general aviation (GA) industry. One difference for
these aircraft is the use of more relaxed dimensional tolerances. This means that the bondline is
frequently thicker than optimum and may vary in thickness over the joint length. Experimental
data indicates that adhesive strength varies with bondline thickness to a greater degree than can
be explained by stress analysis alone. Another challenge is the range of geometric details being
considered. Figure 1 shows a joggled joint being used in the Seawind aircraft. To properly
model this joint, the geometric details of the joggle need to be incorporated. In addition, these
joints are being used in the fuselage where the load state is complex. The joints must be
evaluated for in-plane shear, moment, and transverse shear, as well as the more typical axial
load. The ability of the Materials Sciences Corporation (MSC) codes to model complex
behavior is illustrated in figure 2.

Figure 1. Typical Joggle Lap Joint Used in GA Application

1
 Uniform Disp

(a)

(b)

(c )

Figure 2. SUBLAM-Generated Displacement Plots for Joggled and Straight Overlap Joints
Typical of GA Construction

A final challenge introduced by GA aircraft is the need for easy-to-use tools. The typical GA
manufacturer does not have the engineering resources, or money, to spend a large amount of
time developing joint models. To gain widespread acceptance, any tool should require only the
minimum geometry and material data input to set up the analysis. The tool should reliably
perform an accurate analysis, and present the results so that maximum load validation can be
easily accomplished. This program represents an ideal opportunity to fit between the extremes
of obsolete codes written in the 1970s, and overly expensive and difficult to interpret
general-purpose, finite element (FE) codes.

A major objective of this program was to release a sophisticated software product for joint
analysis along with extensive commercial quality documentation. This objective was met with
the completion of the Bond Master program. Bond Master is a sophisticated analysis tool that
goes well beyond any of the existing bonded joint codes in terms of including all the mechanics
features relevant to a bonded joint. In addition, the code includes an intuitive, easy-to-use front-
end. The front-end guides the user through the development of a model. The front-end uses a
database-centric software model that stores all of the material data, laminate construction
information, and model details in a series of relational databases. The result is that a company
can maintain all of its material data and models in a central file that assures the analyst will have
access to the correct and current data for a project. These features help improve the reliability of
the final structural assessment.

Another objective of this program was to improve the mechanics models used to represent
adhesive response, including failure. The Bond Master code includes a new model that uses the
full stress-state of the adhesive to track adhesive nonlinearity and failure. Several models were
examined in detail to explain the observed thickness effects in adhesive strength. These models
were not included in the final commercial code, but the research provides a basis for future work.

2
The final objective of this program was to verify and validate the code. Extensive analytical
validations were performed by comparing Bond Master to detailed, nonlinear ABAQUS FE
results. In addition, Seawind, Inc. provided experimental data for typical GA joints that became
part of the validation process.

2. WORK PLAN.

This program was divided into two groups of tasks. The first group is related to the development
of models for the prediction of failure in both the adhesive and the adherend. The second group
is related to code development and its incorporation into a commercial product. A description of
each task is given below followed by a summary of the accomplishments.

2.1 TASK 1 ADHESIVE INELASTICITY.

Current bonded joint analysis methods consider only the nonlinearity of the adhesive transverse
shear stiffness. An advanced tool such as SUBLAM gives all the stress components. A J2 type
deformation theory will be implemented for modeling inelastic behavior of the adhesive under
multiaxial stress states. This approach is expected to yield better correlation of predicted and
measured failure loads due to adhesive or adherend failure.

Task 1 was accomplished and is now a user-selected option in the initial code release.
Additional research showed that the best approach to account for material nonlinearity used a
mathematical model based on the deformation theory of the plasticity and the von Mises yield
criterion. Details of the model are given in section.3.2.

In a related activity, test data generated by Wichita State University, through a Federal Aviation
Administration (FAA) sponsorship, will be shipped with the Bond Master code in a reduced
form suitable for analysis. The code includes the parameters for each of several adhesive
models, for each of the materials and environmental conditions tested by Wichita State. New
data reduction techniques had to be developed. These are discussed in section 3.1.

2.2 TASK 2 ADHESIVE THICKNESS EFFECTS.

A one-dimensional model considering the effect of inelastic strain gradient in a damage band,
which grows with straining, was used previously to model the reduction of peak shear stress with
increasing bondline thickness and the drastic reduction in failure strain for large thickness.

Several possible models for explaining the thickness effect were examined in detail during this
program. A summary of this research is provided in section 4. One result of the research was
that the damage mechanics-based model was not sufficiently mature or accurate enough to be
included in a commercial product.

2.3 TASK 3 COMPOSITE ADHEREND FAILURE.

Composite adherend failure is often the mode of failure in typical GA bonded structures. Several
alternative ply failure criteria for combined stresses will be used after analyzing typical joints

3
with due consideration to adhesive inelasticity. Results will be correlated with test data for
selecting one or more of them as needed.

Bond Master includes failure criteria for both the adherends and the bondline. The adherend
failure calculations are described in section 6.1. The failure of nonlinear adhesives is described
in section 6.2. A point-stress criterion involving an empirical offset distance is discussed in
section 6.3. Bond Master has the ability to evaluate interlaminar stress failures in the adherends.

2.4 TASK 4 INELASTIC MIXED MODE FRACTURE.

Calculation of strain energy release rates based on elastic analyses can be performed using the
SUBLAM code. The calculation needs to be modified for inelastic properties and validated.

Based on feedback from potential users, MSC made the decision to focus the Bond Master code
on stress-based failure modes. The core analysis code is capable of efficiently performing a
fracture analysis, but the front-end will not currently create fracture models. This is a reflection
of the current acceptability of fracture methods for joints. Future releases of Bond Master are
likely to incorporate both methods.

2.5 TASK 5 CORE CODE ENHANCEMENT.

Several features were added to the core analysis code, including geometric nonlinearity, and an
improved convergence algorithm for nonlinear solutions. In addition, the mechanics model
being developed in the model development group of the tasks was implemented in the code.

Several major enhancements to the SUBLAM analysis code were accomplished under the Small
Business Innovation Research (SBIR) program. These included:

• The addition of material nonlinearity (section 3.2)

• The improvement of the nonlinear solvers to increase efficiency and accuracy

(section 3.3)

• The addition of geometric nonlinearity (section 5)

• An iterative algorithm for determining failure for nonlinear adhesives (section 3.3)

In addition to the above mentioned work, extensive validation tests were performed during the
program that allowed MSC to improve the code efficiency and accuracy.

2.6 TASK 6 FRONT-END DEVELOPMENT.

A large family of parametric models were developed using procedures similar to those
demonstrated previously.

The front-end development was successfully concluded, and a robust, high-quality product was
created. A description of many of the front-end features can be found in section 7. The front-

4
end allows the user to analyze many common joint configurations using predefined parametric
models. The user simply enters the current dimensions and data into the appropriate fields in a
graphical display. The front-end also manages a system that allows the user to maintain material
properties, common laminates, and existing joint models in an easy to access database. Finally,
the front-end allows the user to view output in a variety of formats, including a printed results
document that summarizes the model input and the results.

2.7 TASK 7 DOCUMENTATION.

A commercial software product must be extensively documented with user instructions and
guidance, reference solutions, example problems, and a thorough description of the models.

The product includes an internal help file (see section 7), along with a separate theoretical
manual (consisting of sections 3 through 6 of this report, plus additional information on the
internal formulation of SUBLAM). It will also be shipped with all the validation information.
Together, these represent a far larger body of knowledge than is typically included with specialty
engineering codes.

2.8 TASK 8 BETA TEST PROGRAM.

MSC solicited beta testing of the code at multiple sites. The feedback from the beta testing was
used to improve the code for future commercialization.

A Beta version of Bond Master was created and sent to several sites that had expressed interest
in the code (table 1). These same sites will be provided with 30-day trial versions of the final
release.

Table 1. Beta Test Sites

Individual Company
Ray Kaiser Northwest Airlines, Inc.
Dennis Roach Sandia National Laboratories
Daniel C. Hammerand Sandia National Laboratories
F. Scapinello Agusta
B. Cole Aurora Flight Sciences
Ray Kaiser Northwest Airlines, Inc.
Jim Krone Cessna Aircraft Company
Daniel Jaredson Saab Ericsson Space
Roderick Martin MERL

2.9 TASK 9 TYPICAL JOINT ELEMENTS.

The goal of testing is to validate the system using typical bonded joints. Examples would be
joggled laps and double laps.

5
A thorough verification and validation program was executed. The results are described in
section 8. The majority of the verification effort was in analytic comparisons to FE results. This
process assures that the code was correctly modeling all of the physics, and could replicate the
stress state within both the adhesive and adherends, to an accuracy comparable with highly
detailed FE models. In addition, comparisons were made to the test data provided to MSC by
Seawind Inc.

3. MATERIAL NONLINEARITY.

It is well known that adhesives used in the structural building exhibit inelastic behavior.
Specifically, local plastic strains are induced even with low external loads. To realistically
model the response of the adhesive joints, it is essential to consider material nonlinearity of the
adhesives. The adhesive is considered an isotropic material. The nonlinear adhesive behavior is
often described via a stress-strain curve from a uniaxial tension or shear test. Several
mathematical models, such as elastic-perfectly plastic, bilinear, and Ramberg-Osgood, are
further employed to characterize this nonlinear behavior for the sake of analytic and numerical
analysis. This section presents the treatments for nonlinear adhesives in Bond Master.

3.1 CURVE FITTING OF ADHESIVE STRESS-STRAIN DATA.

The Bond Master database includes curves for a large number of adhesive materials based on the
data generated at Wichita State University under FAA sponsorship [1]. Generating the curve fits
required the development of some specialized functions in Mathematica. There are several
considerations that went into making the curve fits. First, the raw stress-strain data for a single
test does not form a smooth curve. This is due to the finite precision of the test and the data
recording process. Connecting the individual data points to form a curve would result in a
slightly jagged curve that could potentially cause the numerical solutions within Bond Master to
converge slowly or even fail. Another consideration was a method to handle multiple data sets
for each adhesive material. One method would be to include only a representative curve in the
database. However, there was a desire to include all the available data in some manner. Also,
the methods devised had to be automated and robust for handling a large quantity of data.
Finally, the methods had to be adaptable to a variety of material models typically used to
describe adhesive properties.

The approach chosen was to superimpose all the curves for a particular material and
environmental condition. This resulted in points that were scattered about an implied material
curve. The next step was to fit a curve to the points. This was done by using a least squares fit
routine for a curve constructed from linear segments. The fit routine was programmed into
Mathematica. An example of the process is shown in figure 3. Most data sets are well
represented by 8-10 segments.

It was found that the iteration process in Bond Master could be accelerated if the slope
discontinuities between segments were eliminated. Therefore, Bond Master goes one step
further and smoothes the curves through the use of a cubic spline fit that goes through the linear
segment end points. The cubic spline interpolation uses third-degree polynomials to connect the
data points, which results in smooth curve fits. Specifically, given a set of N data points, {(x0,

6
f(x0)), (x1, f(x1)), (x2, f(x2)),… (xN, f(xN)), it is assumed that f '' ( x0 ) = f '' ( xN ) = 0 , and the rest of
second-order derivatives can be evaluated from:

( xi − xi−1 ) f '' ( xi−1 ) + 2( xi+1 − xi−1 ) f '' ( xi ) + ( xi+1 − xi ) f '' ( xi+1)

6 6 (1)
= [ f ( xi+1) − f (xi )] + [ f (xi−1) − f ( xi )] i = 1, 2, 3," n −1
( xi+1 − xi ) ( xi − xi−1 )

The cubic function for the ith interval can then be expressed as

f '' (xi−1 ) f '' (xi )

fi (x) = (xi − x) +
3
(x − xi−1)3
6(xi − xi−1 ) 6( xi − xi−1)
⎡ f (xi−1 ) f '' (xi−1)(xi − xi−1 ) ⎤ ⎡ f ( xi ) f '' ( xi )( xi − xi−1) ⎤ (2)
+⎢ − (
⎥ i x − x) + ⎢ − ⎥ (x − xi−1)
⎣ (xi − xi−1 ) 6 ⎦ ⎣ (xi − xi−1 ) 6 ⎦
i = 1, 2, 3," n −1

where prime denotes the derivative with respect to x.

1 1

0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2

2 4 6 8 2 4 6 8
(a) (b)

Figure 3. Piecewise Linear Curve Fit to Scattered Data With (a) Four Segments and
(b) Eight Segments

Other material models can be derived from the piecewise fit. For example, the Hart-Smith
approach for constructing an elastic-perfectly plastic approximation is to specify that the energy
under the approximate curve is equal to the actual material curve, and that the plastic stress is
equal to the maximum stress for the material. This is easily constructed by directly integrating
the piecewise curve, and interrogating the curve for a maximum stress value. An example is
shown in figure 4.

7
 1

0.8

0.6

0.4

0.2

2 4 6 8

Figure 4. Elastic-Perfectly Plastic Curve Fit to Scattered Data

Finding a general method for determining an appropriate initial linear approximation to the data
was the most challenging numerical problem. The problem is finding a method for choosing
how many points should be included in the fit for the initial slope. The solution was to employ
the linear segment fit in an iterative manner. The method involves comparing the change in
slope between the first and second segments to a user-established tolerance. If the change in
slope is greater than the tolerance, then the data is subdivided to allow for an additional segment,
and the check is repeated. Once the change in slope is below the tolerance, the slope of the first
segment is the desired initial stiffness. Figure 5 shows a fit to a continuously varying, scattered
set of data, as computed using this algorithm.
2

1.5

0.5

2 4 6 8

Figure 5. Linear Elastic Approximation for Scattered Data

The Ramberg-Osgood fit was performed using the Mathematica nonlinear fit functionality. A
series of approximation equations were used to get accurate initial estimates for the Ramberg-
Osgood parameters so that the nonlinear fit would converge.

8
3.2 NONLINEAR MATERIAL MODELS.

In Bond Master, the material nonlinearity is characterized via three models: elastic-perfectly
plastic, Ramberg-Osgood, and piecewise/data based. Figures 6 through 8 illustrate the three
options for the nonlinear materials.

Figure 6. Example of Elastic-Perfectly Plastic Model

Figure 7. Examples of Ramberg-Osgood Model

9
 Figure 8. Data-Based, Stress-Strain Curve With Linear Fitting and Spline Fitting
(Test data are for EA 9359 adhesive from reference 1.)

In the elastic-perfectly plastic model, the nonlinear material behavior is described by two distinct
regions: linear and perfectly plastic; the corresponding parameters are E0 (or G0) and σ0 (or τ0).
The Ramberg-Osgood model describes the stress-strain relation via the following form:

β β
σ ⎡ ⎛σ ⎞ ⎤ τ ⎡ ⎛τ ⎞ ⎤
ε = ⎢1 + ⎜ ⎟ ⎥ or γ = ⎢1 + ⎜ ⎟ ⎥ (3)
E0 ⎢ ⎝ σ 0 ⎠ ⎥⎦ G0 ⎢ ⎝ τ 0 ⎠ ⎥
⎣ ⎣ ⎦

The piecewise option allows the user to describe the material nonlinear behavior via a set of test
data points. Although linear fitting is the simplest approach, it may require many data points to
obtain a smooth curve, depending on the nature of the data. In Bond Master, the cubic spline
interpolation is used. Figure 8 displays both the linear fitting and spline fitting. As expected, the
difference between the two is more pronounced when the greater changes are present in the test
data. The increasing smoothness of the stress-strain curve may speed up the convergence rate in
an iterative computational process.

3.3 DIRECT ITERATION PROCESS.

In Bond Master, the material nonlinearity is addressed from two aspects, namely, allowing the
variation of the material properties within a single element and modeling material properties as a
function of strain. The former is accomplished by giving the material properties a functional
form that varies with position. Figure 9 demonstrates the approximation of the elastic-perfectly
plastic model using 10-term Legendre polynomials. Appropriate integration of the strain-energy
will account for the material variation in the solution series.

10
 Figure 9. Approximated Elastic-Perfectly Plastic Model With 10-Term Legendre Polynomials

For the latter, the material properties are updated according to the equivalent strain in the
adhesive layer. The solution requires an iteration scheme to achieve equilibrium. A secant
modulus approach is chosen, owing to its simplicity and reliability in terms of convergence even
for highly nonlinear materials. Figure 10 shows the schematic of the iterative process.
Specifically, the linear solution is obtained first (by using linear properties); the equivalent strain
is then computed. Based on the nonlinear stress-strain curve, a secant modulus is determined,
and the corresponding material properties are corrected/updated. With the updated material
properties, another set of solutions is then computed. The process repeats until reaching the
equilibrium state, which is controlled by a convergence criterion.

It is noted that the material nonlinearity of adhesives is often given in two forms: uniaxial tensile
stress-strain curves and shear stress-strain curves. In Bond Master, there are two models to use
these one-dimensional properties, namely, the transverse shear model and the von Mises model.

Figure 10. Schematic of the Iterative Process

11
3.3.1 Transverse Shear Model.

It is believed that the load transfer in the adhesive is primarily shear; hence as a simplified
model, the material nonlinearity can be handled via updating the transverse shear modulus based
on the equivalent shear strain:

γ eff = γ xz2 + γ y2z (4)

Note that there are relatively less terms (in the laminate stiffness matrices) that need to be
updated accordingly in the iterative process; hence, less computational time is needed. However,
when the stress state in the adhesive layer is multiaxial, such model may not be justified.

3.3.2 von Mises Model.

The von Mises option is based on the deformation theory of the plasticity and the von Mises
yield criterion. The mathematical model, which was originally proposed by Szabo, et al. [2], and
its three-dimensional (3-D) extension are outlined below. Consider a typical uniaxial stress-
strain curve (see figure 11). Here, E is the elastic modulus, and Es is the secant modulus; the
superscript e and p denote for elastic and plastic, respectively.

Figure 11. A Typical Uniaxial Stress-Strain Curve [2]

Consider the small strain, the equivalent elastic and plastic strains are given by:

e 2
ε = (ε1e − ε e2 )2 + (ε e2 − ε 3e )2 + (ε 3e − ε1e ) 2
2(1 +ν )
(5)
p 2
ε = (ε1p − ε 2p )2 + (ε 2p − ε 3p )2 + (ε 3p − ε1p ) 2
3

12
where ε1, ε2, and ε3 are the principle strains and ν is the Poisson’s ratio. The total equivalent
strain is defined by:

e p (6)
ε =ε +ε

Defining the average stress as σ ave = (σ x + σ y + σ z ) / 3 , and the deviator stress tensor is given by:

⎧σ x ⎫ ⎧σ x ⎫ ⎧σ ave ⎫
⎪ ⎪ ⎪ ⎪ ⎪ ⎪
⎪σ y ⎪ ⎪σ y ⎪ ⎪σ ave ⎪
⎪σ ⎪ ⎪σ ⎪ ⎪σ ⎪
⎪ z ⎪ ⎪ z ⎪ ⎪ ave ⎪
⎨ ⎬= ⎨ ⎬- ⎨ ⎬ (7)
⎪τ xy ⎪ ⎪τ xy ⎪ ⎪0 ⎪
⎪τ ⎪ ⎪τ ⎪ ⎪0 ⎪
⎪ yz ⎪ ⎪ yz ⎪ ⎪ ⎪
⎪⎩τ zx ⎪⎭ ⎩⎪τ zx ⎪⎭ ⎪⎩0 ⎪⎭

To generalize the uniaxial stress state for which the experimental information is available, the
following assumptions are made:

a. The total strain ε is the sum of the elastic strain εe and the plastic strain εp.

ε=ε e +ε p (8)

The absolute values of the stress components are nondecreasing, and the stress
components remain in a fixed proportion as the deformation progresses.

b. The plastic strain is proportional to the deviator stress tensor; specifically

⎧ε xp ⎫ ⎧σ x ⎫
⎪ p⎪ ⎪ ⎪
⎪ε y ⎪ ⎪σ y ⎪
⎪ p⎪
⎪ε z ⎪ 3 ⎛⎜ 1 1 ⎞ ⎪⎪σ z ⎪⎪
⎨ p⎬ = − ⎟⎨ ⎬ (9)
⎪ε xy ⎪ 2 ⎜⎝ E s E ⎟⎠ ⎪ τ xy ⎪
⎪ p⎪ ⎪τ ⎪
⎪ε yz ⎪ ⎪ yz ⎪
⎪ε ⎪p
⎩⎪ τ zx ⎪⎭
⎩ zx ⎭

13
From equations 7 and 9, one can obtain a relationship between the total strain components and
the stress components for the 3-D case:

⎧ εx ⎫ ⎛ ⎡ 1 −ν −ν 0 0 0 ⎤ ⎡ 2 −1 −1 0 0 0⎤ ⎞ ⎧ σ x ⎫
⎪ ε ⎪ ⎜ ⎢ −ν −ν 0 ⎥⎥ ⎢ ⎟
0 0 0 ⎥⎥ ⎟ ⎪⎪ σ y ⎪⎪
⎪ y⎪ ⎜ ⎢ 1 0 0 ⎢ −1 2 −1
⎪⎪ ε z ⎪⎪ ⎜ 1 ⎢ −ν −ν 1 0 0 0 ⎥ E − Es ⎢ −1 −1 2 0 0 0 ⎥ ⎟ ⎪⎪ σ z ⎪⎪
⎨ ⎬=⎜ ⎢ ⎥+ ⎢ ⎥⎟⎨ ⎬ (10)
⎪ γ xy ⎪ ⎜ E ⎢ 0 0 0 2(1 +ν ) 0 0 ⎥ 2 EEs ⎢0 0 0 6 0 0 ⎥ ⎟ ⎪ τ xy ⎪
⎪ γ yz ⎪ ⎜ ⎢ 0 0 0 0 2(1 +ν ) 0 ⎥ ⎢0 0 0 0 6 0 ⎥ ⎟ ⎪ τ yz ⎪
⎪ ⎪ ⎜⎜ ⎢ ⎥ ⎢ ⎥⎟⎪ ⎪
⎩⎪ γ zx ⎭⎪ ⎝ ⎣ 0 0 0 0 0 2(1 +ν ) ⎦ ⎣0 0 0 0 0 6 ⎦ ⎟⎠ ⎩⎪ τ zx ⎭⎪

The matrix in equation 10 is the elastic-plastic compliance matrix, which can be inverted to the
material stiffness matrix directly. Note that all stress and strain components are related to the
secant modulus, instead of the shear components.

Again, the direct iteration process is employed. The steps are outlined below:

• Compute the total equivalent strain as given in equations 5 and 6, except for the first
iteration. For the first iteration, the total equivalent strain is set to equal the equivalent
elastic strain from the linear solution.

• Determine the secant modulus according to the total equivalent strain.

• Update the stiffness matrix C.

• Obtain the solution with the updated C, until the solution is converged.

Note that the plastic strain components can be evaluated, although its output is not included in
the current release of Bond Master. In addition, a shear stress-strain curve can be converted to
an equivalent stress-strain curve based on the von Mises equivalent stress-strain concept [3].
Specifically, the von Mises stress concept represents the relationship between pure uniaxial
tension and pure shear characteristics of an isotropic material. When the material is subjected to
pure shear loading, the von Mises equivalent stress and strain are defined as

σ eq = σ x = 3τ xy
(11)
2 1
ε eq = ε x = ε xy = γ xy
3 3

Figure 12 displays the shear stress-strain curve and the converted von Mises equivalent stress-
strain curve.

14
 Figure 12. Shear Stress-Strain Curve and von Mises Equivalent Stress-Strain Curve for
AF126 (RT) Adhesive, With Shear Test Data From Reference 1

3.4 DISCUSSION ON SOLUTION CONVERGENCE.

Unlike linear elastic analysis for nonlinear material, the large load can lead to a diverged
solution. The usual reason for divergence is that the material is not capable of taking any
additional load when the entire region becomes plastic. As the load is gradually reduced, the
load corresponding to the first converged solution is the maximum load at which the material
becomes fully plastic. In the case where a large load is applied to the bonded joint, if the initial
determination attempt results in a diverged solution, Bond Master will automatically attempt to
find the first converged solution via adaptively reducing the load. Note that the deformation,
stress, and strain output are with respect to the reduced load, which can be determined using the
total load correction factor.

A single lap joint under large axial tension (figure 13) was selected for the purpose of
demonstration. The properties used were for AF126 adhesive at room temperature and the
piecewise option was used. The yield stress is 6700 psi. The initial applied load of 7200 lb
yielded a diverged solution. After four attempts, a converged solution was obtained, and the
corresponding load was reduced to 6655 lb. Figure 13 displays the transverse shear stress at the
lower interface of the adhesive layer. It is seen that the shear stress in most of the region equals
the yield stress except near the end of the bondline. This indicates that the entire adhesive layer
became plastic at the current load (6655 lb) and would not be capable of taking any load beyond
this point.

15
 Figure 13. Shear Stress Distribution at the Lower Interface of the Bondline for a Single
Lap Joint With Converged Solution Corresponding to the Formation of a Total Plastic
Region Within the Adhesive Layer

4. EXISTING MODELING APPROACHES FOR THE BONDLINE THICKNESS EFFECTS.

4.1 INTRODUCTION.

Experimental studies show that the strength of the adhesive joints depends on the thickness of
the adhesive layer. For some adhesives, joint strength increases with adhesive thickness until an
optimal adhesive thickness; beyond this thickness, joints strength decreases [4]. For example, in
a recent study on the adhesive joints used in GA aircraft [5], it was observed that the peak stress
and the failure strain reduce with thicker adhesive. At the same time, the stress analyses based
on elastic and elastic-plastic models indicate that the joint strength increases as the thickness
increases, owing to the lower maximum normal and shear stresses.

In GA aircraft structures, use of thicker adhesive layers is becoming a standard practice because
of considerable simplification and cost saving in the manufacturing process; hence, it is
important to quantify this effect for the design purpose. An extensive literature review was
conducted on the modeling approaches that deal with thickness dependence, in the hope that a
suitable approach can be identified, and further implemented for the current project. In this
report, five modeling approaches are summarized, and the merits and limitations are discussed.

16
4.2 CONCLUSIONS.

An extensive literature review was conducted on the modeling approaches that deal with the
thickness dependence. Five modeling approaches are presented in this report. It will be seen
that, as these current available models demonstrate some successes in certain cases, further
development and validation are required, especially for elastic-plastic adhesive. Among them,
the cohesive zone model shows some promising results; however, the model was fracture
mechanics-based and to implement a fracture mechanics-based failure criterion was beyond the
focus of the current project.

4.3 EXPERIMENTAL OBSERVATION.

The effect of the bondline thickness has been observed in several experimental studies [4, 5,
and 6]. According to a recent study by Tomlin, et al. [5], the single lap joint strength decreases
as the bondline thickness increases. The same trend was found in two test configurations: thin
adherend (ASTM 3165) and thick adherend (ASTM 5656), and in six adhesives. The thickness
ranges from 0.013 (0.33 mm) to 0.12 (3.05 mm). Figure 14 plots the apparent strength versus
bondline thickness, and figure 15 shows the shear stress-strain curves for various t (bondline
thickness). It is shown that, as t increases, not only the peak stress decreases, but also the failure
strain decreases drastically. In reference 7, a scarf joint with thin adhesive layers was studied.
The result is shown in figure 16, which shows that the joint strength reaches maximum when the
thickness is ~0.004 (0.1 mm) and decreases after 0.004.

Figure 14. Experimental Result from Single Lap Adhesive Joints (ASTM 3165—Thin Adherend
and ASTM 5655—Thick Adherend): Apparent Shear Strength vs Bondline Thickness for
Various Types of Adhesive, Reference 5

17
 Figure 15. Apparent Stress-Strain Curves With Different Bondline Thickness, Reference 6

Figure 16. A Scarf Joint—Failure Load vs Bondline Thickness, Reference 7

4.4 MODELING APPROACHES.

The effect of the bondline thickness has received increasing attention recently. Note that in the
stress analysis of the adhesive joints, it is often assumed that the stress field in the adhesive layer
is uniform through the thickness. The average shear and peel stresses are then used to predict the
joint strength according to point-stress-based criteria. It is well known that the maximum
average stresses decrease with increasing thickness. As an example, a single lap joint with
EA9359 adhesive under axial tension is analyzed. Figure 17(a) and (b) show the average shear
and peel stress distributions along the length of the adhesive layer with t = 0.005″, 0.02″, 0.05″,
and 0.1″. Clearly, using the maximum stress-strain criteria will lead to higher joint strength
prediction, which is contrary to the experimental data. There have been modeling efforts in the

18
literature. These models can be classified into five approaches, namely, strain gradient plasticity
theories, stress singularity model, global yield model, cohesive zone model, and statistical
model. In the following section, each model approach will be outlined, and the focus will be
placed on the assumptions made in the models and their implications.

Figure 17. Average Stresses in the Adhesive Layer of a Single Lap Joint (a) Shear Stress
Distribution Along the Bondline With t = 0.005″, 0.02″, 0.05″, and 0.1″, and (b) Peel Stress
Distribution Along the Bondline With t = 0.005″, 0.02″, 0.05″, and 0.1″

4.4.1 Strain Gradient Plasticity Theories.

In an earlier effort at MSC [7], a strain-softening model is explored to model the response of
bonded joints with a thick adhesive layer, in particular, the decreasing peak stress and failure
strain with increasing bondline thickness. It is thought that the difference between the

19
experimental observation and the prediction based on the stress analyses is caused by stable and
unstable structural response due to the relative sizes of the damage zone or band (compared with
the layer thickness) for a thin and thick adhesive layer, respectively. Based on the concept
proposed by Schrayer and Chen [8], a model of strain softening (associate with localization) is
developed with the assumption that the shear stress is constant along the length of the adhesive
layer. Moreover, the focus is placed on the structural response in the post-peak regime; an
illustrative example, correlating with test data for Hysol EA9394, shows the model is capable of
capturing the post-peak regime. But the reduction of the peak stress is not considered. It is
suggested that a form of initial imperfection may be introduced to account for the reduction of
the peak stress with increase in the bondline thickness [8 and 9].

In the SUBLAM analysis, large elements are intended to be used for modeling adhesive joints.
Within an element, the stresses are not constant along the length of the element. Furthermore,
both shear and peel stresses in the adhesive layer can be rather high. To adequately model and
describe the effect of the adhesive layer thickness, other more complex strain-gradient theories
may be considered; however, the development of the strain-gradient plasticity theory is still an
on-going effort [9].

4.4.2 Stress Singularity Model.

It is also considered that the interface stresses may play an important role in the bondline
thickness effect. At the interfaces between the adhesive and the adherend, stress singularities are
present at the edge of the interface due to material and geometric discontinuity, such as the four
corners shown in figure 18. These singularities complicate the interpretation of the interface
results; consequently, the average stresses or the stresses along the centerline of an adhesive
layer are used for the purpose of simplification. It is argued that the maximum stresses are often
not along the centerline, especially for the thicker adhesive layer, and the interface stresses may
provide the explanation for the thickness effect, if the stress singularities can be handled
properly.

Figure 18. Corner Points on the Interface of Adhesive Layer and Adherends in a
Single Lap Joint

There are two methods to handle the singularity: a stress singularity approach [4] and the stress
at a distance [10]. For the former, the stress field near the singularity can be characterized by a
stress-intensity factor H, which can often be determined via a detailed FE analysis. The stress-
intensity factor H depends on the geometrical configuration of the joint, the material properties,
and the applied load. It is assumed that the failure of the bonded joint occurs when H reaches a
critical value Hc, the bondline toughness. Hc is a material/geometry-related scalar quantity

20
similar to the fracture toughness and needs to be determined from the experiments. The reported
FE results show that for the single lap joint studied, the value of H decreases as the bondline
thickness increases, and for a given Hc, the predicted failure loads correlate well with the
experimental data in general. Note that this approach assumes a small yield in the adhesive layer
after fracture and is only applicable to relatively brittle adhesives. In addition, the evaluation of
the stress-intensity factor H is based on the elastic model, while the adherend in the bonded
joints often exhibits elastic-plastic behavior.

In the stress-at-a-distance approach, a stress value is taken at a short distance from the
singularity, which is not influenced by the singularity stress field, but close enough to be
considered as the approximation of the maximum stresses. A point-stress criterion can then be
applied to predict the joint strength. The reported FE analyses demonstrated that the interface
shear and peel stresses at the short distance from the free edge vary with the bondline thickness,
specifically, the stresses increase with increasing bondline thickness after the thickness is larger
than 0.3 mm. It is concluded that the maximum stress-based prediction will correlate well with
the experimental trend. Clearly, the study demonstrates the importance of the interface stress,
and the limitation of the assumption of constant stresses through the bondline thickness.
However, there are two issues with this approach: the selection of the short distance away from
the singularity is rather random (mesh dependent), and the analysis considered therein is linear-
elastic.

4.4.3 Global Yield Model.

A global yield criterion is introduced by Crocombe [11] to predict the thickness effect. The
theory assumes a complete yielding of the entire adhesive layer for failure, the thickness effect is
explained by the fact that the entire adhesive layer yields faster in a thicker bondline than in a
thinner one. Although the theory is capable of showing that the strength decreases as the
thickness increases, the assumption is not justified because most joints fail before the entire
adhesive yield, especially for relatively brittle adhesives.

4.4.4 Cohesive Zone Model.

Fracture mechanics-based models are common approaches to evaluate the strength or toughness
of the bonded joints with initial damages and cracks. As a recent development, the cohesive
zone models are introduced to accurately describe the response of adhesive layers to mechanical
loads and to simulate crack propagation (see references 12, 13, and 14). In this approach, the
adherend is modeled as elastic-plastic continuum, and the adhesive layer is represented by a
cohesive zone [13] or by two elastic-plastic adhesive layers with an embedded cohesive zone
[14]. Figure 19 shows the schematic of the two representations. The response of the cohesive
zone is governed by the traction separation law, as described in the works by Tvergaard and
Hutchinson [15] and Wei and Hutchinson [16]. In reference 15, the bond toughness Γ is defined
as Γ 0 + Γ p , where Γ0 is the intrinsic work of the fracture associated with the embedded cohesive
zone response and the Γp is the contribution rising from the plastic dissipation and stored elastic
energy within the adhesive layer. Furthermore, Γ0 is a material parameter independent of local
stress state and changes induced by the adhesive layer thickness translate into a change of Γp.
The numerical results demonstrate the effect of the bondline thickness for two adhesive

21
materials, in terms of the dependence of Γp and the height of the plastic zone on the adhesive
layer thickness. This approach shows promising results. The limited parameters and the
capability of handling mix-mode loads are also attractive. However, implementing a fracture
mechanics-based model is beyond the scope of the current project.

Figure 19. Schematic of Two Cohesive Zone Models for a Peel Test of Adhesive Joint
(a) A Single Cohesive Zone and (b) Two Elastic-Plastic Adhesive Layers With an
Embedded Cohesive Zone, Reference 16

4.4.5 Statistical Model for Scaling Effect.

In this approach [17], it is considered that the thickness effect is essentially a scaling effect.
Such an effect is often related to random defects present in the material, for example, voids and
microcracks due to residual stresses (figure 20). The study by Towse, et al. [18] demonstrated
that the failure strain of adhesives is scale sensitive. In reference 17, a probability law of a
reference volume of a material under uniaxial stress is proposed. In particular, a two-parameter
Weibull distribution is adopted and then generalized to a structure subjected to a nonuniform
stress field. Note that the Weibull parameters and the reference volume are determined from the
bulk properties of the adhesive. In the given numerical examples, the experimental trend is
captured.

Although this approach is simple and physically justified, there are limitations. First, the
Weibull distribution and the generalization to nonuniform stress field (the weakest link model)
are applicable to relatively brittle materials. Second, the approach considers a single stress
component, which is not the case for the adhesive layer in a joint.

22
 Figure 20. Defects (Voids) in an Adhesive Material, Reference 17

5. GEOMETRICAL NONLINEARITY.

Large deformation often occurs in thin, nonsymmetric bonded joints. Therefore, a joint analysis
method with geometric nonlinearity is desirable. The geometric nonlinearity option in Bond
Master uses an approximate solution that accounts for large deflections. The solution assumes
that average w deflections for a sublaminate may be large, and therefore contribute to the
y-direction strain as follows

1
w = (ω1 ( y ) + w2 ( y ))
2
2 (12)
∂ν 1 ⎛ ∂w ⎞
εy = + ⎜ ⎟
∂ν 2 ⎝ ∂γ ⎠

The additional contribution is then combined into the system equations. Note that the solution
forces all elements to be treated using the approximate P-element approach. The solution iterates
until an equilibrium state is found. Termination of the iteration is controlled by comparing
relative changes in nodal deflections to the predefined error bound. A single-step solution is
used in the sense that the load is not incremented. The method simply applies the final load state
and attempts to find a stable state. Note that for compressive or shear loads, the method will
cause the apparent stiffness of the model to decrease. Stiffness will go to zero at loads beyond a
buckling value, and a stable solution will not be achieved. Load following is not used in the
solution. This means that the load vectors are fixed in the original, global coordinate system.
Note that the finite deflection strain equations will not be accurate for truly large deflections.
One cannot expect the solution to work for extreme deflections. Figure 21 shows the comparison
of the deflections of a cantilever beam under load rendered from linear analysis, geometrically
nonlinear analysis, and analytic solution. In this example, large deflection is more pronounced
when the ratio of length/thickness is high. It is seen that the SUBLAM nonlinear solution agrees
with the analytic solution in general, although the difference between the two solutions increases
as the L/h increases.

23
 Figure 21. Maximum Deflection of a Cantilever Beam Under Load vs the Length-Thickness
Ratio (Three sets of solutions are rendered from the linear analysis (SUBLAM), the
geometrically nonlinear analysis (SUBLAM), and the analytic solution.)

6. FAILURE OF BONDED COMPOSITE JOINTS.

For bonded joints, the failure can occur either with the adhesive (cohesive failure), at the
adhesive/adherend interface (interface failure), or in the adherends. In general, failure prediction
requires the accurate stress-strain field and appropriate failure criteria associated with specific
failure models. Figure 22 shows failure modes in bonded joints identified by Heslehurst and
Hart-Smith [19]. Bonded joints with composite adherends often fail in complex failure modes
compared to metal adherends. In some cases, composite joints fail progressively after initial
damage occurs in the adherends or in the adhesive. The process of damage growth and coalesces
is relatively complex and often requires specific theories/models that address the process
adequately. Currently, the failure prediction in Bond Master is material strength based, such as
the maximum stress-strain criterion. In conjunction with the results for the stress analysis, the
failure modes in the adherends and the adhesive are determined, and the factor of safety is
computed for the adherends and the adhesive, respectively. In addition, Bond Master provides
an option to determine the exact failure load at which the allowable strain in the adhesive is
reached. The details of the calculation as well as some discussion are presented in the following
sections.

24
 Far Field Interlaminar

Shear Peel

Shear Peel

Figure 22. Typical Failure Modes of the Bonded Joints

6.1 FACTOR OF SAFETY CALCULATION.

Failure of adherends is predicted by using lamina maximum strain-stress failure criterion. In

order to use the lamina maximum strain-stress criterion, the strain state along each ply in the
model has to be known. Recall that elements within Bond Master consist of multiple, stacked
sublaminates, as shown in figure 23. Each sublaminate is composed of one or more plies. The
grouping of the plies into sublaminates is usually determined by the model geometry. Bond
Master outputs the strain-state and displacement fields along the top and bottom of each
sublaminate. Using the results for the top and bottom of a sublaminate, the strain-state can be
computed at any location within the sublaminate. In particular, the strain-state within the
sublaminate can be written in terms of surface strains and surface displacements. For a straight
sublaminate, the strain-state is given by:

25
 ⎧ 0 ⎫
⎪ 1 ⎪
⎪ ε 22 ( x, y ) + ε 22 ( x, y ) + z[ε 22 ( x, y ) − ε 22 ( x, y )] ⎪
2 2 1

⎧ ε11 ⎫ ⎪ 2 h ⎪
⎪ε ⎪ ⎪ ⎪
⎪ 2ε133 ( x, y )hz + (h + 2 z )[ w1 ( x, y ) − w2 ( x, y )] ⎪
⎪ 22 ⎪
⎪⎪ ε 33 ⎪⎪ ⎪ h2 ⎪
⎪ 1 ⎪
⎨ ⎬ ( x, y, z ) = ⎨ γ 23 ( x, y )(h − 2 z )(3h + 2 z ) + γ 23 ( x, y )(h + 2 z ) ⎬
2 2 (13)
⎪ γ 23 ⎪ ⎪ ⎪
⎪ γ 31 ⎪ ⎪ 4h 2 ⎪
⎪ ⎪ ⎪ u1 ( x, y ) − u 2 ( x, y ) ⎪
⎪⎩ γ12 ⎪⎭ ⎪ ⎪
⎪ 1 h ⎪
⎪ γ12 ( x, y ) + γ122
( x, y ) z[γ12 2
( x, y ) − γ112 ( x, y )] ⎪
⎪⎩ + ⎪⎭
2 h

Here, the coordinates are with respect to the ply’s local coordinate system. Note that the
interlaminar stress distribution of each sublaminate can be solved directly from the plate
equilibrium equations.

interface

sublaminate

A single lap joint

Figure 23. Schematic of an Element in a Bonded Joint Model

SUBLAM’s selected displacement field assumes a linear distribution of u and v displacements,

and a quadratic distribution of w displacements. This gives a plate that is shear deformable and
that allows stretching through the thickness. The displacement field for a straight element is

u ( x, y , z ) =
2
[
1 (T )
] [
z
u ( x, y )+ u ( B ) ( x, y ) + u (T ) ( x, y ) − u ( B ) ( x, y )
h
] (14)

v ( x, y , z ) =
2
[
1 (T )
] [
z
v ( x, y )+ v ( B ) ( x, y ) + v (T ) ( x, y ) − v ( B ) ( x, y )
h
] (15)

26
 ⎡⎛ 2 z ⎞ 2 ⎤
w( x, y, z ) =
2
[
1 (T )
] [
z
]
w ( x, y )+ w ( B ) ( x, y ) + w (T ) ( x, y ) − w ( B ) ( x, y ) + Ψw ( x, y ) ⎢⎜ ⎟ − 1⎥ (16)
h ⎣⎢⎝ h ⎠ ⎦⎥

where the superscripts (B) and (T) refer to the z = − h and z = h sides of the sublaminate,
2 2
respectively. The plane z = 0 is defined as passing through the midplane of the sublaminate.
The standard definitions of infinitesimal engineering strains are shown below.

⎧ ∂u ( x, y, z ) ⎫
⎪ ∂x ⎪
⎪ ⎪
⎪ ∂v( x, y, z ) ⎪
⎧ε11 ⎫ ⎪ ∂y ⎪
⎪ε ⎪ ⎪ ⎪
⎪ 22 ⎪ ⎪ ∂w( x, y, z ) ⎪
⎪⎪ε 33 ⎪⎪ ⎪⎪ ∂z ⎪⎪
⎨ ⎬ = ⎨ ∂v( x, y, z ) ∂w( x, y, z ) ⎬ (17)
⎪ε 23 ⎪ ⎪ + ⎪
⎪ε13 ⎪ ⎪ ∂z ∂y ⎪
⎪ ⎪ ⎪ ⎪
⎪⎩ε12 ⎪⎭ ⎪ ∂u ( x, y, z ) + ∂w( x, y, z ) ⎪
⎪ ∂z ∂x ⎪
⎪ ∂u ( x, y, z ) ∂v( x, y, z ) ⎪
⎪ + ⎪
⎪⎩ ∂y ∂x ⎭⎪

By substituting displacement field equations into the strain equations, the strain-state can be
∂Ψw ( x, y )
written in terms of surface displacements, derivatives of surface displacements, , and
∂y
Ψw ( x, y ) . Because SUBLAM uses a plane-strain formulation, derivatives with respect to x are
0. The surface displacements are output by SUBLAM. The remaining unknowns can be solved
for using the strain-state at the top and bottom of the sublaminate. This strain-state can be
equated to the strain equations to solve for the derivatives of the surface displacements,
∂Ψw ( x, y )
, and Ψw ( x, y ) in terms of surface strains. The strain-state at any location within the
∂y
sublaminate can then be written in terms of surface strains and surface displacements resulting in
equation 13. The strain-state within a curved sublaminate can be derived similarly.

Then, a ply-by-ply failure criterion can be applied to the strain and/or stress state to predict
failure of the laminate. For example, given a set of maximum strains,
ε11 , ε11 , ε 22 , ε 22 , ε 33 , ε 33 , γ 23crit , γ 31crit , γ12 crit , the corresponding failure mode of each ply
t crit c crit t crit c crit t crit c crit

can be predicted, and the factor of safety (F.O.S.) is given by

27
 ⎧ ⎧ ε11 t crit
ε11 , ε11 > 0 ⎫
⎪ ⎨ c crit ⎪
⎪ ⎩−ε11 ε11 , ε11 < 0 ⎪
⎪ t crit ⎪
⎪ ⎧⎨ ε 22 ε 22 , ε 22 > 0 ⎪
⎪ ⎩−ε c22 crit ε 22 , ε 22 < 0 ⎪
⎪ ⎪
⎪ ⎧ ε t crit ε , ε > 0 ⎪
F .O.S . = Min ⎨ 33 33 33 ⎬ (18)
⎨ c crit
⎪ ⎩−ε 33 ε 33 , ε 33 < 0 ⎪
⎪ ⎪
⎪ γ 23crit γ 23 ⎪
⎪ ⎪
⎪ γ 31crit γ31 ⎪
⎪ ⎪
⎪⎩ γ12 crit γ12 ⎪⎭

Note that although the strain distribution through the thickness of the sublaminate is known, in
the current Bond Master, the strains at the top and bottom of each ply are computed and the
failure criterion is then applied at both locations. This method is capable of capturing any
peaking strains through the thickness of the plies with the exception of γ23, which may reach the
maximum inside a ply. The effects of missing the peak γ23 may be negligible due to the length
scale of the ply thickness.

The adhesive failure is also predicted by using the maximum strain criterion. In particular,
because the shear strain within the adhesive layer is considerably higher than other strain
components, the F.O.S. is computed based on the shear strain only:

F .O.S .adhesive = γ allowable γ max (19)

where γmax is the maximum average shear strain in the adhesive layer.

For the nonlinear analysis, the F.O.S. is calculated in the same manner as linear F.O.S. The
value of F.O.S. indicates whether the joint will fail at the applied load: failure occurs when
F.O.S.<1. Otherwise, the nonlinear F.O.S. does not contain information about the exact load at
which the joint will fail.

6.2 FAILURE LOAD FOR NONLINEAR ADHESIVE—MAXIMUM STRAIN CRITERION.

It is often of interest to determine the exact load level at which failure occurs. Bond Master
offers such an option for the nonlinear adhesive. The procedure is outlined as follows. For a
bonded joint under certain applied load, the solution is first obtained for the initial load. The
strain within the adhesive layer is then compared with the given allowable strain for the
nonlinear adhesive; then, the load is adjusted according to the difference between the current
strain state and the allowable strain, δγ. The process terminates when (following subscript
should be allowable) δγ γ alloawble is less than 1%. Hence, the load at the time is then the exact
failure load corresponding to the allowable strain, the displacement, stress, and strain fields can
be solved and output accordingly.

28
6.3 STRESS SINGULARITY.

At the interfaces between the adhesive and the adherend, stress singularities present at the edge
of the interface due to material and geometric discontinuity, such as the four corners shown in
figure 18. Moreover, these stress singularities are most pronounced in the elastic analysis.
Figure 24 displays the shear stress distributions at the top and bottom interfaces, along the
bondline of the single lap joint under axial tension. These singularities complicate the
interpretation of the interface results; consequently, the average stresses or the stresses along the
centerline of adhesive layer are often used for the purpose of simplification. However, it is also
argued that the maximum stresses are often not along the centerline, especially for the thicker
adhesive layer; and the interface stresses may be used for the failure prediction, if the stress
singularities can be handled properly.

Figure 24. Stresses Spike at the Corners of the Lower and Upper Interfaces of
the Adhesive Layer

One way of handling the stress singularity along the interface is the stress-at-a-distance
approach, in which a stress value is taken at a short distance from the singularity, which is not
influenced by the singularity stress field, but close enough to be considered as the approximation
of the maximum stresses. The maximum stress-strain criterion can then be applied to predict the
joint strength. In Bond Master, this characteristic distance is referred to as the Edge Offset
Distance, which can be specified by the user.

7. FRONT-END DESCRIPTION FOR Bond Master.

The analysis engine for Bond Master is a general-purpose code called SUBLAM that is similar
to the finite element method. SUBLAM reads a text-based data file to perform a single analysis.
While SUBLAM is reasonably easy to use, it requires experience and knowledge to set up a

29
model. The role of the front-end is to make model creation easy and intuitive. In addition, the
front-end stores and retrieves all the model information across sessions. The key to making the
model creation easy is to parameterize a set of common bonded joint configurations. The user
can simply enter the case-specific data to perform an analysis. The front-end takes care of
setting up an analysis model, using heuristic rules for assuring the accuracy of the model and
results.

Internally, Bond Master breaks joints into a connected series of common topological features.
For example, a joggle is common to a couple of the joint models. Each topological feature has
associated modeling rules. This approach makes it relatively simple to add additional joint types
to the program if there is user interest in expanding the code.

The front-end makes extensive use of a database representation of all the material and model
data. The database uses Microsoft® Access® 2000. The underlying technology for database
access is ADO (Active Data Objects), part of the Microsoft.NET services. For the user, this
approach means that all models may be permanently stored for later modification or review. A
user site may have an established database of material properties that is used for all joint
analysis. The database representation also simplifies the process of setting up a model. A
project is likely to have an established set of structural laminates. The modeling hierarchy
allows the user to access the database of laminates, and quickly connect the laminates to joint
models. The joint model database allows users to return to the model at a later date. In addition,
the database is a convenient mechanism for organizing and storing a series of analyses to
preserve a permanent record for certification or compliance. Because of the selected software
technologies (Microsoft Access and ADO), the front-end is limited to computers using the
Windows operating system.

Five types of adhesive joints are supported in the initial release.

• Basic Single Lap Joint (Single Lap)

• Single Lap Joint with Flush Exterior (Single Joggle)
• Basic Symmetric Lap Splice (Double Lap)
• Symmetric Splice with Flush Exterior (Double Joggle)
• Scarf Joint (Scarf)

Bond Master is preloaded with 73 adhesives. The user can create new adhesives. Bond Master
supports four models of adhesive stress-strain behavior.

• Linear Elastic
• Elastic-Perfectly Plastic
• Ramberg-Osgood
• Piecewise

The four models are illustrated in figure 25.

30
 linear elastic
elastic plastic
ramberg osgood
piecewise
Comparison of
4
Adhesive Models
1 10

8000

6000
shear stress

4000

2000

0
0 0.05 0.1 0.15 0.2 0.25

shear strain

Figure 25. The Four Adhesive Material Models Supported by Bond Master

Bond Master allows the user to define the adherend properties. Adherends can be isotropic,
transversely isotropic, or orthotropic. Bond Master supports simple adherends such as an
aluminum plate or laminated adherends such as an eight-ply E-glass/epoxy construction. The
user specifies the joint geometry and boundary conditions on user friendly forms.

Bond Master produces the following types of plots for all analyses.

• Average YZ shear stress versus distance along bond

• Midplane YZ shear strain versus distance along bond
• Factor of safety versus distance along plate 1
• Factor of safety versus distance along plate 2

The user can request the following plots showing the load conditions throughout the model.

Plot X Axis:

• Distance along any plate or bond

31
Plot Y Axis:

• Shear load
• Twisting moment
• Axial load
• Axial moment
• Vertical shear load
• First moment vertical shear

The user can also request these plots showing stress, strain, and displacement along the edges of
the model’s components.

Plot X Axis:

• Distance along top or bottom of plate 1, 2, or 3 or bond

Plot Y Axis:

• V (displacement along Y)
• W (displacement along Z)
• Epsilon Y
• Tau YZ
• Sigma Z

A series of screen shots from Bond Master are listed below. These shots illustrate the basic
functions of the code. The opening window provides access to menus that control the actions of
the program (figure 26). The functions of the front-end can be categorized as follows:

• The front-end provides help and prompts for the user (figures 27 and 28).

• The definition of lamina materials that will be used in the construction of the adherends
(figures 29 and 30).

• The definition of laminates that form the adherends (figure 31).

• The definition of adhesive properties and models that represent the adhesive (figures
32-34).

• The definition of the joint topologies and geometric details (figures 35-42).

• The definition of loads and boundary conditions (figure 43).

• The selection of user-adjustable solution parameters (figure 44).

• Viewing results (figures 45 and 46).

32
 Figure 26. The Opening Screen Provides Access to the Main Menus

Figure 27. The “?” Symbols on Many Forms Link to Context-Sensitive Help Files

33
 Figure 28. There is a General Help Utility Built Into the Program

Figure 29. Definition of Layer Properties for Incorporation in the Adherend Plate Laminates

34
Figure 30. The Option is Available to Define Strength Properties for the Plate Laminate Layers

Figure 31. General Laminates can be Defined Using the Existing Database of Layer Materials

35
 Figure 32. The Program Comes With a Library of Adhesive Materials
(The user may add additional adhesives or modify the properties.)

Figure 33. Adhesives may be Defined in Terms of Model Parameters or as the Point-Wise Data
for a General Stress-Strain Curve (The database includes model parameters for all the predefined
adhesives, along with the stress-strain curves.)

36
Figure 34. The Plotting Option Allows the User to View Different Adhesive Models for the
Stress-Strain Curves

Figure 35. Typical Starting Form for a Joint Model (Each joint type has its own form. Model
names and comments are also stored in the user database.)

37
 Figure 36. Geometry Data Entry for Single Lap Joint
(Color coding is used as a visual aid for connecting joint elements with the data entry.)

Figure 37. The Ends of the Plates may be Tapered

(A common form is used to define the taper parameters.)

38
 Figure 38. Geometric Inputs for Single Lap Joint With Flush Exterior

Figure 39. The Joggles Require Additional Data to Fully Define the Geometry

39
 Figure 40. Geometric Data Needed for Symmetric Lap Splice

Figure 41. Geometric Data Needed for Symmetric Lap Joint With Flush Exterior

40
 Figure 42. Geometric Data Needed for a Scarf Joint

Figure 43. General Plate Boundary Conditions may be Applied at the Ends of the Adherends
(The user may select either force resultants or displacements. The boundary conditions are
automatically translated into the nodal quantities used by SUBLAM. The application of
rotational constraints requires the use of multipoint constrain equations within SUBLAM.)

41
Figure 44. Additional Solution Parameters are Available for User Control

Figure 45. There is an Extensive Array of Available Solution Plots

42
 Bond Master
version 1.0
Model Title: Single Lap Elastic
Plastic
Date:12/20/2005

SUBLAM Model

Deformations scale factor: 0.13

MODEL Geometry

Geometric Parameter Value

Length of plate 1 10
Length of plate 2 10
Bond width 5
Bond thickness 0.01

Figure 46. Example Report File Generated by Bond Master (Report files may be printed or
exported and stored as Adobe.pdf files. The report elements also may be copied and
pasted into a word processor such as Microsoft Word®.)

43
 Boundary Conditions
Left Side Right Side

U 0.00E+00 Fy 3.00E+03
V 0.00E+00
W 0.00E+00
Θx 0.00E+00
Θy 0.00E+00

Ply Properties
Ply ID Label
1 E-Glass

Ply ID E1 E2 E3 G12 G13 G23 Nu12 Nu13

Nu23
1 3.49E+06 3.49E+06 1.50E+06 5.60E+05 5.45E+05 5.45E+05 0.10 0.31
0.31

Ply ID Alpha 1 Alpha 2 Alpha 3

1 0.00E+00 0.00E+00 0.00E+00

Ply ID Failure Criterion F1T / F1C F2T / F2C F3T / F3C S12 S13 S23
S12Norm
1 Max Strain 1.63E-02 1.63E-02 3.32E-03 1.79E-021.47E-021.47E-02
1.35E-02 1.35E-02 1.00E+00

Laminate Stacking Sequence

[Ply ID / Angle / Thickness]

Plate 1
[1 / 0.0 / 0.024], [1 / 45.0 / 0.024], [1 / 0.0 / 0.024], [1 / 45.0 / 0.024], [1 / 0.0 / 0.024], [1 / 45.0 / 0.024], [1 /
0.0 / 0.024], [1 / 45.0 / 0.024], [1 / 0.0 / 0.024], [1 / 45.0 / 0.024], [1 / 0.0 / 0.024]

Plate 2
[1 / 0.0 / 0.024], [1 / 45.0 / 0.024], [1 / 0.0 / 0.024], [1 / 45.0 / 0.024], [1 / 0.0 / 0.024], [1 / 45.0 / 0.024], [1 /
0.0 / 0.024], [1 / 45.0 / 0.024], [1 / 0.0 / 0.024], [1 / 45.0 / 0.024], [1 / 0.0 / 0.024]

Figure 46. Example Report File Generated by Bond Master (Continued)

44
 Adhesive Properties
Adhesive 1 Elastic Plastic: 3M DP-460 EG-11-RTD

U11P042A Typical response of adhesive. DO NOT USE AS AN ALLOWABLE.

E 3.84E+05
G 1.90E+04
ν 3.50E-01
τ 4.28E+03
α 0.00E+00
Shear Strain Allowable 7.00E-01

Summary
Edge offset 0.00E+00
Peak shear stress in bond 4.39E+03
Peak shear strain in bond 9.52E-01
Factor of safety of bond 7.35E-01
Factor of safety of plate 1 4.24E-01
Failure mode of plate 1 F3T
Factor of safety of plate 2 2.42E-01
Failure mode of plate 2 F3T

Figure 46. Example Report File Generated by Bond Master (Continued)

45
Figure 46. Example Report File Generated by Bond Master (Continued)

46
Figure 46. Example Report File Generated by Bond Master (Continued)

47
 Figure 46. Example Report File Generated by Bond Master (Continued)

8. Bond Master ANALYTICAL VERIFICATION PROBLEMS.

8.1 INTRODUCTION.

This section describes work that was done to validate Bond Master results. Validation was done
by comparing Bond Master results to data generated from another analysis tool, ABAQUS.
Bond Master results were also compared to measured data from the technical literature. Ideally,
Bond Master output would be compared with actual test data in order to verify the analytical
results. Unfortunately, limited bonded joint test data was available for validation purposes.
Relevant test data was available in references 20 and 21. To take advantage of the published test
data, Bond Master verification models were built to match the tested coupons from the literature.

48
The first comparisons were based on the tested coupons described in the literature. The tested
coupons matched three of the five Bond Master joint types: single lap, double lap, and single
joggle. Models were built in Bond Master and ABAQUS to match the coupon materials and
joint dimensions described in reference 20. No test data was found for two of the joints available
in Bond Master, the double joggle and the scarf joints. A double joggle and scarf joint were built
in both Bond Master and ABAQUS. These models used the same adherend and adhesive as the
other three joint models. The joint dimensions were selected to be similar to the first three
models. This strategy provided a simple comparison between Bond Master results and
ABAQUS results. The results for the single lap, double lap, and single joggle could be
confirmed later by test data. The initial Bond Master and ABAQUS runs were made using a
linear adhesive material model.

Additional comparisons were made between the two formulations (transverse shear and von
Mises) that Bond Master uses to model nonlinear adhesives. This study had the same
motivations as the nonlinear adhesive comparison. The two formulations should produce similar
results. Furthermore, if it can be shown that both formulations behave the same, then only one
formulation needs to be used for future comparison to ABAQUS.

8.2 DESCRIPTION OF ABAQUS MODELS.

Since Bond Master uses a two-dimensional (2-D) model, it was appropriate to use a 2-D model
in ABAQUS. A literature search revealed that plane-strain, parabolic elements with the reduced
integration formulation would be a good choice for this analysis. In ABAQUS, this element type
is designated by CPE8R.

Reference 22 described a study of bonded joints in composite structures using ABAQUS CPS8R
elements. The element type CPS8R is a 2-D, plane-stress, 8-noded, biquadratic element with the
reduced integration formulation. Reference 23 compared 2-D plane-stress (ABAQUS CPS8R)
and 2-D plane-strain models (ABAQUS CPE8R) with results from a 3-D analysis. The study
found that results from plane-stress and plane-strain models can be used as the upper and lower
bounds of the 3-D results. Reference 24 stated that plane-stress elements are not considered
suitable for adhesive layers. Reference 24 recommended the use of plane-strain elements for
modeling adhesive layers.

Linear elements were avoided for these models since they occasionally exhibited shear locking.
Shear locking would produce shear stresses that are too high. The use of parabolic elements
eliminated the shear locking problem.

In Bond Master, the joint is modeled in the YZ plane. A different coordinate system was
required for the ABAQUS models. The Bond Master joggle and double joggle models contain
curved sections. Modeling the curved sections in ABAQUS requires the use of cylindrical
coordinate systems. The ABAQUS convention sets axis 1 as the radial dimension, axis 2 as the
tangential direction, and axis 3 as the cylinder axis. Curved elements in Bond Master are
represented with a radial dimension and a tangential dimension. Therefore, the plane-strain
ABAQUS model must lie in the axis 1-axis 2 (or XY) plane. The material properties were

49
adjusted to be consistent for both the Bond Master and ABAQUS models. For all output plots,
the Bond Master coordinate system was used. The ABAQUS results were transformed into the
Bond Master coordinate system.

It was known that the peak stresses in the adhesive would be near the ends of the bond. The
models were meshed to put small elements at these locations. Away from the bond ends, larger
elements could be used without sacrificing accuracy. Geometric biasing was used to put small
elements near the bond ends and larger elements elsewhere.

8.3 Bond Master VERSUS ABAQUS LINEAR ADHESIVE.

The initial set of comparisons had the following characteristics:

• Bond Master results compared to ABAQUS results

• All five Bond Master-supported joints were compared
• Material properties and joint dimensions taken from [20]
• Tensile load
• Linear elastic adhesive material model

All five joint models were constructed to simulate a tensile test. One end of each model was
clamped. The opposite end was loaded so that a tensile load was applied to the joint. Since the
load was applied in the same plane as the rest of the model, many of the calculated quantities
became zero. Plots were generated that compared Bond Master and ABAQUS output for all
nonzero stresses, strains, and displacements.

A large number of plots could be generated for each joint. A limited subset was produced for
this report. Stresses, strains, and displacements were plotted for the adherend closest to the load.
If the displacement results for the adherend close to the load matched well, then it can be inferred
that the displacements for the other adherends in the model also matched well. The displacement
of the clamped adherend must have been close to the correct position in order to put the loaded
adherend into the correct position.

Bond Master and ABAQUS produced closely correlated results for all five joints. The
correlation was especially good for displacements and strains. Some of the other quantities
showed some differences. The following specific differences were noted.

• Some stresses (Tau YZ, Sigma Z) and strains (Epsilon Y, Epsilon Z, and Gamma YZ)
peaked at the ends of the bondline. The two analysis programs predicted different peak
heights.

• Some spikes along the Bond Master results were noted for Tau XY in the scarf model.
The spikes were aligned with the Bond Master element boundaries.

• Gamma YZ on any plate showed poor correlation between the two programs.

50
Double joggle displacements were poorly matched along the curved elements. Other quantities
did not match along the curved elements so the plots were not included here.

Bond Master inputs were chosen to match the data from [21]. The adherends were made of
vinylester DERAKINE MOMENTUM 411-350 matrix resin with E-glass woven fabric, 7781
style with 550 finish. The adhesive was Hysol 9359.3, which was adhesive 20 in the Bond
Master database. Material properties used in the analyses are shown in table 2.

Table 2. Material Property Inputs for Linear Studies

Adherend Structural Properties Classification: Transversely Isotropic

E1 = 1.057 Msi E2=E3 = 3.958 Msi
ν12=ν13 = 0.04 ν23=0.13
G12=G13 = 0.567 Msi G23 = 0.567 Msi
Adherend Failure Criterion Criterion: Maximum Stress
Ply Allowables
F1T: 5000 psi F1C: 20000 psi
F2T: 57000 psi F2C: 47000 psi
S12: 8000 psi S23: 10000 psi
Adhesive Material Properties Adhesive #20 EA 9359.2-11-RTD.
All material parameters taken from Bond Master
database
Linear Elastic:
E = 254773 psi ν = 0.35
Elastic Plastic Model:
G = 37915 psi Tau = 4087 psi
Ramberg-Osgood Model:
G = 57419 psi Tau = 3680 psi
Beta = 7.522
Laminate Properties Angle: 0
Thickness: 0.063
Material: Adherend Properties from above

A Bond Master laminate stiffness can be specified in two ways. The plate material properties
can refer to ply stiffnesses and the user can build up the plies on the laminate form. Alternately,
the plate material properties can refer to the laminate stiffness and the user would enter a single
ply on the laminate form. The Bond Master validation models used the laminate properties
instead of ply properties since the laminate properties were given in reference 19.

For each of the Bond Master models, the degrees of freedom U, V, W, Theta X, and Theta Y
were fixed on the left side. On the right side, Fy was assigned various values and U, W, and
Theta X were fixed.

51
8.4 RESULTS OF VERIFICATION OF LINEAR MODELS USING ABAQUS.

The following section shows illustrations of the geometry for each of the five Bond Master
model types along with the mesh used in the ABAQUS model. Verification of each model is
also presented.

8.4.1 Single Lap Joint, Linear Adhesive, Tensile Load.

The dimensions of the single overlap joint model are given in table 3. Figure 47 shows the Bond
Master model, while figure 48 shows the details of the finite element mesh used to generate the
comparison. The results are given in figure 49.

Table 3. Geometry Inputs for Single Lap Joint

Length of plate 1 3.98 in.

Length of plate 2 3.98 in.
Bond width 1.00 in.
Bond thickness 0.005 in.

Figure 47. Geometry for Single Lap Joint

Figure 48. Mesh of Single Lap Joint Showing Yellow Adherends and Blue Adhesive

52
 BondMaster results BondMaster results
ABAQUS results ABAQUS results

Tau YZ along Bottom of Bond Tau YZ along Top of Bond

4 4
3 10 3 10

4 4
2 10 2 10

1 10
4
1 104

Tau YZ [psi]
Tau YZ [psi]

0 0

4 4
-1 10 -1 10

4 4
-2 10 -2 10

4 4
-3 10 -3 10
-0.2 0 0.2 0.4 0.6 0.8 1 1.2 -0.2 0 0.2 0.4 0.6 0.8 1 1.2

Distance along Bond [inch] Distance along Bond [inch]

BondMaster results BondMaster results

ABAQUS results ABAQUS results

V Displacement along Bottom of Plate 2 W Displacement along Bottom of Plate 2

0.045 0.15

0.04 0.1

0.05
W displacement [inch]
V displacement [inch]

0.035

0
0.03
-0.05

0.025
-0.1

0.02
-0.15

0.015 -0.2
-1 0 1 2 3 4 -1 0 1 2 3 4

Distance along Plate 2 [inch] Distance along Plate 2 [inch]

BondMaster results BondMaster results

ABAQUS results ABAQUS results

Epsilon Y along Bottom of Plate 2 Epsilon Z along Bottom of Plate 2

0.03 0.01

0.025 0.008

0.02 0.006
Epsilon Y [inch/inch]

Epsilon Z [inch/inch]

0.015 0.004

0.01 0.002

0.005 0

0 -0.002

-0.005 -0.004
-1 0 1 2 3 4 -1 0 1 2 3 4

Distance along Plate 2 [inch] Distance along Plate 2 [inch]

Figure 49. Single Lap Joint, Linear Adhesive, Tensile Load

53
 BondMaster results BondMaster results
ABAQUS results ABAQUS results

Tau YZ along Bottom of Plate 2 Sigma Z along Bottom of Plate 2

4 4
1.5 10 5 10

4
1 10 4 10
4

5000 4
3 10

Sigma Z [psi]
Tau YZ [psi]

4
2 10
-5000
4
1 10
4
-1 10

0
-1.5 104

4
-2 10
4 -1 10

4 4
-2.5 10 -2 10
-1 0 1 2 3 4 -1 0 1 2 3 4

Distance along Plate 2 [inch] Distance along Plate 2 [inch]

BondMaster results BondMaster results

ABAQUS results ABAQUS results

Sigma Z along Bottom of Plate 2 Sigma Z along Bottom of Plate 2

4 4
5 10 5 10

4 4
4 10 4 10

4 4
3 10 3 10
Sigma Z [psi]

Sigma Z [psi]

4 4
2 10 2 10

4 4
1 10 1 10

0 0

4 4
-1 10 -1 10

4 4
-2 10 -2 10
0 0.05 0.1 0.15 0.2 0.8 0.85 0.9 0.95 1 1.05

Distance along Plate 2 [inch] Distance along Plate 2 [inch]

BondMaster results BondMaster results

ABAQUS results ABAQUS results

V Displacement along Top of Plate 2 W Displacement along Top of Plate 2

0.042 0.15

0.1
0.04

0.05
W displacement [inch]
V displacement [inch]

0.038

0
0.036
-0.05

0.034
-0.1

0.032
-0.15

0.03 -0.2
-1 0 1 2 3 4 -1 0 1 2 3 4

Distance along Plate 2 [inch] Distance along Plate 2 [inch]

Figure 49. Single Lap Joint, Linear Adhesive, Tensile Load (Continued)

54
 BondMaster results
ABAQUS results

Epsilon Y along Top of Plate 2

0.015

0.01
Epsilon Y [inch/inch]

0.005

-0.005

-0.01
-1 0 1 2 3 4

Distance along Plate 2 [inch]

Figure 49. Single Lap Joint, Linear Adhesive, Tensile Load (Continued)

8.4.2 Single Joggle Joint, Linear Adhesive, Tensile Load.

The dimensions of the joggle joint model are given in table 4. Figure 50 shows the Bond Master
model, while figure 51 shows the details of the finite element mesh used to generate the
comparison. The results are given in figure 52.

Table 4. Geometry Inputs for Single Joggle Joint

Length of plate 1 3.98 in.

Length of plate 2 3.98 in.
Bond width 1.00 in.
Bond thickness 0.005 in.
Radius of joggle in plate 0.0767 in.
Vertical offset of plate 1 0.068 in.

55
 Figure 50. Geometry for Single Joggle Joint

Figure 51. Mesh of Single Joggle Joint Showing Yellow Adherends and Blue Adhesive

56
 BondMaster results BondMaster results
ABAQUS results ABAQUS results

Tau YZ along Bottom of Bond Tau YZ along Top of Bond

4 4
2 10 1 10

4
1.5 10
5000
4
1 10

5000
0
Tau YZ [psi]

Tau YZ [psi]
0

-5000
-5000

-1 104
4
-1 10
4
-1.5 10

4 4
-2 10 -1.5 10
-0.2 0 0.2 0.4 0.6 0.8 1 1.2 -0.2 0 0.2 0.4 0.6 0.8 1 1.2

Distance along Bond [inch] Distance along Bond [inch]

BondMaster results BondMaster results

ABAQUS results ABAQUS results

V Displacement along Bottom of Plate 2 W Displacement along Bottom of Plate 2

0.015 0.06

0.014 0.05

0.013
0.04
W displacement [inch]
V displacement [inch]

0.012
0.03
0.011
0.02
0.01

0.01
0.009

0.008 0

0.007 -0.01
-1 0 1 2 3 4 -1 0 1 2 3 4

Distance along Plate 2 [inch] Distance along Plate 2 [inch]

BondMaster results BondMaster results

ABAQUS results ABAQUS results

Epsilon Y along Bottom of Plate 2 Tau YZ along Bottom of Plate 2

4
0.0035 1 10

0.003
5000
0.0025
Epsilon Y [inch/inch]

0.002
0
Tau YZ [psi]

0.0015

-5000
0.001

0.0005
4
-1 10
0

4
-0.0005 -1.5 10
-1 0 1 2 3 4 -1 0 1 2 3 4

Distance along Plate 2 [inch] Distance along Plate 2 [inch]

Figure 52. Single Joggle Joint, Linear Adhesive, Tensile Load

57
 BondMaster results BondMaster results
ABAQUS results ABAQUS results

Sigma Z along Bottom of Plate 2 V Displacement along Top of Plate 2

4
1.5 10 0.015

0.014
4
1 10

0.013

V displacement [inch]
Sigma Z [psi]

5000
0.012

0.011
0

0.01

-5000
0.009

4
-1 10 0.008
-1 0 1 2 3 4 -1 0 1 2 3 4

Distance along Plate 2 [inch] Distance along Plate 2 [inch]

BondMaster results BondMaster results

ABAQUS results ABAQUS results

W Displacement along Top of Plate 2 Epsilon Y along Top of Plate 2

0.06 0.002

0.05
0.0015

0.04
W displacement [inch]

Epsilon Y [inch/inch]

0.001
0.03

0.02
0.0005

0.01

0
0

-0.01 -0.0005
-1 0 1 2 3 4 -1 0 1 2 3 4

Distance along Plate 2 [inch] Distance along Plate 2 [inch]

Figure 52. Single Joggle Joint, Linear Adhesive, Tensile Load (Continued)

8.4.3 Scarf Joint, Linear Adhesive, Tensile Load.

The dimensions of the scarf joint model are given in table 5. Figure 53 shows the details of the
Bond Master model, while figure 54 shows the details of the finite element mesh used to
generate the comparison. Figure 55 shows a close-up of the scarf termination in the finite
element model. The results are given in figure 56.

58
 Table 5. Geometry Inputs for Scarf Joint

Length of plate 1 2.17024 in.

Length of plate 2 2.17024 in.
Scarf taper ratio 5
Scarf width 0.34 in.
Plate 1 termination thickness 0.01 in.
Plate 2 termination thickness 0.01 in.
Bond thickness 0.005 in.

Figure 53. Geometry for Scarf Joint

Figure 54. Mesh of Scarf Joint Showing Yellow Adherends and Blue Adhesive

Figure 55. Close-Up of Mesh for Scarf Joint Showing Yellow Adherends and Blue Adhesive

59
 BondMaster results BondMaster results
ABAQUS results ABAQUS results

Tau YZ along Top of Bond V Displacement along Bottom of Plate 2

4
4 10 0.024

4
3.5 10 0.022

4
3 10
0.02

V displacement [inch]
4
2.5 10
Tau YZ [psi]

0.018
4
2 10
0.016
4
1.5 10

0.014
1 104

5000 0.012

0 0.01
-0.05 0 0.05 0.1 0.15 0.2 -0.5 0 0.5 1 1.5 2 2.5

Distance along Bond [inch] Distance along Plate 2 [inch]

BondMaster results BondMaster results

ABAQUS results ABAQUS results

W Displacement along Bottom of Plate 2 Epsilon Y along Bottom of Plate 2

0.003 0.025

0.02
0.002
W displacement [inch]

Epsilon Y [inch/inch]

0.015
0.001

0.01

0
0.005

-0.001
0

-0.002 -0.005
-0.5 0 0.5 1 1.5 2 2.5 -0.5 0 0.5 1 1.5 2 2.5

Distance along Plate 2 [inch] Distance along Plate 2 [inch]

BondMaster results
ABAQUS results

Tau YZ along Bottom of Plate 2

4
2 10

4
1.5 10

4
1 10
Tau YZ [psi]

5000

-5000

4
-1 10

4
-1.5 10
-0.5 0 0.5 1 1.5 2 2.5

Distance along Plate 2 [inch]

Figure 56. Scarf Joint, Linear Adhesive, Tensile Load

60
 BondMaster results BondMaster results
ABAQUS results ABAQUS results

Sigma Z along Bottom of Plate 2 Sigma Z along Bottom of Plate 2

4 4
2.5 10 2.5 10

4 4
2 10 2 10

4 4
1.5 10 1.5 10

4 4
1 10 1 10

Sigma Z [psi]
Sigma Z [psi]

5000 5000

0 0

-5000 -5000

4 4
-1 10 -1 10

4 4
-1.5 10 -1.5 10
-0.5 0 0.5 1 1.5 2 2.5 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35

Distance along Plate 2 [inch] Distance along Plate 2 [inch]

BondMaster results BondMaster results

ABAQUS results ABAQUS results

V Displacement along Top of Plate 2 W Displacement along Top of Plate 2

0.024 0.003

0.022 0.002

0.02
W displacement [inch]
V displacement [inch]

0.001

0.018
0
0.016

-0.001
0.014

-0.002
0.012

0.01 -0.003
-0.5 0 0.5 1 1.5 2 2.5 -0.5 0 0.5 1 1.5 2 2.5

Distance along Plate 2 [inch] Distance along Plate 2 [inch]

BondMaster results
ABAQUS results

Epsilon Y along Top of Plate 2

0.006

0.005

0.004
Epsilon Y [inch/inch]

0.003

0.002

0.001

-0.001

-0.002
-0.5 0 0.5 1 1.5 2 2.5

Distance along Plate 2 [inch]

Figure 56. Scarf Joint, Linear Adhesive, Tensile Load (Continued)

61
8.4.4 Double Lap Joint, Linear Adhesive, Tensile Load.

The dimensions of the double lap joint model are given in table 6. Figure 57 shows the Bond
Master model, while figure 58 shows the details of the finite element mesh used to generate the
comparison. Figure 59 is a close-up of the mesh for the joint termination. The results are given
in figure 60.

Table 6. Geometry Inputs for Double Lap Joint

Length of plate 1 3.46 in.

Length of plate 2 3.46 in.
Plate 1 bond width 0.48 in.
Plate 3 bond width 0.48 in.
Bond thickness 0.005 in.
End clearance 0.04 in.

Figure 57. Geometry for Double Lap Joint

Figure 58. Mesh of Double Lap Joint Showing Yellow Adherends and Blue Adhesive

Figure 59. Close-Up of Mesh for Double Lap Joint Showing Yellow Adherends
and Blue Adhesive

62
 BondMaster results BondMaster results
ABAQUS results ABAQUS results

Tau YZ along Bottom of Bond Tau YZ along Top of Bond

4 4
4 10 3 10

4
3 10
4
2 10
4
2 10

4
1 10

Tau YZ [psi]
Tau YZ [psi]

4
1 10

0
0

4
-1 10
4
-1 10
4
-2 10

4 4
-3 10 -2 10
-0.1 0 0.1 0.2 0.3 0.4 0.5 -0.1 0 0.1 0.2 0.3 0.4 0.5

Distance along Bond [inch] Distance along Bond [inch]

BondMaster results BondMaster results

ABAQUS results ABAQUS results

V Displacement along Bottom of Plate 3 W Displacement along Bottom of Plate 3

0.1 0.0002

0.00015
0.09
0.0001
W displacement [inch]
V displacement [inch]

0.08 -5
5 10

0.07 0

-5
-5 10
0.06
-0.0001

0.05
-0.00015

0.04 -0.0002
-0.5 0 0.5 1 1.5 2 2.5 3 3.5 -0.5 0 0.5 1 1.5 2 2.5 3 3.5

Distance along Plate 3 [inch] Distance along Plate 3 [inch]

BondMaster results BondMaster results

ABAQUS results ABAQUS results

Epsilon Y along Bottom of Plate 3 Tau YZ along Bottom of Plate 3

4
0.03 1.2 10

4
1 10
0.025

8000
Epsilon Y [inch/inch]

0.02
Tau YZ [psi]

6000
0.015
4000

0.01
2000

0.005
0

0 -2000
-0.5 0 0.5 1 1.5 2 2.5 3 3.5 -0.5 0 0.5 1 1.5 2 2.5 3 3.5

Distance along Plate 3 [inch] Distance along Plate 3 [inch]

Figure 60. Double Lap Joint, Linear Adhesive, Tensile Load

63
 BondMaster results BondMaster results
ABAQUS results ABAQUS results

Sigma Z along Bottom of Plate 3 V Displacement along Top of Plate 3

4
8 10 0.1

4
6 10 0.09

V displacement [inch]
4
4 10 0.08
Sigma Z [psi]

4
2 10 0.07

0 0.06

4
-2 10 0.05

4
-4 10 0.04
-0.5 0 0.5 1 1.5 2 2.5 3 3.5 -0.5 0 0.5 1 1.5 2 2.5 3 3.5

Distance along Plate 3 [inch] Distance along Plate 3 [inch]

BondMaster results BondMaster results

ABAQUS results ABAQUS results

W Displacement along Top of Plate 3 Epsilon Y along Top of Plate 3

0.0002 0.03

0.00015
0.025

0.0001
W displacement [inch]

Epsilon Y [inch/inch]

0.02
-5
5 10

0 0.015

-5
-5 10
0.01

-0.0001

0.005
-0.00015

-0.0002 0
-0.5 0 0.5 1 1.5 2 2.5 3 3.5 -0.5 0 0.5 1 1.5 2 2.5 3 3.5

Distance along Plate 3 [inch] Distance along Plate 3 [inch]

Figure 60. Double Lap Joint, Linear Adhesive, Tensile Load (Continued)

8.4.5 Double Joggle Joint, Linear Adhesive, Tensile Load.

The dimensions of the double joggle joint model are given in table 7. Figure 61 shows the Bond
Master model, while figure 62 shows the details of the finite element mesh used to generate the
comparison. Figure 63 shows a close-up of the finite element mesh near the joint termination.
The results are given in figure 64.

64
 Table 7. Geometry Inputs for Double Joggle Joint

Length of plate 1 3.46 in.

Length of plate 3 3.46 in.
Plate 1 bond width 0.48 in.
Plate 3 bond width 0.48 in.
Bond thickness 0.005 in.
End clearance 0.04 in.
Radius of joggle in plate 1 0.0767 in.
Vertical offset of plate 1 0.068 in.
Radius of joggle in plate 3 0.0767 in.
Vertical offset of plate 3 0.068 in.

Figure 61. Geometry for Double Joggle Joint

Figure 62. Mesh of Double Joggle Joint Showing Yellow Adherends and Blue Adhesive

Figure 63. Close-Up of Mesh for Double Joggle Joint Showing Yellow Adherends and Blue
Adhesive

65
 BondMaster results BondMaster results
ABAQUS results ABAQUS results

Tau YZ along Bottom of Bond Tau YZ along Top of Bond

4 4
2 10 1.5 10

4
1.5 10
4
1 10
4
1 10

5000
5000

Tau YZ [psi]
Tau YZ [psi]

0
-5000

4
-1 10
-5000
4
-1.5 10

4 4
-2 10 -1 10
-0.1 0 0.1 0.2 0.3 0.4 0.5 -0.1 0 0.1 0.2 0.3 0.4 0.5

Distance along Bond [inch] Distance along Bond [inch]

BondMaster results BondMaster results

ABAQUS results ABAQUS results

V Displacement along Bottom of Plate 3 W Displacement along Bottom of Plate 3

0.08 0.1

0.06

0.04 0.05
W displacement [inch]
V displacement [inch]

0.02

0 0

-0.02

-0.04 -0.05

-0.06

-0.08 -0.1
-0.5 0 0.5 1 1.5 2 2.5 3 3.5 -0.5 0 0.5 1 1.5 2 2.5 3 3.5

Distance along Plate 3 [inch] Distance along Plate 3 [inch]

Figure 64. Double Joggle Joint, Linear Adhesive, Tensile Load

8.5 ALTERNATE BOND MASTER NONLINEAR ADHESIVE FORMULATIONS.

8.5.1 Introduction.

There are two options in Bond Master to handle material nonlinearity: the transverse shear
model and the von Mises model. In the former, the transverse shear secant modulus is used to
carry out the nonlinear analysis and is justified since the adhesive layer is primarily under shear.
In the latter, the equivalent secant modulus is used; the stress state in the adhesive layer can be
multiaxial. Clearly, the two should render similar results when the shear deformation is
dominant in the adhesive layer. For the purpose of validation, an analysis is conducted for a
single lap joint under two loading conditions: axial tension and bending.

The stress results from the two options are compared, and in general, they agree well. Moreover,
the strain components were also plotted to confirm the dominance of the shear deformation. It is

66
verified that the two options are equivalent in the case that the adhesive layer is primarily under
shear.

8.5.2 Problem Description.

A single lap joint is selected for the comparative study. The geometric configuration and
material properties are adopted from reference 3. Figure 65 depicts the schematic of the single
lap joint configuration. The left end of the single lap joint is clamped. Two loading conditions
are applied: axial tensile force P = 1000 lb and vertical point load W = 20 lb. The geometric
parameters are given in table 8. The adherends and adhesive are made of E-glass/vinylester 8
harness satin woven composite and Hysol EA 9359.3, respectively. Note that the adherends are
assumed to be linearly elastic. Their mechanical properties are listed in table 9. The shear
stress-strain curve and the converted von Mises equivalent stress-strain curve (see figure 66) are
given for the two options, respectively.

Figure 65. Schematic of the Single Lap Joint Configuration

Table 8. Dimensions of the Single Lap Joint

Joint Width 2c Adhesive Thickness t1 Adherend Thickness t1 A

(mm/in.) (mm/in.) (mm/in.) (mm/in.)
25.4/1 0.127/0.005 1.6/0.063 75.6/2.98

Table 9. Mechanical Properties of the Adherends and the Adhesive

Ex = Ey (GPa/ Msi) Gxy (GPa/ Msi) υxy

Adherend 27.292/3.958 3.907/0.5666 0.13
Adhesive 2.137/0.3099 0.815/0.118 0.31

67
Figure 66. Shear Stress-Strain Curve and von Mises Equivalent Stress-Strain Curve for EA 9359
Adhesive (Shear data is from reference 1.)

8.5.2.1 Results and Discussion.

8.5.2.1.1 Axial Tensile Load.

The single lap joint under applied axial tensile load is analyzed with both options. Figures 67
and 68 show the shear and peel stresses at the lower interface of the adhesive layer, respectively.
It is seen that the results agree well in general. Both results give the same plastic zone, although
the magnitudes of shear stresses differ in the plastic zone slightly. Figure 69 plots three strains,
γyz, εy, εz, at the lower interface of the adhesive along the bondline. In this case, it is clear that the
shear deformation is dominant; hence, it is expected that the similar results are rendered from the
two options.

68
Figure 67. Shear Stress Distribution at the Lower Interface of the Adhesive Layer, for the Single
Lap Joint Under Axial Tensile Load

Figure 68. Peel Stress Distribution at the Lower Interface of the Adhesive Layer, for the Single
Lap Joint Under Axial Tensile Load

69
 Figure 69. Strain Distribution (γyz, εy, εz) at the Lower Interface of the Adhesive Layer, for the
Single Lap Joint Under Axial Tensile Load

8.5.2.1.2 Bending Load.

The same analyses are conducted for the bending load. Note that much less load can be carried
for the selected joint. Figures 70 and 71 show the shear and peel stresses at the lower interface
of the adhesive layer, respectively. It is shown that the results agree well in general. Both
results give the same plastic zone, although the magnitudes of shear stresses differ in the plastic
zone slightly. Figure 72 plots three strains, γyz, εy, εz, at the lower interface of the adhesive along
the bondline. Again, the adhesive layer is primarily under shear deformation.

Figure 70. Shear Stress Distribution at the Lower Interface of the Adhesive Layer, for the Single
Lap Joint Under Bending Load

70
Figure 71. Peel Stress Distribution at the Lower Interface of the Adhesive Layer, for the Single
Lap Joint Under Bending Load

Figure 72. Strain Distribution (γyz, εy, εz) at the Lower Interface of the Adhesive Layer, for the
Single Lap Joint Under Bending Load

71
8.5.3 Bond Master Versus ABAQUS, Nonlinear Adhesive.

The next set of comparisons introduced the nonlinear adhesive material into the analytical
models.

• Bond Master results compared with ABAQUS results

• All five Bond Master-supported joints were compared
• Material properties and joint dimensions taken from reference 20
• Tensile load
• Nonlinear elastic-perfectly plastic adhesive material model

The joint models were clamped on one end and subjected to a tensile load on the other end.
Since the load was applied in the same plane as the rest of the model, many of the calculated
quantities became zero. The zero quantities (U, Epsilon X, etc.) were ignored. Plots were
generated that compared Bond Master and ABAQUS output for the nonzero stresses, strains, and
displacements. The results are given in figures 73 through 77.

There was generally good correlation between Bond Master and ABAQUS using the nonlinear
adhesive models. Correlation between the programs was not good for the following quantities:

• Correlation was not good for all scarf joint plots.

• The ends of the bonds showed spikes for many different quantities.

• In areas where the adhesive deformed plastically, ABAQUS reported greater strain than
Bond Master.

72
 BondMaster results BondMaster results
ABAQUS results ABAQUS results

V Displacement along Bottom of Bond V Displacement along Top of Bond

0.021 0.023

0.0205 0.0225

0.02 0.022

V displacement [inch]
V displacement [inch]

0.0195 0.0215

0.019 0.021

0.0185 0.0205

0.018 0.02

0.0175 0.0195
-0.2 0 0.2 0.4 0.6 0.8 1 1.2 -0.2 0 0.2 0.4 0.6 0.8 1 1.2

Distance along Bottom of Bond [inch] Distance along Top of Bond [inch]

BondMaster results BondMaster results

ABAQUS results ABAQUS results

W Displacement along Bottom of Bond Epsilon Y along Bottom of Bond

0.15 0.03

0.02
0.1

0.01
W displacement [inch]

Epsilon Y [inch/inch]

0.05

0
0
-0.01

-0.05
-0.02

-0.1
-0.03

-0.15 -0.04
-0.2 0 0.2 0.4 0.6 0.8 1 1.2 -0.2 0 0.2 0.4 0.6 0.8 1 1.2

Distance along Bottom of Bond [inch] Distance along Bottom of Bond [inch]

BondMaster results BondMaster results

ABAQUS results ABAQUS results

Epsilon Y along Top of Bond Tau YZ along Bottom of Bond

0.03 7000

6000
0.02

5000
0.01
Epsilon Y [inch/inch]

4000
Tau YZ [psi]

0
3000
-0.01
2000

-0.02
1000

-0.03 0

-0.04 -1000
-0.2 0 0.2 0.4 0.6 0.8 1 1.2 -0.2 0 0.2 0.4 0.6 0.8 1 1.2

Distance along Top of Bond [inch] Distance along Bottom of Bond [inch]

Figure 73. Single Lap Joint, Nonlinear Adhesive, Tensile Load

73
 BondMaster results BondMaster results
ABAQUS results ABAQUS results

Tau YZ along Top of Bond Sigma Z along Bottom of Bond

4
6000 2 10

5000
4
1.5 10

4000

Sigma Z [psi]
1 10
Tau YZ [psi]

3000

2000
5000

1000

0
0

-1000 -5000
-0.2 0 0.2 0.4 0.6 0.8 1 1.2 -0.2 0 0.2 0.4 0.6 0.8 1 1.2

Distance along Top of Bond [inch] Distance along Bottom of Bond [inch]

BondMaster results
ABAQUS results

Sigma Z along Top of Bond

4
2 10

4
1.5 10

4
Sigma Z [psi]

1 10

5000

-5000
-0.2 0 0.2 0.4 0.6 0.8 1 1.2

Distance along Top of Bond [inch]

Figure 73. Single Lap Joint, Nonlinear Adhesive, Tensile Load (Continued)

74
 BondMaster results BondMaster results
ABAQUS results ABAQUS results

V Displacement along Bottom of Bond W Displacement along Bottom of Bond

0.0085 0.065

0.06

0.008

W displacement [inch]
V displacement [inch]

0.055

0.0075 0.05

0.045

0.007

0.04

0.0065 0.035
-0.2 0 0.2 0.4 0.6 0.8 1 1.2 -0.2 0 0.2 0.4 0.6 0.8 1 1.2

Distance along Bottom of Bond [inch] Distance along Bottom of Bond [inch]

BondMaster results BondMaster results

ABAQUS results ABAQUS results

Epsilon Y along Bottom of Bond Tau YZ along Bottom of Bond

0.02 4000

3500
0.015
3000
Epsilon Y [inch/inch]

2500
Tau YZ [psi]

0.01

2000

0.005
1500

1000
0
500

-0.005 0
-0.2 0 0.2 0.4 0.6 0.8 1 1.2 -0.2 0 0.2 0.4 0.6 0.8 1 1.2

Distance along Bottom of Bond [inch] Distance along Bottom of Bond [inch]

BondMaster results BondMaster results

ABAQUS results ABAQUS results

Sigma Z along Bottom of Bond V Displacement along Top of Bond

4
2 10 0.0088

0.0086
1.5 104
V displacement [inch]

0.0084
4
Sigma Z [psi]

1 10

0.0082

5000
0.008

0
0.0078

-5000 0.0076
-0.2 0 0.2 0.4 0.6 0.8 1 1.2 -0.2 0 0.2 0.4 0.6 0.8 1 1.2

Distance along Bottom of Bond [inch] Distance along Top of Bond [inch]

Figure 74. Single Joggle Joint, Nonlinear Adhesive, Tensile Load

75
 BondMaster results BondMaster results
ABAQUS results ABAQUS results

W Displacement along Top of Bond Epsilon Y along Top of Bond

0.065 0.003

0.0025
0.06
0.002
W displacement [inch]

Epsilon Y [inch/inch]
0.055
0.0015

0.05 0.001

0.0005
0.045
0

0.04
-0.0005

0.035 -0.001
-0.2 0 0.2 0.4 0.6 0.8 1 1.2 -0.2 0 0.2 0.4 0.6 0.8 1 1.2

Distance along Top of Bond [inch] Distance along Top of Bond [inch]

BondMaster results BondMaster results

ABAQUS results ABAQUS results

Tau YZ along Top of Bond Sigma Z along Top of Bond

4
6000 2 10

5000
4
1.5 10

4000

4
Sigma Z [psi]

1 10
Tau YZ [psi]

3000

2000
5000

1000

0
0

-1000 -5000
-0.2 0 0.2 0.4 0.6 0.8 1 1.2 -0.2 0 0.2 0.4 0.6 0.8 1 1.2

Distance along Top of Bond [inch] Distance along Top of Bond [inch]

Figure 74. Single Joggle Joint, Nonlinear Adhesive, Tensile Load (Continued)

76
 BondMaster results BondMaster results
ABAQUS results ABAQUS results

V Displacement along Bottom of Bond W Displacement along Bottom of Bond

0.0066 0.0015

0.001
0.0064

W displacement [inch]
V displacement [inch]

0.0005
0.0062

0.006
-0.0005

0.0058
-0.001

0.0056 -0.0015
-0.05 0 0.05 0.1 0.15 0.2 -0.05 0 0.05 0.1 0.15 0.2

Distance along Bottom of Bond [inch] Distance along Bottom of Bond [inch]

BondMaster results BondMaster results

ABAQUS results ABAQUS results

Epsilon Y along Bottom of Bond Tau YZ along Bottom of Bond

0.1 6000

0.08 5000

4000
Epsilon Y [inch/inch]

0.06
Tau YZ [psi]

0.04 3000

0.02 2000

0 1000

-0.02 0
-0.05 0 0.05 0.1 0.15 0.2 -0.05 0 0.05 0.1 0.15 0.2

Distance along Bottom of Bond [inch] Distance along Bottom of Bond [inch]

BondMaster results BondMaster results

ABAQUS results ABAQUS results

Sigma Z along Bottom of Bond V Displacement along Top of Bond

4
2 10 0.0075

4
1.5 10
V displacement [inch]

4
1 10
Sigma Z [psi]

5000 0.007

-5000

4
-1 10 0.0065
-0.05 0 0.05 0.1 0.15 0.2 -0.05 0 0.05 0.1 0.15 0.2

Distance along Bottom of Bond [inch] Distance along Top of Bond [inch]

Figure 75. Scarf Joint, Nonlinear Adhesive, Tensile Load

77
 BondMaster results BondMaster results
ABAQUS results ABAQUS results

W Displacement along Top of Bond Epsilon Y along Top of Bond

0.0015 0.1

0.001 0.08
W displacement [inch]

Epsilon Y [inch/inch]
0.0005 0.06

0 0.04

-0.0005 0.02

-0.001 0

-0.0015 -0.02
-0.05 0 0.05 0.1 0.15 0.2 -0.05 0 0.05 0.1 0.15 0.2

Distance along Top of Bond [inch] Distance along Top of Bond [inch]

BondMaster results BondMaster results

ABAQUS results ABAQUS results

Tau YZ along Top of Bond Sigma Z along Top of Bond

4
6000 2 10

4
5000 1.5 10

4000 1 104
Sigma Z [psi]
Tau YZ [psi]

3000 5000

2000 0

1000 -5000

4
0 -1 10
-0.05 0 0.05 0.1 0.15 0.2 -0.05 0 0.05 0.1 0.15 0.2

Distance along Top of Bond [inch] Distance along Top of Bond [inch]

Figure 75. Scarf Joint, Nonlinear Adhesive, Tensile Load (Continued)

78
 BondMaster results BondMaster results
ABAQUS results ABAQUS results

V Displacement along Bottom of Bond W Displacement along Bottom of Bond

0.033 0.0001

0.032
-5
5 10

0.031

W displacement [inch]
V displacement [inch]

0
0.03

0.029 -5
-5 10

0.028

-0.0001
0.027

0.026 -0.00015
-0.1 0 0.1 0.2 0.3 0.4 0.5 -0.1 0 0.1 0.2 0.3 0.4 0.5

Distance along Bottom of Bond [inch] Distance along Bottom of Bond [inch]

BondMaster results BondMaster results

ABAQUS results ABAQUS results

Epsilon Y along Bottom of Bond Tau YZ along Bottom of Bond

0.03 -1000

-1500
0.025

-2000
Epsilon Y [inch/inch]

0.02
Tau YZ [psi]

-2500
0.015
-3000

0.01
-3500

0.005
-4000

0 -4500
-0.1 0 0.1 0.2 0.3 0.4 0.5 -0.1 0 0.1 0.2 0.3 0.4 0.5

Distance along Bottom of Bond [inch] Distance along Bottom of Bond [inch]

BondMaster results BondMaster results

ABAQUS results ABAQUS results

Sigma Z along Bottom of Bond V Displacement along Top of Bond

4
2 10 0.032

0.031
1.5 104

0.03
V displacement [inch]

4
Sigma Z [psi]

1 10
0.029

0.028
5000

0.027

0
0.026

-5000 0.025
-0.1 0 0.1 0.2 0.3 0.4 0.5 -0.1 0 0.1 0.2 0.3 0.4 0.5

Distance along Bottom of Bond [inch] Distance along Top of Bond [inch]

Figure 76. Double Lap Joint, Nonlinear Adhesive, Tensile Load

79
 BondMaster results BondMaster results
ABAQUS results ABAQUS results

W Displacement along Top of Bond Epsilon Y along Top of Bond

0.0003 0.006

0.005
0.0002

0.004
W displacement [inch]

Epsilon Y [inch/inch]
0.0001
0.003

0.002
0

0.001

-0.0001
0

-0.0002 -0.001
-0.1 0 0.1 0.2 0.3 0.4 0.5 -0.1 0 0.1 0.2 0.3 0.4 0.5

Distance along Top of Bond [inch] Distance along Top of Bond [inch]

BondMaster results BondMaster results

ABAQUS results ABAQUS results

Tau YZ along Top of Bond Sigma Z along Top of Bond

-500 6000

-1000
4000
-1500

-2000
Sigma Z [psi]

2000
Tau YZ [psi]

-2500

0
-3000

-3500
-2000
-4000

-4500 -4000
-0.1 0 0.1 0.2 0.3 0.4 0.5 -0.1 0 0.1 0.2 0.3 0.4 0.5

Distance along Top of Bond [inch] Distance along Top of Bond [inch]

Figure 76. Double Lap Joint, Nonlinear Adhesive, Tensile Load (Continued)

80
 BondMaster results BondMaster results
ABAQUS results ABAQUS results

V Displacement along Bottom of Bond W Displacement along Bottom of Bond

0.0098 0.095

0.0096

0.09
0.0094

W displacement [inch]
V displacement [inch]

0.0092
0.085
0.009

0.0088
0.08

0.0086

0.0084 0.075
-0.1 0 0.1 0.2 0.3 0.4 0.5 -0.1 0 0.1 0.2 0.3 0.4 0.5

Distance along Bottom of Bond [inch] Distance along Bottom of Bond [inch]

BondMaster results BondMaster results

ABAQUS results ABAQUS results

Epsilon Y along Bottom of Bond Tau YZ along Bottom of Bond

0.025 0

0.02
-1000
Epsilon Y [inch/inch]

0.015
-2000
Tau YZ [psi]

0.01

-3000
0.005

-4000
0

-0.005 -5000
-0.1 0 0.1 0.2 0.3 0.4 0.5 -0.1 0 0.1 0.2 0.3 0.4 0.5

Distance along Bottom of Bond [inch] Distance along Bottom of Bond [inch]

BondMaster results BondMaster results

ABAQUS results ABAQUS results

Sigma Z along Bottom of Bond V Displacement along Top of Bond

4
2 10 0.0098

0.0096
1.5 104

0.0094
V displacement [inch]

4
Sigma Z [psi]

1 10
0.0092

0.009
5000

0.0088

0
0.0086

-5000 0.0084
-0.1 0 0.1 0.2 0.3 0.4 0.5 -0.1 0 0.1 0.2 0.3 0.4 0.5

Distance along Bottom of Bond [inch] Distance along Top of Bond [inch]

Figure 77. Double Joggle Joint, Nonlinear Adhesive, Tensile Load

81
 BondMaster results BondMaster results
ABAQUS results ABAQUS results

W Displacement along Top of Bond Epsilon Y along Top of Bond

0.095 0.025

0.02
0.09
W displacement [inch]

Epsilon Y [inch/inch]
0.015

0.085

0.01

0.08
0.005

0.075 0
-0.1 0 0.1 0.2 0.3 0.4 0.5 -0.1 0 0.1 0.2 0.3 0.4 0.5

Distance along Top of Bond [inch] Distance along Top of Bond [inch]

BondMaster results BondMaster results

ABAQUS results ABAQUS results

Tau YZ along Top of Bond Sigma Z along Top of Bond

4
0 2 10

4
-1000 1.5 10

4
Sigma Z [psi]

-2000 1 10
Tau YZ [psi]

-3000 5000

-4000 0

-5000 -5000
-0.1 0 0.1 0.2 0.3 0.4 0.5 -0.1 0 0.1 0.2 0.3 0.4 0.5

Distance along Top of Bond [inch] Distance along Top of Bond [inch]

Figure 77. Double Joggle Joint, Nonlinear Adhesive, Tensile Load (Continued)

8.6 Bond Master VERSUS ABAQUS, ELASTIC-PLASTIC ADHESIVE, INCREASING

LOAD.

A small study was done to ensure that the correlation between Bond Master and ABAQUS was
maintained as more of the adhesive deformed plastically.

• Bond Master results compared with ABAQUS results

• Single lap joint was compared
• Material properties and joint dimensions taken from reference 19
• Three different tensile loads
• Nonlinear elastic-perfectly plastic adhesive material model

82
Three loads (1010, 1400, and 1800 lb) were applied to the single lap joint to observe the
behavior as the adhesive plastically deformed. As the load increased, the ends of the bond
started to yield.

Figure 78 shows the stress-strain relationship for the elastic-perfectly plastic model for adhesive
20, EA 9359.3-11-RTD. Bond Master calculated the peak strains in the bond for joint loads of
1010, 1400, and 1800 lb. These peak strains were also plotted, as shown on figure 78.

Figure 79 displays the shear stress along the bottom of the adhesive. The Bond Master database
gave a yield shear stress value of 4087 psi for the adhesive in the model. As the load increased,
the ends of the bond started to yield. In the area where the adhesive yielded, the shear stress
leveled off at the yield stress value.

Shear Stress Adhesive 20

Load=1010 pounds
Load=1400 pounds
Stress-Strain
Load=1800 pounds Adhesive 20
Elastic Perfectly Plastic
5000

4000
Shear Stress

3000

2000

1000

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7

Shear Strain

Figure 78. Stress-Strain Curve for Adhesive 20 From the Bond Master Database

83
 Load=1010 lb
Load=1400 lb
Load=1800 lb
Tau YZ along Bottom of Bond
Results from BondMaster
8000

6000

4000
Tau YZ [psi]

2000

-2000
-0.2 0 0.2 0.4 0.6 0.8 1 1.2

Distance along Bottom of Bond [inch]

Figure 79. The Length of Plastic Deformation Along the Bond Increases as the
Applied Load Increases

The comparison between Bond Master and ABAQUS remains good as the adhesive begins to
yield. Figures 80 through 88 show the results comparison as the load increases.

BondMaster results
ABAQUS results

Tau YZ along Bottom of Bond

7000

6000

5000

4000
Tau YZ [psi]

3000

2000

1000

-1000
-0.2 0 0.2 0.4 0.6 0.8 1 1.2

Distance along Bottom of Bond [inch]

Figure 80. The Single Lap Joint is Loaded With 1010 lb (This plot compares the shear stress
along the bottom of the bond as derived by Bond Master and ABAQUS.)

84
Figure 81. The Single Lap Joint is Loaded With 1010 lb (This contour plot from ABAQUS
shows the shear stress in the bond and adjacent adherends.)

Figure 82. The Single Lap Joint is Loaded With 1010 lb (This contour plot from
ABAQUS shows the plastic strain in the adhesive. Only the very end of the bond
shows plastic deformation.)

85
 BondMaster results
ABAQUS results

Tau YZ along Bottom of Bond

8000

6000

4000
Tau YZ [psi]

2000

-2000
-0.2 0 0.2 0.4 0.6 0.8 1 1.2

Distance along Bottom of Bond [inch]

Figure 83. Load on the Single Lap Joint was Increased to 1400 lb (The plastic zone increased in
length as evidenced by the higher length where the shear stress equaled the yield stress.)

Figure 84. Load on the Single Lap Joint was Increased to 1400 lb
(The length along the bond where the shear stress equaled the yield stress has increased.)

86
 Figure 85. Load on the Single Lap Joint was Increased to 1400 lb
(The plastic zone has increased in length.)

BondMaster results
ABAQUS results

Tau YZ along Bottom of Bond

8000

6000

4000
Tau YZ [psi]

2000

-2000
-0.2 0 0.2 0.4 0.6 0.8 1 1.2

Distance along Bottom of Bond [inch]

Figure 86. The Load on the Single Lap Joint was Increased to 1800 lb
(The length of the plastically deformed adhesive has increased as well.)

87
 Figure 87. The Load on the Single Lap Joint was Increased to 1800 lb
(The length increased where the shear stress equaled the yield stress.)

Figure 88. The Load on the Single Lap Joint was Increased to 1800 lb
(The length of the plastically deformed adhesive has increased as well.)

8.7 Bond Master VERSUS ABAQUS, LINEAR ADHESIVE, OUT-OF-PLANE LOAD.

Another study substituted an out-of-plane load for the tensile load, as shown in figure 89.

• Bond Master results compared with ABAQUS results

88
• Single lap joint was compared
• Material properties and joint dimensions taken from reference 20
• Out-of-plane load
• Linear elastic adhesive material model

Figure 89. ABAQUS Model of the Single Lap Joint (The left end of the joint was constrained.
An out-of-plane load was applied to the right end of the joint.)

The magnitude of the out-of-plane load was chosen to yield a minimum factor of safety equal to
one in the Bond Master model. This load was much lower than the failure load for the tensile
load case. This was not surprising since the model was a long (6.96 inches), thin (0.063 inch)
beam. Under tensile loading, Bond Master predicted tensile failure in the loaded adherend.
Under out-of-plane loading, Bond Master predicted compressive failure in the adherend near the
clamped end.

A static linear analysis was run on the single lap joint in both Bond Master and ABAQUS. The
results were postprocessed to plot the following quantities along the top and bottom of the
adhesive bond.

• V (displacement along Y)
• W (displacement along Z)
• Epsilon Y
• Tau YZ
• Sigma Z

The single lap joint model was clamped on one end and subjected to an out-of-plane load on the
other end. Since the load was applied in the same plane as the rest of the model, many of the
calculated quantities became zero. The zero quantities (U, Epsilon X, etc.) were ignored. Plots
were generated that compared the Bond Master and ABAQUS output for the nonzero stresses,
strains, and displacements.

There was excellent agreement between Bond Master and ABAQUS for the single lap joint
under out-of-plane loading. The results plots are given in figure 90. There were some small
discrepancies at the very ends of the adhesive bonds.

89
 BondMaster results BondMaster results
ABAQUS results ABAQUS results

V Displacement along Bottom of Bond V Displacement along Top of Bond

-0.0273 -0.0318

-0.0274 -0.0319

-0.0275

V displacement [inch]
V displacement [inch]

-0.032

-0.0276
-0.0321
-0.0277

-0.0322
-0.0278

-0.0323
-0.0279

-0.028 -0.0324
-0.2 0 0.2 0.4 0.6 0.8 1 1.2 -0.2 0 0.2 0.4 0.6 0.8 1 1.2

Distance along Bottom of Bond [inch] Distance along Top of Bond [inch]

BondMaster results BondMaster results

ABAQUS results ABAQUS results

W Displacement along Bottom of Bond W Displacement along Top of Bond

2.4 2.4

2.2 2.2
W displacement [inch]
W displacement [inch]

2 2

1.8 1.8

1.6 1.6

1.4 1.4
-0.2 0 0.2 0.4 0.6 0.8 1 1.2 -0.2 0 0.2 0.4 0.6 0.8 1 1.2

Distance along Bottom of Bond [inch] Distance along Top of Bond [inch]

BondMaster results BondMaster results

ABAQUS results ABAQUS results

Epsilon Y along Bottom of Bond Epsilon Y along Top of Bond

0.002 0.01

0
0.008

-0.002
Epsilon Y [inch/inch]

0.006
Epsilon Y [inch/inch]

-0.004

-0.006 0.004

-0.008
0.002

-0.01

0
-0.012

-0.014 -0.002
-0.2 0 0.2 0.4 0.6 0.8 1 1.2 -0.2 0 0.2 0.4 0.6 0.8 1 1.2

Distance along Bottom of Bond [inch] Distance along Top of Bond [inch]

Figure 90. Single Lap Joint, Linear Adhesive, Out-of-Plane Load

90
 BondMaster results BondMaster results
ABAQUS results ABAQUS results

Tau YZ along Bottom of Bond Tau YZ along Top of Bond

4 4
1.5 10 1 10

4
1 10
5000

5000

Tau YZ [psi]
Tau YZ [psi]

0 0

-5000
-5000
4
-1 10

4 4
-1.5 10 -1 10
-0.2 0 0.2 0.4 0.6 0.8 1 1.2 -0.2 0 0.2 0.4 0.6 0.8 1 1.2

Distance along Bottom of Bond [inch] Distance along Top of Bond [inch]

BondMaster results
ABAQUS results

Sigma Z along Bottom of Bond

4
1 10

5000

-5000
Sigma Z [psi]

4
-1 10

4
-1.5 10

-2 104

4
-2.5 10

4
-3 10
-0.2 0 0.2 0.4 0.6 0.8 1 1.2

Distance along Bottom of Bond [inch]

Figure 90. Single Lap Joint, Linear Adhesive, Out-of-Plane Load (Continued)

8.8 Bond Master VERSUS ABAQUS, NONLINEAR ADHESIVE, ALUMINUM

ADHEREND.

The effect of material changes was also considered.

• Bond Master results compared with ABAQUS results

• Double lap joint was compared
• Aluminum adherends with AF126 film adhesive
• Tensile load
• Three nonlinear adhesive material models

This study took a slightly different approach to validation. The Bond Master product consists of
an analysis module, SUBLAM, connected to a user-friendly, graphical front-end. Input to the
SUBLAM module is a formatted text file that lists a series of SUBLAM commands. Creation of

91
the text file is tedious and error-prone, which was the motivation for development of the
graphical front-end. The graphical front-end collects the user input and creates the proper
SUBLAM input file. Only the SUBLAM portion of the Bond Master product was validated by
the work described in this section.

8.8.1 Introduction.

Adhesive joints are common in structures made of composites as well as structural repairs, which
allow load transfer mainly through shear. Under load, the adhesive layer often shows an initial
linear response followed by some strain hardening to a peak stress, and then the stress begins to
drop with a low softening modulus and the material experiences large strain, which in turn leads
to separation/failure.

To model adhesives with such nonlinear shear behavior, three nonlinear material options are
incorporated in the current version of SUBLAM, namely, elastic-perfectly plastic, Ramberg-
Osgood, and a piecewise model in which the stress-strain curve is defined by a series of data
points via cubic-spline fitting. For the purpose of validation, a typical joint type with all
adhesive nonlinear options is studied using SUBLAM, namely, the double lap joint. In the
double lap joint study, three nonlinear options are assigned to the adhesive layer; the results are
then compared with the corresponding FE analysis results from ABAQUS.

8.8.2 Conclusions.

The double lap joint with varied adhesive nonlinear models is studied using SUBLAM; the
comparisons are made between the results (the shear and peel stresses at the adhesive interface)
from SUBLAM and the FE analysis. In all cases, the results agree well. It is also seen that the
small difference present at the very end of the bonded region is believed to stem from the free
edges at the end.

8.8.3 Double Lap Joint Problem Description.

A double lap joint under axial load is considered. The joint configuration along with dimensions
is shown in figure 91. The material properties of aluminum adherend are E =10 Msi and ν = 0.3.
Three different material properties that correspond to the three nonlinear options are used for the
adhesive. The details are given in the following sections.

Figure 91. Configuration of the Double Lap Joint

92
8.8.3.1 Case 1, Elastic-Perfectly Plastic.

The parameters are given as G0 = 0.58928 Msi and τ0 =5000 psi, and the shear stress-strain curve
is displayed in figure 92. The elastic properties for the adhesive are E = 1.65 Msi and ν = 0.4.

Figure 92. Shear Stress-Strain Curve for an Elastic-Perfectly Plastic Adhesive

8.8.3.2 Case 2, Piecewise.

In this case, the data points are extracted from the experimental data for AF126 film adhesive at
RT, reported in reference 1. Figure 93 shows the nonlinear stress-strain curve. The elastic
properties are E = 0.34748 Msi and ν = 0.42.

Figure 93. Shear Stress-Strain Curve of AF126 Film Adhesive at RT [1]

93
8.8.3.3 Case 3, Ramberg-Osgood.

A Ramberg-Osgood fit describes stress-strain relation as

τ ⎡ ⎛ τ ⎞ ⎤
β

γ= ⎢1 + ⎜ ⎟ ⎥
G0 ⎢ ⎝ τ 0 ⎠ ⎥
⎣ ⎦

where G0, τ0, and β are fitted parameters. The same AF126 film as in Case 2 is considered; the
Ramberg-Osgood parameters obtained are G0 = 0.118 Msi, τ0 = 4520 psi, and β = 19.5.

8.8.4 Double Lap Joint Results and Discussion.

Three sets of SUBLAM models were generated. Although SUBLAM allows large elements in a
model, six elements are used in the bonded region to obtain better results. Two small elements
are placed at the end of the over lap region to capture the high stress gradient near the free edges.
A finite element model is also generated in ABAQUS, using plane-strain elements. To capture
the behavior of the adhesive layer, there are four elements through the thickness of the adhesive
layer.

In general, the three major failure modes of the adhesive joints are adherend failure, cohesive
failure (failure of adhesive material), and adhesive failure (separation between the adhesive and
the adherend). It is believed that the shear stress in the adhesive layer dictates the cohesive
failure, while the separation is due to the peel stress and the shear stress. Hence, the effect of the
material nonlinearity on both the shear stress and the peel stress is of interest. Due to the
symmetry, only the stress distributions in the bottom adhesive layer are plotted.

For Case 1 (elastic-perfectly plastic option), the shear stress and the peel stress along the top
interface of the adhesive layer from the SUBLAM and the ABAQUS results are shown in figures
94 and 95. Both models agree well, except at the very end of the bonded region. The plastic
regions predicted from both models are very similar, and the profiles for the rest of the region are
almost identical. Note that the small degree of oscillation in the SUBLAM result is because of
the polynomial approximations. In addition, the results from both models capture the rapid
change in the peel stress distribution, which is known to be due to the free edge effect at the end
of the bonded region.

94
 Figure 94. Elastic-Perfectly Plastic Adhesive: Shear Stress Distribution of the Top Interface
Along the Adhesive Bond

Figure 95. Elastic-Perfectly Plastic Adhesive: Peel Stress Distribution of the Top Interface
Along the Adhesive Bond

The AF126 film adhesive is considered to demonstrate the capability of the SUBLAM in
handling material nonlinearity. In the SUBLAM analysis, the material stress-strain relation is
defined via the piecewise option as well as the fitted Ramberg-Osgood model, while the
experimental data is used in the ABAQUS analysis. Again, the results are compared in terms of
the interface stress distribution. The shear stress and the peel stress along the top interface of the
adhesive layer from the SUBLAM and ABAQUS results are shown in figures 96 and 97. The

95
results from the SUBLAM models and the ABAQUS model agree well, except at the very end of
the bonded region. As expected, the SUBLAM models with two different nonlinear options
produce nearly identical results. Specifically, all solutions yield very similar in the nonlinear
regions and the profiles for the rest of the region are almost identical. It should be noted that the
stresses diverge when the distance along the bond approaches the end of adhesive layer.

Figure 96. AF126 Film Adhesive: Shear Stress Distribution on the Top Interface Along
the Adhesive Bond

Figure 97. AF126 Film Adhesive: Peel Stress Distribution on the Top Interface Along
the Adhesive Bond

96
8.9 Bond Master VERSUS ABAQUS, SINGLE LAP JOINT, PLY-BY-PLY FAILURE MODE.

Laminate adherends were considered next.

• Bond Master results compared with ABAQUS results

• Single lap joints were compared
• Fabric laminate with ply-by-ply failure criterion
• Tensile and in-plane shear load
• Linear elastic and Ramberg-Osgood adhesive material models

The single lap joint shown in figure 98 was modeled in Bond Master and ABAQUS, a
commercial FE solver. The joint was analyzed for two loading conditions: an axial tension load
and an in-plane shear load. The joint consists of a joggled laminate adhesively bonded to a flat
laminate. The laminates are quasi-isotropic, consisting of 21 fabric plies each. The plies
alternate between 0 and 45 degrees in the laminate lay-up. The ply properties are listed in
table 10.

Figure 98. Flush Lap Joint (All dimensions are in inches.)

Table 10. Ply Properties

Ply Thickness 0.005 (in.) ν13, ν23 0.25

E11, E22 3 (Msi) ε11, ε22 (ult. tension) 0.01
E33 1.5 (Msi) ε11, ε22 (ult. compression) 0.005
G12 0.6 (Msi) ε12 0.03
G13, G23 0.8 (Msi) ε33 (ult. tension) 0.02
ν12 0.04 ε13, ε23 0.01

Separate FE models were created for each load case. The tension condition was modeled using
bilinear plane-strain elements. The in-plane shear loading was modeled using solid brick
elements.

The shear loading condition could not be modeled using plane-strain elements since the shear
loads are perpendicular to the element faces. An additional degree of freedom is needed. The
plane-strain elements from the tension model were extruded to create solid brick elements. This
created a 3-D model representing a slice of the joint. The nodes on one side of the slice were
tied to the corresponding nodes on the opposite side. This enforced the constraints ε zz = 0 and
ε xx , ε yy , γ xy , γ xz , γ yz constant in the z-direction.

97
In both models, the individual plies were modeled explicitly. As a result, the models have a high
mesh density. The FE models have approximately 7100 elements each. Twenty-one elements
are modeled through the thickness of each laminate, one element per ply thickness. Each
element is assigned the properties of the corresponding ply. A close-up of the plane-strain mesh
is shown in figure 99.

Figure 99. Close-Up of FE Mesh, Lap Joint

The joint was modeled for two different adhesives. The first adhesive is linear elastic. The
second adhesive follows the Ramberg-Osgood rule of plasticity. The adhesive properties are
listed in table 11.

Table 11. Adhesive Properties

G0 (ksi) ν τ0 (ksi) β Adhesive Model

Linear 142 0.35 N/A N/A Linear Elastic τ
Adhesive γ=
G0
Nonlinear 266 0.35 4.2 18.4 Ramberg-Osgood
τ ⎡ ⎛ τ ⎞ ⎤
β

Adhesive γ= ⎢1 + ⎜ ⎟ ⎥
G0 ⎢ ⎝ τ 0 ⎠ ⎥
⎣ ⎦

The maximum strain criterion was used to predict failure on a ply-by-ply level. The following
sections describe how the criterion was applied.

8.9.1 Approach.

Bond Master predicts ply-by-ply failure by computing the strain state on the surfaces of the plies
and then applying the maximum strain criterion to the computed strains.

Bond Master assumes a linear distribution of u and v displacements, and a quadratic distribution
of w displacements through the thickness of each plate. This gives a plate that is shear
deformable and allows stretching through the thickness. The strain field within the plates was
derived from the displacement field.

98
Bond Master computes ply-by-ply factors of safety from the derived strain-state. Bond Master
applies the chosen failure criterion to the strains at the surfaces of each ply. Bond Master then
tabulates the factors of safety.

ABAQUS was used to predict ply failure on an element-by-element basis. ABAQUS computed
the strain at the centroid of each element. Custom code was used to transform the strain into the
material orientation and to compute a factor of safety from the transformed strain.

8.9.2 Flush Lap Joint, Tension Loading.

An axial load of 1000 lbf/in. was applied to the end of the flat laminate. As mentioned, the joint
was modeled for a linear adhesive and again for a nonlinear adhesive. The results for each
adhesive are discussed separately in the following sections.

8.9.2.1 Linear Elastic Adhesive.

The joint was modeled for a linear adhesive. The calculated factors of safety are plotted in
figures 100 and 101. First ply failure is plotted along the lengths of the two adherends. In most
regions, Bond Master’s results are nearly identical to the FE results. In the region of the bond,
Bond Master is conservative. However, the difference between the two results is less than 10%.

7
Finite Element Model
6 BondMaster

5
Factor of Safety

0
0 1 2 3 4
Y-Coordinate (in.)

Figure 100. Flush Lap Joint Tension Loading Linear Elastic Adhesive First Ply
Failure Along Plate 1

99
 3.5
Finite Element Model
3 BondMaster

2.5

Factor of Safety
2

1.5

0.5

0
0 1 2 3 4
Y-Coordinate (in.)

Figure 101. Flush Lap Joint Tension Loading Linear Elastic Adhesive First Ply
Failure Along Plate 2

8.9.2.2 Nonlinear Ramberg-Osgood Adhesive.

The joint was modeled for a nonlinear adhesive that followed Ramberg-Osgood plasticity. The
shear stress along the bond is plotted in figure 102. In most regions, Bond Master’s results are
nearly identical to the FE results. The adhesive goes highly nonlinear in the region where plate 2
terminates. In this region, Bond Master’s results oscillate. Bond Master uses polynomial
approximations to describe the displacement field along the bond. The oscillations are an artifact
of the polynomial approximation.

Typically, a factor of safety is calculated by dividing a material allowable by the stress or strain
in the material. This approach does not yield a true factor of safety when the stresses or strains
do not scale linearly with load. This approach was used in the present analysis, even though the
stresses and strains do not scale linearly due to adhesive nonlinearity. The factors of safety
reported in this section are not true factors of safety. The relative magnitudes of the factors of
safety can be compared. A factor of safety less than 1 indicates material failure. However, the
load to failure does not equal the applied load times the factor of safety.

The factors of safety along plates 1 and 2 are plotted in figures 103 and 104. The factors of
safety computed by Bond Master agree well with the FE results. The two results diverge slightly
in the region where the adhesive becomes highly nonlinear.

100
 6000
Finite Element Model
5000 BondMaster
Adhesive Shear Stress (psi)

4000

3000

2000

1000

0
0 0.5 1 1.5 2
-1000
Y-Coordinate (in.)

Figure 102. Tension Loading Nonlinear Ramberg-Osgood Adhesive Shear Stress Along Bond

7
Finite Element Model
6 BondMaster

5
Factor of Safety

0
0 1 2 3 4
Y-Coordinate (in.)

Figure 103. Tension Loading Nonlinear Ramberg-Osgood Adhesive First Ply

Failure Along Plate 1

101
 3.5
Finite Element Model
3 BondMaster

2.5
Factor of Safety

1.5

0.5

0
0 1 2 3 4
Y-Coordinate (in.)

Figure 104. Flush Lap Joint Tension Loading Linear Elastic Adhesive First Ply
Failure Along Plate 2

8.9.3 Flush Lap Joint, In-Plane Shear Loading.

The flat laminate was loaded in in-plane shear to a load of 4000 lbf/in. The loading is shown in
figure 105. As mentioned, the joint was modeled for a linear adhesive and again for a nonlinear
adhesive. The results for each adhesive are discussed separately in the following sections.

Figure 105. In-Plane Shear Loading

8.9.3.1 Linear Elastic Adhesive.

The joint was modeled for a linear adhesive. The factors of safety for first ply failure are plotted
in figures 106 and 107. Bond Master’s results are nearly identical to the FE results. The results
diverge in regions of the plate terminations. The divergence is due to differences in the
computed values for XY and XZ shear strains between Bond Master and the FE results.

102
 1
0.9 Finite Element Model
BondMaster
0.8
Factor of Safety 0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0 1 2 3 4
Y-Coordinate (in.)

Figure 106. In-Plane Shear Loading Linear Elastic Adhesive First Ply Failure Along Plate 1

1
Finite Element Model
0.9
BondMaster
0.8
0.7
Factor of Safety

0.6
0.5
0.4
0.3
0.2
0.1
0
0 1 2 3 4
Y-Coordinate (in.)

Figure 107. In-Plane Shear Loading Linear Elastic Adhesive First Ply Failure Along Plate 2

8.9.3.2 Nonlinear Ramberg-Osgood Adhesive.

The joint was modeled for a nonlinear adhesive, which followed Ramberg-Osgood plasticity.
The shear stress along the bond is plotted in figure 108. The adhesive goes highly nonlinear over
approximately 35% of the bond length. Bond Master’s computed stress is nearly identical to the
FE stress.

103
The factors of safety are plotted in figures 109 and 110. The results diverge in the regions of the
plate terminations. This trend was also observed for the linear adhesive. Otherwise, the two
results are in good agreement.

8000
Finite Element Model
7000 BondMaster
6000
Adhesive Shear Stress (psi)

5000

4000

3000

2000

1000

0
0 0.5 1 1.5 2
-1000
Y-Coordinate (in.)

Figure 108. In-Plane Shear Loading Nonlinear Ramberg-Osgood Adhesive

Shear Stress Along Bond

3
Finite Element Model
2.5 BondMaster

2
Factor of Safety

1.5

0.5

0
0 1 2 3 4
Y-Coordinate (in.)

Figure 109. In-Plane Shear Loading Nonlinear Ramberg-Osgood Adhesive

First Ply Failure Along Plate 1

104
 2.5
Finite Element Model
BondMaster
2

Factor of Safety
1.5

0.5

0
0 1 2 3 4
Y-Coordinate (in.)

Figure 110. In-Plane Shear Loading Nonlinear Ramberg-Osgood Adhesive

First Ply Failure Along Plate 2

8.9.4 Scarf Joint.

Laminate adherends were considered in a second joint type.

• Bond Master results compared with ABAQUS results

• Scarf joints were compared
• Fabric laminate with ply-by-ply failure criterion
• Tensile load
• Linear elastic and Ramberg-Osgood adhesive material models

8.9.4.1 Approach.

The goal of the analysis is to compare ply-by-ply factors of safety calculated by Bond Master
with finite element results. The scarf joint shown in figure 111 was modeled in Bond Master and
ABAQUS, a commercial FE solver. The joint was modeled for an axial tension load. The
laminates are near quasi-isotropic, consisting of seven fabric plies each. The plies alternate
between 0 and 45 degrees in the laminate lay-up. There is one additional 0 degree ply at the
mid-plane of each laminate. The ply properties are listed in table 12.

105
 Figure 111. Scarf Joint (All dimensions are in inches.)

Table 12. Ply Properties

Ply Thickness 0.015 (in.) ν13, ν23 0.25

E11, E22 3 (Msi) ε11, ε22 (ult. tension) 0.01
E33 1.5 (Msi) ε11, ε22 (ult. compression) 0.005
G12 0.6 (Msi) ε12 0.03
G13, G23 0.8 (Msi) ε33 (ult. tension) 0.02
ν12 0.04 ε13, ε23 0.01

The joint was modeled using bilinear plane-strain elements. The individual plies were modeled
explicitly. As a result, the model has a high mesh density. The FE model has approximately
2100 elements. Seven elements are modeled through the thickness of each laminate, one element
per ply thickness. Each element is assigned the properties of the corresponding ply. A close-up
of the mesh is shown in figure 112.

Figure 112. Close-Up of FE Mesh, Scarf Joint

Real tapers involve discrete drop-offs of plies. The discrete drop-offs were not modeled in the
FE model. Instead, the tapers were treated as a proportional change in the thickness in all the
laminate’s plies. Bond Master also uses this approach to approximate a taper, so the results of
the two analysis methods should be directly comparable. The effect of this approach on the
accuracy of the results was not studied. This approach is most accurate for a thick laminate with
well-dispersed plies, and a well-dispersed pattern of drop-offs.

106
The FE mesh has notches at the ends of the bondline. The notches represent regions too thin to
contain a ply. In a manufactured scarf joint, these regions would either be filled with resin or
eliminated by overlapping a ply onto the laminate surface. The FE mesh was created with the
assumption that these regions are nonstructural. Any resin pooling in these regions is assumed to
have no effect on joint strength. Bond Master also uses this assumption.

The joint was modeled for two different adhesives. The first adhesive is linear elastic. The
second adhesive is elastic-perfectly plastic. The adhesive properties are listed in table 13.

Table 13. Adhesive Properties

G0 (ksi) ν τ0 (ksi) Adhesive Model

Linear 142 0.35 N/A Linear Elastic τ = G0 γ
Adhesive
Nonlinear 189 0.35 4.28 Elastic-Perfectly Plastic ⎡ τ0 ⎤
Adhesive ⎢γ < G G0 γ ⎥
τ=⎢ ⎥
0
⎢ τ0 ⎥
⎢γ≥ τ0 ⎥
⎣ G0 ⎦

The maximum strain criterion was used to predict failure on a ply-by-ply level. The following
sections describe how the criterion was applied.

Bond Master predicts ply-by-ply failure by computing the strain-state on the surfaces of the plies
and then applying the maximum strain criterion to the computed strains.

Bond Master assumes a linear distribution of u and v displacements, and a quadratic distribution
of w displacements through the thickness of each plate. This gives a plate that is shear
deformable and allows stretching through the thickness. The strain field within the plates was
derived from the displacement field.

Bond Master computes ply-by-ply factors of safety from the derived strain-state. Bond Master
applies the chosen failure criterion to the strains at the surfaces of each ply. Bond Master then
tabulates the factors of safety.

ABAQUS was used to predict ply failure on an element-by-element basis. ABAQUS computed
the strain at the centroid of each element. Custom code was used to transform the strain into the
material orientation and to compute a factor of safety from the transformed strain.

8.9.4.2 Results for Linear Elastic Adhesive.

An axial load of 2700 lbf/in. was applied to the end of the flat laminate. The joint was modeled
for a linear adhesive. The calculated factors of safety are plotted in figures 113 and 114. First
ply failure is plotted along the lengths of the two adherends. Bond Master’s results are nearly
identical to the FE results.

107
 2
1.8 Finite Element Model
BondMaster
1.6
1.4

Factor of Safety
1.2
1
0.8
0.6
0.4
0.2
0
0 0.5 1 1.5 2
Y-Coordinate (in.)

Figure 113. Scarf Joint Linear Elastic Adhesive First Ply Failure Along Plate 1

2
1.8 Finite Element Model
BondMaster
1.6
1.4
Factor of Safety

1.2
1
0.8
0.6
0.4
0.2
0
0 0.5 1 1.5 2
Y-Coordinate (in.)

Figure 114. Scarf Joint Linear Elastic Adhesive First Ply Failure Along Plate 2

8.9.4.3 Results for Elastic Perfectly-Plastic Adhesive.

The joint was modeled for an elastic-perfectly plastic adhesive. The shear stress along the bond
is plotted in figure 115. In most regions, Bond Master’s results are nearly identical to the FE
results. The adhesive goes highly nonlinear in the region where plate 2 terminates. In this
region, Bond Master’s results oscillate. Bond Master uses polynomial approximations to
describe the displacement field along the bond. The oscillations are an artifact of the polynomial
approximation.

Typically, a factor of safety is calculated by dividing a material allowable by the stress or strain
in the material. This approach does not yield a true factor of safety when the stresses or strains
do not scale linearly with load. This approach was used in the present analysis, even though the
adhesive is nonlinear. Therefore the factors of safety reported in this section are not true factors
of safety; however, the relative magnitudes can be compared. A factor of safety less than 1

108
indicates material failure. However, the load to failure does not equal the applied load times the
factor of safety.

The factors of safety along plates 1 and 2 are plotted in figures 116 and 117. Bond Master’s
results are nearly identical to the FE results.

4400
Finite Element Model
4200 BondMaster
Adhesive Shear Stress (psi)

4000

3800

3600

3400

3200

3000
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Y-Coordinate (in.)

Figure 115. Scarf Joint Elastic-Perfectly Plastic Adhesive Shear Stress Along Bond

2
1.8 Finite Element Model
BondMaster
1.6
1.4
Factor of Safety

1.2
1
0.8
0.6
0.4
0.2
0
0 0.5 1 1.5 2
Y-Coordinate (in.)

Figure 116. Scarf Joint Elastic-Perfectly Plastic Adhesive First Ply Failure Along Plate 1

109
 2
Finite Element Model
1.8
BondMaster
1.6
1.4

Factor of Safety
1.2
1
0.8
0.6
0.4
0.2
0
0 0.5 1 1.5 2
Y-Coordinate (in.)

Figure 117. Scarf Joint Elastic-Perfectly Plastic Adhesive First Ply Failure Along Plate 2

8.10 Bond Master VERSUS TEST DATA, LINEAR, AND NONLINEAR ADHESIVE.

Bond Master results were compared with the test data from the literature.

• Bond Master results compared with test data

• Single lap, double lap, and single joggle joints were compared
• Materials properties and joint dimensions taken from reference 21
• Tensile load
• Linear elastic and elastic-perfectly plastic adhesive material models

Reference 19 describes some physical testing that was done on adhesive lap joints. The
measured data reported in reference 19 were compared to Bond Master results. The joint naming
terminology is slightly different between the test data and Bond Master.

Test Data Term Bond Master Term

Single lap joint Basic Single Lap Joint

Joggle Single Lap Joint with Flush Exterior
L-section Basic Single Lap Joint (with stiffened Plate #1)
Double strap joint Basic Symmetric Lap Splice

Figures 118 through 121 show force versus displacement data for the four joint configurations
listed above. During testing [19], each joint was subjected to an increasing tensile load until the
joint failed. Each joint was modeled in Bond Master so that the joint materials and geometry
matched the test coupons. For each joint model, a number of different loads were tried in Bond
Master. Bond Master calculates a factor of safety for the adhesive and the adherends. The
failure load was found when the lowest factor of safety equaled 1.

110
Each graph contains four plots:

• Test data from reference 19

• Bond Master results with the linear elastic adhesive model

• Bond Master results using the elastic plastic adhesive model

• Bond Master results using the elastic plastic adhesive model with the nonlinear geometry
switch on

Only one plot is displayed for the three nonlinear adhesive models. This was done because the
Bond Master results were similar for the elastic-plastic, Ramberg-Osgood, and piecewise
models. The results are summarized in tables 14 and 15.

measured data
linear elastic adhesive
elastic plastic adhesive
nonlinear geometry
LOAD VS. DISPLACEMENT
4
SINGLE LAP JOINT
1 10

8000

6000
load [N]

4000

2000

-2000
-0.5 0 0.5 1 1.5 2 2.5

displacement [mm]

Figure 118. Bond Master Matches the Failure Load of the Single Lap Joint but
Overpredicts the Stiffness

111
 measured data
linear elastic adhesive
elastic plastic adhesive
nonlinear geometry LOAD VS. DISPLACEMENT
JOGGLE
4000

3500

3000

2500
load [N]

2000

1500

1000

500

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8

displacement [mm]

Figure 119. Bond Master Accurately Predicts Failure Load and Stiffness of the Joggle Joint

measured data
linear elastic adhesive
elastic plastic adhesive
nonlinear geometry
LOAD VS. DISPLACEMENT
4
L-SECTION
1 10

8000

6000
load [N]

4000

2000

0
0 0.5 1 1.5

displacement [mm]

Figure 120. Bond Master Results Agree Well With Measured Data for the Single
Lap Joint With an L-Section

112
 measured data
linear elastic adhesive
elastic plastic adhesive
nonlinear geometry LOAD VS. DISPLACEMENT
4 DOUBLE STRAP JOINT
1.2 10

4
1 10

8000
load [N]

6000

4000

2000

0
0 0.5 1 1.5 2 2.5 3 3.5

displacement [mm]

Figure 121. Bond Master Slightly Over Predicts the Failure Load and Stiffness for the
Double Strap Joint

Table 14. Data From Figures 118 Through 121 in Tabular Form

Single Double
Lap Joint Joggle L-Section Strap Joint
Failure Load [N] 9892 3935 8269 10834
Measured Data
Failure Displacement [mm] 2.1 0.6 1.1 3.0
Measured Data
Failure Load [N] 9430 3737 7740 10987
Linear Elastic
Failure Displacement [mm] Linear Elastic 1.07 0.74 1.40 1.85
Failure Load [N] 8941 3737 7740 11610
Elastic Plastic
Failure Displacement [mm] 1.04 0.77 1.48 1.97
Elastic Plastic
Failure Load [N] 4404 2002 3692 7962
Nonlinear Geometry
Failure Displacement [mm] 0.50 0.41 0.70 1.35
Nonlinear Geometry

113
 Table 15. Failure Modes

Failure Predicted Failure Predicted

Using Bond Master Using Bond Master
Failure of Tested With Linear Elastic With Elastic Plastic
Coupons Adhesive Model Adhesive Model
Single Lap Joint Tensile failure Tensile failure Tensile failure
Plate 2 Plate 1 or 2 Plate 1 or 2
Joggle Cracks initiated by Shear failure Shear failure
peel stress Plate 1 Plate 1
Plate 1
Double Strap Joint Tensile failure Tensile failure Shear failure
Plate 3 Plate 2 Bond

8.11 Bond Master VERSUS TEST DATA, DIFFERENT ADHESIVE THICKNESSES.

Reference 21 also described testing of joints with various adhesive thicknesses. Equivalent Bond
Master models were built and compared with the test data.

• Bond Master results compared with test data for three adhesive thicknesses
• Single lap, double lap, and single joggle joints were compared
• Materials properties and joint dimensions taken from reference 19
• Tensile load
• Linear elastic and elastic-perfectly plastic adhesive material models

Figures 122 through 124 show force versus displacement data for three different adhesive
thicknesses in the L-section model. Each figure shows the following three plots:

• Test data from reference 19

• Bond Master results with the linear elastic adhesive model
• Bond Master results using the elastic plastic adhesive model

The results are summarized in tables 16 and 17.

114
 measured data
linear elastic
elastic plastic
L-SECTION LOAD VS DISPLACEMENT
4
ADHESIVE THICKNESS = 0.127 mm
1 10

8000

6000
load [N]

4000

2000

0
0 0.2 0.4 0.6 0.8 1 1.2

displacement [mm]

Figure 122. Bond Master Shows Good Agreement With the Measured Data for the L-Section
Joint With 0.127-mm (0.005-in.) Thick Adhesive

measured data
linear elastic
elastic plastic
L-SECTION LOAD VS DISPLACEMENT
4
ADHESIVE THICKNESS = 0.635 mm
1 10

8000

6000
load [N]

4000

2000

0
0 0.2 0.4 0.6 0.8 1 1.2 1.4

displacement [mm]

Figure 123. Bond Master Over Predicts the Failure Load of the L-Section Joint With
0.635-mm (0.025-in.) Thick Adhesive

115
 measured data
linear elastic
elastic plastic
L-SECTION LOAD VS DISPLACEMENT
ADHESIVE THICKNESS = 2.54 mm
8000

7000

6000

5000
load [N]

4000

3000

2000

1000

0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6

displacement [mm]

Figure 124. Bond Master Over Predicts the Failure Load of the L-Section Joint With
2.54-mm (0.100-in.) Thick Adhesive

Table 16. Data From Figures 122 Through 124 in Tabular Form

L-Section L-Section L-Section

Adhesive Thickness = Adhesive Thickness = Adhesive Thickness =
0.127 mm 0.635 mm 2.54 mm
Failure Load [N] 8340 5310 3850
Measured Data
Failure Displacement [mm] 1.12 0.78 0.65
Measured Data
Failure Load [N] 8879 8683 7731
Linear Elastic
Failure Displacement [mm] 1.17 1.22 1.42
Linear Elastic
Failure Load [N] 7228 8745 7749
Elastic Plastic
Failure Displacement [mm] 0.97 1.27 1.50
Elastic Plastic

116
 Table 17. Failure Modes

Failure Predicted Failure Predicted

Using Bond Master Using Bond Master
Failure of With Linear Elastic With Elastic Plastic
Tested Coupons Adhesive Model Adhesive Model
L-section Tensile failure Tensile failure Shear failure
Adhesive Thickness Plate 2 Plate 2 Bond
= 0.127 mm
L-section Tensile failure Tensile failure Tensile failure
Adhesive Thickness Plate 2 Plate 2 Plate 2
= 0.635 mm
L-section Tensile failure Tensile failure Tensile failure
Adhesive Thickness Plate 2 Plate 2 Plate 2
= 2.54 mm

8.11.1 Discussion of Results.

In some cases, the force displacement curves generated by Bond Master differed from the
measured data. Some of the differences along with possible explanations for the discrepancies
are discussed in the following sections.

8.11.1.1 Single Lap Joint.

It is noted that the single lap joint representation in Bond Master is different than the test
specimen, see figure 125. Although adjusting the offset distance improves the prediction
slightly, it is thought that the fundamental difference is owing to the configuration. Clearly, the
bonded region of the single lap joint subject to axial tension is under tensile and bending load,
and the configuration of the test specimen offer higher bending rigidity.

Figure 125. Two Single Lap Joint Configurations

A detailed SUBLAM model was generated and analyzed for the comparison. Figure 126 shows
the deformed shape for both configurations at P=100 lb. It is seen that, for the Bond Master
version, the deformation of the adherend under bending is rather large. On the other hand, the
test specimen version is much smaller. The experiment indicates that the specimens failed at the
bonded region instead of breaking at the small notch area (~0.063″ according to ASTM 3165-
00); hence, it is justified that the failure prediction can be based on the behavior in the bonded
region only.

117
The Bond Master single lap joint model assumes that each adherend is a constant thickness
throughout its length. The test coupons did not conform to this assumption. In the bonded
region, the test specimen was one layer thick. Outside the bonded region, the test specimen was
two layers thick. The Bond Master model could have been constructed with either thickness. A
two-layer thickness makes the Bond Master model more likely to fail in the bond region,
replicating the experimental results. The two-layer thickness was chosen to try to predict the
failure load. However, the extra adherend thickness makes the model stiffer than the test
specimens. Using an offset distance = 0.05″, the predicted failure load is close to the test
average. The predicted coupon stiffness is greater than the tested coupon as expected.

Figure 126. Deformation of Two Single Lap Joint Configurations

8.11.1.2 Joggle Lap Joint.

Two parameters were adjusted in an attempt to improve the failure prediction: the geometry in
the curved region and the offset distance. Bond Master yielded an excellent result with an inner
radius = 0.15″ and an offset distance = 0.1″.

8.11.1.3 Double Strap Joint.

Bond Master slightly overpredicted the failure load and joint stiffness.

8.11.1.4 L-section Joint.

Bond Master overpredicted the failure load for larger adhesive thicknesses.

In any of these cases, discrepancies could be due to flaws in the Bond Master theory or
implementation. The ply allowables were estimated from knowledge of the ply constituents. If
these estimates were incorrect, then the Bond Master results would be change.

Bond Master does not contain logic that would account for the large drop in failure load as the
adhesive thickness goes up. This missing logic probably accounts for the results discrepancy in
the L-section model.

118
9. REFERENCES.

1. Tomblin, J., Seneviratne, W., Escobar, P., and Yap, Y., “Shear Stress-Strain Data for
Structural Adhesives,” FAA report DOT/FAA/AR-02/97, November 2002.

2. Actis, R.L. and Szabo, B.A., “Analysis of Bonded and Fastened Repairs by the p-Version
of the Finite-Element Method,” Computers & Mathematics With Applications, Vol. 46,
2003, pp. 1-141.

3. Von Tooren, M.J.L., Gleich, D.M., and Beukers, A., “Experimental Verification of a
Stress Singularity Model to Predict the Effect of Bondline Thickness on Joint Strength,”
Journal of Adhesion Science and Technology, Vol. 18, No. 4, 2004, pp. 395-412.

4. Objois, A., Gibert, Y., and Fargette, B., “Theoretical and Experimental Analysis of the
Scarf Joint Bonded Structure: Influence of the Adhesive Thickness on the Fine
Mechanical Behaviour,” Journal of Adhesion, Vol. 70, 1999, pp. 13-32.

5. Tomblin, J.S., Yang, C.Y., and Harter, P., “Investigation of Thick Bondline Adhesive
Joints,” FAA report DOT/FAA/AR-01/33, June 2001.

6. Yang, C.Y., Tomlin, J.S., and Guan, Z.D., “Analytical Modeling of ASTM Lap Shear
Adhesive Specimens,” FAA report DOT/FAA/AR-02/130, February 2003.

7. Flanagan, G., “Evaluation of Composite Joints in General Aviation (GA) Structures,”

MSC TPR 3017/GE02, Technical Progress Report, April 2003.

8. Schreyer, H.L. and Chen, Z., “One-Dimensional Softening With Localization,” Journal
of Applied Mechanics, Vol. 53, December 1986, pp. 791-797.

9. Niordson, C.F. and Hutchison, J.W., “On Lower Order Strain Gradient Plasticity
Theory,” European Journal of Mechanics, A/ Solids, Vol. 22, 2003, pp. 771-778.

10. Gleich, D.M., Van Tooren, M.J.L., and Beukers, A., “Analysis and Evaluation of
Bondline Thickness Effects on Failure Load in Adhesively Bonded Structures,” Journal
of Adhesion Science and Technology, Vol. 15, No. 9, 2001, pp. 1091-1101.

11. Crocombe, A.D., “Global Yielding as a Failure Criteria for Bonded Joints,” International
Journal of Adhesion & Adhesives, Vol. 9, 1989, pp. 145-153.

12. Yang, Q.D. and Thouless, M.D., “Mixed-Mode Fracture Analyses of Plastically
Deforming Adhesive Joints,” International Journal of Fracture, Vol. 110, 2001, pp. 175-
187.

13. Kafkalidis, M.S., Thouless, M.D., Yang, Q.D., and Ward, M.S., “The Effects of
Geometry and Material Properties on the Fracture of Single Lap-Shear Joints,”
International Journal of Solids and Structure, Vol. 39, 2002, pp. 4367-4383.

119
14. Pardoen, T., Ferracin, T., Landis, C.M., and Delannay, F., “Constraint Effects in
Adhesive Joint Fracture,” Journal of the Mechanics and Physics of Solids, Vol. 53, 2005,
pp. 1951-1983.

15. Tvergaard, V. and Hutchinson, J.W., “On the Toughness of Ductile Adhesive Joints,”
Journal of Mechanics and Physics of Solids, Vol. 44, 1996, pp. 789-800.

16. Wei, Y. and Hutchinson, J.W., “Interface Strength, Work of Adhesion and Plasticity in
the Peel Test,” International Journal of Fracture, Vol. 93, 1998, pp. 315-333.

17. Hadi-Ahmed, R., Foret, G., and Ehrlacher, A., “Probabilistic Analysis of Failure in
Adhesive Bonded Joints,” Mechanics of Materials, Vol. 33, 2001, pp. 77-84.

18. Towse, A., Potter, K., Wisnom, M.R., and Adams, R.D., “Specimen Size Effects in the
Tensile Failure Strain of an Epoxy Adhesive,” Journal of Material Sciences, Vol. 33,
1998, pp. 4307-4314.

19. Heslehurst, R.B. and Hart-Smith, L., “The Science and Art of Structural Adhesive
Bonding,” SAMPE Journal, Vol. 38, 2002, pp. 60-71.

20. Taib, A.A., Boukhili, R., Achiou, S., and Boukehili, H., “Bonded Joints With Composite
Adherends, Part II, Finite Element Analysis of Joggle Lap Joints,” International Journal
of Adhesion and Adhesives, 2005.

21. Taib, A.A., Boukhili, R., Achiou, S., Gordon, S., and Boukehili, H., “Bonded Joints With
Composite Adherends, Part I, Effect of Specimen Configuration, Adhesive Thickness,
Spew Fillet and Adherend Stiffness on Fracture,” International Journal of Adhesion and
Adhesives, 2005.

22. Hoyt, D.M., Ward, S., and Minguet, P., “Strength and Fatigue Life Modeling of Bonded
Joints in Composite Structure,” American Society of Composites, 2000.

23. Krueger, R., Paris, I., O'Brien, T.K., and Minguet, P., “Comparison of 2D Finite Element
Modeling Assumptions With Results From 3D Analysis for Composite Skin-Stiffener
Debondings,” NASA Langley Research Center, Hampton, Virginia, 2001.

24. Duncan, B. and Crocker, L., “Characterization of Flexible Adhesives for Design,”
Measurement Good Practice Guide No. 45, National Physical Laboratory, Teddington,
UK, July 2001.

120