Академический Документы
Профессиональный Документы
Культура Документы
by
KYLE J. TOUSIGNANT
Kyle J. Tousignant
Doctor of Philosophy
Department of Civil Engineering
University of Toronto
2017
ABSTRACT
An experimental and finite element (FE) research program was conducted to investigate the “fit-for-
purpose” design of welds in hollow structural section (HSS) connections using weld effective lengths, and the
applicability of the North American directional strength-increase factor for fillet welds to HSS. Weld
effective lengths for circular hollow section (CHS) connections are studied for the first time.
Experimental testing of a 10-m span, simply-supported, rectangular hollow section (RHS) Warren truss,
with fillet- and groove-welded overlapped K-connections, was performed by applying a single quasi-static
point load to truss panel points to cause sequential rupture of nine test welds to overlapping branches. The
weld in-situ strength, effect of key connection parameters, and safety (reliability) index of AISC formulae for
weld design in RHS overlapped K-connections were determined. Modifications to the AISC formulae and a
simplified method for modelling overlap-jointed RHS trusses under service load conditions are proposed.
The North American directional strength-increase factor for fillet welds to HSS (covering both RHS and
CHS) was then studied through 33 experimental tests and 73 non-linear FE analyses of weld-critical HSS-to-
rigid plate connections. The safety index was determined for prominent North American design codes, and a
recommendation to prohibit the use of the directional strength-increase factor for the design of fillet welds to
all HSS was made. More accurate design formulae for fillet welds to HSS are proposed.
Twelve large-scale experiments and 256 non-linear FE tests were then conducted to determine weld
effective lengths for CHS-to-CHS X- (or Cross-) connections, which are shown to decrease as branch-to-
chord width ratio, chord wall slenderness, and branch-to-chord thickness ratio increase. North American fillet
weld design formulae were found to already provide adequate reliability provided that the directional
strength-increase factor is not used. A new fillet weld design approach for CHS X-connections, using weld
effective lengths, is proposed.
ii
ACKNOWLEDGEMENTS
First and foremost, I would like to thank my supervisor, Professor Jeffrey A. Packer, for his support,
guidance and dedication to this project and my professional development. I would like to also thank the
technical staff at the University of Toronto Structural Testing Facility: Mr. R. Basset, Mr. J. MacDonald, Mr.
M. Sun, Mr. G. Buzzeo, Mr. B. Cook, Mr. M. Fiss, and Mr. A McClenaghan.
Financial support for this project was provided by the American Institute of Steel Construction (AISC),
the Natural Sciences and Engineering Research Council of Canada (NSERC), and the Steel Structures
Education Foundation (SSEF). Hollow structural sections were donated by Atlas Tube, Harrow, Ontario, and
fabrication services were donated by Walters Inc., Hamilton, Ontario.
Additional thanks are extended to Kubes Steel, Hamilton, Ontario, for fabricating the CHS-to-rigid plate
connection specimens, and to my colleagues Mr. C. Ritchie and Mr. P. Oatway for their ongoing interest in
this work. The laboratory contributions of Dr. G. S. Frater, Dr. M. Sun, Mr. P. Oatway, Ms. J. Lu, and Mr. F.
Wei are gratefully acknowledged.
iii
TABLE OF CONTENTS
Abstract .................................................................................................................................................. ii
Acknowledgements ............................................................................................................................................ iii
Table of Contents ............................................................................................................................................... iv
List of Tables ................................................................................................................................................. ix
List of Figures ............................................................................................................................................... xiii
Abbreviations and Symbols ............................................................................................................................. xix
Chapter 1: Introduction ............................................................................................................................... 1
1.1. Hollow Structural Sections ....................................................................................................... 1
1.2. Weld Effective Lengths ............................................................................................................ 2
1.3. Research Program Goal and Overview ..................................................................................... 3
Chapter 2: Weld Effective Lengths in RHS Overlapped K-Connections ................................................... 6
2.1. Introduction ............................................................................................................................... 6
2.2. Experimentation ........................................................................................................................ 7
2.2.1. Scope............................................................................................................................ 7
2.2.2. Truss Design ................................................................................................................ 7
2.3. Geometric and Mechanical Properties .................................................................................... 11
2.4. Instrumentation and Loading Strategy .................................................................................... 13
2.4.1. Instrumentation .......................................................................................................... 13
2.4.2. Loading Strategy ........................................................................................................ 14
2.5. Results ..................................................................................................................................... 16
2.6. Evaluation of AISC 360 (2010) .............................................................................................. 20
2.6.1. AISC 360 (2010) Provisions for Weld Effective Lengths in RHS Overlapped
K-Connections ........................................................................................................... 20
2.6.2. Safety Level Inherent in AISC 360 (2010) ................................................................ 22
2.7. Recommendation .................................................................................................................... 23
2.7.1. Background ................................................................................................................ 23
2.7.2. Proposal ..................................................................................................................... 23
2.7.3. Safety Level Inherent in Recommendation................................................................ 25
2.7.4. Comments .................................................................................................................. 25
2.8. Summary ................................................................................................................................. 27
iv
Table of Contents v
C.1. Fillet Weld Resistance According to EN1993-1-8 (2005) for a Skewed-T Fillet Weld ....... 191
C.2. CHS-to-Rigid Plate Fabrication Drawings ........................................................................... 193
C.3. Weld Measurements for CHS-to-Rigid plate Connections ................................................... 197
C.4. Theoretical Throat Dimension of a Skewed-T Fillet Weld ................................................... 204
C.5. Material Property Test Results .............................................................................................. 206
C.6. Photographs of Experimental Tests and Failure Modes for CHS-to-Rigid Plate Tests ........ 207
C.7. Examples of Weld Strength Calculation Method for Fillet-Welded CHS-to-Rigid Plate
Connections........................................................................................................................... 212
C.8. Approach to Calculating Loading Angle for CHS-to-Plate Connections ............................. 214
C.9. Template Length Formula for CHS-to-Plate Connections.................................................... 216
C.10. Summary of Predicted Weld Strengths for HSS-to-Rigid Plate Connections ...................... 217
Fillet Welds in HSS-to-Rigid Plate Connections: Finite Element Modelling....................... 219
D.1. Supplementary Figures ......................................................................................................... 219
D.2. Summary of Actual and Predicted Strengths According to Code Design Methods for
FE CHS- and RHS-to-Rigid Plate Tests ............................................................................... 223
Fillet Weld Effective Lengths in CHS X-Connections: Experimentation ............................ 225
E.1. CHS-to-CHS X-Connection Fabrication Drawings .............................................................. 225
E.2. Template Length Formula for CHS-to-CHS Connections .................................................... 233
E.3. Measured Fillet Weld Dimensions for CHS-to-CHS X-Connections................................... 236
E.4. 3D Solidworks Weld Profiles for All Test Welds................................................................. 243
E.5. Post-Rupture Macroetch Measurements ............................................................................... 246
E.6. Material Property Test Results.............................................................................................. 247
E.7. Photographs of Experimental Tests and Failure Modes ....................................................... 248
E.8. Summary of Actual and Predicted Strengths According to Code Design Methods for
FE CHS X-Connection Tests ................................................................................................ 253
Fillet Weld Effective Lengths in CHS X-Connections: Finite Element Modelling .............. 254
F.1. Complete Parametric Study Results ...................................................................................... 254
F.2. Correlation Plots and Reliability Analysis Parameters for Code Design Methods ............... 259
Table 2.1. Measured properties of nine RHS overlapped K- (test) connections ................................................ 8
Table 2.2. Average effective weld throat thickness for individual weld elements ........................................... 12
Table 2.3. All-weld-metal tensile coupon test results for RHS truss tests ....................................................... 13
Table 2.4. Actual and predicted nominal weld strength for each test connection (actual-to-predicted
ratio shown in parentheses) ......................................................................................................... 19
Table 2.5. Mean actual-to-predicted electrode strength (ρM) and its associated variation (VM) amongst
typical weld metal ....................................................................................................................... 23
Table 3.1. Measured and calculated section properties for the RHS truss members ....................................... 31
Table 3.2. Mean percentage errors for theoretical (predicted) axial forces ...................................................... 37
Table 3.3. Comparison of experimental and theoretical (predicted) maximum truss deflections .................... 39
Table 4.1. Comparison of fillet weld effective throats required to develop the yield resistance of a
connected RHS branch member wall (updated from McFadden et al. 2013) ............................. 42
Table 4.2. Measured geometric properties, failure loads, and failure modes for θ = 90° RHS-to-rigid
plate specimens ........................................................................................................................... 51
Table 4.3. Measured geometric properties, failure loads, and failure modes for θ = 60° RHS-to-rigid
plate specimens ........................................................................................................................... 52
Table 4.4. Measured geometric properties, failure loads, and failure modes for θ = 90° and θ = 60°
CHS-to-rigid plate specimens ..................................................................................................... 52
Table 4.5. Measured material properties for all 33 HSS-to-rigid plate connections ........................................ 54
Table 5.1. FE mesh sensitivity study results .................................................................................................... 74
Table 5.2. Comparison of experimental and FE results for HSS-to-rigid plate tests ....................................... 77
Table 5.3. Non-dimensional parameters and FE parametric study results (PFE/AwFEXX and failure mode)
for RHS-to-rigid plate connections ............................................................................................. 81
Table 5.4. Non-dimensional parameters and FE parametric study results (PFE/AwFEXX and failure mode)
for CHS-to-rigid plate connections ............................................................................................. 81
Table 5.5. Reliability analysis parameters with respect to 21 weld-critical RHS-to-rigid plate FE test
results with θ = 90°...................................................................................................................... 89
Table 5.6. Reliability analysis parameters with respect to 20 weld-critical CHS-to-rigid plate FE test
results θ = 90°.............................................................................................................................. 89
Table 6.1. Measured properties of 12 CHS X- (test) connections.................................................................... 95
Table 6.2. Weld dimensions and predicted fracture loads for test joints according to existing AWS
D1.1-15 provisions for weld effective lengths in CHS X-connections ..................................... 102
Table 6.3. All-weld-metal tensile coupon test results for CHS X-connection tests ....................................... 103
ix
List of Tables x
Table 6.4. Residual chord deformation (at start of test) and chord deformation at rupture for all 12 tests.... 107
Table 6.5. Reliability analysis parameters for 12 CHS X-Connections ......................................................... 111
Table 6.6. Weld effective length ratios for CHS X-connections determined using three possible
methods ..................................................................................................................................... 117
Table 7.1. Geometric properties of CHS X- (test) connections and comparison of experimental and FE
results ........................................................................................................................................ 124
Table 7.2. Mesh sensitivity study results for test 102-273-90a ...................................................................... 129
Table 7.3. Effect of X-connection model size (scalability) ............................................................................ 132
Table 7.4. Yield stress and ultimate stress of materials used in the parametric models................................. 137
Table 7.5. Reliability analysis parameters for proposed procedure with ϕ = 0.80, 0.75 and 0.67 ................. 144
Table B.17. Mean percentage errors in axial force predictions in test 4.2 ..................................................... 188
Table C.1. List of CHS-to-rigid plate test specimen fabrication drawings .................................................. 193
Table C.2. Weld measurements for CHS-to-rigid plate test no. 22 ............................................................. 197
Table C.3. Weld measurements for CHS-to-rigid plate test no. 23 ............................................................. 198
Table C.4. Weld measurements for CHS-to-rigid plate test no. 24 ............................................................. 198
Table C.5. Weld measurements for CHS-to-rigid plate test no. 25 ............................................................. 199
Table C.6. Weld measurements for CHS-to-rigid plate test no. 26 ............................................................. 199
Table C.7. Weld measurements for CHS-to-rigid plate test no. 27 ............................................................. 200
Table C.8. Weld measurements for CHS-to-rigid plate test no. 28 ............................................................. 200
Table C.9. Weld measurements for CHS-to-rigid plate test no. 29 ............................................................. 201
Table C.10. Weld measurements for CHS-to-rigid plate test no. 30 ............................................................. 201
Table C.11. Weld measurements for CHS-to-rigid plate test no. 31 ............................................................. 202
Table C.12. Weld measurements for CHS-to-rigid plate test no. 32 ............................................................. 202
Table C.13. Weld measurements for CHS-to-rigid plate test no. 33 ............................................................. 203
Table C.14. Application of component approach to CSA S16-01 with and without the sinθ factor for
test no. 33 .................................................................................................................................. 212
Table C.15. Application of component approach to AISC 360 (2016), CSA S16 (2014), and EN1993-
1-8 (2005) for test no. 33 ........................................................................................................... 213
Table C.16. Summary of actual and predicted strengths for all experimental RHS- and CHS-to-rigid
plate connections according to AISC 360-16, CSA S16-14, and CSA S16-01 ......................... 217
Table C.17. Summary of actual and predicted strengths for RHS- and CHS-to-rigid plate connections
according to EN1993-1-8:2005 Directional Method ................................................................. 218
Table D.1. Summary of actual and predicted nominal strengths for 21 FE RHS-to-rigid plate
connections according to AISC 360-16, CSA S16-14, and EN1993-1-8:2005 Directional
and Simplified Methods (θ = 90° connections) ......................................................................... 223
Table D.2. Summary of actual and predicted nominal strengths for 20 FE CHS-to-rigid plate
connections according to AISC 360-16, CSA S16-14, EN1993-1-8:2005 Directional and
Simplified Methods (θ = 90° connections)................................................................................ 224
Table E.1. List of CHS-to-CHS X-connection fabrication drawings ........................................................... 225
Table E.2. Weld measurements for 102-273-90a......................................................................................... 236
Table E.3. Weld measurements for 102-273-90b ........................................................................................ 237
Table E.4. Weld measurements for 102-406-90a......................................................................................... 237
Table E.5. Weld measurements for 102-406-90b ........................................................................................ 238
Table E.6. Weld measurements for 127-273-90a......................................................................................... 238
Table E.7. Weld measurements for 127-273-90b ........................................................................................ 239
Table E.8. Weld measurements for 127-406-90a......................................................................................... 239
Table E.9. Weld measurements for 127-406-90b ........................................................................................ 240
Table E.10. Weld measurements for 102-406-60a......................................................................................... 240
Table E.11. Weld measurements for 102-406-60b ........................................................................................ 241
xiii
List of Figures xiv
Fig. 3.8. Comparison of experimental and theoretical bending moment distributions for test 4.1 ................. 38
Fig. 4.1. Fillet-welded (a) lap-splice, (b) cruciform, and (c) HSS-to-rigid plate test specimens .................... 43
Fig. 4.2. Stress components in the plane of the weld effective throat ............................................................ 49
Fig. 4.3. Connection specimens (with RHS or CHS members) ...................................................................... 50
Fig. 4.4. Examples of fillet weld throat measurements from the macroetch examinations ............................ 53
Fig. 4.5. Typical load-strain curves from four strain gauges on one side of RHS (Group B strain
gauges, shown for test no. 5) ....................................................................................................... 55
Fig. 4.6. Instrumentation and test set-up ......................................................................................................... 56
Fig. 4.7. RHS- and CHS-to-rigid plate specimens at failure .......................................................................... 57
Fig. 4.8. Comparison of actual and nominal strengths per AISC 360-16 without the sinθ factor .................. 58
Fig. 4.9. Comparison of actual and nominal strengths per AISC 360-16 with the sinθ factor ....................... 58
Fig. 4.10. Comparison of actual and nominal strengths ................................................................................... 59
Fig. 4.11. Comparison of actual and nominal strengths per CSA S16-14 with the sinθ factor ........................ 59
Fig. 4.12. Comparison of actual and nominal strengths per CSA S16-01 without the sinθ factor ................... 60
Fig. 4.13. Comparison of actual and nominal strengths per CSA S16-01 with the sinθ factor ........................ 60
Fig. 4.14. Comparison of actual strengths and design strengths per EN1993-1-8 (2005): Directional
Method ........................................................................................................................................ 62
Fig. 4.15. Effect of weld size on actual-to-nominal strength ratio per AISC 360-16 without the sinθ
factor............................................................................................................................................ 64
Fig. 4.16. Effect of branch slenderness on actual-to-nominal strength ratio per AISC 360-16 without
the sinθ factor .............................................................................................................................. 64
Fig. 5.1. HSS-to-rigid plate connection specimens (with RHS or CHS members) ........................................ 66
Fig. 5.2. FE CHS-to-rigid plate connection model with θ = 60° .................................................................... 68
Fig. 5.3. FE RHS-to-rigid plate connection model with θ = 60° .................................................................... 69
Fig. 5.4. Effective weld dimensions in FE models (shown for a model with θ = 90°) ................................... 69
Fig. 5.5. Post-necking generated T- T curve using the Ling (1996) procedure (shown for CHS 127.4 ×
11.55 TC) .................................................................................................................................... 71
Fig. 5.6. FE CHS TC (a) geometry and meshing, and (b) comparison of deformed shape and results at
fracture ........................................................................................................................................ 72
Fig. 5.7. Comparison of numerical and experimental - curves (shown for CHS 127.4 × 11.55 TC).......... 72
Fig. 5.8. Mesh arrangements shown on a 1/4 θ = 90° FE CHS-to-rigid plate connection.............................. 73
Fig. 5.9. Load-displacement curves for CHS test no. 26 with fine, medium and coarse mesh, and four
CHS through-thickness elements ................................................................................................ 74
Fig. 5.10. Progression of fracture in FE RHS-to-rigid plate test no. 13 ........................................................... 78
Fig. 5.11. Comparison of RHS-to-rigid plate FE and experimental load-displacement responses .................. 79
Fig. 5.12. Comparison of CHS-to-rigid plate FE and experimental load-displacement responses .................. 79
Fig. 5.13. Engineering stress-strain curves for the cold-formed HSS, end plate, and weld metal used in
the FE parametric study............................................................................................................... 80
Fig. 5.14. Effect of the ratio tw/tb on fillet weld strength in θ = 90° HSS-to-rigid plate connections: (a)
for RHS branches; (b) for CHS branches .................................................................................... 83
Fig. 5.15. Examples of branch rotation in fillet-welded joints to RHS ............................................................ 84
Fig. 5.16. Effect of the ratios Bb/tb and Db/tb on fillet weld strength in θ = 90° HSS-to-rigid plate
connections: (a) for RHS branches; (b) for CHS branches ......................................................... 84
Fig. 5.17. Effect of θ on fillet weld strength in HSS-to-rigid plate connections .............................................. 85
Fig. 5.18. Evaluation of design equations against all 90° (FE and experimental) HSS-to-rigid plate test
results: (a) for RHS branches; (b) for CHS branches .................................................................. 88
Fig. 6.1. Variation of X-connection stress distribution ................................................................................... 94
Fig. 6.2. Connection layout and nomenclature ............................................................................................... 96
Fig. 6.3. Local dihedral angle curves for test joints, with subtended angle measured from the crown
heel .............................................................................................................................................. 97
Fig. 6.4. Fit-up of branch to chord after profiling and tack welding .............................................................. 98
Fig. 6.5. Vector calculus method used to determine weld lengths ............................................................... 100
Fig. 6.6. 3D Solidworks models of weld profile and weld dimensions ........................................................ 101
Fig. 6.7. Typical testing arrangement (shown for test 127-273-90a)............................................................ 104
Fig. 6.8. Strain gauges near weld toe (and weld fracture) in test 127-273-90a ............................................ 105
Fig. 6.9. Typical weld fractures .................................................................................................................... 106
Fig. 6.10. Typical load versus chord deformation relationships ..................................................................... 107
Fig. 6.11. Procedure for calculation of load versus deformation for second welds tested ............................. 107
Fig. 6.12. Typical strain distributions adjacent to test weld (θ = 90° connections) ........................................ 108
Fig. 6.13. Typical strain distributions adjacent to test weld (θ = 60° connections) ........................................ 108
Fig. 6.14. Effect causing compressive strains at the crown (ρ = 0° and 180° points) (θ = 90°
connections) .............................................................................................................................. 109
Fig. 6.15. Correlation of existing AWS D1.1-15 provisions with the test results, with weld effective
lengths ....................................................................................................................................... 112
Fig. 6.16. Correlation of existing AWS D1.1-15 provisions (excluding weld effective lengths) and
AISC 360-16 provisions with test results .................................................................................. 112
Fig. 6.17. Correlation of CSA S16-14 provisions with test results................................................................. 113
Fig. 6.18. Comparison of lw/πDb using Eq. (6.6) (AWS 2015) and the vector-calculus method .................... 114
Fig. 6.19. Comparison of lw/πDb using Eq. (6.7) (AWS 2015) and the vector-calculus method .................... 115
Fig. 7.1. CHS X-connection general configuration and non-dimensional parameters ................................. 120
Fig. 7.2. Load transfer efficiency across the weld of a θ = 90° CHS-to-CHS T-connection, adapted
from Marshall (1992) ................................................................................................................ 123
Fig. 7.3. FE CHS-to-CHS X-connection geometry, mesh layout, and boundary conditions with θ =
90° ............................................................................................................................................. 125
Fig. 7.4. FE CHS-to-CHS X-connection geometry, mesh layout, and boundary conditions with θ <
90° (θ = 60° shown) .................................................................................................................. 126
Fig. 7.5. Comparison of typical experimental (solid line) and FE (dashed line) stress-strain curves........... 127
Fig. 7.6. Mesh layouts used in the mesh sensitivity study ............................................................................ 128
Fig. 7.7. Comparison of FE and experimental mesh sensitivity parameters for test 102-273-90a ............... 129
Fig. 7.8. Ratios of Paꞌ/PFE and a/ FE for the training set for different values of ef.weld ................................ 131
Fig. 7.9. Comparison of experimental and FE load-deformation curves ...................................................... 133
Fig. 7.10. Correlation of experimental and FE results .................................................................................... 133
Fig. 7.11. Comparison of typical experimental and FE longitudinal strain distributions adjacent to the
weld ........................................................................................................................................... 134
Fig. 7.12. Compatible values of and θ to keep Ѱ between 60° and 120° along the entire weld length,
determined using methodology by Luyties & Post (1988) ........................................................ 136
Fig. 7.13. Engineering stress-strain curves for materials used in the parametric models ............................... 137
Fig. 7.14. Load-deformation curves for typical FE tests ................................................................................ 138
Fig. 7.15. Effect of chord slenderness and branch-to-chord diameter ratio on effective length ..................... 139
Fig. 7.16. Effect of the product (D/t) on effective length ............................................................................. 140
Fig. 7.17. Effect of branch inclination angle on effective length ................................................................... 140
Fig. 7.18. Effect of branch-to-chord thickness ratio on effective length ........................................................ 141
Fig. 7.19. Weld effective lengths in CHS-to-CHS X-connections with 0.10 ≤ ≤ 0.50 and 10 ≤ D/t ≤
50 according to Eq. (7.14) ......................................................................................................... 142
Fig. 7.20. Correlation of proposed design method with all test results .......................................................... 144
Fig. A.1. Truss elevation, joint layout and typical connection detail (drawing T-1) .................................... 160
Fig. A.2. Web members 11-1, 1-4, 4-5 and 5-8 cut details (according to drawing T-1) .............................. 161
Fig. A.3. Web members 8-9, 9-7, 7-6 and 6-3 cut details (according to drawing T-1) ................................ 162
Fig. A.4. Web member 3-2 and 2-12 cut details (according to drawing T-1)............................................... 163
Fig. A.5. Upper reinforcement details .......................................................................................................... 164
Fig. A.6. Lower reinforcement details .......................................................................................................... 165
Fig. A.7. Weld-critical test joints 1, 2, 5, 6 and 9 (according to drawing T-1) ............................................. 166
Fig. A.8. Weld-critical test joints 3, 4, 7 and 8 (according to drawing T-1) ................................................. 167
Fig. A.9. Weld dimensions for (a) fillet weld elements and (b) PJP weld elements .................................... 168
Fig. A.10. TC test results for RHS branch and chord materials ..................................................................... 173
Fig. A.11. All-weld-metal TC test results (E71T-1C electrode)..................................................................... 173
Fig. A.12. Truss load point, reaction, and out-of-plane support ................................................................... 174
Fig. A.13. Weld-critical overlapped K-connections with instrumentation immediately prior to testing........ 175
Fig. A.14. Weld-critical overlapped K-connections with instrumentation immediately after testing ............ 176
Fig. A.15. Distribution of normal strain around branch perimeter for test K-90-0.50a .................................. 177
Fig. A.16. Distribution of normal strain around branch perimeter for test K-30-0.50b ................................. 177
Fig. A.17. Distribution of normal strain around branch perimeter for test K-90-0.71 ................................... 178
Fig. A.18. Distribution of normal strain around branch perimeter for test K-60-0.71a .................................. 178
Fig. A.19. Distribution of normal strain around branch perimeter for test K-60-0.71b ................................. 179
Fig. A.20. Distribution of normal strain around branch perimeter for test K-30-0.71 ................................... 179
Fig. B.1. Truss member and panel point designations used in Appendix B ................................................. 183
Fig. B.2. LVDTs transverse to the chord longitudinal axis (a) along the chord and (b) at the support ........ 183
Fig. B.3. Experimental bending moment distributions for test 4.2 ............................................................... 189
Fig. B.4. Comparison of experimental and theoretical bending moment distributions for test 4.2 .............. 190
Fig. C.1. Stress components in the plane of the weld effective throat for Ѱ = θ = 90° ................................ 191
Fig. C.2. Ninety degree CHS-to-rigid plate test specimens .......................................................................... 194
Fig. C.3. Sixty degree CHS-to-rigid plate test specimens ............................................................................ 195
Fig. C.4. All-weld-metal TC test specimen and trial weld specimens for CHS-to-rigid plate tests ............. 196
Fig. C.5. Skewed-T fillet weld profile dimensions....................................................................................... 204
Fig. C.6. TC test results for CHS branch materials ...................................................................................... 206
Fig. C.7. All-weld-metal TC test results (E71T-1C-H8 electrode) .............................................................. 206
Fig. C.8. TC test results for end-plate material (heat no. 4506365) ............................................................. 206
Fig. C.9. Photographs of test 22 ................................................................................................................... 207
Fig. C.10. Photographs of test 23 ................................................................................................................... 207
Fig. C.11. Photographs of test 24 ................................................................................................................... 208
Fig. C.12. Photographs of test 25 ................................................................................................................... 208
Fig. C.13. Photographs of tests 26 and 27 ...................................................................................................... 209
Fig. C.14. Photographs of test 28 ................................................................................................................... 209
Fig. C.15. Photographs of test 29 ................................................................................................................... 210
Fig. C.16. Photographs of tests 30 and 31 ...................................................................................................... 210
Fig. C.17. Photographs of tests 32 and 33 ...................................................................................................... 211
Fig. D.1. FE CHS-to-rigid plate connection model with θ = 90° ................................................................. 219
Fig. D.2. FE RHS-to-rigid plate connection model with θ = 90° ................................................................. 220
Fig. D.3. Graphical comparison of FE versus predicted ultimate load (showing FE failure modes) ........... 220
Fig. D.4. Non-uniform FE strain distributions adjacent to the weld in RHS-to-rigid plate tests.................. 221
Fig. D.5. Comparison of Pa/Py and PFE/Py to tw/tb for FE and experimental HSS-to-rigid plate tests .......... 222
Fig. E.1. Test 127-406-90a/b specimen ........................................................................................................ 226
Fig. E.2. Test 127-273-90a/b specimen ........................................................................................................ 227
Fig. E.3. Test 102-406-90a/b specimen ........................................................................................................ 228
Fig. E.4. Test 102-273-90a/b specimen ........................................................................................................ 229
Fig. E.5. Test 127-406-60a/b specimen ........................................................................................................ 230
Fig. E.6. Test 102-406-60a/b specimen ........................................................................................................ 231
Fig. E.7. All-weld-metal TC test specimen, trial weld specimens, and parts list for CHS X-connection
tests ............................................................................................................................................ 232
Fig. E.8. CHS-to-CHS template geometry view no. 1 ................................................................................. 233
xix
Abbreviations and Symbols xx
VG = coefficient of variation of ρG
VM = coefficient of variation of ρM
VP = coefficient of variation of ρP
VR = coefficient of variation of ρR
Vrw = design resistance of the fillet weld
Xu = electrode ultimate strength per CSA S16
Z = plastic section modulus of the HSS
a = distance from the chord face of the SGs measured along the RHS web centreline;
regression constant
b = distance from the heel of the SGs measured along the RHS chord centreline;
regression constant
beoi = effective length of the weld to the chord
beov = effective length of the weld to the overlapped branch
d = greatest perpendicular dimension measured from a line flush to the base metal
surface to the weld surface
e = eccentricity in a truss connection, positive being away from the branches;
loading eccentricity of the fillet weld
g = gap between web members at a truss panel point (ignoring welds) on the chord face
hg = length of outside RHS corner radius contour
h = length of inside RHS corner radius contour
i = subscript/term used to identify the overlapping branch member; subscript/term used
to identify weld elements; symbol denoting ρ or ρ + Δρ
j = subscript/term used to identify the overlapped branch member
kFE = finite element initial stiffness
ka = actual (experimental) initial stiffness
l = length of the chord
lb = web (or branch) member length between node points; length of the branch
le = weld effective length
le,i = effective length of weld to the overlapping branch
lh = weld leg measured along the plate; weld leg measured along the HSS chord
lt = template length
lv = weld leg measured along the HSS branch
lw = total weld length
p = projected length of the overlapping branch on the connecting face of the chord
q = overlap length, measured along the connecting face of the chord beneath the region
of overlap of the branches
r = radius of gyration of the HSS
ro = average outside corner radius of RHS
ri = average inside corner radius of RHS
t = wall thickness of the HSS chord member
tb = wall thickness of the HSS branch member
tbi = wall thickness of the overlapping RHS branch member
tbj = wall thickness of the overlapped RHS branch member
tp = plate thickness
tw = weld effective throat dimension
w = weighting factor
x = distance between the collar and the weld toe along the branch
Δmax = maximum truss deflection
Δρ = subtended angle increment
α = chord length parameter (=2l/D)
αR = coefficient of separation (taken to be 0.55)
= ratio of overall branch width to chord width for RHS; ratio of overall branch
diameter to chord diameter for CHS; theoretical angle between the plane of the
effective throat and the fusion face corresponding to lv
w = correlation factor for fillet welds per EN1993-1-8
+
= safety index
= theoretical angle between the plane of the effective throat and the fusion face
corresponding to lh; half diameter-to-thickness ratio of the chord (=D/2t)
M2 = partial safety factor of 1.25 for the resistance of weld in EN1003-1-8:2005
= chord deformation
FE = finite element chord deformation at rupture
a = actual (experimental) chord deformation at rupture
= engineering strain
T = true strain
Tꞌ = true strain at the start of coupon necking
avg = average strain
bottom = in-plane strain on the bottom face of the RHS
e = equivalent (von Mises) strain
ef = equivalent (von Mises) strain at rupture for failure criterion
ef,plate = equivalent (von Mises) strain at rupture for plate failure criterion
ef,weld = equivalent (von Mises) strain at rupture for weld failure criterion
rup = strain at rupture
top = in-plane strain on the top face of the RHS
u = ultimate engineering strain
y = yield strain
θ = angle of loading measured from the weld longitudinal axis for fillet weld strength
calculation (in degrees); branch inclination angle
θ1 = angle of loading (in degrees) of the weld element under consideration
θ2 = angle of loading (in degrees) of the weld element in the joint that is nearest to 90˚
θi = included angle between the overlapping branch and chord
θj = included angle between the overlapped branch and chord
ρ = subtended angle around the branch, measured from heel
ρG = mean ratio of the measured-to-nominal weld throat area
ρM = mean ratio of the measured-to-nominal electrode ultimate strength
ρP = mean test-to-predicted capacity ratio
ρR = bias coefficient for the resistance
= engineering stress
T = true stress
Tꞌ = true stress at the start of coupon necking
u = ultimate engineering stress
𝜎⊥ = normal stress perpendicular to the plane of the throat
𝜎∥ = normal stress parallel to the axis of the weld
= branch-to-chord thickness ratio
𝜏⊥ = shear stress (in the plane of the throat) perpendicular to the axis of the weld
𝜏∥ = shear stress (in the plane of the throat) parallel to the axis of the weld
ϕ = resistance factor (associated with the LRFD or LSD method)
ϕ+ = adjustment factor for +
Ψ = local dihedral angle (angle between the base metal fusion faces)
(a) Retractable stadium roof at the Rogers Centre in (b) Beijing National Stadium in Beijing, China
Toronto, Canada
1
Chapter 1: Introduction 2
A simple and cost-effective way to make connections between HSS members is to weld them directly
together. Directly welded HSS connections prevent the need for additional material, such as gusset plates,
and time-intensive fabrication. For these so-called standard truss connections, smaller HSS branch members
are connected to larger HSS chord members by means of a weld around the perimeter of the intersection of
the two members. These welds experience non-uniform loading as a result of the HSS walls being non-
uniformly flexible. Stiffer elements in the walls, such as corners in a rectangular hollow section (RHS) chord,
attract more load. The stresses in the weld at the location of these stiffer elements are thus higher than the
average stress in the weld, which can lead to premature rupture. In many cases, conventional design equations
for predicting the strength of such welds to HSS have been questioned (Davies & Packer 1982; Frater &
Packer 1992a, 1992b; Packer & Cassidy 1995; McFadden & Packer 2014; Tousignant & Packer 2014).
In 1989, the International Institute of Welding (IIW) recommended that welds to HSS branches be
designed to develop the yield capacity of the member, so that they are capable of resisting, by default, any
arrangement of loads in the member. This remains today as one of two acceptable design methods for welds
between HSS. The second method, recognized by the International Organization for Standardization (ISO
2013), uses weld effective lengths to explicitly take into account the weld non-uniform loading.
No supporting research on weld-critical joints between CHS has hitherto been performed and, as a result,
most specifications are reluctant to provide a “fit-for-purpose” approach to weld design for CHS connections.
Both validation of the AISC 360-10 formulae for weld effective lengths in RHS-to-RHS overlapped K-
connections, and a weld design procedure for CHS-to-CHS connections based on actual branch member
forces, are needed.
1. to complete the recommendations for weld effective lengths in RHS truss connections by
experimentally evaluating the proposed formulae for weld effective lengths in RHS-to-RHS
overlapped K-connections;
2. to evaluate the extent to which fillet welds to RHS and CHS branch members behave differently with
regard to the weld effective length phenomenon, and to evaluate the safety of conventional design
equations in this case; and
3. to determine the effect of non-uniform flexibility on the strength of welds in CHS-to-CHS
connections, and develop an accurate (yet safe) design procedure for fillet welds in such connections.
This research first investigates the validity of the AISC 360-10 formulae for weld effective lengths in
RHS-to-RHS overlapped K-connections (Chapter 2), by testing one large-scale, 10-m span, simply-
supported, Warren truss. Nine overlapped connections, within the truss, were designed to be weld-critical and
sequentially failed by producing an axial force distribution with a point load, applied quasi-statically, to
strategic panel points. The structural reliability (or safety index) of the AISC (2010) specification formulae
for the effective length of welds in RHS overlapped K-connections was determined. The results indicate that
these provisions are conservative. Hence, a modification to the current requirement that limits the effective
width of transverse weld elements is proposed. The proposal establishes a more economical and yet still safe
weld design method for welds in such connections.
Elastic tests were also performed on the truss, using the same loading scheme, to obtain experimental
axial force, bending moment, and truss deflection patterns (Chapter 3). A total of 53 axial forces and bending
moments were measured in the chord and web members during two tests, and deflections at panel points
along the bottom chord were measured during a broader series of nine tests. The theoretical behaviour given
by practical frame analysis models, consisting of: (a) all joints pinned and concentric; (b) all joints rigid and
concentric; (c) pin-ended webs connected eccentrically to continuous chords; and (d) rigid-ended webs
connected eccentrically to continuous chords, is compared to the experimental results. Recommendations are
thus made for the analysis of welded RHS trusses with overlapped K-connections.
This research then examines the structural reliability of the (1.00+0.50sin1.5θ) directional strength-
increase factor for fillet welds used in North America (Chapter 4). Contemporary design procedures for fillet
welds to HSS in several prominent design codes are first discussed. Then, the structural reliability associated
with the (1.00+0.50sin1.5θ) factor is examined through a series of laboratory tests on fillet-welded connections
between HSS and rigid end plates. A total of 33 connection specimens, with the welds designed to be the
critical elements, was tested to failure by application of axial tension loading to the HSS members. The
experimentally obtained weld strengths were compared to the predicted nominal strengths. It was found that
the use of the directional strength-increase factor leads to unsafe strength predictions, particularly for large
weld sizes. Hence, it is determined that restrictions on the application of the (1.00+0.50sin1.5θ) factor for
welds to HSS members need to be considered. It was also found that Eurocode 3 fillet weld design provisions
give conservative strength predictions.
To extend these results, Chapter 5 presents a finite element (FE) investigation on the behaviour of fillet-
welded HSS-to-rigid plate connections. The FE models are validated by comparison of load-displacement
response and fracture load with results from the previous 33 experimental tests, and a parametric study is
performed. Seventy-three FE tests are therein conducted to evaluate the effect of weld size, HSS branch wall
slenderness and branch inclination angle, on fillet weld strength. The inherent problem with single-sided fillet
welds to a tension-loaded element is illustrated. A reliability analysis is then performed, and it is determined
that the (1.00 + 0.50sin1.5θ) is unsafe for both RHS and CHS connections, when considered separately,
especially for connections with large welds. Hence, an alternative yet safe method for estimating the strength
of fillet welds to HSS, based on weld size, is proposed. An expression for the fillet weld size required to
develop the yield strength of a 90° HSS branch member is derived.
Chapter 6 presents the first-ever laboratory-based test program conducted to assess the performance of
fillet welds in connections to a CHS chord. Six large-scale, fillet-welded CHS-to-CHS X-connections,
designed to be weld-critical, were tested under quasi-static tension applied to the ends of each branch.
Fracture of 12 test welds (two per connection) was obtained. By using the carefully measured mechanical and
geometric properties of the fillet welds and CHS members, the safety index of the existing AWS (2015)
specification provisions for weld effective lengths in CHS-to-CHS X-connections was determined. For the
range of parameters studied, the AWS formulae are found to be very conservative, hence a comprehensive
parametric FE modelling study was conducted to develop more liberal recommendations (Chapter 7).
In Chapter 7, non-linear FE models of CHS-to-CHS X-connections, with weld fracture, are validated by
comparison of spot strains, load-deformation response, and fracture load with the previous 12 experimental
tests. Two hundred and fifty-six weld-critical connections, with varied branch-to-chord diameter ratio, chord
wall slenderness, branch inclination angle, and branch-to-chord thickness ratio, are then analysed under
quasi-static branch tension. The effect of these parameters on fillet weld strength is illustrated, and the safety
index of North American specification provisions for weld effective lengths is confirmed. An alternative
method for estimating fillet weld strength, with specific weld effective lengths, is proposed.
Recommendations for a new design approach that meets the minimum target safety index in North America
are made.
Chapter 8 presents a summary of this research, its impact on engineering design practice, and
recommendations for future research.
2.1. INTRODUCTION
It is well known that the differences in relative stiffness of rectangular hollow structural section (HSS)
walls cause non-uniform load transfer along lines of welds at a branch connection. Historically, international
design recommendations have required that these welds be designed to develop the yield strength of the
member, such that they may resist any arrangement of loads in the branch. This requirement is almost
exclusively based on old recommendations from the International Institute of Welding (IIW 1989). Designing
welds to branches to develop the yield strength of the member is justifiable in situations when there is low
confidence in the design forces, or if plastic stress redistribution is required in the connection (Packer et al.
2009). This design method is not always merited, and its requirement for large weld sizes is excessively
conservative in many situations.
Extensive laboratory tests have been performed at the University of Toronto on welds in both isolated
rectangular hollow section (RHS) connections and complete trusses (Frater & Packer 1992a, 1992b; Packer &
Cassidy 1995; McFadden et al. 2013; McFadden & Packer 2014) which have led to the development and
international recognition (IIW 2012; ISO 2013) of a more modern design approach based on actual branch
member forces to achieve more appropriate and economical weld sizes. This so-called “fit-for-purpose”
approach makes use of effective weld properties to account for the non-uniform loading of the weld
perimeter.
In the 2010 edition of AISC Specification for Structural Steel Buildings (AISC 2010), a detailed design
method considering effective weld properties for predominantly statically loaded RHS-to-RHS connections
was given in Section K4: Welds of Plates and Branches to Rectangular HSS. Table K4.1 Effective Weld
Properties for Connections to Rectangular HSS contained formulae to determine the effective length of welds
for axially-loaded RHS connections and the effective elastic section modulus of welds subject to bending.
The design methods in Table K4.1 for welds in axially-loaded T-, Y-, X- and gapped K-connections were
based on experimental data from full-scale tests on connections in which failure occurred by shear rupture
along a plane through the weld (herein called “weld-critical connections”) (Frater & Packer 1992a, 1992b;
Packer & Cassidy 1995). However, at the time that AISC 360-10 was published, no such data was available
6
Chapter 2: Weld Effective Lengths in RHS Overlapped K-Connections 7
to substantiate the design methods given, in the same document, for welds in unreinforced RHS moment T-
connections and axially-loaded RHS overlapped K-connections.
In order to evaluate the adequacy of these design methods, AISC initiated a two-phase study at the
University of Toronto. The first phase of the study investigated the strength and behaviour of welds in
unreinforced RHS moment T-connections. The results of this phase have been published by McFadden &
Packer (2014). Phase two of the study is presented herein.
2.2. EXPERIMENTATION
2.2.1. SCOPE
An experimental program was developed to test large-scale RHS overlapped K-connections in order to
verify, or adjust, the weld effective length rules defined by Equations K4-10 to K4-12 in Table K4.1 of AISC
360 (2010). Nine overlapped, 60° K-connections within one large-scale, 10-m span, simply-supported
Warren truss, were designed to be weld-critical under the application of tension to the overlapping branch.
Key parameters, such as the branch member overlap (Ov), the branch-to-chord width ratio ( -ratio) and the
chord wall slenderness (B/t), were investigated, and varied within the Limits of Applicability of Section K2.3
of the Specification (AISC 2010). The non-uniform distribution of normal strain in the branch, near the
connection, was measured with strain gauges oriented along the longitudinal axis of the member at uniform
spacing around its perimeter, and the weld strength was obtained directly from strain gauges in the constant
stress region of the branch. To induce weld rupture, a single point load was applied to various truss panel
points in a quasi-static manner. The loading strategy was carefully planned to accentuate the force in the
critical web member(s) and resulted in all nine joints failing by shear rupture along a plane through the weld.
2.2.2.1. General
Connections were welded using a semi-automatic flux-cored-arc-welding (FCAW) process with full-CO2
shielding gas, and fabricated from members conforming to CSA G40.20/G40.21 Class C (CSA 2013a) and
ASTM A1085 (ASTM 2015a). The experimental test designations, and a summary of the key test parameters
for each joint (Ov, , and B/t), are given in Table 2.1.
The amount of overlap was varied, from 30% to 90%, and chord member sections that were selected
produced relatively rigid and flexible connections. Connections were made to an HSS 178 × 178 × 13 that
were more rigid ( = 0.71 and B/t = 14.2), and to an HSS 254 × 254 × 9.5 chord were more flexible ( = 0.50
and B/t = 27.5). Web members (HSS 127 × 127 × 8.0) were specified to minimize the ratio: predicted weld
strength/connection resistance and to also allow, by virtue of matched-width web members, either side of the
truss connection to be designated as the overlapping (or “test”) branch. The latter detail was intended to
support the design of a loading sequence to achieve sequential rupture of welds within the truss (see Section
2.4.2).
Complete truss testing has been the preferred approach for testing welds in K-connections because it
correctly accounts for connection boundary conditions (i.e. member continuity and truss deflection effects)
(Frater & Packer 1992a, 1992b, 1992c). The truss layout and its dimensions are shown in Fig. 2.1. In Fig. 2.1,
load locations for the nine tests are numbered around the truss and connection numbers (and designations) are
shown inside of the truss. The fabrication drawings for the truss, in U.S. customary units (as they were
provided to the fabricator), are provided in Appendix A.1.
Fig. 2.2. Terminology for RHS overlapped K-connections and weld details (including labelling convention)
for test joints
All critical test welds (to the overlapping branches at the connections), with the exception of weld
element c, were performed in the horizontal position. Weld element c was performed in the flat position.
Minimum weld sizes, as specified in Table 5.7 and Table 3.5 of AWS D1.1 (2015) and Table J2.4 and Table
J2.3 of AISC 360 (2016) for fillet welds and PJP flare-bevel-groove welds, respectively, were used to ensure
enough heat input during welding to establish a sound weld. The “hidden toe” of the overlapped branch
(beneath c in Fig. 2.2) was always welded to the chord, and the remainder of the welds in the truss were sized
so as to not fail before yielding of the attached branch member. Welding (or not welding) of the hidden toe of
the overlapped (or through) branch is believed to be an issue that affects failure of the welded joint to the
overlapped branch (when it is loaded in tension). Since the experiments herein were designed to examine
welding of the tension-loaded overlapping branch, it was decided to adopt the conservative practice of fully
welding the non-critical through branch. Section 2.7.4 provides further comments on this matter. Fig. 2.3
shows the specified weld sizes and the associated welding symbols in a typical connection detail for a joint
with Ov = 30%.
Fig. 2.3. Typical connection detail drawing (shown for test K-30-0.50a or K-30-0.50b)
During fabrication, the backing bar in detail a’ of K-60-0.50 was pried, about the tack weld (see Fig. 2.2),
away from the inside face of the HSS branch. The member itself was the last one to be fitted into the truss,
and was hammered into place. It is believed that during hammering, catching of the backing bar, which made
contact with the chord, caused it to be pried. This complication was not identified until after welding of the
opposite side (details b and b’) was complete. The resulting “gap” was filled with weld metal and welding of
the test joint proceeded. Based on non-destructive (ultrasound) test results, this was not a cause for rejection;
but it is speculated that the strength of the joint was nevertheless affected by this defect. The joint is
identified in the subsequent analysis (Figs. 2.13 – 2.18) by a red data point.
All of the test welds were ground (long after welding) to reduce the weld throat dimension to below the
minimum sizes specified by AISC and AWS. This was necessary to obtain a weld-fracture failure mode.
Since the code provisions are based on achieving the necessary heat input at the time of welding, and
minimum sizes were at that time provided, the soundness of the welds was likely unaffected.
tw tbi d (2.1)
where d = greatest perpendicular dimension measured from a line flush to the overlapping branch member
surface to the weld surface; and tbi = average measured thickness of the overlapping branch member.
For the fillet welds, tw was determined by making a negative mould of each fillet-weld element at
numerous locations along its length, as shown in Fig. 2.4. The mould was cut normal to the axis of the weld
root, scanned, and digitally measured; the effective throat was taken as the minimum distance between the
root and face of the diagrammatic weld. Fig. 2.4 also shows a cut of the mould, and a typical weld throat
measurement. Over 180 weld dimensions were taken for the nine connections (five along each of the four
sides of the connection), and the average measured values for the weld throat dimension are shown in Table
2.2. The individual measurements of weld dimensions for the fillet and PJP welds are provided in Appendix
A.2.
Table 2.2. Average effective weld throat thickness for individual weld elements
Test Measured weld throat dimension, tw
a a’ b b’ c d
mm mm mm mm mm mm
K-90-0.50a 3.45 3.18 3.12 3.58 3.76 4.27
K-90-0.50b 4.60 3.45 3.81 3.66 3.84 4.27
K-60-0.50 2.67 3.81 2.39 3.56 3.86 4.22
K-30-0.50a 3.35 4.60 2.95 3.96 4.34 3.63
K-30-0.50b 3.28 4.29 3.05 4.29 4.01 3.53
K-90-0.71 3.18 3.89 3.18 3.81 3.63 3.84
K-60-0.71a 3.99 3.51 3.86 3.12 3.78 3.86
K-60-0.71b 3.43 3.56 3.23 3.76 3.84 3.96
K-30-0.71 4.57 4.93 3.40 4.78 4.27 3.78
T2 joint 4 † 4.50 4.50 4.39 4.39 7.19 6.71
T2 joint 6 † 7.11 7.11 6.50 6.50 9.09 10.59
Note: b and b’ are analogous to a and a’ (see Fig. 2.2), but on the opposite side of the overlapping branch.
†
Tests by Frater (1991).
Mechanical properties of the as-laid welds were determined by tensile coupon (TC) tests (three total) as
specified by AWS D1.1 (2015). The results are shown in Table 2.3. The average yield stress (by 0.2% strain
offset) was 563 MPa and the average ultimate strength (FEXX) was 619 MPa with 27.5% elongation at rupture.
The measured ultimate strength was 26.3% stronger than the nominal strength of the electrode used (AWS
E71T-1C). Stress-strain curves generated from the all-weld metal TC tests are provided in Appendix A.3.
Table 2.3. All-weld-metal tensile coupon test results for RHS truss tests
All-weld-metal coupon Yield stress Young’s modulus, E FEXX Rupture strain, εrup a
designation
MPa MPa MPa %
[i] 558 202000 629 27.0
[ii] 561 200000 612 26.4
[iii] 567 219000 617 29.2
Average 563 207000 619 27.5
a
Rupture strain determined by re-joining the fractured coupon and measuring: change in gauge length / initial gauge
length
Pa Ab E avg (2.2)
where Ab = cross-sectional area of the branch, determined by weighing the cross-section; E = elastic modulus
of the RHS, determined by tensile coupon tests in accordance with ASTM A370 (ASTM 2017); and ɛavg =
average strain measured on opposite faces of the RHS.
The non-uniform normal strain distribution around the branch perimeter, adjacent to the weld, was
measured using SGs oriented along the longitudinal axis of the member, and 25 mm away from the weld toe
[in order to avoid the strain concentrations caused by the notch effect (Packer & Cassidy 1995)]. Since the
strain distribution is theoretically symmetric about the y-y axis of the member (for plane-frame behaviour),
SGs were only installed on half of the member (along Hbi and half of Bbi on two sides), with an additional SG
at mid-height on the opposite side to confirm the assumption of symmetry. The SG spacing is shown in Fig.
2.5.
Fig. 2.5. Spacing of strain gauges around the branch footprint adjacent to the welded connection
Fig. 2.6. Laboratory testing arrangement for full-scale RHS overlapped K-connection experiments
2.5. RESULTS
All of the test welds failed in a brittle manner, by fracture along a plane through the weld, which occurred
almost simultaneously at all locations around the branch perimeter. Failure was sudden, and accompanied by
a dynamic “jolt” (caused by the release of strain energy) which, in some tests, displaced SGs from the branch
member surface (see Fig. 2.7a).
Fig. 2.8 shows the relationship between the applied MTS load and the load in the branch member
(measured by SGs in the position shown) for test K-30-0.50a. By virtue of a constant slope (indicating a
linear variation in average strain), it can be seen that the member itself remained elastic throughout the entire
load range. The branch load at rupture was hence calculated using Eq. 2.2.
Fig. 2.8. Typical MTS load versus branch load magnitude relationship (shown for test K-30-0.50a)
Figs. 2.9 - 2.11 show the variation in strain measured by 13 different SGs adjacent to the test welds at the
initial unloaded stage and at 50%, 80% and 100% of the weld rupture load for three typical test welds. The
variations in strain for the remaining test welds are provided in Appendix A.6.
For the joints tested, it was found that the magnitude of strain decreased as a function of the distance
from the toe of the connection – believed to be caused by differences in the relative stiffness of the chord ( =
0.50 and B/t = 27.5) and the overlapped branch ( = 1.00 and Bbj/tbj = 16.3), that results in the latter attracting
more load. As Ov increases, this change becomes less pronounced and is accompanied, generally, by a higher
average failure stress in the weld. The magnitude of strain along the branch transverse faces is seen to
decrease towards the mid-wall locations (SGs 1 and 13) except for the final stage of stress redistribution
(Figs. 2.9 and 2.11). This variation is expectedly more pronounced when the branch lands on a flexible chord
( = 0.50, B/t = 27.5), and for low values of Ov. The less sudden change in strain approaching the mid-wall
along the toe (in Figs 2.9 - 2.11) is due to the more uniform transverse stiffness of the overlapped branch ( =
1.00 and B/t = 16.3).
In Fig. 2.9, some of the weld to the heel actually remains in compression for the entire load range (branch
in tension). The strain at the toe in Fig. 2.9 may have been exacerbated by the proximity of the weld to the
hidden toe of the overlapped branch which was itself welded and thus increased the stiffness of the
connection and the weld effective length at this location.
Fig. 2.9. Typical distribution of normal strain around branch perimeter for specimens with Ov = 30%, =
0.50 and B/t = 27.5
Fig. 2.10. Typical distribution of normal strain around branch perimeter for specimens with Ov = 60%, =
0.50 and B/t = 27.5
Fig. 2.11. Typical distribution of normal strain around branch perimeter for specimens with Ov = 90%, =
0.50 and B/t = 27.5
The distributions of strain around the branch members adjacent to the test welds show that longitudinal
welds to overlapped K-connections can be regarded as completely effective at resisting the applied load when
Ov = 60% and 90%. For Ov = 30% (Fig. 2.9), the strain along the longitudinal weld (SGs 4 - 10) can be seen
to be more non-uniform. The transverse welds are always only partially effective and generally become less
effective as the -ratio decreases, as Ov decreases, and as B/t increases. These trends are verified by the actual
rupture loads (Pa), which are given in Table 2.4.
Table 2.4. Actual and predicted nominal weld strength for each test connection (actual-to-predicted ratio
shown in parentheses)
Pnw
kN
Test Pa AISC 360-10 Modified AISC 360-10 Without weld effective
kN lengths
K-90-0.50a 1232 832 (1.48) 876 (1.41) 965 (1.28)
K-90-0.50b 1277 872 (1.46) 916 (1.39) 1010 (1.26)
K-60-0.50 * 596 614 (0.97) 685 (0.87) 894 (0.67)
K-30-0.50a 765 383 (2.00) 454 (1.69) 850 (0.90)
K-30-0.50b 738 378 (1.95) 449 (1.64) 832 (0.89)
K-90-0.71 1139 885 (1.29) 930 (1.22) 1014 (1.12)
K-60-0.71a 974 649 (1.50) 712 (1.37) 894 (1.09)
K-60-0.71b 863 663 (1.30) 725 (1.19) 912 (0.95)
K-30-0.71 1054 463 (2.28) 529 (1.99) 947 (1.11)
T2 joint 4 † 1686 778 (2.17) 854 (1.97) 1143 (1.48)
T2 joint 6 † 1668 1165 (1.43) 1272 (1.31) 1681 (0.99)
Note: Underlined values are strength predictions which exceed the measured strength (i.e. nominally unsafe).
*
Imperfect weld root detail (see 2.2.2.2).
†
Tests by Frater (1991); failed by a combined mechanism of weld fracture and premature branch yielding.
where Fnw = nominal strength of weld metal; tw = weld effective throat around the perimeter of the branch;
and le = effective length of fillet and groove welds. An LRFD resistance factor, ϕ, equal to 0.75 and 0.80,
applies for fillet welds and PJP groove welds, respectively.
In Table J2.5 of AISC 360-10, Fnw is specified as 0.60FEXX for both fillet and PJP groove welds. In the
case of the former, it implies that the failure mode is by shear rupture on the effective throat; however, for
PJP groove welds (sides a’ and b’), it is an arbitrary reduction factor that has been in effect since the early
1960s to compensate for the notch effect of the unfused area of the joint and does not imply that the tensile
failure mode is by shear stress on the effective throat (per AISC 360-10 Specification Commentary to
Chapter J). Since a CP detail was provided (see Section 2.2.2.2) in order to establish a high degree of
certainty with respect to the fusion area (and the weld throat dimension) in this region, a more suitable term
of 1.00FEXX has been used herein for Fnw for groove welds.
The formulae for le are given in Table K4.1 of AISC 360 (2010) and are as follows:
O H Ov
1 v bi
20v H bi
le,i beoi beov (2.4)
50 100 sin i
100 sin i j
O H O H bi
le,i 2 1 v bi v beoi beov (2.5)
100 sin i 100 sin i j
O H Ov H bi
le,i 2 1 v bi Bbi beov (2.6)
100 sin i 100 sin i j
where i = subscript used to refer to the overlapping branch; j = subscript used to refer to the overlapped
branch; Hb = overall height of the branch member measured in the plane of the connection; and θ = included
angle between the branch and the chord (= 60° for all test connections). The weld effective length dimensions
when 50% ≤ Ov are shown in Fig. 2.12.
The terms beoi and beov are empirically derived from laboratory tests in the 1960s and 1970s (Davies &
Packer 1982) and quantify the effective widths of weld to the branch face, normal (transverse) to the plane of
the connection:
10 Fy t
beoi Bbi Bbi (2.7)
B / t Fybi tbi
10 Fybj tbj
beov Bbi Bbi (2.8)
Bbj / tbj Fybi tbi
where B = overall width of the RHS chord, normal to the plane of the connection; Bb = overall width of the
RHS branch, normal to the plane of the connection; t = wall thicknesses of the chord; tb = wall thicknesses of
the branch; Fy = yield stress of the chord; and Fyb = yield stress of the branch.
AISC 360 (2010) also limits the values of beoi/2 and beov/2 through a notwithstanding clause, which states,
“When Bbi/B > 0.85 or θi > 50°, beoi/2 shall not exceed 2t and when Bbi/Bbj > 0.85 or (180° – θi – θj) > 50°,
beov/2 shall not exceed 2tbj.” Thus, for the RHS overlapped K-connections tested, the upper limits of beoi = 4t
and beov = 4tbj apply.
R exp R VR
(2.9)
where αR = coefficient of separation taken as 0.55 (Ravindra & Galambos 1978); ρR = bias coefficient for
resistance; VR = associated coefficient of variation (COV) of ρR; and ϕ +
= adjustment factor for +
that is
needed when +
≠ 3.0 (Fisher et al. 1978). The bias coefficient for resistance (ρR) and its associate COV (VR)
are given by Eqs. (2.10) and (2.11):
R M G P (2.10)
where ρM = mean ratio of actual-to-nominal electrode strength; ρG = mean ratio of actual-to-nominal weld
throat area; ρP = mean test-to-predicted capacity ratio (with predicted capacity calculated using actual
measured properties); and VM, VG, and VP = COV of ρM, ρG, and ρP, respectively.
A formula to calculate ϕ + based on the reliability index ( +) was derived by Franchuk et al. (2002):
The mean actual-to-nominal electrode strength (ρM) and its associated COV (VM) were taken from a
database of 708 tests summarized in Table 2.5. The data from recent University of Toronto test programs in
Table 2.5 are average values from tensile coupon tests done by McFadden & Packer (2014), those presented
earlier in this Chapter (recall that the measured electrode strength was 26.3% stronger than the specified
nominal strength), and those presented later in this thesis, in Chapters 4 and 7. The composite mean and COV
of all test data were used for ρM and VM, respectively.
The mean measured-to-nominal weld throat area (ρG) and its associated COV (VG) were taken as 1.03 and
0.10, respectively, as justified by Callele et al. (2009). These factors account for the fact that larger weld
throats are typically provided by convexity of the weld face.
The mean test-to-predicted capacity ratio (ρP) was taken as the average over all tests of actual
(experimental) strength divided by the predicted strength, calculated using the measured values of tw, the
weld lengths, and FEXX.
Table 2.5. Mean actual-to-predicted electrode strength (ρM) and its associated variation (VM) amongst typical
weld metal
Study Number of tests ρM VM
Lesik & Kennedy (1990) 672 1.12 0.077
Callele et al. (2009) 32 1.15 0.080
Recent University of Toronto test programs (including
4 1.21 0.039
current work) a
Composite/total values: 708 1.12 0.121
a
nominal electrode strength is assumed to be 490 MPa for all electrodes tested.
In the evaluation that follows, correlation plots are produced using the measured ultimate weld strengths
(rupture loads) from the nine tests (including test K-60-0.50, with the imperfect root detail), and the results
from two similar connection tests that were conducted at the University of Toronto (the details of which
appear at the bottom of Table 2.1) (Frater 1991).
The implied resistance factor, ϕ, is equal to 0.883 for the AISC (2010) specification provisions and is
larger than the necessary resistance factors for fillet welds and PJP groove welds (0.75 and 0.80, respectively)
indicating an excessive level of safety for the AISC formulae. Fig. 2.13 shows the correlation of the predicted
nominal strengths with the experimental results.
2.7. RECOMMENDATION
2.7.1. BACKGROUND
By means of 12 full-scale experiments on isolated T-connections, conducted during phase one of the
research program, excessive safety was found to exist in the AISC 360 (2010) formula for the effective
elastic section modulus for in-plane bending for RHS moment T-connections (McFadden & Packer 2014).
The authors proposed a change to the current requirement that restricts the effective widths of welds to the
branch face from two times the chord wall thickness (2t) to a more reasonable limit of Bb/4. Their proposal
increases the effective length of the transverse weld elements in most RHS connections and was shown to
also be applicable to the formulae for the effective length of welds in axially-loaded RHS T- and X- (or
Cross) connections.
2.7.2. PROPOSAL
Since the same pattern is observed for RHS overlapped K-connections, it is proposed that the existing
formulae for the effective length of welds be modified in the same manner, by changing the requirement:
“When Bbi/B > 0.85 or θi > 50°, beoi/2 shall not exceed 2t and when Bbi/Bbj > 0.85 or (180° – θi – θj) > 50°,
beov/2 shall not exceed 2tbj.”
to:
“When Bbi/B > 0.85 or θi > 50°, beoi/2 shall not exceed Bbi/4 and when Bbi/Bbj > 0.85 or (180° – θi – θj) >
50°, beov/2 shall not exceed Bbi/4.”
This change produces the correlation with the test data given by Fig. 2.14.
Fig. 2.13. Correlation with all test results for AISC Fig. 2.14. Correlation with all test results for modified
360-10 provisions AISC 360-10 provisions
Fig. 2.15. Correlation with all test results without using weld effective length rules
2.7.4. COMMENTS
It is worth noting that if no effective length rules are applied, and the total weld length is used to
determine the strength of the welded joint to the overlapping branch, then the correlation with the test data
shown in Fig. 2.15 results. The implied resistance factor, ϕ, is equal to 0.564 which is less than the necessary
resistance factors for fillet and PJP welds, illustrating that such an approach provides an insufficient safety
margin. If historical tests (by Frater 1991) are omitted from the analysis, a marginal reduction to the inherent
safety factors is found; however, the previous discussion still applies, and the recommendation is found to be
safe. Correlations to this effect are given in Figs 2.16 – 2.18. All of the above still conservatively considers
test K-60-0.50 in the analysis.
Fig. 2.16. Correlation with current test results for Fig. 2.17. Correlation with current test results for
AISC 360-10 provisions modified AISC 360-10 provisions
Fig. 2.18. Correlation with current test results without using weld effective length rules
If the hidden toe of the overlapped branch was not welded, a smaller effective length would likely result
at the toe of the overlapping branch (beov) in connections with Ov = 30% since the restraint to transverse
deformation (stiffness) would be lost. This stands to reduce the mean test-to-predicted capacity ratio (ρP) in
Eq. (2.10), and hence reduce ρR in Eq. (2.9); however, by virtue of the test-to-predicted capacity ratio being
already higher for connections with Ov = 30% relative to the other connections, there would be some
counteracting decrease in VP, and hence VR, and thus a minimal effect on the reliability of the proposed
changes.
2.8. SUMMARY
Based on the results from nine full-scale tests on weld-critical RHS overlapped K-connections, and the
measured strength of two overlapped K-connection tests from a previous experimental program (Frater
1991), it has been found that:
1. The distribution of normal strain adjacent to the welded joint in RHS overlapped K-connections
is highly non-uniform.
2. As the overlap increases, stiffening the joint, the distribution of normal strain adjacent to the
welded joint becomes more uniform.
3. Transverse welds are only partially effective and generally become less effective as the -ratio
decreases, as Ov decreases, and as B/t (of the landing surface) increases.
The effective length rules defined by Equations K4-10 to K4-12 and given in Table K4.1 of AISC 360
(2010) for welds in RHS overlapped K-connections are quite conservative. It is recommended to modify the
requirement (AISC 360-10):
“When Bbi/B > 0.85 or θi > 50°, beoi/2 shall not exceed 2t and when Bbi/Bbj > 0.85 or (180° – θi – θj) > 50°,
beov/2 shall not exceed 2tbj.”
to:
“When Bbi/B > 0.85 or θi > 50°, beoi/2 shall not exceed Bbi/4 and when Bbi/Bbj > 0.85 or (180° – θi – θj) >
50°, beov/2 shall not exceed Bbi/4.”
to increase the predicted strength of welded joints in RHS overlapped K-connections. This modification is
adopted from McFadden & Packer (2014) and has been shown to still be conservative yet generally provide a
more economical design approach for RHS T-, Y- and X- (or Cross-) connections subject to branch axial load
or branch bending. Using this recommendation would thus establish consistent rules across AISC 360 for the
design of welded truss connections between RHS.
2.8.1. FOOTNOTE
The above recommendation was duly implemented in Table K5.1 of AISC 360-16.
3.1. INTRODUCTION
3.1.1. RECOMMENDATIONS FOR RHS TRUSS ANALYSIS
International (Packer et al. 2009) and Canadian (Packer & Henderson 1997) design guides recommend
that the static design of standard (e.g., Warren and Pratt) RHS trusses, having either gapped or overlapped
connections (Figs. 3.1a,b), be carried out on the basis of a force distribution obtained from an elastic analysis
of the truss assuming either: (a) all of the members are pin-connected, or (b) web members are pin-connected
to continuous chords, with the distance between the intersection of the web and chord centrelines (+e or –e in
Figs. 3.1a,b) modelled as an extremely stiff, or rigid, link.
Fig. 3.1. RHS (a) overlapped and (b) gapped K-connections, adapted from Packer & Henderson (1997)
Moments resulting from connection noding eccentricity (e) must be taken into account when designing
the chords (as beam-columns). Using method (a), these “eccentricity moments” can be easily overlooked,
since it is necessary to calculate and distribute the total panel point moment (i.e. the sum of horizontal web
member force components × e) to the chord (which is assumed to primarily resist these moments) in
28
Chapter 3: Analysis of RHS Trusses with Overlapped K-Connections 29
proportion to the chord stiffness on either side of the connection. This process is automated using method (b),
which was initially suggested in the 1980s with the popularity of personal computers (Packer & Davies 1989;
Wardenier & Packer 1992; Packer & Henderson 1992; Packer et al. 1992).
demonstrating it to be a suitable analysis method for overlap-jointed trusses. This Chapter evaluates several
truss analysis models that are available for obtaining axial force and bending moment distributions, including
(a) and (b), for the design of overlap-jointed RHS trusses.
3.2. EXPERIMENTATION
A ten-metre span, simply-supported, overlap-jointed Warren truss (Fig. 3.2) (the same truss as in Chapter
2), with web members at 60° to the chord, was designed and fabricated using RHS with a minimum nominal
yield stress, Fy = 345 MPa. The truss chord members were comprised of two RHS sections, as noted in Table
3.1. Matched web members were overlapped at their connection to the chord, with 30% ≤ Ov ≤ 90% (see Fig.
3.1), and produced noding eccentricities (e) that ranged from 0 to -0.45 times the chord height (H). All of the
connections were designed within the parametric range of validity given by Packer & Henderson (1997).
The truss was tested under “service-load conditions” by applying a single quasi-static point load with a
universal test frame to one of seven top chord panel points (bold numbers 4 to 10 in Fig. 3.2). This was
facilitated by translating, rotating, or inverting the truss to invert the arrangement shown in Fig. 3.2. The
compression chord was braced against lateral instability by two frames that were post-tensioned to the
laboratory strong-floor, and supported out-of-plane by polytetrafluoroethylene (PTFE) plates. The PTFE
plates allowed vertical movement of the truss within the set-up, and two rollers provided the vertical support
to the truss at its ends. Centreline dimensions of the truss and e values at the connections are shown in Fig.
3.2. Note that the labelling convention in this figure bares no relation to that given in Fig. 2.1 for the weld
tests.
The actual (mechanical and geometrical) properties of the chord and web members were determined, for
use later in the analysis, as follows: the cross-sectional areas (A for the chord, and Ab for the webs) were
determined by cutting a prescribed length of the RHS, weighing it, and then using a density of 7850 kg/m3
(CISC 2016) to calculate the area; the Young’s modulus (E) was determined from tensile coupon tests in
accordance with ASTM A370 (2017); and the moment of inertia (I) was determined from formulae given in
Annex B of EN10219-2 (CEN 2006), using the measured widths, heights, thicknesses, and corner radii of the
RHS members. The results of the tensile coupon tests for the RHS truss members are provided in Appendix
A.3. The measured and calculated section properties for the members are shown in Table 3.1. Additional
measurements of the RHS, and the EN10219-2 (CEN 2006) section property formulae, are provided in
Appendices B.1 and B.2, respectively.
Table 3.1. Measured and calculated section properties for the RHS truss members
A or Ab E I
RHS Designation mm2 × 103 MPa × 106 mm4
HSS 127 × 127 × 8.0 (webs 1 – 10) 3626 193.0 8.25
HSS 178 × 178 × 13 (top chord 11 – 15; posts 22 and 23) 7776 193.0 33.9
HSS 254 × 254 × 9.5 (bottom chord 16 – 21) 8804 191.7 85.6
Instrumentation consisting of electrical resistance strain gauges (SGs) and linearly varying differential
transformers (LVDTs) was used to record the member axial forces and bending moments, and truss
deflection patterns. A total of 106 SGs (shown by symbols in Fig. 3.3) were installed on the top and bottom
of each member, centred in the plane of the truss, and oriented along the longitudinal axis of the members to
measure the in-plane strain profiles. Gauges were located (in pairs) at mid-length of each member and at a
distance from the connections (as shown in Fig. 3.3). LVDTs were placed along the underside of the bottom
chord at every panel point, and inside of the chord at corner joints (1 and 13, or 2 and 12, in Fig. 3.2). This
entailed a total of seven LVDTs when the truss was upright, and a total of six LVDTs when the truss was
inverted (see Fig. 3.3). Photographs of the LVDT orientations are provided in Appendix B.3. Two load cells
situated below the rollers were used to measure the reaction forces during the tests. These load cells are
apparent in the photographs of Appendix A.4. An overall view of the arrangement is shown in Fig. 3.4.
3.3. RESULTS
The SG data and the actual member properties (Table 3.1) were used to determine the experimental axial
force (Pe) and bending moment (Me) at each pair of SGs in the truss, from strain distributions such as the one
shown in Fig. 3.5. According to the SG output, the truss remained linear elastic. Hence, Pe and Me for strain
distributions such as the one shown in Fig. 3.5 were calculated from Eqs. (3.1) and (3.2).
Fig. 3.5. Potential measured in-plane strain distribution for truss members
top bottom
Pe E A (3.1)
2
top - bottom
Me E I (3.2)
H
For a test designated as 4.1, the point load was applied to panel point 4 (Fig. 3.2), and the experimental
axial force and bending moment distributions at the peak load are shown in Fig. 3.6. A similar test, 4.2, gave
comparable Pe and Me distributions (see Appendices B.5 and B.6). [For test numbers, the first digit refers to
the loaded panel point, and the second digit refers to the first or second time that the panel point was loaded
in a series of tests (see Table 3.3 in Section 3.4.1.3)]. It may be noted that the large variations in axial force
along a member length are due to experimental error, and are not atypical of measurements obtained using
strain gauges (Frater & Packer 1992c). The axial force used for each segment is noted in Appendix B.5.
Note: bending moment diagrams are plotted on the side of the member subject to tensile bending stress
Fig. 3.6. Experimental (a) axial force and (b) bending moment distributions for test 4.1
In eight total tests, the panel point deflections and support settlements were determined from the LVDT
data. The maximum truss defection (Δmax) for each test is given in Table 3.3, and has been calculated by
taking into account the support settlements. Measured deflections at each panel point, and for each test, are
tabulated in Appendix B.4.
Methods (a) and (b) above are the same ones recommended by Packer & Henderson (1997), and have
been described in Section 3.1.1. Method (c) entails using fixed-end web members that are capable of
transmitting axial, in-plane bending and shear forces, thereby making it possible to plot a bending moment
distribution for both the web and chord members. In method (c), joints are modelled concentrically regardless
of the e value at the joint. Method (d) uses fixed-end web members that are connected to continuous chords,
with the distance between the intersection of the web and chord centrelines (+e or –e) modelled as a rigid
two-joint link.
Fig. 3.7. Assumptions in models (a) P, (b) PR, (c) R, and (d) RR with respect to noding at interior joints,
and at corner joints
The true effective depth (distance between chord centrelines) has been modelled for methods (a) and (c),
meaning that the true web-to-chord angle (θi = θj = 60°) has not been maintained. This assumption, for
“forced” concentric analyses, is claimed to give better results than the alternative (i.e. modelling the true web-
to-chord angles) (Philiastides 1988). Using methods (b) and (d), both the true effective depth and the true
web-to-chord angles have been maintained, since system lines for the models coincide with the member
centrelines. Assumptions with respect to noding at the corner joints are shown in Figs. 3.7e,f. One should
note that for concentrically-noded trusses (i.e. e = 0 at all connections), models R and RR are identical.
FE modelling of the truss (e.g. Coutie & Saidani 1989, 1991) (to explicitly take into account the semi-
rigid connection behaviour) was also considered; however, this was deemed an unlikely approach to the
design of planar HSS trusses. Such an approach is common as a research exercise only, or for the design of
high-value tubular frames (e.g. large-span 3D roof structures, offshore structures).
Table 3.2. Mean percentage errors for theoretical (predicted) axial forces
Test no. Model RR Model PR Model R Model P
4.1 Top chord (11 – 15) +5% +4% +5% +8%
Bottom chord (17 – 20) +7% +10% +8% +15%
Webs (1 – 10) +6% +6% +6% +7%
will be larger, and may not redistribute as easily as in a gap-jointed truss since connections are more rigid.
Designers should therefore be mindful of cases where web moments could become significant. The
experimental bending moments for test 4.2, and a comparison to theoretical bending moments, are provided
in Appendix B.6.
Note: bending moment diagrams are plotted on the side of the member subject to tensile bending stress
Fig. 3.8. Comparison of experimental and theoretical bending moment distributions for test 4.1
Table 3.3. Comparison of experimental and theoretical (predicted) maximum truss deflections
Applied load Δmax Model RR Model PR Model R Model P
Test no. kN mm mm mm mm mm
Upright 4.2 490 4.50 4.42 (-2% a) 4.57 (+2%) 4.72 (+5%) 5.02 (+10%)
6.1 497 6.49 6.93 (+6%) 7.05 (+8%) 7.37 (+12%) 7.69 (+16%)
6.2 497 6.46 7.00 (+6%) 7.12 (+7%) 7.41 (+11%) 7.73 (+15%)
8 497 6.61 6.93 (+7%) 7.05 (+8%) 7.37 (+12%) 7.69 (+16%)
10 490 4.32 4.36 (+3%) 4.60 (+6%) 4.70 (+8%) 5.00 (+14%)
Inverted 7.1 483 6.21 6.57 (+5%) 6.67 (+7%) 6.95 (+11%) 7.22 (+14%)
7.2 480 6.35 6.54 (+3%) 6.63 (+4%) 6.91 (+8%) 7.19 (+12%)
9 496 6.00 5.84 (-3%) 5.95 (-1%) 6.18 (+3%) 4.44 (+3%)
3.5. SUMMARY
The experimental behaviour of a large-scale overlap-jointed RHS truss has been compared with the
theoretical behaviour given by practical frame analysis models, consisting of:
The following conclusions have been reached for the analysis of RHS trusses with overlapped K-connections:
1. Model P provides a conservative (over-)prediction of the axial force distributions and truss
deflections; however, since it neglects continuity bending moments it will be a poor indicator of
the actual truss behaviour.
2. Model R is not suited for analysis since it has been shown to unsafely (under-)predict member
axial forces and, since it neglects eccentricity bending moments, it will be a poor indicator of
both the magnitude and sense of bending in chords.
3. Model PR gives generally good predictions for axial forces in chords and webs, bending
moments in the chords, and for the maximum truss deflection.
4. Model RR may be used for analysis since it provides even better predictions for the axial forces,
bending moments, and maximum truss deflection than model PR, but it does produce a bending
moment distribution for the web members which may cause difficulties using conventional truss
design recommendations, as these are based on axially-loaded web members.
The above conclusions are based on limited experimentation and pertain to RHS trusses of similar
geometry and, in particular, low span-to-depth ratios. It is hence recommended that the analysis of RHS
trusses with overlapped connections be done on the basis of a model with rigid-ended web members
connected eccentrically to continuous chords (Model RR). This is a logical extension of the recommendation
by Frater & Packer (1992c) to use Model PR for RHS trusses with gapped K-connections, wherein
connections are more flexible (due to chord face deformation) resulting in larger overall truss deformations
and different magnitudes of axial forces/bending moments in members.
4.1. INTRODUCTION
4.1.1. WELD DESIGN PHILOSOPHIES FOR HSS CONNECTIONS
For steel hollow structural section (HSS) connections, current codes, standards, specifications and design
guides (Wardenier et al. 2008; Packer et al. 2009; Packer et al. 2010; ISO 2013) acknowledge two design
philosophies for proportioning welds:
(i) The weld can be proportioned so that it develops the yield strength of the connected branch
member. By setting the design strength of a fillet-welded joint equal to that of the yield strength
of the connected branch member, the required effective weld throat dimension (tw) can be
calculated in terms of the connected branch wall thickness (tb). This will produce an upper limit
on the weld size and hence be a conservative design procedure. Assuming an axially-loaded 90˚
T-connection between rectangular hollow sections (RHS) made to ASTM A500 Grade C (ASTM
2013) with matching electrodes, the results of Method (i) for the design of fillet welds in various
steel specifications/codes (CSA 2001; CEN 2005; AISC 2016; AWS 2015; CSA 2014) are listed
in Table 4.1 (updated from McFadden et al. 2013). Clearly, there is quite a disparity in fillet weld
design criteria in these steel specifications/codes.
(ii) As an alternative to Method (i), the weld can be proportioned so that it resists the actual forces in
the connected branch member. This approach involves a consideration of the effective length of
the weld group, since extensive research (Frater & Packer 1992a, 1992b; Packer & Cassidy 1995;
Packer & Sun 2011) has proven that the connected branch wall and the adjacent weld are
generally loaded in a highly non-uniform manner around the branch, because the base to which
the weld is attached is flexible in a typical HSS-to-HSS connection.
Method (i) is appropriate if there is low confidence in the design forces in the branch, or if there is
uncertainty regarding Method (ii), or if plastic stress redistribution is required in the connection. Method (i)
enables a prequalified weld size to be readily specified. However, Method (ii) generally allows for weld
“downsizing” and hence is popular. It is particularly appropriate if the branch forces are low relative to the
branch member capacity. AISC 360 (AISC 2016) has adopted Method (ii) in Chapter K for welded
41
Chapter 4: Fillet Welds in HSS-to-Rigid Plate Connections: Experimentation 42
connections to RHS by specifying various weld effective lengths for different connection types and loading
situations.
Table 4.1. Comparison of fillet weld effective throats required to develop the yield resistance of a connected
RHS branch member wall (updated from McFadden et al. 2013)
Specification or code tw
AISC 360-16 Table J2.5 (AISC 2016) 1.43 tb
AWS D1.1-15 Clause 9.6.1.3 and Fig. 9.10 (AWS 2015) 1.07 tb
CAN/CSA S16-01 Clause 13.13.2.2 (CSA 2001) 1.14 tb
CSA S16-14 Clause 13.13.2.2 (CSA 2014) 0.95 tb
EN1993-1-8: Directional method (CEN 2005) 1.28 tb
EN1993-1-8: Simplified method (CEN 2005) 1.57 tb
Note: This table assumes an axially-loaded 90˚ T-connection between RHS made to ASTM A500 Grade C with
matching electrodes.
The discrepancy between the tw values in Table 4.1 is due to simplified models for fillet weld design, as
well as use (or non-use) of the the (1.00+0.50sin1.5θ) directional strength-increase factor, discussed herein.
fillet weld gradually increased to 1.50 times as the loading angle increased from 0˚ to 90˚. Based on the
experimental results, Miazga & Kennedy (1989) proposed a method to predict the strength of fillet welds of
different orientations based on a maximum shear stress failure criterion. Later, Lesik & Kennedy (1990)
extended the work of Miazga & Kennedy (1989) and proposed a simplified equation which is a function only
of the loading angle [i.e. the (1.00+0.50sin1.5θ) directional strength-increase factor adopted in current
American (AISC 2016) and Canadian (CSA 2014) specifications]. It takes the form of a multiplier that is
applied to the longitudinal fillet weld strength.
The test program of Miazga & Kennedy (1989) included connection specimens lap-spliced by fillet welds
(Fig. 4.1a) using the shielded metal arc welding (SMAW) process, which is not commonly used in industry
for high-production welding. To re-evaluate the effectiveness of the (1.00+0.50sin1.5θ) factor on the more
prevalent flux-cored arc welding (FCAW) process, a series of investigations was conducted by Ng et al.
(2004a, 2004b) and Deng et al. (2006). Their reliability analyses showed that the design equations in the
American and Canadian standards provide an adequate level of safety for both welding processes.
Fig. 4.1. Fillet-welded (a) lap-splice, (b) cruciform, and (c) HSS-to-rigid plate test specimens
The tests performed by Ng et al. (2004a, 2004b) and Deng et al. (2006) consisted of concentrically loaded
fillet-welded connections with all welds having the same loading orientation. However, fillet-welded
connections commonly include welds at different orientations to the applied load, and the interaction between
fillet welds of different loading angles remained unknown. Hence, Callele et al. (2009) tested 19 lap splice
connections with multiple weld segments of different orientations. It was still found that the deformation
capacity of the fillet welds decreased as the angle of loading increased (i.e. maximum deformation capacity
for a weld element loaded longitudinally; minimum deformation capacity for a weld element loaded
transversely). As a result of such incompatibility, the transverse weld prevents the longitudinal weld from
reaching its full capacity before failure of the joint takes place. Hence, the tested weld groups possessed
capacities significantly lower than the sum of the individual weld segment strengths. Therefore, Callele et al.
(2009) proposed a simple method to account for this phenomenon conservatively by reducing the capacities
of the more ductile welds by 0 to 15%. For example, for a weld group containing longitudinal and transverse
welds, the longitudinal weld can only develop 85% of its full capacity before joint failure. This method has
been adopted by current American and Canadian steel design specifications. In order to investigate the
response of eccentrically loaded fillet welds, where the load is not in the plane of the weld group, Kanvinde et
al. (2009) performed 60 bend tests on cruciform connection specimens (Fig. 4.1b). It was found that a bearing
mechanism between the connected plates, which is not present for concentrically loaded joints, made an
obvious contribution to the strength of fillet-welded joints under out-of-plane eccentric loading. Hence, the
authors proposed a design approach which incorporated this beneficial effect.
Another important observation, based on the experimental evidence in the above research, is that the
actual weld fracture plane does not always coincide with the theoretical throat. Since the theoretical effective
throat thickness of a fillet weld is commonly defined, in various design specifications, as the height of the
largest triangle that can be inscribed within the fusion faces and the weld surface (i.e. the shortest distance
from the root to the face of the weld), the use of the theoretical effective throat thickness generally produces a
conservative strength prediction.
However, the application of this (1.00+0.50sin1.5θ) factor, also known as the “sinθ” factor, in the design
of fillet welds in HSS connections (Fig 4.1c) has been questioned since:
1. Unlike lap splice connections, fillet welds in many HSS connections have the welded attachment
loaded in tension or bending, rather than in shear.
2. Since welding can only be performed on one side of the hollow section wall, fillet welds to HSS
members will be subject to a local eccentricity. For example, tension loading in an attached wall
will produce additional tensile stress at the root of the weld (Fig. 4.1c). In fact, relevant codes and
standards recognize that eccentric loading on a fillet weld, causing tension at the weld root, may
reduce weld capacity. For example, CSA W59 (CSA 2013b) Clause 4.1.3.3.2 even states that …
“Single fillet and single partial joint penetration groove welds shall not be subjected to bending
about the longitudinal axis of the weld if it produces tension at the root of the weld”. EN 1993-1-
8 (CEN 2005) Clause 4.12 states that such local eccentricity, producing tension at the root of the
weld, should be taken into account, but it specifically notes that … “Local eccentricity need not
be taken into account if a weld is used as part of a weld group around the perimeter of a structural
hollow section”. The basis for this Eurocode waiver is unknown. AWS D1.1 (2015) Clause 2.6.2
states that, in the design of welded joints, the calculated stresses shall include those due to
eccentricity caused by alignment of the connected parts, size and type of welds, but this Section
pertains to connections which are “non-tubular”.
3. By the same token as 2., the HSS wall is not externally restrained from rotation.
4. It has been shown experimentally that the inclusion of the sinθ factor in the strength calculation
is generally unsafe, when applied to RHS-to-RHS fillet-welded connections and used in
conjunction with AISC 360-10 (and AISC 360-16) Chapter K weld effective lengths/properties,
because target structural reliability levels are not achieved (Packer & Sun 2011; McFadden &
Packer 2013; McFadden et al. 2013; McFadden & Packer 2014; Tousignant & Packer 2015a).
Thus, AISC does not permit the use of the sinθ factor when the “effective length method” is used
for proportioning fillet welds in hollow section connections (AISC 360-16 Commentary on K5).
The purpose of this Chapter is to determine if the sinθ factor is even applicable to fillet welds between an
HSS branch and a rigid base, where all of the weld length is effective (i.e. the AISC effective length method
is not applicable). Hence, in this investigation all connection specimens were made by welding HSS to rigid
steel plates to remove the effect of any flexibility of the surface on which the fillet weld lands.
EN1993-1-8 (CEN 2005), which was developed based on a von Mises hypothesis and verified experimentally
by assessing the strength of fillet welds loaded at different angles. Tests in the above research showed that the
strength of fillet welds under combined stresses, due to load applied at different angles, can be roughly
represented by an ellipsoid in the 𝜎⊥ , 𝜏⊥ , 𝜏∥ space. Recent European research by Björk et al. (2012, 2014) has
extended the fillet weld design rules to connections made of high and ultra-high-strength steel.
In Section J of AISC 360, unless overmatched weld metal is used, the design strength (Vrw = Pnw) of a
single fillet weld is based on the assumed single limit state of shear rupture along the plane of the weld
effective throat. It is taken as the product of the nominal stress of the weld metal (Fnw) and the weld effective
throat area (Aw) with a resistance factor ( = 0.75) applied. The nominal strength (Pnw) is determined as
follows:
where θ = angle of loading measured from the weld longitudinal axis (in degrees).
As a special application of Section J2.4(b), Section J2.4(b)(2) gives provisions for concentrically loaded
connections with weld elements of multiple orientations (longitudinal and transverse to the direction of
applied load). It specifies that the nominal strength (Pnw) of concentrically loaded joints with both
longitudinal and transverse fillet welds be determined as the greater of Eqs. (4.2a) and (4.2b). This provision
is to account for the deformation incompatibility between longitudinal and transverse fillet welds.
where Pnwl = total nominal strength of longitudinally loaded fillet welds; and Pnwt = total nominal strength of
transversely loaded fillet welds with Fnw calculated using Eq. (4.1b).
In the 2001 Canadian steel structures design standard, the design strength (Vrw = Pnw = 0.67Pnw) of a
fillet weld was taken as the lesser of two limit states: (i) shear rupture along the fusion face with the base
metal using Eq. (4.3a), and (ii) shear rupture along the plane of the weld effective throat using Eq. (4.3b),
which permits the use of a sinθ factor.
where Am = area of fusion face between weld and base metal; and Fu = ultimate tensile strength of base metal.
It should be noted that the design resistance (Vrw) calculated using Eqs. (4.1a) and (4.1c) per AISC 360-16 is
the same as that calculated using Eq. (4.3b) per CSA S16-01, because the terms ( = 0.75)(0.60FEXX) and (
= 0.67)(0.67FEXX) both equal 0.45FEXX.
0.85 1 / 600
Mw (4.4b)
0.85 2 / 600
where θ and θ1 = angle of loading (in degrees) of the weld element under consideration; θ2 = angle of loading
(in degrees) of the weld element in the joint that is nearest to 90˚; Mw = strength reduction factor to allow for
the variation in deformation capacity of weld elements with different orientations, similar to the 0.85 factor in
Eq. (4.2b).
0.5
2 3( 2 2 ) Fu / ( w M 2 ) (4.5a)
where 𝜎⊥ = normal stress perpendicular to the throat; 𝜏⊥ = shear stress (in the plane of the throat)
perpendicular to the axis of the weld; 𝜏∥ = shear stress (in the plane of the throat) parallel to the axis of the
weld; 𝑀2 = partial safety factor for the resistance of weld equal to 1.25; and 𝑤 = correlation factor for fillet
welds.
Fig. 4.2. Stress components in the plane of the weld effective throat
Fu
Vrw ( )twlw (4.6a)
3 w M 2
Fu
Vrw ( )twlw (4.6b)
2 w M 2
where Vrw = design resistance of the fillet weld. Intermediate steps in the simplification are shown in
Appendix C.1.
Thus, Eurocode 3 (CEN 2005) uses a relationship between the strength of a transverse weld to a
longitudinal weld of (√ /√ ) = . , which is significantly less than the 1.50 factor used in current North
American specifications [Eq. (4.1c) or Eq. (4.4a)].
The Simplified Method is an alternative to the Directional Method for fillet weld design. This method is
independent of the orientation of the weld throat plane with respect to the applied force. In fact, it is a
conservative alternative to Eq. (4.5a). The Simplified Method assumes that all welds are loaded in pure shear
parallel to the axis of the weld (Fig. 4.2b) and the welds can then be proportioned using Eq. (4.6a).
4.3. EXPERIMENTATION
Since a primary objective of this study is to determine if the sinθ factor is applicable when the entire
length of a fillet weld in an HSS connection is effective (i.e. the effect of any flexibility of the surface on
which the fillet weld lands is removed), all connection specimens were made by welding either circular
hollow sections (CHS) or RHS to a rigid steel plate. For this work, 12 CHS-to-rigid plate connections with
different weld sizes, and angles between the HSS and plate of 60˚ and 90˚ (Fig. 4.3), were designed and
fabricated to be weld-critical under the applied tension loads. The fabrication drawings for these specimens
are provided in Appendix C.2. A further 21 tests on RHS-to-rigid plate connections conducted by Frater
(1986) and Oatway (2014) are also discussed. A total of 33 HSS-to-rigid plate connections have thus been
tested, and are analyzed herein.
skew-T fillet weld gauge. For each RHS connection specimen, the weld cross-sectional dimensions were
carefully measured at 20 locations around the footprint of the branch. For each CHS connection specimen,
the weld cross-sectional dimensions were measured at uniform increments of 25-30 mm around the footprint
of the branch. The averages of the theoretical effective throat thickness of the weld (tw), and the weld leg
length measured along the branch (lv) and along the plate (lh) are listed in Tables 4.2 - 4.4. Complete
measurements of the test welds in the CHS-to-rigid plate connections are provided in Appendix C.3.
Table 4.2. Measured geometric properties, failure loads, and failure modes for θ = 90° RHS-to-rigid plate
specimens
RHS Dimensions Exp.
Test no. θ Bb × Hb ×tb tp lw Average of tw Average of lv Average of lh Pa Fail.
° mm×mm×mm mm mm mm mm mm kN Mode
1† 3.59 5.45 4.78 831 W
† 127.0 × 127.0 × 7.78
2 5.91 8.30 8.43 1166 W
90˚ (r0 = 15.88 mm) 25.0 481
3† (ri = 8.98 mm) 5.29 7.73 7.25 1235 W
†
4 5.61 9.65 6.90 1311 W
†
5 6.38 9.28 8.78 2433 W
† 177.8 × 177.8 × 12.53
6 8.61 13.28 11.30 2574 W
90˚ (r0 = 35.0 mm) 25.0 651
7† (ri = 23.4 mm) 7.02 10.38 9.53 2525 W
†
8 7.35 12.23 9.20 2302 W
††
9 4.20 5.30 6.90 1020 W
††
10 3.57 4.40 6.10 960 W
††
11 3.24 3.90 5.80 840 W
12†† 127.6 × 127.6 × 9.54 4.74 6.30 7.20 1140 W
13 †† 90˚ (r0 = 19.08 mm) 19.0 478 5.69 8.10 8.00 1200 WP
†† (ri = 9.54 mm)
14 5.44 7.50 7.90 1207 WP
††
15 7.71 10.80 11.00 1494 P
16†† 9.33 13.20 13.20 1578 P
††
17 10.82 15.30 15.30 1788 P
Note 1: ro = average outside corner radius of RHS; ri = average inside corner radius of RHS.
Note 2: W = weld failure; WP = mixed failure mode of weld failure and partial plate failure; P = end-plate rupture along
at least one weldment.
†
Tests by Oatway (2014).
††
Tests by Frater (1986); ri and ro are taken as 1.0tb and 2.0tb, respectively, for all RHS used by Frater (1986).
Table 4.3. Measured geometric properties, failure loads, and failure modes for θ = 60° RHS-to-rigid plate
specimens
RHS Dimensions lw Average of tw Average of lv Average of lh Exp.
θ Bb × Hb ×tb tp mm mm mm mm Pa Fail.
Test no. ° mm×mm×mm mm a&b c d a&b c d a&b c d a&b c d kN Mode
18†† 137 119 119 6.10 4.20 8.30 8.00 8.40 10.8 9.40 8.60 8.80 1131 WP
†† 127.0 × 127.0 × 7.78
19 137 119 119 3.80 2.70 5.20 5.60 5.40 6.10 5.30 5.60 6.00 982 W
60˚ (r0 = 15.88 mm) 19.0
††
20 (ri = 8.98 mm) 137 119 119 4.90 3.80 8.10 8.30 8.60 11.2 6.00 6.80 8.50 1270 WP
††
21 137 119 119 9.20 6.80 12.4 12.5 13.2 14.5 13.5 13.9 14.2 1534 WP
Note 1: Locations a, b, c and d are indicated in Fig. 4.3.
Note 2: ro = average outside corner radius of RHS; ri = average inside corner radius of RHS.
Note 3: W = weld failure; WP = mixed failure mode of weld failure and partial plate failure; P = end-plate rupture along
at least one weldment.
††
Tests by Frater (1986); ri and ro are taken as 1.0tb and 2.0tb, respectively, for all RHS used by Frater (1986).
Table 4.4. Measured geometric properties, failure loads, and failure modes for θ = 90° and θ = 60° CHS-to-
rigid plate specimens
CHS Dimensions Exp.
θ Db ×tb tp lw Average of tw Average of lv Average of lh Pa Fail.
Test no. ° mm×mm×mm mm mm mm mm mm kN Mode
22 4.81 7.04 6.60 1261 W
90 167.9 × 6.70 25.0 528
23 6.63 9.64 9.13 1279 W
24 6.87 9.89 9.54 1459 W
90 127.4 × 11.55 25.0 401
25 7.98 11.23 11.34 1597 W
26 6.38 8.85 9.22 841 W
90 101.0 × 7.34 25.0 318
27 6.16 9.23 8.28 864 W
28 5.32 6.13 7.41 1450 W
60 167.9 × 6.70 25.0 569
29 5.73 6.88 7.71 1331 W
30 5.21 8.06 8.54 1109 W
60 127.4 × 11.55 25.0 432
31 6.78 10.50 11.13 1479 W
32 5.39 6.93 7.75 776 W
60 101.0 × 7.34 25.0 342
33 4.98 6.71 7.32 803 W
Note: W = weld failure; WP = mixed failure mode of weld failure and partial plate failure; P = end-plate rupture along at
least one weldment.
After testing and noting the actual load at failure, each connection was cut (where possible) normal to the
longitudinal axis of the weld at several locations around the footprint of the branch (two cuts per side for the
RHS, and at the locations of the weld cross-sectional dimension measurements for the CHS). After surface
polishing, all cross-sections were subjected to macroetch examinations in accordance with ASTM E340-15
(ASTM 2015b) using a 10% nital etchant solution to observe the weld profile. These cross-sections were then
scanned and input into AutoCAD so that the weld cross-sectional dimensions tw, lv and lh could be re-
measured with greater accuracy (Fig. 4.4). For these measurements, the throat thickness of the fillet weld was
taken as the height of the largest triangle that could be inscribed within the fusion faces and the weld surface.
Fig. 4.4. Examples of fillet weld throat measurements from the macroetch examinations
For the RHS, credence was given to the externally measured values of lv and lh using the fillet weld gauge
(presented in Tables 4.2 and 4.3), since all macroetch measurements showed good agreements with the
external measurements. The values of tw in Table 4.2 were then determined from the weld leg measurements
using Eq. (4.7), which takes into account the effect of unequal weld leg sizes and the local dihedral angle
(angle between the base metal fusion faces), ψ, on the orientation of the weld throat plane. A full derivation
of this equation is provided in Appendix C.4.
lv lh sin
tw (4.7)
l lh2 2lv lh cos
2
v
where lv = weld leg measured along the HSS branch; and lh = weld leg measured along the plate.
For the CHS, credence was given to the macroetch measurements of lv, lh and tw, since there was a
sizeable discrepancy with some of the external measurements. The values of tw, lv and lh in Table 4.4 and
Appendix C.3 hence correspond to these measurements.
The total weld lengths (lw in Tables 4.2 - 4.4) were determined based on the HSS perimeters and were
hence measured along the root of the weld considering the angle between the HSS and plate, and the corner
radii of the RHS.
Table 4.5. Measured material properties for all 33 HSS-to-rigid plate connections
HSS End-plate Weld metal
Fyb Fu Fyb/Fu rup Fyp Fup Fyp/Fup rup Fyw FEXX Fyw/FEXX rup
Test No. MPa MPa MPa MPa MPa MPa
1-4 412 478 0.862 0.33 383 563 0.680 0.24 563 619 0.910 0.28
5-8 380 489 0.777 0.33 383 563 0.680 0.24 563 619 0.910 0.28
b a
9 - 11, 17 & 19 426 500 0.852 - 351 558 0.629 0.36 634 687 0.923 0.24
b a
12 - 16, 18, 20 & 21 426 500 0.852 - 351 558 0.629 0.36 641 739 0.867 0.24
22, 23, 28, 29 421 501 0.840 0.31 409 566 0.723 0.24 501 571 0.877 0.26
24, 25, 30, 31 431 488 0.883 0.35 409 566 0.723 0.24 501 571 0.877 0.26
26, 27, 32, 33 385 450 0.856 0.35 409 566 0.723 0.24 501 571 0.877 0.26
a
Coupons fractured at the gauge lines giving incorrect elongations.
b
Excludes results for corner coupons published by Frater (1986).
Matching electrodes with a nominal tensile strength of 490 MPa were used for all fillet welds. To
determine the material properties of the as-laid weld metals, all-weld-metal tensile coupons were created in
accordance with AWS D1.1 (2015). The averages of the measured yield stresses (Fyw, determined by the
0.2% strain offset method) and ultimate strengths (FEXX) of the as-laid weld metals are also given in Table
4.5. The results of the TC tests for the CHS, end plate, and weld metal used for the CHS-to-rigid plate tests
(nos. 22 - 33) are provided in Appendix C.5.
4.3.3. INSTRUMENTATION
For all RHS connection specimens, a group of four strain gauges (Group A) were mounted on the four
faces of the RHS well above the intermediate plate. The purpose of these strain gauges was to measure any
strain difference between opposite RHS faces during testing, thereby monitoring any unintended bending
moments. It was found that no bending moment was applied to any connection, hence all specimens were
loaded in pure tension during testing. In order to further verify that the weld elements were uniformly loaded
along their lengths, another group of eight strain gauges (Group B) were placed on two adjacent faces of the
RHS just above the intermediate plate for all connection specimens. Typical load-strain curves at four
different locations along one side of an RHS are shown in Fig. 4.5. Since the strains at four points along each
RHS wall remained almost identical, with increasing branch load, it was deduced that all welds were
uniformly loaded during each RHS connection test; hence, all of the weld length could be considered
effective. Fig 4.5 shows that, at high applied loads, strains along the RHS branch face start to deviate from
each other. This is due to progressive yielding of the branch (due to varying amounts of cold working along
the face), and the weld (due to the weld size being close to – but not exactly – uniform). Additionally, this is
caused by variations in the extent of the welding heat affected zone, but it has little effect on the findings.
For all CHS connection specimens, Group A and B consisted of four or eight strain gauges mounted with
uniform spacing around the CHS perimeter either well above the intermediate plate (Group A) or just above
the intermediate plate (Group B). Similarly, it was found that all welds were uniformly loaded during the
tests. Linearly varying differential transformers (LVDTs) were also used to measure the load-displacement
behaviour of the connection region (see Fig. 4.6a,b).
Fig. 4.5. Typical load-strain curves from four strain gauges on one side of RHS (Group B strain gauges,
shown for test no. 5)
(b) Instrumentation for CHS-to-rigid plate tests (c) overall test set-up (shown for a CHS-to-rigid plate test)
(a) RHS-to-rigid plate (shown for test no. 1) (b) CHS-to-rigid plate (shown for test no. 22)
of this procedure for several code methods applied to a 60° CHS-to-rigid plate connections are provided in
Appendix C.7.
Fig. 4.8. Comparison of actual and nominal strengths Fig. 4.9. Comparison of actual and nominal strengths
per AISC 360-16 without the sinθ factor per AISC 360-16 with the sinθ factor
The predicted nominal strengths of the test welds per AISC 360 with the sinθ factor are compared to the
actual strengths at failure in Fig. 4.9. For the 90˚ connections, the nominal strengths of the test welds were
determined using Eq. (4.1a,c). For the 60˚ RHS connections, the nominal strengths of the test welds were
computed using Eq. (4.2b) with the 1.5 factor for Pnwt. Eq. (4.1a,b) are used to calculate Pnwt since the
directional strength-increase is considered by the 1.5 factor already. Eq. (4.1a,c) are used to calculate Pnwl to
account for the directional strength-increase factor for the 60˚ oblique welds. Also, Pnwl is multiplied by a
0.85 factor [similar to the Mw factor in Eq. (4.3a) per CSA S16-14] to account for the difference in
deformation capacity between the oblique and transverse weld elements. (In theory, the 0.85 factor should
really be higher since the 0.85 applies to longitudinal welds). For the 60˚ CHS connections, Eq. (4.1a-c) were
used to compute the strength of the weld component tributary to each measurement. These strengths were
then added together to calculate the nominal joint strength. The 0.85 factor in Eq. (4.2b), which could
technically apply - since the connection encompasses weld components with multiple orientations - was
deemed too general for this situation, and omitted from the analysis.
Fig. 4.10. Comparison of actual and nominal strengths Fig. 4.11. Comparison of actual and nominal strengths
per CSA S16-14 without the sinθ factor per CSA S16-14 with the sinθ factor
sinθ factor). The predicted nominal strength of each welded joint per CSA S16-01 with the sinθ factor was
determined by repeating the above procedures, with the sinθ factor in Eq. (4.3b) included. An example of this
procedure applied to a 60° CHS-to-rigid plate connection is provided is Appendix C.7.
All predictions per CSA S16-01 are compared to the actual strengths in Figs. 4.12 and 4.13. When the
sinθ factor is not used (Fig. 4.12), the predicted nominal strengths of nearly all of the 90° HSS connections
(16 of 17 RHS connections and 6 of 6 CHS connections) are governed by the limit state of shear rupture
along the plane of the weld effective throat, and all but one of the 60° HSS connections are governed by the
limit state of shear rupture along the fusion face with the base metal at some location along the weld length.
When the sinθ factor is included in the calculation (Fig. 4.13), the nominal strengths of all connections are
governed by fusion face rupture at some location along the weld length.
Fig. 4.12. Comparison of actual and nominal strengths Fig. 4.13. Comparison of actual and nominal strengths
per CSA S16-01 without the sinθ factor per CSA S16-01 with the sinθ factor
internal weld geometry measurement via sectioning on all connection specimens. Eq. (4.5b) was satisfied in
all cases. The correlation factor for fillet welds, w, was taken as 0.9 according to Table 4.1 in EN1993-1-8
(CEN 2005) for both cold-formed hollow sections (to EN10219) and hot-finished hollow sections (to
EN10210), for grade S355 (HSS with a nominal yield strength of 355 MPa). Since the target safety
(reliability) index for this Eurocode method is unknown, a comparison is performed against the limit states
design resistance, including the partial safety factor, 𝑀2 .
P cos
(4.8a)
t w lw
P sin cos
(4.8b)
t w lw
P sin sin
(4.8c)
t w lw
P 0.5
sin 2 cos 2 3(sin 2 sin 2 cos 2 ) Fu / ( w M 2 )
t w lw (4.8d)
Fu t w lw Pnw
P Vrw (4.8e)
( w M 2 ) sin 2 cos 2 3(sin 2 sin 2 cos 2 ) 0.5
M2
Design strengths (Vrw) calculated with Eq. (4.8e) are compared to the actual strengths in Fig. 4.14, which
shows that EN1993-1-8 produces safe predictions for all tested weld joints. The average actual-to-design
strength ratios for the RHS and CHS connections are 1.46 and 1.53, respectively (with an average of 1.48
overall). The predicted and/or design strengths discussed in this section (Section 4.4) are summarized for
each joint in Appendix C.10.
Fig. 4.14. Comparison of actual strengths and design strengths per EN1993-1-8 (2005): Directional Method
Fig. 4.15. Effect of weld size on actual-to-nominal Fig. 4.16. Effect of branch slenderness on actual-to-
strength ratio per AISC 360-16 without the sinθ factor nominal strength ratio per AISC 360-16 without the
sinθ factor
4.6. SUMMARY
In this chapter, 33 HSS-to-rigid plate weld-critical connections that were tested to failure under axial
tension loading have been compared. The design methods in CSA S16-01, EN1993-1-8:2005, AISC 360-16,
and CSA S16-14 for fillet welds to HSS members were evaluated by comparing the actual strengths of the
fillet welds to the predicted strengths. It can be concluded from this work that:
1. The Directional Method in Eurocode 3 produces safe strength predictions for fillet welds to HSS.
Hence, the Simplified Method is even more conservative.
2. When the (1.00+0.50sin1.5θ) directional strength increase factor is not included in the strength
calculation of fillet welds to HSS, the equations in both the current American and Canadian
specifications meet the target safety index according to AISC ( +
≥ 4.0).
3. Restrictions need to be placed in current North American steel design codes on the application of
such a directional strength increase factor for fillet welds in HSS connections. It should be noted
that the directional strength-increase factor was developed based on tests on fillet welds in lap
splice connections. According to this investigation of fillet welds in HSS connections, strength
calculation including a directional strength increase factor leads to predictions which do not have
a sufficient safety margin, even when it is not used in conjunction with the “effective length
method” of AISC 360 Chapter K.
4. The relative strength (per unit throat thickness) of small fillet welds is generally greater than for
large fillet welds.
5. CHS-to-rigid plate specimens generally exhibited higher average strengths than did RHS-to-rigid
plate specimens. This is believed to be due to the greater restraint from local rotation offered by
the constant curvature of the CHS wall. RHS, on the other hand, have less restraint from rotation
along the flats, especially for slender sections, which results in greater local rotation and higher
tension at the weld root. Additionally, non-uniform stiffness of the branch parallel to the branch
longitudinal axis (resulting in stress concentrations in the corners) may play a role. This is
investigated in Chapter 5.
4.6.1. FOOTNOTE
It is worth noting that the above tests were performed on connections between HSS and rigid end-plates.
In typical connections that occur between HSS and flexible end-plates, where the end plates are fastened to
adjacent structural components using fasteners or anchors, weld effective lengths would re-occur.
5.1. INTRODUCTION
When proportioning welds to resist the applied forces in axially-loaded branch members of hollow
structural section (HSS) connections, weld effective lengths are required (ISO 2013). Weld effective lengths
account for the non-uniform loading of the weld due to differences in the relative flexibility of the HSS chord
face, around the weld perimeter. Weld effective lengths for truss connections between rectangular hollow
sections (RHS) have been researched at the University of Toronto over the last several decades by Frater &
Packer (1992a, 1992b), Packer & Cassidy (1995), McFadden & Packer (2014), and Tousignant & Packer
(2015). AISC 360 Section K5 (AISC 2016), and Packer & Henderson (1997), present weld effective length
rules for designers derived from this research. McFadden & Packer (2014) and Chapter 2 have shown that the
fillet weld “directional strength-increase factor” (1.00+0.50sin1.5θ) in Section J2.4 of AISC 360 (AISC 2016),
and its equivalent of (1.00+0.50sin1.5θ)Mw in Clause 13.13.2.2 of CSA S16 (CSA 2014), does not provide
adequate structural reliability (or safety) when used in conjunction with the rules for weld effective lengths.
AISC (2011) and CISC (2016) hence disallow the use of the directional strength-increase, or “sinθ factor”,
when these effective length rules are used to design fillet welds in HSS-to-HSS connections.
Fig. 5.1. HSS-to-rigid plate connection specimens (with RHS or CHS members)
Chapter 4 presented a large number of laboratory tests on HSS-to-rigid plate connections (Fig. 5.1) to
investigate the applicability of the sinθ factor to single-sided fillet welds to HSS, joined to a rigid end-plate.
These experiments removed the influence of a flexible landing surface for the fillet weld, and hence removed
the weld effective length phenomenon. It was shown that HSS-to-rigid plate fillet welds still did not provide
adequate structural reliability if the sinθ factor was implemented.
66
Chapter 5: Fillet Welds in HSS-to-Rigid Plate Connections: Finite Element Modelling 67
The research herein is a finite element (FE) extension of the experimental work to determine (for HSS-to-
rigid end-plate connections):
(a) the extent to which fillet welds to RHS and circular hollow sections (CHS) are similar;
(b) the effect of relative weld size (tw/tb), branch wall slenderness (Bb/tb and Db/tb), and branch
inclination angle (θ) on the weld strength; and
(c) if alternate expressions are more appropriate for estimating the strength of fillet welds to HSS.
5.3.1. GENERAL
Each FE connection was comprised of a single tension-loaded branch welded to a rigid end-plate, with
fixed restraints applied to the nodes on the “underside” of the end-plate. When θ = 60°, one half of each FE
connection could be modelled using symmetry boundary conditions parallel to the axis of the toe and heel of
the connection (Figs. 5.2 and 5.3). When θ = 90°, an additional plane of symmetry, orthogonal to the first
plane and through the centre of the branch, allowed one quarter of each connection to be modelled (see
Appendix D.1). It was shown that these FE models provided identical load-displacement responses and the
same weld rupture loads as FE models that included two concentric, tension-loaded, branches (welded to
opposite sides of the end-plate). Since two branches were used for the experimental tests in the manner
described, it was necessary to demonstrate that the above models were equivalent, prior to evaluating them.
Fig. 5.4. Effective weld dimensions in FE models (shown for a model with θ = 90°)
To ensure that load transfer was only through the fillet weld, a gap was used to separate the branch and
the end-plate at their interface, as shown in Fig. 5.4. From preliminary analyses of the connections, the size of
the gap was determined to have a significant effect on the non-linear response of the fillet weld. To take into
account (and mitigate) this effect: (a) the size of the gap can be minimized (close to the Boolean tolerance of
the FE program), and (b) the calculation of the weld throat area (Aw) can be done based on the portion of the
fillet above the gap, as shown in Fig. 5.4. In applying (a) and (b), the load-displacement response and
ultimate fracture strain of the FE models were found to closely agree with the experimental tests. Generally,
however, when welds are much larger than the Boolean tolerance of the FE program, provision (b) is not
necessary, since the size of the gap resulting from provision (a) will be small compared to weld leg
dimensions (lv and lh).
Models were analyzed under static incremental displacements applied to the end of the branch, in the
theoretical constant stress region identified by Mehrotra & Govil (1972) – a distance of 3Bb or 3Db from the
weld toe. It was therefore only necessary to model the portion of the branch within this distance.
T (1 ) (5.1)
T ln(1 ) (5.2)
It is well-known that different material properties are obtained in the HSS corners (higher yield stress,
higher ultimate stress, lower ductility) relative to the flats, where coupon measurements were taken; however,
the corner area as a proportion of the total HSS cross-sectional area is typically low, so these differences will
be diminished. Furthermore, the objective of the FE modelling exercise was primarily to capture the weld
behaviour and fracture, so the refinement in HSS modelling was considered to be of low influence on the
results.
Eqs. (5.1) and (5.2) are not valid past the coupon necking point, since the stress distribution in the
material is no longer uniaxial. Therefore, an iterative approach given by Ling (1996) was used to generate the
remainder of the T- T curve. Ling’s (1996) expression for the T- T curve in the necked region is given by Eq.
(5.3):
T
T T w(1 T T ) (1 w) T
(5.3)
T
T
where 𝜎𝑇′ is the true stress at which necking starts, ε′𝑇 is the true strain at which necking starts, and w is a
weighting factor.
Ling’s (1996) approach is based upon weighting an approximate lower- and upper-bound to the T- T
curve past the coupon necking point. The lower bound (w = 0.0) represents a power law (Hollomon 1945),
and the upper bound (w = 1.0) is linear (Fig. 5.5).
Fig. 5.5. Post-necking generated T- T curve using the Ling (1996) procedure (shown for CHS 127.4 × 11.55
TC)
The weighting constant, w, is determined by selecting a trial value of w to generate points on the T- T
curve in the necked region, running an incremental load-step analysis of a TC up to fracture, and comparing
the - curve from the model, which is generated from applied loads and nodal displacements, to
experiments. The nodal displacements, and hence the FE strains, were calculated over a “virtual” 50-mm
gauge length from nodes on the exterior of the TC. This process is repeated until a value of w is found that
results in acceptable agreement between the experimental and FE - curves.
For this work, TCs were modelled using average measured dimensions of the coupon widths and
thicknesses (or diameters, for screw-type TCs), and nominal values of the grip dimensions, including the
radius of the machined fillet and the length of the reduced section outside of the TC gauge length (ASTM
2017). CHS TCs were modelled with curved geometries, as shown in Fig. 5.6, to replicate both the
experimental tests, and the in-situ condition of the material. ANSYS SOLID45 8-noded brick elements, with
non-linear material and geometric (large deformation) properties, were used.
Fig. 5.6. FE CHS TC (a) geometry and meshing, and (b) comparison of deformed shape and results at
fracture
The clip gauge used to measure strain during the experimental tests was removed at ≅ 0.20, to avoid
damage; however, the elongation of the coupon and the load at rupture were measured and recorded. In the
region of the - curve between first necking and the point where the clip gauge was removed, where Ling’s
(1996) approach was used, the experimental and numerical results always showed good agreement (see Fig.
5.7). At the point of coupon rupture, agreement varied. In two previous FE studies (Voth & Packer 2012a;
Martinez-Saucedo et al. 2006), it was determined that rupture in large-scale HSS connections typically occurs
at strains well below those present at the point of coupon rupture, due to boundary conditions and
confinement in connections that are different from those in TCs. The gap between the HSS and the end-plate
also creates a crack-like feature at the root of the weld, which contributes to earlier rupture in the HSS
connections relative to coupon specimens. Therefore, the variation in agreement between the numerical and
experimental - curves at the point of coupon rupture was deemed inconsequential and, for the same reason,
a fracture criterion was not calibrated for the TCs.
Fig. 5.7. Comparison of numerical and experimental - curves (shown for CHS 127.4 × 11.55 TC)
Fig. 5.8. Mesh arrangements shown on a 1/4 θ = 90° FE CHS-to-rigid plate connection
Fig. 5.9. Load-displacement curves for CHS test no. 26 with fine, medium and coarse mesh, and four CHS through-
thickness elements
A medium mesh and SOLID45 elements (using reduced integration and hourglass control) were selected
for the models based on the results of this study and previous HSS connection FE studies (Voth & Packer
2012a, 2012b) that also obtained good results with these parameters. Seven elements along the weld face, six
elements along the vertical weld leg, and six elements along the horizontal weld leg (see Fig. 5.4, shown
previously), were found to be suitable for achieving good resolution of the weld response under applied
loads. For the HSS, four through-thickness elements, biased towards the outside face of the branch in contact
with the weld, were used to capture local stresses and bending effects due to eccentric loading. Three
elements were used through the plate thickness; however, the number of plate through-thickness elements
was found not to affect the results.
The element “death feature” was thus programmed for elements in both the weld and the end-plate, as a
loop within each load step, to determine the FE ultimate load (PFE), the failure mode (W, P, or WP), and the
sequence of failure (the location of first-cracking and subsequent crack propagation).
Table 5.2. Comparison of experimental and FE results for HSS-to-rigid plate tests
Failure Mode Ultimate Load
Branch Test mean
Type No. θ Ab Bb/tb or Db/tb tw/tb Exp. FE Pa PFE Pa /PFE (A/P) COV
° mm2 kN kN
1 0.46 W W 831 841 0.99
2 0.76 W W 1166 1252 0.93
RHS 90 3563 16.3
3 0.68 W W 1235 1147 1.08
Oatway
(2014)
2160 13.8
27 0.84 W W 864 877 0.99
28 0.79 W W 1450 1207 1.20
3393 25.1
29 0.86 W W 1331 1324 1.01
30 0.45 W W 1109 1278 0.87
CHS 60 4204 11.0 1.02 0.12
31 0.59 W W 1479 1601 0.92
32 0.73 W W 776 763 1.02
2160 13.8
33 0.68 W W 803 743 1.08
Note: W = weld failure; WP = mixed failure mode of weld failure and partial plate failure; P = end-plate rupture along at
least one weldment.
a
ri and ro were taken as 1.0tb and 2.0tb, respectively, for all RHS used by Frater (1986).
Frater’s (1986) tests showed that it is possible for cracks that begin in the weld to propagate into the
plate, or vice versa. Since the FE models were terminated once PFE was reached (to avoid problems that
would later occur with convergence), it is not possible to determine if this phenomenon would have occurred
in the numerical tests. However, to assess failure progression in the FE models from first cracking (the death
of the first element) to the ultimate load (PFE), a list containing the numbers of “killed” elements was output
from each load step. In the 90° RHS-to-rigid plate connections, the first weld elements to fail were generally
close to the RHS corners, whereas the first plate elements to fail were near the middle of the RHS wall. These
failure locations, and the progression of joint failure for test no. 13, which failed in the weld and the end-
plate, are illustrated in Fig. 5.10. The above behaviour suggests that non-uniform stress distributions in RHS
welded joints can still be caused by a difference in relative flexibility of the branch parallel to its surface,
around its perimeter, leading to weaker welds (per unit length) for RHS-to-rigid plate connections versus
CHS-to-rigid plate connections.
The first weld elements to fail in the θ = 60° RHS-to-rigid plate FE tests were located at the toe of the
connection; and when end-plate failure did occur numerically the first end-plate elements to fail were also
located at the toe [and not at the heel, as Frater (1986) observed]. Therefore, more experimental tests may be
necessary to understand the conditions that lead to end-plate rupture in skewed HSS-to-rigid plate
connections (θ < 90°). It is also worth noting that while element failure typically occurred over several load
steps in the RHS-to-rigid plate FE tests (i.e. a crack formed in the weld near the RHS corner, and
propagated), it usually occurred in a single load step in the CHS-to-rigid plate FE tests.
therein with respect to the elastic limits, between the experimental and FE results, are believed to be due to
residual stresses in the RHS, which were not modelled.
• two on RHS-to-rigid plate specimens with Bb/tb = 12.5 and 50, and θ = 60° and tw/tb = 0.50;
• two on RHS-to-rigid plate specimens with Bb/tb = 12.5 and 50, and θ = 75° and tw/tb = 0.50;
• two on CHS-to-rigid plate specimens with Db/tb = 12.5 and 50, and θ = 60° and tw/tb = 0.50; and
• two on CHS-to-rigid plate specimens with Bb/tb = 12.5 and 50, and θ = 75° and tw/tb = 0.50.
These eight analyses were then compared to four of the previous FE analyses on specimens with θ = 90°
and tw/tb = 0.50 (two RHS-to-rigid plate specimens with Bb/tb = 12.5 and 50, and two CHS-to-rigid plate
specimens with Db/tb = 12.5 and 50). A total of 73 (65 + 8) analyses were hence performed.
The HSS width, or diameter, and the end-plate thickness were kept constant for all joints (Bb = 200 mm,
Db = 168 mm, and tp = 25 mm, respectively), and fillet welds were modelled with equal-sized legs. A 0.25-
mm gap between the branch member and the end-plate was used to restrict load transfer to the weld, as
discussed in Section 5.3.1. All models used SOLID45 8-noded hexahedral elements and were permitted to
fail either by weld or end-plate rupture, using the fracture criterion discussed in Section 5.3.4.
The – curves for each of the materials (cold-formed HSS, end-plate, and weld metal), developed from
experimental TC tests, are shown in Fig. 5.13. These material properties were used for all connections
irrespective of geometry and branch type. The material properties were: yield stress of HSS (Fyb) = 421 MPa,
ultimate stress of HSS (Fu) = 501 MPa, yield stress of end-plate (Fyp) = 409 MPa, ultimate stress of end-plate
(Fup) = 566 MPa, yield stress of weld metal (Fyw) = 501 MPa, and ultimate stress of weld metal (FEXX) = 571
MPa. These values were chosen to provide a similar level of base metal and weld metal over-strength (the
difference between the actual and minimum specified values used in design), to represent the standard
assumption in connection design that matched electrodes are used. Outside and inside corner radii of the RHS
(ro and ri) were taken as 2.0tb and 1.0tb, respectively.
Fig. 5.13. Engineering stress-strain curves for the cold-formed HSS, end plate, and weld metal used in the FE
parametric study
Early on in the parametric study, it was found that significant yielding of the branch occurred in CHS-to-
rigid plate connections with tw/tb > 1.06, preventing the analyses from completing. The range of tw/tb was
therefore reduced to 0.35 ≤ tw/tb ≤ 1.06 for the CHS specimens. Tables 5.3 and 5.4 show the non-dimensional
connection parameters (Bb/tb or Db/tb, and tw/tb) for the θ = 90° models in the parametric study.
Table 5.3. Non-dimensional parameters and FE parametric study results (PFE/AwFEXX and failure mode) for
RHS-to-rigid plate connections
(Bb = 200 mm) tw/tb
tb Bb/tb
mm 0.35 0.50 0.71 0.90 1.06 1.41
4.00 50 0.80 W 0.76 W 0.72 W 0.65 W 0.65 W 0.59 W a
5.00 40 0.80 W 0.77 W 0.73 W 0.69 W 0.67 W 0.59 P a
6.67 30 0.82 W 0.77 W 0.74 W 0.70 W 0.67 P 0.56 P a
10.00 20 0.84 W 0.80 W 0.76 W 0.71 P 0.64 P 0.54 P a
16.00 12.5 0.87 W 0.83 W 0.72 P 0.63 P 0.56 P 0.49 P a
22.00 9.1 0.88 W 0.84 W 0.64 P 0.57 P 0.53 P -a
a
Connections noted thus experienced nominal branch yielding.
Note: W = rupture through the weld, around the entire branch perimeter; P = rupture only in the end-plate.
Table 5.4. Non-dimensional parameters and FE parametric study results (PFE/AwFEXX and failure mode) for
CHS-to-rigid plate connections
(Db = 168 mm) tw/tb
tb Db/tb
mm 0.35 0.50 0.71 0.90 1.06 1.41
3.36 50 0.85 W 0.87 W 0.81 W 0.76 W 0.73 W a -a
4.20 40 0.89 W 0.86 W 0.82 W 0.76 W 0.73 W a -a
5.59 30 0.88 W 0.88 W 0.82 W 0.78 W 0.74 W a -a
8.40 20 0.90 W 0.88 W 0.85 W 0.80 W a 0.76 W a -a
13.44 12.5 0.93 W 0.89 W 0.87 W 0.79 P a 0.71 P a -a
18.48 9.1 0.92 W 0.90 W 0.83 P 0.71 P 0.65 P a -a
a
Connections noted thus experienced nominal branch yielding.
Note: W = rupture through the weld, around the entire branch perimeter; P = rupture only in the end-plate.
tw
tb t
w 0.035
Bb Bb (5.4)
tb
For weld rupture to govern over plate rupture in CHS-to-rigid plate connections:
tw
tb t
w 0.072 (5.5)
Db Db
tb
When tw/Bb > 0.035 or tw/Db > 0.072, plate rupture can be expected to govern. The use of non-
dimensional parameters to study the behaviour of HSS connections is a common practice which allows
results to be generalized for all sizes of HSS; thus, Eqs. (5.4) and (5.5), and the remainder of the results, are
believed to be valid for smaller or larger sizes of Bb or Db than those selected in this study.
All of the end-plate ruptures in the FE RHS-to-rigid plate tests were observed at mid-wall of the RHS, a
critical location validated by experimental tests (Frater 1986). Furthermore, the values of tw/Bb for the
experimental RHS-to-rigid plate tests that failed by end-plate rupture satisfied Eq. (5.4). Since end-plate
rupture did not occur in any of the experimental CHS-to-rigid plate tests, more work is necessary to validate
Eq. (5.5).
the longitudinal axis of the weld increases. This moment causes bending of the branch wall inward and
tension near the weld root, which is well-known to be a situation that can lead to premature weld failure. This
detrimental feature of “single-sided fillet welds” is inherent to all HSS fillet-welded joints. Fig. 5.15, from FE
analysis, illustrates this rotation of the branch wall due to eccentricity in two fillet-welded RHS joints. In Fig.
5.15, the von Mises equivalent stress is overlaid atop the deformed joint at failure. This effect is greater for
RHS than CHS, which have less restraint from rotation along the flats, especially for slender sections.
Fig. 5.14. Effect of the ratio tw/tb on fillet weld strength in θ = 90° HSS-to-rigid plate connections: (a) for
RHS branches; (b) for CHS branches
Fig. 5.16. Effect of the ratios Bb/tb and Db/tb on fillet weld strength in θ = 90° HSS-to-rigid plate connections:
(a) for RHS branches; (b) for CHS branches
Connections with slender branches and high values of tw/tb, especially RHS connections, exhibited
significant rotations of the branch wall adjacent to the weld inwards at failure, as shown previously in Fig.
5.15. When this rotation was small (i.e. branches were stocky, and tw/tb was small), PFE/AwFEXX was closer to
the value predicted using the sinθ factor, thus illustrating that fillet-welded RHS- and CHS-to-rigid plate
connections, and other single-sided fillet welded connections, merit special attention. This is effect is also
believed to explain the higher strength of CHS connections compared to RHS connections.
B t
For RHS: Pnw 0.954 0.00193 b 0.210 w Aw FEXX (5.6)
tb tb
D t
For CHS: Pnw 1.009 0.00137 b 0.197 w Aw FEXX (5.7)
tb tb
which provide R2 values of 0.97 and 0.93, respectively, and mean actual-to-predicted (A/P) ratios and COVs
of 1.00 and 0.02 (both are the same) when compared to the finite element results.
It is also possible to neglect the effect of branch wall slenderness and still produce a sufficiently accurate
equation for Pnw, using linear regression, as follows:
t
For RHS: Pnw 0.924 0.262 w Aw FEXX (5.8)
tb
t
For CHS: Pnw 0.984 0.226 w Aw FEXX (5.9)
tb
Eqs. (5.8) and (5.9) provide R2 values of 0.83 and 0.77, mean A/P ratios of 1.02 and 1.00, and COVs of
0.04 and 0.03, respectively, when compared to the finite element results. The correlations of Eqs. (5.8) and
(5.9) against both finite element and experimental results are shown as dashed lines in Fig. 5.14 (presented
earlier). While these are simple equations for predicting the strength of fillet welds to HSS, they are not well-
suited for design because they are quadratic with respect to tw (since Aw = twlw). By simplifying the constants
in Eqs. (5.8) and (5.9), and substituting Pr/Py for tw/tb (where Pr = required weld strength, in units of force, and
Py = yield force of the branch), Eqs. (5.8) and (5.9) can be rewritten as:
P
For RHS: Pnw 0.90 0.25 r Aw FEXX (5.10)
Py
P
For CHS: Pnw 1.00 0.25 r Aw FEXX (5.11)
Py
In a worst-case scenario, the required weld strength Pr will equal the weld strength (Pa or PFE), which
becomes disproportionately larger as tw increases. The branch yield load Py is linearly related to tb. By
comparing the relationship between the plotted points and the diagonal lines on Figs. 5.14 and 5.18, it can be
seen that substituting tw/tb with Pr/Py still provides a good prediction of the weld rupture load. Importantly, it
also allows a solution to be found for tw using a linear equation. Graphical comparisons of Pa/Py and PFE/Py to
tw/tb are provided in Appendix D.1.
Comparing Eqs. (5.10) and (5.11) to the FE results yields mean A/P ratios of 1.03 (both are the same),
and COVs of 0.04 and 0.05, respectively. Comparing the same equations to the weld-critical experimental
results yields mean A/P values of 1.08 and 1.04, and COVs of 0.17 and 0.10, respectively. The correlation of
Eq. (5.10) against the FE and experimental RHS results, and the correlation of Eq. (5.11) against the FE and
experimental CHS results, is shown in Fig. 5.18.
Fig. 5.18. Evaluation of design equations against all 90° (FE and experimental) HSS-to-rigid plate test
results: (a) for RHS branches; (b) for CHS branches
(a) AISC 360 (2016) and CSA S16 (2014) (with and without the sinθ factor);
(b) EN 1993-1-8 (CEN 2005) Directional Method and Simplified Method; and
The predicted nominal weld strengths (Pnw) for each of the θ = 90° connections that failed by weld
rupture were calculated as described in Sections 4.4.1 (for AISC) 4.4.2 (for CSA) and 4.4.4 (for EN1993-1-
8), and are given in Appendix D.2.
Because the product 0.67(ϕ=0.67) = 0.60(ϕ=0.75), the predicted weld resistances – and, therefore, the
resulting reliability index – are identical for AISC 360 (AISC 2016) and CSA S16 (CSA 2014), both with and
+
without the sinθ factor. The following calculations for are therefore presented only for AISC 360 (AISC
2016) with respect to (a) above.
Tables 5.5 and 5.6 summarize the key parameters of the reliability analysis. In both tables, Eqs. (5.10) or
(5.11) have been assessed using ϕ = 0.75 and 0.67, to determine their reliability when used with the resistance
factors for weld metal in AISC 360 (AISC 2016) and CSA S16 (CSA 2014), respectively.
Table 5.5. Reliability analysis parameters with respect to 21 weld-critical RHS-to-rigid plate FE test results
with θ = 90°
AISC 360-16 EN1993-1-8:2005 Eq. (5.10)
0.60 0.60(1.0+0.50sin1.5θ) Directional Method Simplified Method ϕ = 0.75 ϕ = 0.67
ϕ 0.75 0.75 1/ M2 = 0.80 1/ M2 = 0.80 0.75 0.67
ρM 1.12 1.12 1.12 1.12 1.12 1.12
VM 0.12 0.12 0.12 0.12 0.12 0.12
ρG 1.03 1.03 1.03 1.03 1.03 1.03
VG 0.10 0.10 0.10 0.10 0.10 0.10
ρP 1.27 0.85 1.11 1.36 1.03 1.03
VP 0.09 0.09 0.09 0.09 0.05 0.05
ρR 1.47 0.98 1.28 1.57 1.19 1.19
VR 0.18 0.18 0.18 0.18 0.16 0.16
ϕ+ 0.84 1.02 0.92 0.84 0.91 0.86
+
5.00 a 2.85 b 3.90 4.99 4.05 4.67
a +
identical to CSA S16-14 method without the sinθ factor.
b +
identical to CSA S16-14 method with the sinθ factor.
Table 5.6. Reliability analysis parameters with respect to 20 weld-critical CHS-to-rigid plate FE test results θ
= 90°
AISC 360-16 EN1993-1-8:2005 Eq. (5.11)
0.60 0.60(1.0+0.50sin1.5θ) Directional Method Simplified Method ϕ = 0.75 ϕ = 0.67
ϕ 0.75 0.75 0.80 0.80 0.75 0.67
ρM 1.12 1.12 1.12 1.12 1.12 1.12
VM 0.12 0.12 0.12 0.12 0.12 0.12
ρG 1.03 1.03 1.03 1.03 1.03 1.03
VG 0.10 0.10 0.10 0.10 0.10 0.10
ρP 1.43 0.95 1.24 1.52 1.03 1.03
VP 0.06 0.06 0.06 0.06 0.04 0.04
ρR 1.65 1.10 1.43 1.75 1.19 1.19
VR 0.17 0.17 0.17 0.17 0.16 0.16
ϕ+ 0.78 0.95 0.86 0.78 0.91 0.86
+
5.87 a 3.57 b 4.69 5.87 4.09 4.72
a +
identical to CSA S16-14 method without the sinθ factor.
b +
identical to CSA S16-14 method with the sinθ factor.
By examining the -values in Tables 5.5 and 5.6, it can be seen that the sinθ factor is still unsafe
(provides less than the target value) when applied to fillet welds in HSS-to-rigid plate connections,
irrespective of whether the branch is an RHS or CHS member, and even when a wider range of connection
parameters (weld size and branch slenderness) is considered. By examining ρP and +
in both tables, it can be
seen that Eqs. (5.10) and (5.11) provide accurate (ρP close 1) and safe predictions ( +
≥ 4.0 and 4.5) of the
fillet weld ultimate strength in RHS- and CHS-to-rigid plate connections.
1 Fyb
tw tb (5.12)
0.65 FEXX
will develop the branch yield strength. Therefore, for a 350W RHS branch member (Fyb = 350 MPa) welded
all around to a plate with a matched electrode (FEXX = 490 MPa):
tw 1.10tb (5.13)
1 Fyb (5.14)
tw tb
0.75 FEXX
(5.15)
tw 0.95tb
A more reasonable estimate for inclined branches (θ < 90°) can be obtained by taking into account the
longer weld length due to the branch angle, for example, by dividing Eq. (5.15) for CHS branches by the
weld length factor (Ka) given by the American Welding Society (AWS 2015):
1 1 / sin
Ka (5.16)
2
5.9. SUMMARY
Finite element models were developed for fillet-welded RHS-to-rigid plate and CHS-to-rigid plate
connections and these were validated against 33 laboratory tests on such welded joints. A parametric study
was then performed in which a total of 73 FE RHS- and CHS-to-rigid plate connections were tested to failure
under axial tension loading. The parametric study investigated the effect of the ratio of the weld throat
dimension to the branch wall thickness (tw/tb), the branch wall slenderness (Db/tb and Bb/tb) and the branch
inclination angle (θ) on the weld strength. It was found that:
1. As tw/tb increases, the average stress on the weld throat area at failure significantly decreases.
2. As Db/tb and Bb/tb increase, the average stress on the weld throat area at failure slightly decreases.
3. The branch inclination angle (θ) has a negligible effect on the weld strength per unit throat area;
however, the longer weld length that results from a reduction in branch angle increases the
absolute strength of the weld.
Equations were then developed to estimate the strength of fillet welds in RHS- and CHS-to-rigid plate
connections. When subjected to a reliability analysis with respect to the weld-critical FE results, these
equations were found to provide an adequate level of safety for use in North America. It is shown, using these
equations, that a weld throat dimension equal to 1.10tb (for an RHS branch) and 0.95tb (for a CHS branch)
will develop the yield strength of the connected branch member at 90° to a rigid plate, for 350 MPa yield-
strength HSS with matching electrode. It was hence found that welds in RHS-to-rigid plate connections were
generally not as strong as welds in CHS-to-rigid plate connections (per unit length) due to non-uniform
membrane stresses parallel to the branch, which peak at the RHS corners, and less restraint from local
rotations in RHS flats versus curved CHS walls.
The design methods for fillet welds to HSS members given in AISC 360 (AISC 2016), CSA S16 (CSA
2014), and EN1993-1-8 (CEN 2005) were evaluated with respect to North American safety index
requirements, looking at fillet welds in RHS- and CHS-to-rigid plate connections independently, and using a
reliability analysis that included the mean values and variations in material, geometric, and professional
factors. The sinθ factor in AISC 360 (AISC 2016) and CSA S16 (CSA 2014) was found to provide
inadequate levels of the safety index when used to design fillet welds to HSS.
For all HSS connections, including HSS-to-HSS connections where the effective length concept is used,
and even HSS connections in which the welds are fully effective, it is therefore recommended that the
provisions of AISC 360 (AISC 2016) and CSA S16 (CSA 2014) for the design of fillet welds be used without
the sinθ factor [i.e. taking θ = 0° in the term (1.00 + 0.50sin1.5θ)], or that the alternate method, developed
herein, be used with North American resistance factors.
6.1. INTRODUCTION
When welding to hollow structural sections (HSS), welds can be proportioned: (a) to achieve the capacity
of the connected branch member wall; or (b) to be “fit for purpose” (ISO 2013). By designing welds as fit for
purpose – to resist the actual forces present in the branch member – smaller, more appropriate weld sizes
typically result.
In order to account for the non-uniform loading of the weld perimeter due to differences in the relative
flexibilities of the chord loaded normal to its surface, and the branch(es) carrying membrane stresses parallel
to its surface, weld effective properties – including weld effective lengths and weld effective section moduli –
are used. These properties are determined by discounting segments of the weld which do not contribute to its
overall resistance.
Over the last 30 years, much research has been conducted at the University of Toronto to determine weld
effective lengths for rectangular hollow section (RHS) connections, including gapped K-connections, T-, Y-
and X- (or Cross-) connections, moment-loaded T-connections, and overlapped K-connections (Frater &
Packer 1992a, 1992b; Packer & Cassidy 1995; McFadden & Packer 2014; and the work of Chapter 2).
Recommendations based on this research have been adopted as code in North America, by the American
Institute of Steel Construction (AISC) in Section K5: Welds of Plates and Branches to RHS of their latest
specification, AISC 360-16.
Since the addition of Section K5 (formerly Section K4, in the 2010 specification), weld effective
properties for circular hollow section (CHS) connections have been an issue faced by code writers, including
AISC and the American Welding Society (AWS), since load transfer around a welded CHS joint can be
highly non-uniform (Marshall 1992) (e.g. Fig. 6.1).
93
Chapter 6: Fillet Weld Effective Lengths in CHS X-Connections: Experimentation 94
While AISC 360-16 is noticeably silent regarding weld effective lengths for CHS connections, AWS
D1.1 Structural Welding Code – Steel (AWS 2015) implies, in Clause 9.6.1.3(4), that the weld effective
length in axially loaded CHS connections is equal to 1/1.5 of the total weld length under factored loads,
regardless of the joint geometry. While believed to be conservative, this rule is not supported by experimental
evidence.
A laboratory-based test program was hence conducted to assess the performance of welds in CHS
connections. For the first time, weld-critical tests (where failure occurs by weld fracture) have been
completed on fillet welds in full-scale CHS connections, and the structural reliability (safety index) of the
existing AWS, AISC and CSA specification provisions for the design of such welds is determined. The effect
of key connection parameters on the weld strength is also investigated, and an empirical method to quantify
the weld effective length is proposed. The results of this chapter are vital for determining a strategy for the
fit-for-purpose design of welds in CHS connections is both accurate (reasonably predicts the correct failure
load) and safe (meets or marginally exceeds target reliability indices provided by design codes).
depends on branch-to-chord diameter ratio ( = Db/D), chord wall slenderness (D/t), and branch inclination
angle (θ), the chord and branch members were selected to cover a wide range of these values, as shown in
Table 6.1. These values were selected to be within limits for fillet welds to develop the full throat thickness,
as given by AWS D1.1-15 (see Section 6.2.1). Four connections had branches at θ = 90° to the chord, and
two connections had branches at θ = 60° to the chord. The connection layout is shown in Fig. 6.2, which
includes definitions for the nomenclature used herein. Appendix E.1 provides specimen fabrication drawings.
Db × tb Ab b Fyb c
D×t Ab Fy c
curves for connections with different values, calculated using the same method, have also been included in
Fig. 6.3. These curves can be compared to those given in Informative Annex O of AWS (2015) to verify the
approach used.
Fig. 6.3. Local dihedral angle curves for test joints, with subtended angle measured from the crown heel
The branches were cut to a minimum branch length (lb) of 6Db, to avoid shear lag effects at mid-length,
from both ends (Mehrota & Govil 1972), and profiled to saddle perfectly onto the chords, without edge
bevelling (Fig. 6.4). The branches were capped by a tee connection through which load was applied. The tee
connection was designed, using Section 7.6 of Packer & Henderson (1997), to develop the member capacity.
The chords were cut to an overall chord length (l) to avoid end effects at the connection (van der Vegte &
Makino 2010). To economize on material, they were left unrestrained (uncapped) at both ends. The average
measured material properties for the CHS branch and chord members were determined from three tensile
coupon tests per section, performed according to ASTM A370 (ASTM 2017) while maintaining the curved
shape. One tensile coupon was taken from each CHS directly across from the weld seam, and the other two
were taken from the CHS faces orthogonal to the weld seam.
(a) Typical 90° connection (shown for 102-273-90a) (b) Typical 60° connection (shown for 102-406-60b)
Fig. 6.4. Fit-up of branch to chord after profiling and tack welding
D D
Pi lt , b sin i , b cos i (6.1a)
2 2
where:
D (1 cos i) D D Db sin i
2 2
(6.1b)
lt b
2 tan 2sin
⃗ by taking the square root of the sum of squares of its three vector
Step 3: Compute the magnitude of 𝑉
components. This is the approximation to the weld length between 𝑃⃗𝜌 and 𝑃⃗𝜌+𝛥𝜌 . The smaller Δρ is, the closer
the approximation will be to the actual weld length.
⃗ to the
Step 4: Increment ρ by Δρ, and repeat Steps 1 to 3, adding the new result for the magnitude 𝑉
previous results. For the total weld length, continue repeating Steps 1 to 3 until ρ is equal to 360° - Δρ. For an
interval length along the weld, continue repeating Steps 1 to 3 until ρ is exactly Δρ less than ρ at the end of
the interval.
The weld lengths herein were calculated as described above with Δρ equal to 1°. This generally gave the
same answer for total weld length as an exact solution based on calculus, but could be more easily applied to
a range of different joint geometries. Lie et al. (2001) also provided equations to describe the geometry of
butt welds in HSS Y-connections. If used to calculate Ѱ or lw for the current tests, the results would match the
Authors’ because both approaches are based strictly on the connection geometry.
After being laid, welds were ground into a triangular shape, with a near-uniform throat size (tw) around
each joint, and flat weld faces. Flat weld faces allowed tw to be obtained from a 3D model of the weld’s exact
geometry, as shown in Fig. 6.6. Using this approach, the orientation of tw and the weld legs (lv and lh) must be
established correctly: in the plane of Ѱ, perpendicular to the weld root, between tangents to the outside
surfaces of the branch and the chord. The computer-aided design (CAD) program Solidworks was employed
to exact this requirement.
First, local components of lv and lh in a plane containing the branch axis and the normal to the branch face
were measured at uniform increments of ρ along the weld length. The component of lv parallel to the branch
was first measured by wrapping a mat board collar with reinforced edges around the branch at a fixed
distance (or “offset”) from the root of the weld at the heel of the connection. The offset distance (Fig. 6.5a)
was then measured using a Mitutoyo Digimatic calliper (with a specified resolution of 0.01 mm). The
distance between the collar and the weld toe along the branch (x) was then measured, using the same calliper,
at uniform increments of ρ along the weld length. The component of lv parallel to the branch at ρ could then
be calculated by subtracting this measured distance (x) from the theoretical distance between the collar and
the weld root (= lt + offset). Historically, lt refers to the template length, which is the length (parallel to the
branch) of a steel-cutting template that was wrapped around a CHS branch and used as a guide for profiling
with a torch. The component of lh perpendicular to the face of the branch at ρ was measured by laying a
standard fillet weld gauge along the axis of the branch and measuring the distance to the weld toe. The weld
profile around the entire joint was then modelled in Solidworks using these measurements and the measured
values of Db and D. Finally, sections were taken through the weld in the plane of Ѱ using Solidworks
geometry tools, and lv, lh and tw were precisely measured, as shown in Fig. 6.6.
The weld area (Aw) was determined by summing up: tw × tributary weld length (Fig. 6.6) around the entire
joint. The measured fillet-weld geometric properties are shown in Table 6.2, and the individual measurements
along each weld, made externally and with Solidworks, are provided in Appendix E.3. Appendix E.4
provides the 3D Solidworks models of the weld profiles generated for all 12 test joints.
Table 6.2. Weld dimensions and predicted fracture loads for test joints according to existing AWS D1.1-15
provisions for weld effective lengths in CHS X-connections
Test Average measured weld dimensions Pnwa
lv lh tw lw Aw
mm mm mm mm mm kN
102-273-90a 6.86 6.17 4.08 322 1312 303
102-273-90b 7.23 6.65 4.37 322 1405 324
102-406-90a 5.16 5.78 3.56 320 1139 263
102-406-90b 4.54 5.08 3.14 320 1004 232
127-273-90a 5.94 5.93 3.63 406 1475 340
127-273-90b 7.05 6.06 4.00 406 1625 375
127-406-90a 4.83 5.03 3.16 403 1273 294
127-406-90b 5.60 5.19 3.47 403 1410 323
102-406-60a 5.83 5.59 3.58 345 1235 285
102-406-60b 6.29 5.83 3.79 345 1307 302
127-406-60a 5.68 8.01 3.95 434 1716 396
127-406-60b 5.39 6.00 3.38 434 1468 339
a
Nominal predicted fracture load according to existing AWS (2015) specification provisions, calculated using Eqs.
(6.2a,b) and (6.3), using Aw and FEXX determined from tensile coupon tests (= 577 MPa).
Table 6.3. All-weld-metal tensile coupon test results for CHS X-connection tests
All-weld-metal coupon Yield stress Young’s modulus, FEXX Rupture strain,
designation E εrupa
MPa MPa MPa %
[i] 510 189000 573 30.6
[ii] 520 201000 576 27.4
[iii] 521 235000 581 26.4
Average 517 208000 577 28.1
a
Rupture strain determined by re-joining the fractured coupon and measuring: change in gauge length / initial gauge
length
Four linear strain gauges (SGs), equally spaced around the perimeter of the branch at mid length (≥ 3Db
from the welded test joint and the end), and oriented along its longitudinal axis, were used to measure the
uniformity of load being applied to the branch. Equal strains were typically measured at all four locations
over all tests, indicating that predominantly axial load was applied.
Seven additional SGs, with the same orientation, were used around half the weld perimeter (i.e. on one
side of the branch only, due to symmetry) to measure non-uniform loading of the weld perimeter (Fig. 6.8).
For this purpose, SGs were centred 20 mm away from the weld toe, to avoid stress concentrations that occur
there due to the notch effect (Packer & Cassidy 1995), and located at ρ = 0°, 30°, 60°, 90°, 120°, 150°, and
180°. A single SG in the saddle position on the opposite side (at ρ = 270°) was used to verify symmetry of the
strain distribution about the plane of the connection. In all, 12 welds were tested to rupture (two per
connection). All welds failed in a brittle manner, by fracture along a plane through the weld. Four typical
failures are shown in Fig. 6.9, for both 90° connections (Figs. 6.9a,b) and 60° connections (Figs. 6.9c,d).
Additional photographs of the test set-up and failure mode for each test are provided in Appendix E.7.
Fig. 6.8. Strain gauges near weld toe (and weld fracture) in test 127-273-90a
After the first test weld (e.g. side a) ruptured in each connection, the branch was re-positioned within the
UTM and tack welded back in place. The entire connection was then removed from the UTM, and fully re-
welded (nominally in the flat position) to ensure that separation of the same branch did not occur again. The
connection was re-placed in the UTM, and tested until rupture of the second test weld (e.g. side b) occurred.
Chord deformation ( ) was continuously monitored throughout both tests with three LED targets: one on each
branch, 50-mm above the crown; and one at the connection work point on the chord face parallel to the plane
of the connection (Figs. 6.7, 6.10). The value of , which is defined as the outward displacement, normal to
the chord, of a single branch from the chord centreline (Packer et al. 2012), was taken as the normal
component of half of the displacement between the LEDs on each branch (Fig. 6.10). It therefore represents
the average deformation on both sides of the connection. The actual weld fracture loads (and hence Pa and
Paꞌ, given in Table 6.1) were obtained from load cells in-line with the UTM actuator, and verified by
comparison with forces computed from average SG readings of strain at mid-length of the branch and the
measured branch cross-sectional area (Ab) and Young’s modulus (E).
applied load response for the second test was then obtained by extending the corresponding first curve past its
point of rupture until it met the second curve. This part of the curve is illustrated by the dashed black line in
Fig. 6.11. The chord deformation at rupture ( a) for the six second welds tested (e.g. side b) could then be
obtained. Despite having only small fillet welds, chord plastification in excess of the 3%D deformation limit
(Lu et al. 1994) occurred before rupture in seven out of the 12 tests (Table 6.4).
Fig. 6.10. Typical load versus chord deformation Fig. 6.11. Procedure for calculation of load versus
relationships deformation for second welds tested
Table 6.4. Residual chord deformation (at start of test) and chord deformation at rupture for all 12 tests
Test Residual chord deformation (as percent of D) a/D
% %
102-273-90a 0 3.23
102-273-90b 1.732 3.68
102-406-90a 1.577 4.70
102-406-90b 0 3.52
127-273-90a 0 2.06
127-273-90b 0.914 2.07
127-406-90a 0 2.61
127-406-90b 0.962 2.78
102-406-60a 0 3.34
102-406-60b 1.759 3.63
127-406-60a 0 2.34
127-406-60b 1.712 3.60
Note: residual chord deformations are equal to zero for first weld tested on each connection
(a) Test 127-273-90a ( = 0.47, θ = 90o) (b) Test 127-406-90a ( = 0.31, θ = 90o)
Fig. 6.12. Typical strain distributions adjacent to test weld (θ = 90° connections)
(a) Test 102-406-60a ( = 0.25, θ = 60o) (b) Test 127-406-60a ( = 0.31, θ = 60o)
Fig. 6.13. Typical strain distributions adjacent to test weld (θ = 60° connections)
It is shown that, for θ = 90° connections (Figs. 6.12a,b), the tensile strain (and hence tensile load)
decreases as a function of distance away from the saddle (ρ = 90° point). The tensile strain is therefore
smallest at the crown (ρ = 0° and 180° points), with much of the weld even remaining in compression for the
entire tension load range. This phenomenon equates to a non-uniform loading of the weld perimeter – which
is expectedly more pronounced for connections with higher -values, where stiff membrane action dominates
load transfer at the saddle. It can thus be concluded that weld effective lengths are present in CHS-to-CHS
connections. An illustration of the effect that causes compression at the crown is shown in Fig. 6.14.
Fig. 6.14. Effect causing compressive strains at the crown (ρ = 0° and 180° points) (θ = 90° connections)
The largest tensile strains for θ = 60° connections were initially measured at the saddle (Figs. 6.13a,b).
As the load increased, the strain adjacent to the saddle, on the heel side of the connection, began to increase at
a faster rate than the strain adjacent to the saddle on the toe side of the connection. This is due to secondary
bending effects from connection flexibility and joint rotation, which may not exist in real structures where the
chord ends are prevented from rotating.
where le = weld effective length (AWS 2015). An LRFD resistance factor for fillet welds, ϕ, equal to 0.80, is
then applied to determine the design strength.
In Clause 9.5.4, simplified equations are given to compute weld lengths for CHS connections under axial
load, which can be traced back to Appendix C of British Standard 449 (BSI 1959). These factors can be
shown to calculate the total weld length, rather than the effective weld length (see Section 6.5.4). A branch
stress/load factor of 1.50 is specified by AWS in Clause 9.6.1.3(4) “Uneven Distribution of Load (Weld
Sizing)”, for design using the LRFD method. This factor, established in the 1980s, is used to prevent
progressive weld failure due to non-uniform load transfer across the weld when welds are designed as fit-for-
purpose. In modern day LRFD, the approach is to apply a reduction to the resistance of the weld, by
calculating a weld effective length, rather than to increase the design load. Hence, it is deduced that the weld
effective length implied by AWS Clause 9.6.1.3(4) is the inverse of the stress/load factor:
1 2
le lw lw (6.3)
1.5 3
+
The implied safety index, , is equal to 7.0 for the existing AWS D1.1-15 specification provisions,
which is much larger than the minimum target safety index of 4.0 in North America. However, it should be
noted that ϕ = 0.80 in AWS (2015) is believed to be an error, as it should match AISC 360 (2016) where ϕ =
0.75. This indicates that a high level of conservatism is present in the AWS formulae. Fig. 6.15 shows the
correlation of the existing AWS predicted nominal strengths with the experimental results. The actual
strength in Fig. 6.15 is taken as the greatest load sustained by the weld (Paꞌ). On average, the actual strength
is 2.13 times larger than that predicted by AWS.
If, instead, no effective length rules are applied, and the total weld length is used to determine the
strength of the welded joint, then the correlation in Fig. 6.16 results. The implied safety index is then 4.9.
+
The mean experimental-to-predicted strength is 1.42. As > 4.0, it can be concluded that, for the range of
parameters studied, weld effective lengths are not required in conjunction with the AWS (2015) code design
method evaluated. The predicted strengths for each connection according to AWS, and the code methods that
follow, are provided in Appendix E.8.
Fig. 6.15. Correlation of existing AWS D1.1-15 provisions with the test results, with weld effective lengths
Fig. 6.16. Correlation of existing AWS D1.1-15 provisions (excluding weld effective lengths) and AISC 360-
16 provisions with test results
(AISC Clause J2.4b and AWS Clause 2.6.4.2), because it has been shown in Chapters 4 and 5 to be generally
unsafe for the design of fillet welds in HSS connections.
The ultimate strength of fillet welds in the Canadian steel code, CSA S16-14, is also based on the limit
state of shear rupture along the weld effective throat; however, CSA gives a different equation than AWS and
AISC for the nominal strength (Pnw) of fillet welds (Clause 13.13.2.2):
where Aw = twlw.
An LRFD resistance factor for fillet welds, ϕ, equal to 0.67, is then applied to determine the design
strength. As discussed for AISC Clause J2.4b (AISC 2016), the above equation also excludes the
1.00+0.50sin1.5θ directional strength-enhancement factor.
CSA gives a higher nominal strength than AISC for fillet welds (0.67 versus 0.60 for the shear strength
+
factor) and a proportionally lower resistance factor (ϕ = 0.67 versus 0.75). The reliability index, , implied
+ +
by CSA Clause 13.13.2.2 is therefore the same as implied by AISC Clause J.2.4a ( = 5.2, in Table 6.5,
which is greater than 4.5, the target safety index per Annex B of CSA S16-14). Fig. 6.17 shows the
correlation of the CSA predicted nominal strengths with the experimental results. On average, the
experimental strength is only 1.27 times larger than that predicted with Eq. (6.4), using the measured values
of Aw and FEXX. This value is the closest to unity amongst all methods investigated (AWS with/without weld
effective lengths, AISC, and CSA). Still, more work is needed to determine the effect of connection
parameters outside of the range studied on the weld strength.
lw Db Ka (6.5)
2
1 3 2 1 3
2
2
1 1
Ka 3 2
(6.6)
2 sin 3 2 2 2 sin 3 2
Eq. (6.6) gives the projection of the weld root along an inclined branch onto a cylindrical surface. It takes
into account both the branch-angle and beta-ratio distortion of the weld length. If one considers a CHS branch
welded to a flat plate at θ = 90°, with lw then equal to πDb, branch-angle distortion is the transformation of the
circular weld into an ellipse caused by a change in θ. Beta-ratio distortion occurs when the flat plate is
replaced by a cylindrical surface, causing the plane of the weld to distort into a saddle shape. Despite its
complex appearance, Eq. (6.6) is only an approximation to the weld length. Note that when equals zero and
θ equals 90°, such as the case for welding a CHS branch at 90° to a flat plate in the example above, Ka does
not equal exactly 1.00 using Eq. (6.6). Instead, it equals 0.99. Hence, as part of a comprehensive evaluation
of the AWS code, it is necessary to evaluate the error associated with this method to calculate lw.
Fig. 6.18. Comparison of lw/πDb using Eq. (6.6) (AWS 2015) and the vector-calculus method
Fig. 6.18 shows the relationship between lw/πDb (= Ka) determined using Eq. (6.6) (shown by the solid
red lines) and lw/πDb determined from the vector-calculus method (shown by the data points), as used herein,
for a range of values. It is shown that Eq. (6.6) is conservative as a design tool, i.e. it under-predicts the
weld length. The maximum error is only 0.6% over the range of parameters studied (for = 0.10 and θ just
less than 90°). Thus, despite its complexity, Eq. (6.6) gives a predicted weld length very close to the actual
weld length.
AWS notes that the following formula for Ka may be conservatively used instead of Eq. (6.6) to calculate
lw for CHS connections:
1 1 / sin
Ka (6.7)
2
Fig. 6.19 (analogous to Fig. 6.18) shows the relationship between lw/πDb determined using Eq. (6.7) and
lw/πDb determined from the vector-calculus method. It is shown that Eq. (6.7) is even more conservative than
Eq. (6.6) as a design tool. The maximum error is still only 1.9% over the range of parameters studied (for =
0.50 and θ = 90°), which is expectedly larger than the error associated with Eq. (6.6). However, Eq. (6.7) is
always conservative.
Fig. 6.19. Comparison of lw/πDb using Eq. (6.7) (AWS 2015) and the vector-calculus method
One could argue that, since these lower-bound approximations give actual-to-predicted weld lengths
greater than 1.00, mean values and variations in the actual-to-predicted weld length should be included in the
factors ρG and VG in the reliability analysis. Including these variations would marginally increase the
reliability index ( +). However, with CAD now widely used for design and analysis, it seems increasingly
likely that designers will opt to find the total weld length using software, rather than using Eqs. (6.6) or (6.7).
It was therefore deemed prudent to omit variations in the actual-to-predicted weld length in the factors ρG and
VG, because including them would be un-conservative. Eqs. (6.6) and (6.7) are still, however, useful (and
conservative) design tools.
• Method 1: The ratio of le to lw is taken as the ratio of nominal-to-peak elastic strain. The nominal
elastic strain was calculated by multiplying strain measured in the branch adjacent to the weld (as shown in
Figs. 6.12a – d and 6.13a,b when P = 0.25Pa), by the weld length tributary to the measurement, then dividing
the sum of the results by the sum of the tributary weld lengths. Since the strain is elastic, the result of using
Paꞌ instead of Pa in this process is not significant. Either the entire weld length, or a length of weld between a
plane or planes of symmetry, should be instrumented with SGs when using this method. Compressive strains
in the branch adjacent to the weld were taken as zero strain (rather than negative strain), because they do not
increase the weld effective length. Wang et al. (2015) used this method to study weld effective lengths for
CHS branch-to-RHS chord connections. Caulkins (1968) used a similar method to study welds in CHS T-
connections. His method used forces from finite element models instead of elastic strains. Due to
proportionality, however, it yields the same results. The appeal of this method is that weld effective lengths
can be determined from elastic tests, and it logically takes into account stress concentrations along the weld
length. Moreover, weld-critical tests, which are difficult to achieve, are not needed. This method is based
explicitly on elastic load/stress distribution, and it does not take into account load/stress re-distribution in the
weld before rupture. It is therefore likely to be a lower-bound.
• Method 2: the weld effective length is empirically determined by comparison of actual-to-predicted
strengths, with Pnw calculated using an accurate predictor such as Eq. (6.8), with actual values of tw, tb, Db, Aw,
and FEXX (Chapter 5):
D t
Pnw 1.009 0.00137 b 0.197 w Aw FEXX
tb tb (6.8)
Eq. (6.8) was developed from regression of a large database of weld-critical CHS-to-rigid plate
connection finite element results (see Section 5.6). As such, it is tailored to the unique loading on single-sided
fillet welds to CHS branches (which produces tension at the weld root under branch axial tension), and
moreover takes into account the principal influential geometric parameters of the CHS member and weld
joint. It is used to predict the strength of a fillet weld to a CHS branch when the weld is fully effective (i.e. le
= lw). The ratio of le to lw for each test is hence the ratio of Paꞌ (Table 6.1) to the predicted value using Eq.
(6.8). Method 2 takes into account load/stress re-distribution in the weld before rupture, unlike Method 1. It
also utilizes an accurate formula for predicting the nominal weld strength [Eq. (6.8)]. Since Eq. (6.8) was
developed from extensive testing and finite element analysis on CHS connections, Method 2 is believed to be
the most accurate way to determine the true weld effective lengths.
Table 6.6 gives the values of le/lw computed for each of the 12 fillet welds in the CHS X-connections
tested herein, using Methods 1 to 3.
Table 6.6. Weld effective length ratios for CHS X-connections determined using three possible methods
Weld effective length ratio, le/lw
Test Method 1 Method 2 Method 3
102-273-90a 0.29 1.01 1.48
102-273-90b 0.29 0.96 1.39
102-406-90a 0.28 1.03 1.54
102-406-90b 0.28 1.03 1.55
127-273-90a 0.28 0.82 1.28
127-273-90b 0.28 0.75 1.16
127-406-90a 0.27 0.81 1.26
127-406-90b 0.27 0.74 1.15
102-406-60a 0.33 1.13 1.69
102-406-60b 0.33 1.08 1.59
127-406-60a 0.32 0.83 1.28
127-406-60b 0.32 1.07 1.67
With Method 1, le/lw is always less than 1.0, and moreover less than 2/3 [Clause 9.6.1.3(4) of AWS
(2015)]. Since le = (2/3)lw has already been shown to be very conservative (Section 6.5.2), it can be concluded
that Method 1 is even more conservative and inaccurate. The aim in assessing the values of le/lw in Table 6.6
is to achieve a weld effective length ratio between 2/3 and 1.0.
With Method 2, le/lw is between 0.75 and 1.13 (Table 6.6). Generally, fillet welds in connections with
similar parameters ( , D/t, ) have similar ratios of le/lw, in accordance with expectations based on previous
research (Caulkins 1968).
For Method 3, using Eqs. (6.2a,b), the weld effective length is always greater than the real length, which
cannot be true. This outcome is not unexpected, since Eq. (6.2) (and AISC 360-16 Clause J2.4a) has already
been shown to be conservative for CHS connections with fully effective welds in Chapter 5.
Method 2 is therefore a logical procedure to determine weld effective lengths from tests. It is thus
deduced that weld effective lengths for CHS vary as a function of connection parameters, and rupture tests
(experimental and/or numerical) over a broad range of geometric parameters are necessary to reasonably
determine weld effective lengths.
6.7. SUMMARY
Based on 12 careful laboratory tests on CHS X-connections under branch axial tension, which all failed
by rupture along a plane through the connecting fillet weld, it is shown that weld effective lengths exist in
CHS-to-CHS connections, and that the existing AWS (2015) code provisions for weld effective lengths in
such connections, given by Clause 9.6.1.3(4), are very conservative. Furthermore, it is shown that the current
AWS (2015), AISC (2016), and CSA (2014) specification provisions provide adequate structural reliability
( +
≥ 4.0 or 4.5) without weld effective lengths (i.e. using the total weld length to determine the weld
strength), assuming the fillet weld directional strength enhancement factor is not used. This is due to the
simplicity of the fillet weld nominal strength formula, because weld lengths were shown (by strain
distributions) to be less than 100% effective. These conclusions are currently limited to 0.25 ≤ ≤ 0.47, 23 ≤
D/t ≤ 34, 0.6 ≤ ≤ 1.0 and θ = 60° or 90°.
A systematic approach to calculating the total weld length in CHS-to-CHS connections has also been
presented. This approach is based on vector calculus and can be easily programmed to allow designers,
fabricators, and researchers of tubular structures to accurately calculate weld lengths. The approximations
given in AWS (2015) Clause 9.5.4 for the total weld length (lw) in CHS connections were compared to the
vector-calculus approach and found to be useful lower-bound design tools for connections with 0.1 ≤ ≤ 0.5
and 60° ≤ θ ≤ 90°.
It is shown that the ratio of the weld effective length to the total weld length (le/lw) is not constant for all
CHS connections, as the current AWS specification currently implies via Clause 9.6.1.3(4), and varies as a
function of connection parameters. It is also concluded that rupture tests on weld-critical connections are
necessary to reasonably determine weld effective lengths.
A comprehensive parametric modelling study, using finite element methods, and a secondary reliability
analysis are henceforth conducted, to determine: (a) if these findings are applicable to a wider range of fillet-
welded CHS X-connections, and (b) the effect of connection parameters , θ, D/t, and on the weld strength.
7.1. INTRODUCTION
To date, only one experimental study (discussed in Chapter 6) has been conducted to evaluate the
performance of welds in hollow structural section (HSS) connections to a circular hollow section (CHS)
chord. Without experimental evidence, specifications have been reluctant to provide a “fit-for-purpose”
approach to weld design that takes into account the non-uniform contribution of the weld perimeter by using
weld effective lengths. Currently, only AWS D1.1 Structural Welding Code – Steel (AWS 2015) gives such a
method, in Clause 9.6.1.3(4), but it is badly defined. Instead, welds in connections to a CHS chord are
routinely sized to develop the yield capacity of the connected branch member. This is done by meeting
prescriptive requirements for the weld throat dimension (tw) as a function of the branch thickness (tb). In rare
cases, it is suggested that welds be designed for the connection capacity, instead of the yield capacity of the
branch (AWS 2015). According to ISO (2013), this approach is not common.
In the lone experimental study (Chapter 6) that evaluated the performance of welds in CHS-to-CHS X-
connections (Fig. 7.1), linear strain gauge (SG) measurements adjacent to the weld showed that load transfer
was highly non-uniform, and peaked in the saddle position. This indicates that a weld effective length
120
Chapter 7: Fillet Weld Effective Lengths in CHS X-Connections: Finite Element Modelling 121
phenomenon exists in CHS connections. A reliability analysis of the experimental results determined that the
weld effective length provisions given by Clause 9.6.1.3(4) of AWS (2015) are highly conservative. For the
range of parameters studied, the AWS (2015), AISC (2016), and CSA (2014) provisions for fillet weld design
+
exceeded the minimum safety index in North America ( > 4.0) even without weld effective lengths.
This chapter presents a parametric modelling study that was performed, using finite element (FE)
methods, to determine: (a) the effect of key connection parameters on weld strength in CHS X-connections;
(b) if these findings are applicable to a wider range of fillet-welded CHS X-connections; and (c) a better
method for estimating fillet weld strength, using weld effective lengths.
where le = weld effective length, and FEXX = ultimate strength of weld metal (a minimum specified tensile
strength is used). An LRFD resistance factor for fillet welds, ϕ, equal to 0.80, is then applied to determine the
design strength of the fillet weld (ϕPnw).
Equations for the total weld length (lw) are given in AWS D1.1-15 Clause 9.5.4, in the following form:
lw Db Ka (7.3)
where Db = diameter of the CHS branch, and Ka = weld length factor, giving the ratio of the total weld length
(taken along the weld root) to the branch circumference. The simplest Ka factor used by AWS is:
1 1 / sin
Ka (7.4)
2
2
le lw (7.5)
3
tw 1.07tb (7.6)
tw t (7.7)
Fleischer & Herion (2015) aimed to reduce the prequalified weld throat from tw = 1.10tb by taking into
account the actual weld length in CHS connections. The actual weld length is a function of the branch
inclination angle (θ) and the branch-to-chord diameter ratio ( = Db/D) (Fig. 7.1). To take the actual weld
length into account, the authors borrowed a more accurate Ka value from AWS D1.1-15 Clause 9.5.4 (Eq.
6.10, in Chapter 6). They then computed the weld strength using the Simplified Method of EN1993-1-8 (CEN
2005), which is a conservative (lower-bound) design alternative to the Directional Method. It is easier to
apply to CHS connections because it does not require stress components on the throat plane to be determined.
Using the CIDECT methodology (i.e. setting weld strength equal to branch capacity), it was shown that the
prequalified fillet weld size could theoretically be reduced for some joints. The required fillet weld throat was
a complex function of , the branch slenderness ratio (Db/tb), and θ (Fig. 7.1). Design charts were hence given
for simplicity. This work was still based on the design philosophy of developing the capacity of the
connected member walls. Moreover, weld effective lengths were not determined.
Fig. 7.2. Load transfer efficiency across the weld of a θ = 90° CHS-to-CHS T-connection, adapted from
Marshall (1992)
In the work by Caulkins (1968) and Marshall (1992), load transfer efficiency was defined as the ratio of
nominal-to-peak elastic load at the connection. This ratio was believed to be related to the weld effective
length. In another study by Wang et al. (2015), load transfer efficiency was defined as the ratio of nominal-to-
peak elastic strain measured adjacent to the weld. This study investigated weld effective lengths for
connections between CHS branches and RHS chords. The definition of load transfer efficiency is marginally
different, but the methods provide the same results because load and strain are linearly related in the elastic
range. Chapter 6 showed that these methods provide a very conservative lower-bound when used to explicitly
determine le.
Table 7.1. Geometric properties of CHS X- (test) connections and comparison of experimental and FE results
Test θ
CHS branch member CHS chord member tw Pa a P aꞌ b a
c
PFE d FE
d
Paꞌ/PFE a/ FE
Db × tb D×t
° mm × mm mm × mm mm kN kN % kN %
102-273-90a 102.0 × 7.34 273.5 × 11.69 4.08 672 672 3.23 655 2.72 1.03 1.19
102-273-90b 102.0 × 7.34 273.5 × 11.69 4.37 678 678 3.68 690 3.03 0.98 1.21
102-406-90a 102.0 × 7.34 406.5 × 12.34 3.56 608 608 4.70 543 3.31 1.12 1.42
102-406-90b 102.0 × 7.34 406.5 × 12.34 3.14 540 540 3.52 495 2.87 1.09 1.23
90
127-273-90a 127.4 × 11.55 273.5 × 11.69 3.63 653 653 2.06 762 2.67 0.86 0.77
127-273-90b 127.4 × 11.55 273.5 × 11.69 4.00 609 653 2.07 811 2.98 0.80 0.69
127-406-90a 127.4 × 11.55 406.5 × 12.34 3.16 557 557 2.61 631 3.21 0.88 0.81
127-406-90b 127.4 × 11.55 406.5 × 12.34 3.47 556 557 2.78 617 3.01 0.90 0.92
102-406-60a 102.0 × 7.34 410.0 × 12.21 3.58 721 721 3.34 640 2.62 1.13 1.28
102-406-60b 102.0 × 7.34 410.0 × 12.21 3.79 538 721 3.63 672 2.84 1.07 1.28
60
127-406-60a 127.4 × 11.55 410.0 × 12.21 3.95 761 761 2.34 903 3.58 0.84 0.65
127-406-60b 127.4 × 11.55 410.0 × 12.21 3.38 798 850 3.60 798 2.87 1.06 1.25
a
Actual (experimental) force in branch member at weld fracture.
b
Greatest load sustained by the weld.
c
Actual (experimental) chord deformation at weld fracture (as % of measured CHS chord dimeter D) averaged over both
branches.
d
PFE and FE are analogous to Paꞌ and a, but refer to the FE values at weld fracture.
The 12 connections were modelled using the commercially available software package ANSYS 14.0
(Swanson Analysis Systems 2011). Although it was possible to model just one eighth of the non-inclined (θ =
90°) connections due to symmetry about three principal planes passing through the connection work point,
one half of each connection was modelled instead, for all connections, to accommodate the inclined branch
cases. This simplified the parametric programming of the models, which was later done using ANSYS batch
files. Symmetry boundary conditions were applied in the plane of the connection (i.e. along the cut face). To
replicate the experimental connection of the branches to the chord, a 0.25-mm gap was modelled between the
two members. This gap ensured that the applied load was only transferred through the fillet weld (i.e. it
prevented direct load transfer between the branch and chord members). The general geometry (showing the
gap), mesh layout, and boundary conditions are shown in Fig. 7.3 (for a non-inclined connection) and Fig. 7.4
(for an inclined connection).
All FE analyses were conducted by applying static incremental displacements (non-linear time-step
analysis) to the ends of the specimen (e.g. the ends of each branch, or for coupon tests, the ends of each
coupon). Large deformation allowance (non-linear geometry) and non-linear material properties (for each
different material) were included.
Fig. 7.3. FE CHS-to-CHS X-connection geometry, mesh layout, and boundary conditions with θ = 90°
Fig. 7.4. FE CHS-to-CHS X-connection geometry, mesh layout, and boundary conditions with θ < 90° (θ =
60° shown)
T (1 ) (7.8)
T ln(1 ) (7.9)
After necking: an iterative approach based on weighting an approximate lower- and upper-bound to the
true stress versus true strain response was used to determine the ordinates on the curve:
T
T T w(1 T T ) (1 w) T
(7.10)
T
T
where 𝜎𝑇′ = true stress at the start of necking, ε′𝑇 = true strain at the start of necking, and w = weighting factor
(Ling 1996).
The weighting factor in Eq. (7.10) was derived for each different material by matching the engineering
stress-strain curve of a tensile coupon modelled in ANSYS to the average engineering stress-strain curve
from the experimental tests. The FE coupon was modelled using the average measured geometry from the
tests, with the original curved shape for CHS coupons. The engineering strains for the FE coupons were
calculated over a 50-mm gauge length from nodes on the exterior of the coupon to closely mimic the
experimental method using a clip gauge. Typical engineering stress-strain curves and the corresponding true
stress-strain curves derived using the above process are shown in Figs. 7.5a,b. In Figs. 7.5a,b the solid line is
the average experimental curve and the dashed line is the FE-generated curve. The true stress-strain curves
continue indefinitely along the x-axis since a fracture criterion was not calibrated for the coupons. Material
properties for the 12 X-connections are given in Chapter 6.
Fig. 7.5. Comparison of typical experimental (solid line) and FE (dashed line) stress-strain curves
During the experimental tensile coupon tests, it was necessary to remove the clip gauge shortly after
necking to prevent damage to it. The comparison of the FE and experimental engineering stress-strain curves
therefore involved the pre-necked portion and a small variable amount of the post-necked portion of the
curve, and the ordinates at rupture (shown with ‘x’ and ‘o’ markers in Fig. 7.5a). The engineering rupture
stress was determined by dividing the load just before rupture by the initial cross-sectional area of the
coupon, and the rupture strain was determined by rejoining the fractured coupon, measuring it to determine
the final gauge length, and dividing the final gauge length by the initial gauge length. At rupture, the necked
shape of the FE coupons closely resembled that of the experimental coupons, with necking at the mid-point.
The same elements ultimately used for the CHS X-connection models (eight-noded solid brick elements)
were used.
The performance of each mesh layout and element type was evaluated by comparing the load-
deformation response of the FE connection to the experimental test (see Chapter 6) (Fig. 7.7a). Table 7.2
compares the values of the initial stiffness (ka = actual (experimental) initial stiffness, kFE = FE initial
stiffness) and the load at a chord deformation ( ) equal to 3%D (the connection plastification limit) (Lu et al.
1994) (P3%D,a = actual load at connection plastification limit, P3%D,FE = FE load at connection plastification
limit). The value of was taken as the chord face deformation (normal to the chord, relative to the chord
centre line) but averaged for both sides of the connection. Displacement was measured between points 50 mm
away from the crown, as show in Fig. 7.7a. In general, as the number of nodes increased by using a finer
mesh, a higher-order element, or more CHS through-thickness elements, the stiffness of the connection
increased and so did the solution time. Fig. 7.7b demonstrates that the solution typically converged with four
elements through the CHS thickness.
Fig. 7.7. Comparison of FE and experimental mesh sensitivity parameters for test 102-273-90a
The best overall agreement with the experimental results was obtained using a medium-density mesh
(Mesh layout A or B), with four elements through the CHS branch and chord member thickness and eight-
noded solid brick elements (SOLID45 in ANSYS). Elements used had reduced integration and hourglass
control. To decide between Mesh layouts A and B, analyses of connections with different branch angles,
values, weld sizes, and CHS member thicknesses were undertaken to foresee any problems associated with
these mesh types. Mesh layout A was found to be capable of mapping to a wider range of connection
geometries (i.e. some models could not be generated using mesh layout B), and was therefore used.
(b) Tests 102-406-90b and 127-406-90a (c) Tests 102-406-60a and 127-406-60a
Fig. 7.8. Ratios of Paꞌ/PFE and a/ FE for the training set for different values of ef.weld
(a) Tests 102-273-90a, 102-406-90b, and 127-273-90a (b) Tests 127-406-90a, 102-406-60a, and 127-406-60a
Experimental spot strain measurements from SGs located 20 mm from the weld (around one half of the
weld, due to symmetry about the plane of the connection), oriented along the axis of the branch, indicated
that elastic load transfer peaked at the saddle point of the connection (subtended angle, ρ = 90°) (see Chapter
6). The sign convention for ρ is shown on Fig. 7.6b (clockwise from the heel). Typical measurements of
elastic strain parallel to the branch along a line intersecting these gauges in the FE models (indicated by a
dashed line in Fig. 7.6b) are shown in Figs. 7.11a,b.
(a) Test 127-273-90a ( = 0.47, θ = 90o, = 0.99) (b) Test 102-406-60a ( = 0.25, θ = 60o, = 0.60)
Fig. 7.11. Comparison of typical experimental and FE longitudinal strain distributions adjacent to the weld
In Figs. 7.11a,b, elastic branch strains from experiments are represented by solid black lines at an applied
load (P) equal to 0.25Pa (i.e. 25% of the load at which the weld actually ruptured in the experiments, during
the same monotonic loading phase), with the FE strains (solid red lines) predicting the same trends and values
as shown by experimental strains. It is shown that agreement between the FE and experimental strain
distributions under elastic load (P = 0.25Pa) was generally good. At ultimate, the FE strain distributions
showed the same trend; however, they poorly predicted the exact spot strain values measured in the
experiments. This is believed to be due to progressive and non-uniform yielding of the weld along its length,
due to variations in the experimental weld geometry that were not captured in the FE models. Nonetheless,
the FE models provided further evidence that load peaked in the saddle position, and revealed that weld
rupture initiated in the saddle position (i.e. the first killed elements were at this location). As load increased
further, failure propagated in both directions towards the crown points.
(a) the local dihedral angle (Ѱ) limits imposed by AWS (2015) Fig. 9.10 and Table 9.5 (60° ≤ Ѱ ≤
120°);
(b) the limits of applicability of connection design formulae in AISC (2016) Table K3.1, which are
given in AISC (2016) Table K3.1A; and
(c) the range of standard CHS sections available for designers in Table 1-13 of the Steel
Construction Manual (AISC 2011).
The parameters varied were: the branch inclination angle (θ = 60°, 70°, 80°, and 90°); the chord
slenderness (D/t = 10, 20, 30, 40, and 50); the branch-to-chord dimeter ratio ( = 0.10, 0.20, 0.30, 0.40, and
0.50); and the branch-to-chord thickness ratio ( = 0.20, 0.40, 0.60, 0.80, and 1.00). Although a total of 500
permutations exist for the values given, there are several practical limitations that must be considered. First,
available CHS sections limit branch slenderness ratios (Db/tb) to between about 10 and 50. Secondly, not all
combinations of and θ produce Ѱ between 60° and 120° (to qualify as a fillet weld) along the entire weld
length. Fig. 7.12 shows the results of a study conducted to determine compatible values of and θ that meet
this requirement. It is shown that permissible values of range from 0.50 (for 90° connections) to 0.28 (for
60° connections).
A comprehensive parametric study was performed by modelling up to 0.30 for 60° connections and
up to 0.50 for all other branch angles. A total of 256 CHS X-connection models was analyzed.
Fig. 7.12. Compatible values of and θ to keep Ѱ between 60° and 120° along the entire weld length,
determined using methodology by Luyties & Post (1988)
Fig. 7.13. Engineering stress-strain curves for materials used in the parametric models
Table 7.4. Yield stress and ultimate stress of materials used in the parametric models
Yield stress Ultimate stress, Fu (or FEXX for weld metal)
MPa MPa
Weld 517 577
Chord 460 540
Branches 431 488
Both welds were modelled as the same size, but the fracture criterion was only assigned to the upper
(test) weld (e.g. as indicated previously in Figs. 7.3 and 7.4). The other weld, the branches, and the chord
were not permitted to fracture. The model was loaded by applying uniform incremental displacements to the
ends of each branch. The applied load (P) was obtained by summing up nodal forces parallel to the branch at
the end of one branch, and multiplying by two to account for the half model. Chord deformation ( ) was
obtained as described in Section 7.3.2, with the vertical displacement taken between nodes on two branches at
50 mm from the crown (Fig. 7.7a, shown previously). No axial load was applied to the chord.
normalized by the weld area and the electrode ultimate strength (P/AwFEXX), for typical tests. It is shown that
deformation ( /D) and strength (P/AwFEXX) varied widely across the FE tests at rupture. A complete list of
results from the parametric study are provided in Appendix F.1.
To determine the weld effective length (le), first the strength of the fillet weld in each connection was
predicted using Eq. (7.11), which was developed from regression of a large database of weld-critical tests on
CHS connections that had fully effective fillet welds (Chapter 5):
D t
Pnw 1.009 0.00137 b 0.197 w Aw FEXX (7.11)
tb tb
The fraction of the weld length that is effective (le/lw) was then found by dividing the weld FE strength by
the predicted weld strength from Eq. (7.11), corresponding to Method 2 in Chapter 6 (Section 6.6).
Determined in this manner, 138 (out of 256) FE connections had weld effective lengths less than 1.0. The
smallest weld effective length was 0.58 times the total weld length (for D/t = 50, = 0.50, = 1.0 and θ =
80°), and the largest weld effective length was theoretically 1.22 times the total weld length (for D/t = 50, =
0.10, = 0.2 and θ = 90°). The fact that weld effective lengths were sometimes greater than 1.0 is believed to
be due, in part, to the scatter associated with the best-fit prediction equation (Eq. 7.11), and also due to the
effect of secondary forces on the weld due to local bending of the chord.
Fig. 7.15. Effect of chord slenderness and branch-to-chord diameter ratio on effective length
le C
b (7.12)
lw ( ) a
where a, b, and C = regression constants, and = half-diameter-to-thickness ratio of the CHS chord (= D/2t).
The half-diameter-to thickness-ratio of the CHS chord is a standard parameter for HSS connection design.
The arrangement of variables in Eq. (7.12) was determined empirically, with efforts made to take into
account the relationships discussed in Section 7.6. The values of a, b, and C were determined by least-squares
regression of the 138 FE results with le/lw < 1, since above this value the relationship changes and it is
inapplicable for design. The “best-fit” equation is given by Eq. (7.13):
le 1.786
(7.13)
lw 0.1007 ( )0.440
Eq. (7.13) gives a mean value of actual-to-predicted rupture strength of 1.001 over the 138 tests with a
COV of 0.030. Eq. (7.13) with = 0.2 and Eq. (7.13) with = 1.0 are plotted as dashed red lines on Fig. 7.16.
These two curves represent the extremes of the data, and also show that Eq. (7.12) as a basis for the
regression is appropriate because it describes the predominant data trends. For a lower-bound to Eq. (7.13),
the value of can be taken as 1.0 (the maximum in this study and a recommended maximum in practice)
which is synonymous with taking b equal to zero in Eq. (7.12). Eq. (7.14) is proposed as an even more
simple equation for determining the weld effective length (le) as a ratio of the total weld length in CHS X-
connections:
le 2
1 (7.14)
lw
Eq. (7.14) is plotted as a solid line in Fig. 7.16. It implies that weld effective lengths do not exist in CHS
X-connections when ≤ 4. Eq. (7.14), plotted in Fig. 7.19, embodies the trend established by Caulkins
(1968) and Marshall (1992) for the range of and D/t in this study (0.10 ≤ ≤ 0.50 and 10 ≤ D/t ≤ 50). As
and D/t increase, the load transfer efficiency (or for this study, le/lw) decreases. The rate of decrease is greater
for lower values of and D/t, when > 4.
Fig. 7.19. Weld effective lengths in CHS-to-CHS X-connections with 0.10 ≤ ≤ 0.50 and 10 ≤ D/t ≤ 50
according to Eq. (7.14)
where Fnw = nominal weld strength. An LRFD or LSD resistance factor for fillet welds, ϕ, is then applied to
determine ϕPnw. Note that Eqs. (7.1) and (7.2) can be written in this form.
Table 7.5 gives the results of the reliability analysis using three different values of ϕ. The safety indices
calculated for the proposed design method for each value of ϕ can be compared to target values given in the
+ +
United States ( = 4.0, per Section B3.1 of the AISC 260-16 Commentary) and Canada ( = 4.5, per Annex
B of CSA S16-14).
The implied safety index ( +) ≥ 4.0 when ϕ = 0.80 (as used by AWS D1.1-15), +
≥ 4.0 when ϕ = 0.75 (as
used by AISC 360-16), and +
≥ 4.5 when ϕ = 0.67 (as used by CSA S16-14). This indicates that the method
meets U.S. and Canadian safety indices. Fig. 7.20 shows the correlation of the predicted nominal strengths
using Steps 1) – 5) for the proposed procedure with the experimental and FE results. On average, the test
capacity is only 1.12 times larger than the predicted rupture load, with a COV of 0.08.
Table 7.5. Reliability analysis parameters for proposed procedure with ϕ = 0.80, 0.75 and 0.67
ϕa 0.80 0.75 0.67
ρM 1.12 1.12 1.12
VM 0.12 0.12 0.12
ρG 1.03 1.03 1.03
VG 0.10 0.10 0.10
ρP 1.12 1.12 1.12
VP 0.08 0.08 0.08
ρR 1.29 1.29 1.29
VR 0.18 0.18 0.18
ϕ+ 0.91 0.89 0.84
+
4.0 4.3 5.0
a
resistance factors for fillet welds according to AWS (2015), AISC (2016), and CSA (2014) (see Chapter 6).
Fig. 7.20. Correlation of proposed design method with all test results
7.8.1.3. Safety Level Inherent in AWS (2015), AISC (2016) and CSA (2014)
The reliability analysis was repeated, using all 268 numerical and FE tests, to determine the implied
+
safety index, , for the current AWS D1.1-15 provisions, with and without the (2/3)lw weld effective length
factor, and the AISC 360-16 and CSA S16-14 provisions, as presented in Chapter 6. The analysis determined
that +
= 6.2 ≥ 4.0 for AWS (2015) with le = (2/3)lw, +
= 4.3 ≥ 4.0 for AWS (2015) with le = lw, and +
= 4.6
for both AISC (2016) and CSA (2014) > 4.0 and 4.5. A table summarizing the reliability analysis parameters,
and correlation plots for each method, is provided in Appendix F.2. It can therefore still be concluded that
weld effective lengths are never required for fillet-welded CHS X-connections in conjunction with these code
design methods. The mean test-to-predicted capacity ratio (ρP) and Vp for these methods (ρP = 2.07 for AWS
D1.1-15 with weld effective lengths, ρP = 1.38 for AWS D1.1-15 without weld effective lengths and AISC
360-16, ρP = 1.24 for CSA S16-14, and Vp = 0.19 for all methods) indicate that the proposed procedure
(Section 7.8.1.1) is much more accurate (ρP in Table 7.5 closer to 1) and precise (VP in Table 7.5 closer to 0)
than the code methods for predicting fillet weld strength in CHS X-connections.
7.9. SUMMARY
Non-linear FE models with weld fracture were developed for fillet-welded CHS X-connections. These
models were validated by comparison with 12 weld-critical tests on CHS X-connections. A parametric study
was then performed in which 256 FE models with 60° ≤ θ ≤ 90°, 10 ≤ D/t ≤ 50, 0.10 ≤ ≤ 0.50, and 0.20 ≤
≤ 1.00 were analysed. All models were shown to fail by weld fracture. Based on the parametric study, the
following can be deduced:
1. The weld effective length in CHS X-connections decreases as the branch slenderness (D/t),
branch-to-chord dimeter ratio ( ), and branch-to-chord thickness ratio ( ) increase.
2. The branch inclination angle θ has an insignificant effect on the weld effective length.
3. Theoretically, the weld effective length can be as low as 0.58 times the total weld length.
4. The weld length is 100% effective for ≤ 4.
5. For CHS X-connections load transfer peaks at the saddle position.
6. Stress re-distribution occurs prior to weld rupture, even in connections with small welds.
Based on a reliability analysis to determine the safety index ( +) for 268 experimental and FE tests
covering the same range of parameters, the following can also be deduced:
7. The existing AWS (2015) specification provisions for fillet welds in CHS X-connections in
+
Clause 9.6.1.3(4), with le = (2/3)lw, are very conservative (safety index, = 6.2 > 4.0).
8. The existing AWS (2015) specification provisions for fillet welds in CHS X-connections in
+
Clause 9.6.1.3(4), with le = lw, are also conservative (safety index, = 4.3 > 4.0).
9. The existing AISC (2016) provisions in Clause J2.4a and the existing CSA (2014) provisions in
+
Clause 13.13.2.2, with le = lw, are conservative (safety index, = 4.6 for both > 4.0 and 4.5).
The evaluations of AWS, AISC and CSA fillet weld design provisions assume that the (1+0.50sin 1.5θ)
directional strength-enhancement factor is not used (AWS D1.1-15 Clause 2.6.4.2, AISC 360-16 Clause
J2.4b, and CSA S16-14 Clause 13.13.2.2), because it has been shown, in Chapters 4 and 5, to be generally
unsafe for the design of fillet welds in HSS connections.
An alternative method for the design of fillet welds, based on rational weld effective lengths as a function
of non-dimensional connection parameters, was proposed. This method was shown be more accurate and
precise for predicting fillet weld strength in CHS X-connections than AWS (2015), AISC (2016) and CSA
(2014). A reliability analysis with respect to all experimental and FE tests showed that this proposed method
provides an adequate level of safety for use with AWS, AISC and CSA codes.
147
Chapter 8: Conclusions 148
Fracture-critical welds in CHS-to-CHS connections have been studied for the first time, and it has been
determined that effective lengths therein are a function of branch-to-chord width ratio ( ), chord wall
slenderness (2 ), and branch-to-chord thickness ratio (Chapters 6 and 7). Welds are proven to be fully
effective for ≤ 4. A weld effective length design approach has been established to provide a more
economical and yet still safe weld design method for welds in CHS-to-CHS X-connections; however it is
unlikely that this approach will replace those given by the American Welding Society (AWS), AISC, and
CSA at this time. It has hence been shown that these current code methods provide sufficient reliability, even
without weld effective lengths, as long as the (1.00+0.50sin1.5θ) factor is not used. The experimental and FE
portions of the CHS-to-CHS X-connection test program have been submitted as papers for review
(Tousignant & Packer 2017b, 2017c, 2017d).
• The range of RHS overlapped K-connections in this study was limited, and covered only
connections with θbi = θbj = 60°. Although this is deemed a critical case by AISC, it is advisable
to evaluate and improve upon the weld effective length formulae for such connections over a
wider range of branch angles.
• Similarly, the previous work done on weld effective lengths in RHS-to-RHS moment T-
connections (McFadden & Packer 2014) could be expanded using FE analysis.
• As stated in Section 5.3.4, in spite of the importance of fracture to structural engineering, there is
no single widely accepted micro-mechanics-based model to predict ductile fracture in steel
components. Previous researchers have proposed several fracture models for monotonic and
cyclic loading [e.g. stress-modified critical strain model, void growth model (Kanvinde &
Deierlein 2006); cyclic void growth model (Kanvinde & Deierlein 2007); Xue model (Xue
2007); modified Xue model (Xue 2009); and simplified damage plasticity model (Ma et al.
2015)], but most of these models require complex calibration. They are therefore difficult to
implement in the analysis of large-scale tests. Confidence in modelling ductile fracture in such
tests could be increased by:
o Compiling/developing a database of large-scale fracture-critical experimental tests,
complete with ancillary coupon tests of the materials; and
o Evaluating the many models on the same set of tests, over a wide range of parameters, to
assess their relative practical merits.
• Research on analytical methods for welds that explicitly take into account secondary bending
moments, including not only local eccentricity, but also the secondary moments produced by
rotation of the attached base metal would be highly informative. Limited data for simple tests on
fillet welds between plates at 90°, given by Chen et al (2001), could be used as a starting point;
however, more such tests on a wider range of plate and fillet weld parameters would be desirable.
Although this research is simple in theory, the results of it may help to further explain the trends
observed in the current work concerning relative weld strength with respect to weld size, branch
slenderness, branch-to-chord thickness ratio, and branch-to-chord width/diameter ratio.
American Institute of Steel Construction (AISC) 2010. ANSI/AISC 360-10. Specification for structural steel
American Institute of Steel Construction (AISC) 2011. Steel construction manual, 14th ed., Chicago, IL, USA.
American Institute of Steel Construction (AISC) 2016. ANSI/AISC 360-16. Specification for structural steel
ASTM International (ASTM) 2013. ASTM A500-13. Standard specification for cold-formed welded and
seamless carbon steel structural tubing in rounds and shapes. West Conshohocken, PA, USA.
ASTM International (ASTM) 2015a. ASTM A1085-15. Standard specification for cold-formed welded steel
ASTM International (ASTM) 2015b. ASTM E340-15. Standard test method for macroetching metals and
ASTM International (ASTM) 2017. ASTM A370-17. Standard test methods and definitions for mechanical
American Welding Society (AWS). 2015. AWS D1.1/D1.1M:2015. Structural welding code – Steel, 23rd ed.,
Björk, T., Toivonen, J. & Nykänen, T. 2012. Capacity of fillet welded joints made of ultra high-strength steel.
Björk, T., Penttilä, T. & Nykänen, T. 2014. Rotation capacity of fillet weld joints made of high-strength steel.
Boresi, A. P. & Schmidt, R.J. 2003. Advanced mechanics of materials, 6 th ed. NJ, USA: John Wiley and
Sons, Inc.
150
References 151
British Standards Institution (BSI) 1959. BS 449:1959. Specification for the use of structural steel in
Butler, L. J. & Kulak, G. L. 1971. Strength of fillet welds as a function of direction of load. Welding Journal,
Callele, L. J., Driver, R. G. & Grondin, G. Y. 2009. Design and behavior of multi-orientation fillet weld
Canadian Institute of Steel Construction (CISC). 2016. Handbook of steel construction, 11 th ed., Toronto,
Canadian Standards Association (CSA) 2001. CAN/CSA S16-01. Limit states design of steel structures.
Canadian Standards Association (CSA) 2013a. CSA G40.20-13/G40.21-13. General requirements for rolled
or welded structural quality steel/structural quality steel. Toronto, Canada: Canadian Standards
Association.
Canadian Standards Association (CSA) 2013b. CSA W59-13. Welded steel construction (metal arc welding).
Canadian Standards Association (CSA) 2014. CSA S16-14. Design of steel structures. Toronto, Canada:
Caulkins, D. W. 1968. CDG Report 15. Parameter study for FRAMETI elastic stress in tubular joints.
Chen, Y., Shen, Z., Zheng, Q. & Chen, C. 2001. Experimental study on the performance of single weld joints
Computers & Structures Inc. 2014. SAP2000 ver. 16.1. Walnut Creek, CA, USA.
Coutie, M. G. & Saidani, M. 1989. The use of finite element techniques for the analysis of RHS structures
with flexible joints. In Niemi, E. & Mäkeläinen, P. (Eds.) Tubular Structures III; Proc. Intern. Symp.,
Coutie, M. G. & Saidani, M. 1991. Comparison of the theoretical behaviour of two rectangular hollow
section trusses. In Wardenier, J. & Shahi, E. P. (Eds.) Tubular Structure IV; Proc. Intern. Symp., Delft 26-
Czechowski, A., Gasparski, T., Zycinski, J., & Brodka, J. 1984. IIW Doc. XV-562-84. Investigation into the
static behaviour and strength of lattice girders made of RHS. Paris, France: International Institute of
Welding.
Davies, G. & Packer, J. A. 1982. Predicting the strength of branch plate - RHS connections for punching
Deng, K., Grondin, G. Y. & Driver, R. G. 2006. Effect of loading angle on the behavior of fillet welds.
European Committee for Standardization (CEN) 2005. EN 1993-1-8:2005. Eurocode 3: Design of steel
European Committee for Standardization (CEN). 2006. EN 10219-2:2006. Cold formed welded structural
hollow sections of non-alloy and fine grain steels: Tolerances, dimensions and sectional properties.
Brussels, Belgium.
Fisher, J. W., Galambos, T. V., Kulak, G. L. & Ravindra, M. K. (1978). Load and resistance factor design
criteria for connectors. Journal of the Structural Division, American Society of Civil Engineers 104(9):
1427-1441.
Fleischer, O. & Herion, S. 2015. Reduction of fillet weld sizes. In Batista, E., Vellasco, P. & Lima, L. (eds.),
Tubular Structures XV; Proc. Intern. Symp., Rio de Janeiro 27-29 May 2015. Rotterdam: Balkema, 473-
480.
Franchuk, C. R., Driver, R. G. & Grondin, G.Y. 2002. Block shear failure of coped steel beams. Proc. Annual
Conf. of the Canadian Society for Civil Engineering, Montreal, 5-8 June 2002. 1000-1009.
Frater, G. S. 1986. Weldment design for hollow structural section joints. M.A.Sc. thesis. Toronto, Canada:
University of Toronto.
Frater, G. S. 1991. Performance of welded rectangular hollow structural section trusses. PhD thesis. Toronto,
Frater, G. S. & Packer, J. A. 1992a. Weldment design for RHS truss connections. I: Applications. Journal of
Frater, G. S. & Packer, J. A. 1992b. Weldment design for RHS truss connections. II: Experimentation.
Frater, G. S., & Packer, J. A. 1992c. Modelling of hollow structural section trusses. Canadian Journal of
Hollomon, J. H. (1945). Tensile deformation. Transactions of the Metallurgical Society of American Institute
International Institute of Welding (IIW) 1980. IIW Doc. XV-467-80. Deformation curves of fillet welds.
London, England.
International Institute of Welding (IIW) 1989. IIW Doc. XV-701-89. Design recommendations for hollow
International Institute of Welding (IIW) 2012. IIW Doc. XV-1402-12. Design recommendations for hollow
International Organization for Standardization (ISO). 2013. ISO 14346:2013 (E). Static design procedure for
Jensen, C. D. 1934. Combined stresses in fillet welds. Welding Journal, American Welding Society 13(2): 17-
21.
Kanvinde, A. M. 2017. Predicting fracture in civil engineering steel structures: State of the art. Journal of
Kanvinde, A. M. & Deierlein, G. G. 2006. Void growth model and stress modified critical strain model to
predict ductile fracture in structural steels. Journal of Structural Engineering, American Society of Civil
Kanvinde, A. M. & Deierlein, G. G. 2007. Cyclic void growth model to assess ductile fracture initiation in
structural steels due to ultra low cycle fatigue. Journal of Structural Engineering, American Society of
Kanvinde, A. M., Grondin, G. Y., Gomez, I. R. & Kwan, Y. 2009. Experimental investigation of fillet-
welded joints subjected to out-of-plane eccentric loads. Engineering Journal, American Institute of Steel
Kato, B. & Morita, K. 1974. Strength of transverse fillet welded joints. Welding Journal, American Welding
Calculation of welds under consideration of constant deformation energy. Vorbericht 2, Kongress Int. Ver.
Lesik, D. F. & Kennedy, D. J. 1990. Ultimate strength of fillet welded connections loaded in plane. Canadian
Lie, S. T., Lee, C. K. & Wong S. M. 2001. Modelling and mesh generation of weld profile in tubular Y-joint.
Ling, Y. 1996. Uniaxial true stress-strain after necking. AMP Journal of Technology 5(1): 37-48.
Lu, L. H., de Winkel, G. D., Yu, Y. & Wardenier, J. 1994. Deformation limit for the ultimate strength of
hollow section joints. In Grundy, P., Holgate, A. & Wong, B. (Eds.), Tubular Structures VI; Proc. Intern.
Luyties, W. H. & Post, J. W. 1988. Local dihedral angle equations for tubular joints and related applications.
Ma, X., Wang, W., Chen, Y. & Qian, X. 2015. Simulation of ductile fracture in welded tubular connections
using a simplified damage plasticity model considering the effects of stress triaxialty and Lode angle.
Marshall, P.W. 1992. Design of welded tubular connections – Basis and use of AWS code provisions.
Martinez-Saucedo, G., Packer, J. A. & Willibald, S. 2006. Parametric finite element study of slotted end
McFadden, M. R. & Packer, J. A. 2013. Effective weld properties for RHS-to-RHS moment T-connections.
Phase 1 Report to the American Institute of Steel Construction. Toronto, Canada: University of Toronto.
McFadden, M. R. & Packer, J. A. 2014. Effective weld properties for hollow structural section T-connections
under branch in-plane bending. Engineering Journal, American Institute of Steel Construction 51(4): 247-
266.
McFadden, M. R., Sun, M. & Packer, J. A. 2013. Weld design and fabrication for RHS connections. Steel
Mehrotra, B. L. & Govil, A. K. 1972. Shear lag analaysis of rectangular full-width tube connections. Journal
Miazga, G. S. & Kennedy, D. J. 1989. Behaviour of fillet welds as a function of the angle of loading.
Miller, D. K. 2002. Designing fillet welds for skewed T-joints – Part 1. Welding Innovation 19(1): 7-11.
Moore, A. M., Rassati G. A. & Swanson J. A. 2010. An experimental analysis of strength and ductility of
high strength fasteners. Engineering Journal, American Institute of Steel Construction 47(3): 161-174.
Ng, A. K. F., Deng, K., Grondin, G. Y. & Driver, R. G. 2004a. Behavior of transverse fillet welds:
experimental program. Engineering Journal, American Institute of Steel Construction 41(2): 39-54.
Ng, A. K. F., Driver, R. G. & Grondin, G. Y. 2004b. Behavior of transverse fillet welds: parametric and
reliability analyses. Engineering Journal, American Institute of Steel Construction 41(2): 55-67.
Oatway, P. 2014. Fillet-welded end-plate connections to square and rectangular HSS. M.Eng. thesis. Toronto,
Packer, J. A. & Cassidy, C. E. 1995. Effective weld length for HSS T, Y, and X connections. Journal of
Packer, J. A., Choo, Y. S., Shen, W., Wardenier, J., van der Vegte, G. J., & Mustard, T. 2012. CIDECT
Report 5BW-2/12. Axially loaded T and X joints of elliptical hollow sections. Geneva, Switzerland:
CIDECT.
Packer, J. A. & Davies, G. 1989. On the use and calibration of design standards for SHS joints. The
Packer, J. A. & Frater, G. S. 2005. Recommended effective throat sizes for flare groove welds to HSS.
Packer, J. A. & Henderson, J. E. 1992. Design guide for hollow structural section connections, 1st ed.,
Packer, J. A. & Henderson, J. E. 1997. Hollow structural section connections and trusses – A design guide,
Packer, J. A., Sherman, D. R. & Leece, M. 2010. Hollow structural section connections, AISC Steel Design
Guide No. 24. Chicago, IL, USA: American Institute of Steel Construction.
Packer, J. A. & Sun, M. 2011. Weld design for rectangular HSS connections. Engineering Journal, American
Packer, J. A., Sun, M., & Tousignant, K. 2016. Experimental evaluation of design procedures for fillet welds
to hollow structural sections. Journal of Structural Engineering, American Society of Civil Engineers
Packer, J. A., Wardenier, J., Kurobane, Y., Dutta, D. & Yeomans, N. 1992. Design guide for rectangular
hollow section (RHS) joints under predominantly static loading, CIDECT Design Guide No. 3, 1st ed.,
Packer, J. A., Wardenier, J., Zhao, X. L., van der Vegte, G. J. & Kurobane, Y. 2009. Design guide for
rectangular hollow section (RHS) joints under predominantly static loading, CIDECT Design Guide No.
Philiastides, A. 1988. Fully overlapped rolled hollow section welded joints in trusses. PhD Thesis.
Ravindra, M. K. & Galambos, T. V. 1978. Load and resistance factor design for steel. Journal of the
Swanson Analysis Systems. 2011. ANSYS ver. 14.0. Houston, TX, USA.
Tousignant, K. & Packer, J. A. 2014. Design of welds in rectangular HSS overlapped K-connections. Report
to the American Institute of Steel Construction. Toronto, Canada: University of Toronto, 283 p.
Tousignant, K. & Packer, J. A. 2015a. Weld effective lengths for rectangular HSS overlapped K-connections.
Tousignant, K. & Packer, J. A. 2015b. Investigation of weld effective length rules for RHS overlapped K-
connections. In Batista, E., Vellasco, P. & Lima, L. (Eds.), Tubular Structures XV; Proc. Intern. Symp.,
Tousignant, K. & Packer, J.A. 2016a. Analysis of rectangular hollow section trusses with overlapped K-
connections. 5th International Structural Specialty Conference, Canadian Society of Civil Engineering,
for fillet welds to CHS. Connections VIII; Proc. Intern. Workshop on Connections in Steel Structures,
Tousignant, K. & Packer, J. A. 2017a. Numerical investigation of fillet welds in HSS-to-rigid end-plate
connections. Journal of Structural Engineering, American Society of Civil Engineers 143(12): 04017165-1
– 04017165-16.
Tousignant, K. & Packer, J. A. 2017b. Fillet weld effective lengths in CHS X-connections. I: Experiments.
Tousignant, K. & Packer, J. A. 2017c. Fillet weld effective lengths in CHS X-connections. II: Finite element
modelling, parametric study and design. Journal of Constructional Steel Research, submitted for review.
Tousignant, K. & Packer, J. A. 2017d. Investigation of weld effective length rules for CHS X-connections.
Tubular Structures XVI; Proc. Intern. Symp., Melbourne 4-6 December 2017, accepted for publication.
van der Vegte, G.J. & Makino, Y. 2010. Further research on chord length and boundary conditions of CHS T-
Voth, A. P. & Packer, J. A. 2012a. Branch plate-to-circular hollow structural section connections. I:
Voth, A. P. & Packer, J. A. 2012b. Branch plate-to-circular hollow structural section connections. II: X-type
parametric numerical study and design. Journal of Structural Engineering, American Society of Civil
Vreedenburgh, G. G. J. 1954. New principles for the calculation of welded joints. Welding Journal, American
Wang, W., Gu, Q., Ma, X. & Wang, J. 2015. Axial tensile behavior and strength of welds for CHS branches
Wardenier J., Kurobane, Y., Packer, J. A., van der Vegte, G. J. & Zhao, X. L. 2008. Design guide for circular
hollow section (CHS) joints under predominantly static loading, CIDECT Design Guide No. 1, 2nd ed.
Wardenier, J., and Packer, J. A. 1992. Connections between hollow sections. In Constructional Steel Design
Xue, L. 2007. Damage accumulation and fracture initiation in uncracked ductile solids subject to triaxial
Xue, L. 2009. Stress based fracture envelope for damage plastic solids. Engineering Fracture Mechanics
76(3): 419-438.
Appendix A provides the supplementary data for Chapter 2: Weld Effective Lengths in RHS Overlapped
K-Connections.
159
Appendix A: Weld Effective Lengths in RHS Overlapped K-Connections 160
Fig. A.1. Truss elevation, joint layout and typical connection detail (drawing T-1)
Fig. A.2. Web members 11-1, 1-4, 4-5 and 5-8 cut details (according to drawing T-1)
Fig. A.3. Web members 8-9, 9-7, 7-6 and 6-3 cut details (according to drawing T-1)
Fig. A.4. Web member 3-2 and 2-12 cut details (according to drawing T-1)
A.2. MEASURED WELD DIMENSIONS FOR FILLET AND PJP WELDS IN RHS
OVERLAPPED K-CONNECTIONS
The measured weld dimensions for each of the fillet and partial joint penetration (PJP) welds in the nine
RHS overlapped K-connection test joints are provided below. The average values along each weld element
are also given. Note that the apparent fillet weld leg along the overlapping branch (Fig. A.9a) is provided for
the fillet welds in Tables A.2 – A.10. The actual leg dimension (lv) is equal to the values shown multiplied by
1/sinѰ. The local dihedral angle (Ѱ) = 90° for weld elements a and b, and Ѱ = 60° for weld elements c and d.
Fig. A.9. Weld dimensions for (a) fillet weld elements and (b) PJP weld elements
A.3. TENSILE COUPON TEST RESULTS FOR RHS AND WELD METAL
Results of tensile coupon tests conducted for the RHS and weld metal in Chapter 2: Weld Effective
Lengths in RHS Overlapped K-Connections are given below.
(a) HSS 254.4 × 254.4 × 9.24 (heat no. 813767) (b) HSS 178.7 × 178.7 × 12.53 (heat no. L62502)
Fig. A.10. TC test results for RHS branch and chord materials
Fig. A.13. Weld-critical overlapped K-connections with instrumentation immediately prior to testing
Fig. A.14. Weld-critical overlapped K-connections with instrumentation immediately after testing
Fig. A.15. Distribution of normal strain around branch perimeter for test K-90-0.50a
Fig. A.16. Distribution of normal strain around branch perimeter for test K-30-0.50b
Fig. A.17. Distribution of normal strain around branch perimeter for test K-90-0.71
Fig. A.18. Distribution of normal strain around branch perimeter for test K-60-0.71a
Fig. A.19. Distribution of normal strain around branch perimeter for test K-60-0.71b
Fig. A.20. Distribution of normal strain around branch perimeter for test K-30-0.71
Appendix B provides the supplementary data for Chapter 3: Analysis of RHS Trusses with Overlapped
K-Connections.
180
Appendix B: Analysis of RHS Trusses with Overlapped K-Connections 181
BH 3 B 2t H 2t
3
2I
S (B.2)
H
I
r (B.3)
A
BH 2 B 2t H 2t
2
Z 4 Ag hg 4 A h (B.4)
4 4
where:
1 1 4
I g ro (B.5)
3 16 3 12 3
Ag 1 ro2 (B.6)
4
H 10 3
hg ro (B.7)
2 12 3
1 1 4
I ri (B.8)
3 16 3 12 3
A 1 ri 2 (B.9)
4
H 2t 10 3
h ri (B.10)
2 12 3
e = eccentricity v v v v
Fig. B.1. Truss member and panel point designations used in Appendix B
(a) (b)
Fig. B.2. LVDTs transverse to the chord longitudinal axis (a) along the chord and (b) at the support
Table B.3. Panel point deflections in test 4.1 (applied load = 483.3 kN to panel point 4)
Deflection (mm)
Panel point number: 3 5 7 9 11 % error a
Experimental -2.789 -5.207 -4.887 -3.609 -2.106 -
Model RR -2.398 -4.366 -3.876 -2.612 -1.006 -19%
Model PR -2.53 -4.51 -3.991 -2.697 -1.055 -15%
Model R -2.559 -4.655 -4.107 -2.769 -1.059 -12%
Model P -2.857 -4.955 -4.299 -2.908 -1.137 -5%
a
Spurious results omitted from analysis in Chapter 3.
Table B.4. Panel point deflections in test 4.2 (applied load = 489.6 kN to panel point 4)
Deflection (mm)
Panel point number: 3 5 7 9 11 % error
Experimental -1.849 -4.496 -4.115 -2.715 -1.031 -
Model RR -2.429 -4.423 -3.927 -2.646 -1.019 -2%
Model PR -2.563 -4.569 -4.043 -2.732 -1.069 2%
Model R -2.592 -4.716 -4.161 -2.805 -1.073 5%
Model P -2.894 -5.020 -4.355 -2.946 -1.152 10%
Table B.5. Panel point deflections in test 4.3 (applied load = 485.3 kN to panel point 4)
Deflection (mm)
Panel point number: 3 5 7 9 11 % error a
Experimental -2.710 -4.943 -4.630 -3.324 -1.882 -
Model RR -2.408 -4.384 -3.892 -2.623 -1.010 -13%
Model PR -2.541 -4.528 -4.007 -2.708 -1.059 -9%
Model R -2.570 -4.674 -4.124 -2.780 -1.064 -6%
Model P -2.869 -4.975 -4.316 -2.920 -1.142 1%
a
Spurious results omitted from analysis in Chapter 3.
Table B.6. Panel point deflections in test 10 (applied load = 490.4 kN applied to panel point 10)
Deflection (mm)
Panel point number: 3 5 7 9 11 % error
Experimental -0.927 -2.549 -3.907 -4.322 -2.316 -
Model RR -1.034 -2.634 -3.922 -4.435 -2.435 3%
Model PR -1.081 -2.723 -4.047 -4.599 -2.572 6%
Model R -1.081 -2.788 -4.141 -4.699 -2.567 8%
Model P -1.159 -2.929 -4.336 -5.003 -2.871 14%
Table B.7. Panel point deflections in test 6.1 (applied load = 496.5 kN applied to panel point 6)
Deflection (mm)
Panel point number: 3 5 7 9 11 % error
Experimental -2.352 -5.648 -6.489 -4.611 -1.719 -
Model RR -2.467 -6.062 -6.929 -4.803 -1.856 6%
Model PR -2.575 -6.188 -7.050 -4.923 -1.932 8%
Model R -2.598 -6.456 -7.367 -5.085 -1.955 12%
Model P -2.766 -6.769 -7.690 -5.279 -2.088 16%
Table B.8. Panel point deflections in test 8 (applied load = 497.4 kN applied to panel point 8)
Deflection (mm)
Panel point number: 3 5 7 9 11 % error
Experimental -1.787 -4.750 -6.612 -6.077 -2.441 -
Model RR -1.896 -4.815 -6.998 -6.113 -2.443 6%
Model PR -1.973 -4.926 -7.124 -6.258 -2.553 7%
Model R -1.992 -5.111 -7.408 -6.472 -2.569 11%
Model P -2.125 -5.305 -7.734 -6.784 -2.739 15%
Table B.9. Panel point deflections in test 6.2 (applied load = 496.5 kN applied to panel point 6)
Deflection (mm)
Panel point number: 3 5 7 9 11 % error
Experimental -2.326 -5.614 -6.456 -4.612 -1.711 -
Model RR -2.467 -6.026 -6.929 -4.803 -1.856 7%
Model PR -2.575 -6.188 -7.050 -4.923 -1.923 8%
Model R -2.598 -6.456 -7.367 -5.085 -1.955 12%
Model P -2.766 -6.769 -7.690 -5.279 -2.088 16%
Table B.10. Panel point deflections in test 9 (applied load = 495.6 kN applied to panel point 9)
Deflection (mm)
Panel point number: 4 6 8 10 % error
Experimental -2.989 -4.942 -5.999 -4.202 -
Model RR -2.513 -4.587 -5.844 -4.208 -3%
Model PR -2.578 -4.679 -5.954 -4.335 -1%
Model R -2.652 -4.845 -6.175 -4.444 3%
Model P -2.760 -5.003 -6.442 -4.695 7%
Table B.11. Panel point deflections in test 5 (applied load = 494.6 kN applied to panel point 5)
Deflection (mm)
Panel point number: 4 6 8 10 % error a
Experimental 0.164 -8.242 -6.273 -3.462 -
Model RR -4.202 -5.794 -4.579 -2.489 -42%
Model PR -4.314 -5.887 -4.662 -2.555 -40%
Model R -4.456 -6.154 -4.848 -2.624 -34%
Model P -4.707 -6.425 -5.006 -2.732 -28%
a
Spurious results omitted from analysis in Chapter 3.
Table B.12. Panel point deflections in test 7.1 (applied load = 482.9 kN applied to panel point 7)
Deflection (mm)
Panel point number: 4 6 8 10 % error a
Experimental -3.460 -6.178 -6.210 -3.634 -
Model RR -3.666 -6.520 -6.570 -3.650 5%
Model PR -3.752 -6.609 -6.667 -3.742 7%
Model R -3.866 -6.920 -6.950 -3.840 11%
Model P -4.009 -7.193 -7.222 -3.893 14%
Table B.13. Panel point deflections in test 7.2 (applied load = 480.4 kN applied to panel point 7)
Deflection (mm)
Panel point number: 4 6 8 10 % error a
Experimental -3.585 -6.326 -6.353 -3.597 -
Model RR -3.647 -6.486 -6.540 -3.630 3%
Model PR -3.732 -6.575 -6.632 -3.723 4%
Model R -3.846 -6.884 -6.911 -3.820 8%
Model P -3.988 -7.156 -7.185 -3.963 12%
Table B.15. Mean percentage errors in axial force predictions in test 4.1
Mean percentage error
Location Model: RR PR R P
Top chord 5% 4% 5% 8%
Webs 7% 10% 8% 15%
Bottom chord 6% 6% 6% 7%
Overall 6% 8% 7% 11%
Overall standard deviation 3.7% 4.0% 6.2% 7.3%
Table B.17. Mean percentage errors in axial force predictions in test 4.2
Mean percentage error
Location Model: RR PR R P
Top chord 8% 8% 9% 11%
Webs 9% 12% 11% 17%
Bottom chord 10% 10% 9% 11%
Overall 9% 11% 10% 14%
Overall standard deviation 3.9% 3.5% 6.0% 6.8%
Note: bending moment diagrams are plotted on the side of the member subject to tensile bending stress
Note: bending moment diagrams are plotted on the side of the member subject to tensile bending stress
Fig. B.4. Comparison of experimental and theoretical bending moment distributions for test 4.2
Appendix C provides the supplementary data for Chapter 4: Fillet Welds in HSS-to-Rigid Plate
Connections: Experimentation.
Fig. C.1. Stress components in the plane of the weld effective throat for Ѱ = θ = 90°
EN1993-1-8 gives the following as one possible design formula for the resistance of a fillet weld:
0.5
2 3( 2 2 ) Fu / ( w M 2 ) (C.1)
191
Appendix C: Fillet Welds in HSS-to-Rigid Plate Connections: Experimentation 192
P cos
(C.2a)
t w lw
P sin cos
(C.2b)
t w lw
P sin sin
(C.2c)
t w lw
P 0.5
sin 2 cos 2 3(sin 2 sin 2 cos 2 ) Fu / ( w M 2 )
t w lw (C.3a)
Fu t w lw Pnw
P Vrw (C.3b)
( w M 2 ) sin cos 3(sin sin cos )
2 2 2 2 2 0.5
M2
For the case of an equal-legged longitudinally-loaded (θ = 0°) fillet weld in a Ѱ = 90° connection ( = 45°),
Eq. (C.3b) simplifies to:
Fu
Vrw ( )twlw (C.4a)
3 w M 2
For the same case (Ѱ = 90° and = 45°), but transversely-loaded (θ = 90˚), Eq. (C.3b) simplifies to:
Fu
Vrw ( )twlw (C.4b)
2 w M 2
Fig. C.4. All-weld-metal TC test specimen and trial weld specimens for CHS-to-rigid plate tests
The following derivation was performed in order to calculate the theoretical throat dimension (tw) of a
skewed-T fillet weld from the measured weld leg dimensions (lv and lh) and the local dihedral angle. The
result is a general equation applicable to all welded joints (not just tubular joints), which can be simplified to
produce more specific and practical formulae for design of skewed-T joints (i.e. joints with lv = lh), as given
by Miller (2002). A general equation such as the one provided below is not given therein. The dimensions
referenced below are shown in Fig. C.5 (Note: Ѱ = + ).
tw
cos (C.5)
lv
tw
cos cos( ) (C.6a)
lh
tw
cos cos sin sin (C.6b)
lh
Substituting Eq. (C.5) into Eq. (C.6b), expanding, and then simplifying:
2
tw t t
cos w sin 1 w (C.7a)
lh lv lv
tw t
2 cos w
tw lh lv (C.7b)
1
lv sin
2 2
tw tw2 t
2 2cos cos 2 w (C.7c)
t
1 w h lv
l lv lh
lv sin 2
2 2 2
t t t2 t
sin sin w w 2cos w cos 2 w
2 2 (C.7d)
lv lh lv lh lv
1 2cos 1
sin 2 tw2 2 2 (C.7f)
lh lv lh lv
sin 2
tw2
1 2cos 1 (C.7g)
2 2
lh lv lh lv
lv lh sin
tw (C.8)
l lh2 2lv lh cos
2
v
(a) HSS 167.9 × 6.70 (heat no. 703172) (b) HSS 127.4 × 11.55 (heat no. 1421012)
Fig. C.7. All-weld-metal TC test results (E71T-1C- Fig. C.8. TC test results for end-plate material (heat
H8 electrode) no. 4506365)
(a) Experimental test set-up (post weld-rupture) (b) Weld rupture failure
(a) Weld rupture failure in test 26 (b) Weld rupture failure in test 27
(a) Experimental test set-up (showing re-welded connection (b) Weld rupture failure
on opposite side)
(a) Weld rupture failure in test 30 (b) Weld rupture failure in test 31
(a) Weld rupture failure in test 32 (b) Weld rupture failure in test 33
Table C.14. Application of component approach to CSA S16-01 with and without the sinθ factor for test no.
33
ρ Weld rupture strength Base metal rupture strength CSA S16-01 component strength e
with sinθ a without sinθ b with lh c with lv d without sinθ with sinθ
° kN kN kN kN kN kN
0 113.9 76.0 96.4 63.9 63.9 63.9
30 96.7 65.7 97.3 57.6 57.6 57.6
60 76.1 53.4 78.7 57.2 53.4 57.2
90 78.2 55.7 76.7 57.6 55.7 57.6
120 78.8 55.3 85.1 54.8 54.8 54.8
150 61.9 42.1 77.9 48.8 42.1 48.8
180 57.6 38.5 65.9 44.2 38.5 44.2
210 58.8 40.0 69.3 54.9 40.0 54.9
240 70.0 49.1 73.3 55.0 49.1 55.0
270 71.3 50.8 69.4 50.7 50.7 50.7
300 67.5 47.4 62.7 58.0 47.4 58.0
330 114.8 78.0 96.9 89.9 78.0 89.9
f
Pnw = 631 693
a
Eq. (4.3b) with Aw = tw × tributary weld length for each component.
b
Eq. (4.3b) without the sinθ factor, with Aw = tw × tributary weld length.
c
Eq. (4.3a) with Am = lh × tributary weld length.
d
Eq. (4.3a) with Am = lv × tributary weld length.
e
Minimum of columns 3-5 for CSA S16-01 without the sinθ factor, or the minimum of columns 2,4 and 5 for CSA S16-
01 with the sinθ factor.
f
Sum of component strengths.
Table C.15. Application of component approach to AISC 360 (2016), CSA S16 (2014), and EN1993-1-8
(2005) for test no. 33
ρ AISC 360-16 CSA S16-14 EN1993-1-8 e
without sinθ a with sinθ b
without sinθ c with sinθ d
Directional method
° kN kN kN kN kN
0 68.1 102.0 76.0 113.1 74.4
30 58.8 86.6 65.7 94.1 58.3
60 47.8 68.1 53.4 72.7 46.3
90 49.9 70.0 55.7 74.3 48.9
120 49.5 70.6 55.3 75.4 47.4
150 37.7 55.4 42.1 60.3 35.0
180 34.5 51.6 38.5 57.2 32.9
210 35.8 52.7 40.0 57.2 33.7
240 44.0 62.7 49.1 66.9 42.4
270 45.5 63.9 50.8 67.8 44.7
300 42.5 60.5 47.4 64.6 42.8
330 69.9 102.8 78.0 111.8 75.6
f
Pnw = 584 847 652 915 582
a
Eq. (4.1b) with Aw = tw × tributary weld length.
b
Eq. (4.1c) with Aw = tw × tributary weld length, and θ for each component.
c
Eq. (4.4a) with Aw = tw × tributary weld length for each component, but omitting the sinθ and Mw factors.
d
Eq. (4.4a) with Aw = tw × tributary weld length, and θ for each component, and Eq. (4.4b) with θ1 = θ and θ2 = 90°.
e
Eq. (4.8e) with tw, lw = tributary weld length, θ and for each component.
f
Sum of component strengths.
D D
Pi lt , b sin i , b cos i (C.9a)
2 2
where:
Db
lt 1 cos i (C.9b)
2 tan
the weld axis at point 𝑃⃗𝜌 (the point of interest), which lies between 𝑃⃗𝜌+𝛥𝜌 and 𝑃⃗𝜌−𝛥𝜌 . The smaller Δρ is, the
closer the approximation will be to the actual weld axis.
Step 4: For axial load, the line of action of the applied force (𝐹 ) will fall along the branch axis. Thus:
Step 5: Using the dot product, calculate the angle between the line of action of the applied force and the
weld axis (θ):
V F
cos 1 (C.11)
V F
Step 6: Increment ρ by Δρ, and repeat Steps 1 to 5 to determine the variation in θ with ρ. This variation
can then be transformed to determine the variation in θ with respect to the weld length by correlating the
point ρ to the subtended weld length. This approach is explained in Chapter 6 for CHS-to-CHS connections,
but for CHS-to-plate connections, it is analogous. The average loading angle (between any two points, a and
b) can be calculated by numerically integrating the resulting θ(lw) curve between a and b, and dividing by (b –
a), according to Eq. (C.12). This approach, as written, can also be applied to CHS-to-CHS connections, by
substituting the appropriate formula for lt.
1 b
b a a
avg d (lw ) (C.12)
The template length (lt) can be determined for an arbitrary point “A” along the intersection surface of the
tube and the plate as follows:
Step 1: Construct ΔABC (Fig. C.3), where AB is along the plate surface and ∠BAC is equal to the branch
inclination angle (θ). The length AC is equal to lt, and can be determined thus (as a function of the subtended
angle, ρ):
BC
lt AC (C.13)
tan
D D D
BC b b cos b 1 cos (C.14)
2 2 2
Db
lt 1 cos (C.15)
2 tan
217
Appendix D: Fillet Welds in HSS-to-Rigid Plate Connections: Finite Element Modelling 218
Table C.17. Summary of actual and predicted strengths for RHS- and CHS-to-rigid plate connections
according to EN1993-1-8:2005 Directional Method
Test no. EN1993-1-8:2005 Directional Method
Pa Pnw (nominal strength) Vrw (design strength)
kN kN (A/P) kN (A/P)
1 831 670 1.24 536 1.55
2 1166 1063 1.10 850 1.37
3 1235 1059 1.17 847 1.46
4 1311 1132 1.16 906 1.45
5 2433 1661 1.46 1329 1.83
6 2574 2304 1.12 1844 1.40
7 2525 2175 1.16 1740 1.45
8 2302 2104 1.09 1683 1.37
9 1020 737 1.38 590 1.73
10 960 619 1.55 495 1.94
11 840 555 1.51 444 1.89
12 1140 856 1.33 685 1.66
13 1200 1065 1.13 852 1.41
14 1207 1002 1.20 802 1.51
15 1494 1432 1.04 1146 1.30
16 1578 1741 0.91 1392 1.13
17 1788 2019 0.89 1615 1.11
18 1131 1271 0.89 1017 1.11
19 982 793 1.24 635 1.55
20 1270 1182 1.07 946 1.34
21 1534 1877 0.82 1502 1.02
22 1261 1017 1.24 814 1.55
23 1279 1395 0.92 1116 1.15
24 1459 1069 1.36 855 1.71
25 1597 1226 1.30 981 1.63
26 841 710 1.18 568 1.48
27 864 706 1.22 565 1.53
28 1450 1198 1.21 958 1.51
29 1331 1339 0.99 1071 1.24
30 1109 811 1.37 649 1.71
31 1479 1120 1.32 896 1.65
32 776 644 1.20 515 1.51
33 803 582 1.38 466 1.72
ρP = 1.19 Mean = 1.48
VP = 0.155
Appendix D provides the supplementary data for Chapter 5: Fillet Welds in HSS-to-Rigid Plate
Connections: Finite Element Modelling.
Fig. D.3. Graphical comparison of FE versus predicted ultimate load (showing FE failure modes)
(a) Test with Bb/tb =12.5 and tw/tb = 0.90 (b) Test with Bb/tb =30.0 and tw/tb = 0.90
Fig. D.4. Non-uniform FE strain distributions adjacent to the weld in RHS-to-rigid plate tests
Fig. D.5. Comparison of Pa/Py and PFE/Py to tw/tb for FE and experimental HSS-to-rigid plate tests
Table D.1. Summary of actual and predicted nominal strengths for 21 FE RHS-to-rigid plate connections
according to AISC 360-16, CSA S16-14, and EN1993-1-8:2005 Directional and Simplified Methods (θ = 90°
connections)
Pnw
tw/tb Bb/tb AISC 360-16 CSA S16-14 EN 1993-1-8:2005
Pa without sinθ a with sinθ b without sinθ c with sinθ d Directional e Simplified f
kN kN (A/P) kN (A/P) kN (A/P) kN (A/P) kN (A/P) kN (A/P)
0.35 50.0 508 381 1.33 571 0.89 425 1.19 638 0.80 438 1.16 357 1.42
0.50 50.0 679 539 1.26 808 0.84 602 1.13 902 0.75 619 1.10 505 1.34
0.71 50.0 911 762 1.20 1143 0.80 851 1.07 1276 0.71 875 1.04 715 1.27
0.90 50.0 1049 970 1.08 1455 0.72 1083 0.97 1624 0.65 1114 0.94 910 1.15
1.06 50.0 1247 1143 1.09 1714 0.73 1276 0.98 1914 0.65 1313 0.95 1072 1.16
0.35 40.0 631 474 1.33 711 0.89 529 1.19 794 0.79 545 1.16 445 1.42
0.50 40.0 855 670 1.28 1006 0.85 749 1.14 1123 0.76 770 1.11 629 1.36
0.71 40.0 1147 948 1.21 1422 0.81 1059 1.08 1588 0.72 1089 1.05 890 1.29
0.90 40.0 1380 1207 1.14 1810 0.76 1348 1.02 2022 0.68 1387 1.00 1132 1.22
1.06 40.0 1578 1422 1.11 2134 0.74 1588 0.99 2382 0.66 1634 0.97 1334 1.18
0.35 30.0 855 628 1.36 942 0.91 701 1.22 1052 0.81 721 1.19 589 1.45
0.50 30.0 1142 888 1.29 1332 0.86 991 1.15 1487 0.77 1020 1.12 833 1.37
0.71 30.0 1546 1256 1.23 1883 0.82 1402 1.10 2103 0.74 1443 1.07 1178 1.31
0.90 30.0 1872 1598 1.17 2397 0.78 1785 1.05 2677 0.70 1836 1.02 1499 1.25
0.35 20.0 1291 927 1.39 1391 0.93 1036 1.25 1553 0.83 1066 1.21 870 1.48
0.50 20.0 1748 1312 1.33 1967 0.89 1465 1.19 2197 0.80 1507 1.16 1230 1.42
0.71 20.0 2338 1855 1.26 2782 0.84 2071 1.13 3107 0.75 2131 1.10 1740 1.34
0.35 12.5 2088 1444 1.45 2166 0.96 1612 1.29 2419 0.86 1659 1.26 1355 1.54
0.50 12.5 2836 2042 1.39 3063 0.93 2280 1.24 3420 0.83 2346 1.21 1916 1.48
0.35 9.1 2831 1931 1.47 2896 0.98 2156 1.31 3234 0.88 2218 1.28 1811 1.56
0.50 9.1 3804 2730 1.39 4095 0.93 3049 1.25 4573 0.83 3137 1.21 2561 1.49
ρP = 1.27 0.85 1.14 0.76 1.11 1.36
VP = 0.09 0.09 0.09 0.09 0.09 0.09
a
Eq. (4.1b) with Aw = tw[4(Bb-4tb)+4πtb].
b
Eq. (4.1c) with Aw = tw[4(Bb-4tb)+4πtb] and θ = 90°.
c
Eq. (4.4a) with Aw = tw[4(Bb-4tb)+4πtb], but omitting the sinθ and Mw factors.
d
Eq. (4.4a) with Aw = tw[4(Bb-4tb)+4πtb], and θ = θ1 = θ2 = 90°.
e
Eq. (4.6b).
f
Eq. (4.6a).
Table D.2. Summary of actual and predicted nominal strengths for 20 FE CHS-to-rigid plate connections
according to AISC 360-16, CSA S16-14, EN1993-1-8:2005 Directional and Simplified Methods (θ = 90°
connections)
Pnw
tw/tb Db/tb AISC 360-16 CSA S16-14 EN 1993-1-8:2005
Pa without sinθ a with sinθ b without sinθ c without sinθ a with sinθ b without sinθ c
kN kN (A/P) kN kN (A/P) kN kN (A/P) kN kN (A/P) kN
0.35 50.0 304 215 1.42 322 0.94 240 1.27 360 0.84 247 1.23 202 1.51
0.50 50.0 441 304 1.45 456 0.97 339 1.30 509 0.87 349 1.26 285 1.55
0.71 50.0 578 430 1.35 644 0.90 480 1.20 720 0.80 494 1.17 403 1.43
0.90 50.0 693 547 1.27 820 0.84 611 1.13 916 0.76 628 1.10 513 1.35
0.35 40.0 400 269 1.49 403 0.99 300 1.33 450 0.89 308 1.30 252 1.59
0.50 40.0 544 380 1.43 570 0.96 424 1.28 636 0.86 436 1.25 356 1.53
0.71 40.0 732 537 1.36 806 0.91 600 1.22 899 0.81 617 1.19 504 1.45
0.90 40.0 868 683 1.27 1025 0.85 763 1.14 1145 0.76 785 1.11 641 1.35
0.35 30.0 524 358 1.47 536 0.98 399 1.31 599 0.87 411 1.28 336 1.56
0.50 30.0 741 506 1.47 759 0.98 565 1.31 847 0.87 581 1.28 474 1.56
0.71 30.0 982 715 1.37 1073 0.92 799 1.23 1198 0.82 822 1.19 671 1.46
0.90 30.0 1176 910 1.29 1366 0.86 1017 1.16 1525 0.77 1046 1.12 854 1.38
0.35 20.0 804 537 1.50 806 1.00 600 1.34 899 0.89 617 1.30 504 1.60
0.50 20.0 1117 759 1.47 1139 0.98 848 1.32 1272 0.88 873 1.28 712 1.57
0.71 20.0 1520 1074 1.42 1611 0.94 1199 1.27 1799 0.84 1234 1.23 1008 1.51
0.35 12.5 1329 859 1.55 1289 1.03 959 1.39 1439 0.92 987 1.35 806 1.65
0.50 12.5 1805 1215 1.49 1823 0.99 1357 1.33 2035 0.89 1396 1.29 1140 1.58
0.71 12.5 2499 1718 1.45 2578 0.97 1919 1.30 2878 0.87 1974 1.27 1612 1.55
0.35 9.1 1817 1181 1.54 1772 1.03 1319 1.38 1979 0.92 1357 1.34 1108 1.64
0.50 9.1 2495 1671 1.49 2506 1.00 1866 1.34 2799 0.89 1920 1.30 1567 1.59
ρP = 1.43 0.95 1.28 0.85 1.24 1.52
VP = 0.06 0.06 0.06 0.06 0.06 0.06
a
Eq. (4.1b) with Aw = twπDb.
b
Eq. (4.1c) with Aw = twπDb and θ = 90°.
c
Eq. (4.4a) with Aw = twπDb, but omitting the sinθ and Mw factors.
d
Eq. (4.4a) with Aw = twπDb, and θ = θ1 = θ2 = 90°.
e
Eq. (4.6b).
f
Eq. (4.6a).
Appendix E provides the supplementary data for Chapter 6: Fillet Weld Effective Lengths in CHS X-
Connections: Experimentation.
225
Appendix E: Fillet Weld Effective Lengths in CHS X-Connections: Experimentation 226
Fig. E.7. All-weld-metal TC test specimen, trial weld specimens, and parts list for CHS X-connection tests
The template length (lt) can be determined, for an arbitrary point “A” along the intersection surface of the
tubes, as follows:
Step 1: Construct ΔABC (Fig. E.8), where AB is perpendicular to the centreline of the branch and ∠BCA
is equal to the branch inclination angle (θ).
Step 2: Construct ΔCDE where CE is perpendicular to the centreline of the cord and ∠CDE is equal to θ.
Step 3: Define M at the crown of the chord along the line ECB; define Q such that ΔCDE and ΔMEQ are
similar triangles; define N at the heel of the connection; and define K as the point along the branch centreline
such that ∠MKN is equal to 90°.
lt KM EM BC CE (E.1)
Db
Db (E.2)
KM 2
tan 2 tan
D
D (E.3)
EM 2
sin 2sin
Db
AB cos (E.4a)
2
AB Db cos
BC (E.4b)
tan 2 tan
D
CD cos P (E.5a)
2
CD D cos P
CE (E.5b)
sin 2sin
D D
sin P b sin (E.6c)
2 2
Db
sin P sin (E.6b)
D
2
D
cos P 1 sin 1 b sin (E.6c)
2
D
Hence:
D 2 Db sin
2
(E.7)
CE
2sin
Substitute Eqs. (E.2), (E.3), (E.4b) and (E.7) into Eq. (E.1):
D 2 Db sin
2
Db D Db cos (E.8)
lt
2 tan 2sin 2 tan 2sin
Db (1 cos ) D D Db sin
2 2
(E.9)
lt
2 tan 2sin
Fig. E.11. 3D Solidworks models of weld profile for 102-series tests with θ = 90°
Fig. E.12. 3D Solidworks models of weld profile for 127-series tests with θ = 90°
Fig. E.13. 3D Solidworks models of weld profile for tests with θ = 90°
Table E.15. Comparison of average weld throat dimensions obtained from post-rupture macro-etch
examinations and external measurements
Test Average weld throat dimension (tw) A/B
Solidworks measurements Macro-etch measurements
prior to testing (A) after testing (B)
mm mm
102-273-90b 4.37 4.13 1.06
102-406-90a 3.56 3.54 1.01
127-273-90b 4.00 4.01 1.00
127-406-90b 3.47 3.82 0.91
102-406-60b 3.79 3.86 0.98
127-406-60b 3.38 4.31 0.78
(a) HSS 273.5 × 11.69 (heat no. 778039) (b) HSS 406.5 × 12.34 (heat no. 823306)
(c) HSS 410.0 × 12.21 (heat no. 777481) (d) HSS 102.0 × 7.34 (heat no. 819720-2)
Fig. E.14. TC test results for CHS chord and branch materials
(a) Weld rupture failure in test 102-273-90b (b) Weld rupture failure in test 102-406-90a
(a) Weld rupture failure (b) Overwelding of test joint (after test)
(a) Weld rupture failure in test 127-273-90a (b) Weld rupture failure in test 127-273-90b
(a) Weld rupture failure in test 127-406-60a (b) Weld rupture failure in test 127-406-60b
Table E.16. Summary of actual and predicted strengths for all experimental CHS X-connections according to
AWS D1.1-15, AISC 360-16, and CSA S16-14
Pnw
Test AWS D1.1-15 AISC 360-16 CSA S16-14
Paꞌ le = (2/3)lw le = lw le = lw le = lw
kN kN (A/P) kN (A/P) kN (A/P) kN (A/P)
102-273-90a 672 303 2.22 454 1.48 454 1.48 507 1.33
102-273-90b 678 324 2.09 486 1.39 486 1.39 543 1.25
102-406-90a 608 263 2.31 394 1.54 394 1.54 440 1.38
102-406-90b 540 232 2.33 348 1.55 348 1.55 388 1.39
127-273-90a 653 340 1.92 511 1.28 511 1.28 570 1.15
127-273-90b 653 375 1.74 563 1.16 563 1.16 628 1.04
127-406-90a 557 294 1.90 441 1.26 441 1.26 492 1.13
127-406-90b 557 323 1.73 484 1.15 484 1.15 540 1.03
102-406-60a 721 285 2.53 428 1.69 428 1.69 477 1.51
102-406-60b 721 302 2.39 453 1.59 453 1.59 505 1.43
127-406-60a 761 396 1.92 594 1.28 594 1.28 663 1.15
127-406-60b 850 339 2.51 508 1.67 508 1.67 568 1.50
ρP = 2.13 1.42 1.42 1.27
VP = 0.13 0.13 0.13 0.13
Appendix F provides the supplementary data for Chapter 7: Fillet Weld Effective Lengths in CHS X-
Connections: Finite Element Modelling.
254
Appendix F: Fillet Weld Effective Lengths in CHS X-Connections: Finite Element Modelling 255
Fig. F.1. Correlation of existing AWS D1.1-15 provisions with all test results, with weld effective lengths
Fig. F.2. Correlation of existing AWS D1.1-15 provisions (excluding weld effective lengths) and AISC 360-
16 provisions with all test results
Fig. F.3. Correlation of CSA S16-14 provisions with all test results