SLOTTED END CONNECTIONS TO HOLLOW SECTIONS

SLOTTED END CONNECTIONS TO HOLLOW SECTIONS by G. Martinez-Saucedo and J. A. Packer Department of Civil Engineering, University of Toronto, Canada FINAL REPORT TO CIDECT ON PROGRAMME 8G CIDECT Report 8G-10/06 August 2006

ABSTRACT This Report deals with the structural behaviour and design of concentrically-aligned single gusset plate welded connections to the ends of steel hollow section members. Such connections are commonly found in diagonal brace members of steel framed buildings and also in roof truss web-to-chord member connections. The types of sections considered are circular hollow sections and elliptical hollow sections, with the plate either slotted and welded into the tube or the tube welded into a slotted plate. In addition, the presence (or lack) of an open slot at the end of a slotted tube connection - a fabrication method particularly favoured in North America - is evaluated within the scope of this work. Under quasi-static loading, the behaviour of the connection has been rigorously studied under both axial tension and axial compression loadings, by both large-scale laboratory experiments and numerical (finite element) analysis. In addition, an exhaustive review and analysis of all prior international work in this field has been made. Non-linear finite element models, validated for all 13 laboratory test specimens, formed the basis of an extensive parametric study resulting in a further 891 "numerical tests" to supplement the data base of experiments by the author and other international researchers. In tension the tube failure modes of circumferential fracture (with or without the presence of shear lag) and tear out (or "block shear" failure) were clearly identified by both experimental and numerical investigations and the parameters influencing these limit states were thus clarified. As a result, new unified design provisions for such connections in tension are presented, which are shown to be a significant improvement over current international design provisions. In compression, the tube failure mode of local buckling governed throughout the connection study and the influence of the shear lag phenomenon - hitherto completely disregarded by all design provisions under compression loading - has been highlighted. A new static design method for slotted end connections in compression is hence advocated, which is shown to be applicable to circular, elliptical, square and rectangular hollow sections. Guidance on the proportioning of the longitudinal fillet welds, so that these do not govern the connection capacity, is also presented. The above static design recommendations, which now more truly reflect the actual connection performance, allow connections to be designed in a more efficient manner. ii

TABLE OF CONTENTS ABSTRACT...ii TABLE OF CONTENTS...iii NOTATION...x CHAPTER 1:INTRODUCTION... 1-1 1.1 Project overview... 1-2 CHAPTER 2:LITERATURE REVIEW... 2-1 2.1 The shear lag phenomenon... 2-1 2.2 Tear-out failure... 2-4 2.3 International specifications... 2-6 2.4 Summary of Chapter 2... 2-10 CHAPTER 3:EXPERIMENTAL PROGRAM... 3-1 3.1 Material properties... 3-1 3.1.1 Stub column tests... 3-4 3.2 Test specimens and instrumentation... 3-6 3.3 Experimental test results... 3-10 3.3.1 Slotted CHS connection - slot end not filled (type A)... 3-10 3.3.2 Slotted CHS connection - slot end filled (with a weld return) (type B)... 3-13 3.3.3 Slotted EHS connection - slot end not filled (gusset plate oriented to give a large eccentricity)... 3-15 3.3.4 Slotted EHS connection - slot end not filled (gusset plate oriented to give small eccentricity)... 3-18 3.3.5 Slotted gusset plate to tube connections in tension... 3-21 3.3.5.1 Slotted gusset plate to CHS connection (type C)... 3-21 3.3.5.2 Slotted gusset plate to EHS connection (gusset plate oriented to give a large eccentricity)... 3-24 3.3.6 Connections under compression load... 3-27 3.3.6.1 Slotted CHS to gusset plate connection - slot end not filled... 3-28 3.3.6.2 Slotted gusset plate to CHS connection... 3-30 3.4 Summary of this experimental program... 3-32 iii

CHAPTER 4:EVALUATION OF EXPERIMENTS AGAINST DESIGN PROVISIONS... 4-1 4.1 Experimental program by British Steel (1992)... 4-2 4.2 Experimental program by Korol et al. (1994)... 4-3 4.3 Experimental program by Zhao and Hancock (1995)... 4-4 4.4 Experimental program by Cheng et al. (1996)... 4-7 4.5 Experimental program by Zhao et al. (1999)... 4-8 4.6 Experimental program by Wilkinson et al. (2002)... 4-10 4.7 Experimental program by the Authors... 4-10 4.8 Experimental program by Ling (2005)... 4-12 4.9 Summary of Chapter 4... 4-14 CHAPTER 5:FE MODELLING OF CONNECTIONS... 5-1 5.1 Material properties... 5-1 5.2 Connection modelling... 5-4 5.2.1 Element selection... 5-6 5.2.2 Analysis considerations... 5-6 5.3 Evaluation of FE models against experimental results... 5-8 5.3.1 Slotted CHS connection - slot end not filled (Type A)... 5-9 5.3.2 Slotted CHS connection - slot end filled (weld return) (Type B)... 5-12 5.3.3 Slotted EHS connection - slot end not filled (gusset plate oriented to give a large eccentricity)... 5-16 5.3.4 Slotted EHS connection - slot end not filled (gusset plate oriented to give small eccentricity)... 5-20 5.3.5 Slotted gusset plate to tube connections in tension... 5-23 5.3.5.1 Slotted gusset plate to CHS connection (Type C)... 5-23 5.3.5.2 Slotted gusset plate to EHS (gusset plate oriented to give a large eccentricity)... 5-26 5.3.6 Connections under compression load... 5-30 5.3.6.1 Slotted CHS to gusset plate connection - slot end not filled... 5-30 5.3.6.2 Slotted gusset plate to CHS connection... 5-32 5.4 Summary of Chapter 5... 5-35 CHAPTER 6:PARAMETRIC FINITE ELEMENT ANALYSIS... 6-1 6.1 Parametric analysis results of slotted CHS connection - slot end not filled... 6-1 iv

6.2 Parametric analysis results of slotted CHS connection - slot end filled (weld return)... 6-7 6.3 Parametric analysis results of slotted EHS connection - slot end not filled (gusset plate oriented to give a large eccentricity)... 6-11 6.4 Parametric analysis results of slotted EHS connection - slot end not filled (gusset plate oriented to give small eccentricity)... 6-14 6.5 Slotted gusset plate to tube connection in tension... 6-17 6.5.1 Parametric analysis results of slotted gusset plate to CHS connection... 6-17 6.5.2 Parametric analysis results of slotted gusset plate to EHS connection (gusset plate oriented to give a large eccentricity)... 6-23 6.6 Connections under compression load... 6-29 6.6.1 Parametric analysis results of slotted CHS connection - slot end not filled... 6-29 6.6.2 Parametric analysis results of slotted gusset plate to CHS connection... 6-32 6.7 Weld design... 6-35 6.8 Summary of Chapter 6... 6-38 CHAPTER 7: ANALYSIS OF FE AND EXPERIMENTAL RESULTS... 7-1 7.1 CHS connections in tension - CF failure... 7-1 7.1.1 Shear lag equations suggested for CSA design provision format... 7-1 7.1.1.1 Equation suggested for slotted CHS to gusset plate connections... 7-1 7.1.1.2 Equation suggested for slotted gusset plate to CHS connections based on ultimate strength... 7-3 7.1.1.3 Equation suggested for slotted gusset plate connections based on deformation limit (0.03D)... 7-5 7.1.2 Shear lag equations suggested for AISC design provision format... 7-6 7.1.2.1 Equation suggested for slotted CHS to gusset plate connections... 7-7 7.1.2.2 Equation suggested for slotted gusset plate to CHS connections based on ultimate strength... 7-8 7.1.2.3 Equation suggested for slotted gusset plate connections based on deformation limit (0.03D)... 7-9 7.2 EHS connections in tension - CF failure... 7-11 v

7.2.1 Shear lag equations suggested for CSA design provision format... 7-11 7.2.1.1 Equation suggested for slotted EHS to gusset plate connections... 7-11 7.2.1.2 Equation suggested for slotted gusset plate to EHS connections based on ultimate strength... 7-12 7.2.1.3 Equation suggested for slotted gusset plate connections based on deformation limit (0.03D 2 )... 7-14 7.2.2 Shear lag equations suggested for AISC design provision format... 7-15 7.2.2.1 Equations suggested for slotted EHS to gusset plate connections... 7-15 7.2.2.2 Equations suggested for slotted gusset plate to EHS connections based on ultimate strength... 7-17 7.2.2.3 Equation suggested for slotted gusset plate to EHS connections based on deformation limit (0.03D 2 )... 7-18 7.3 CHS and EHS connections in tension - TO failure... 7-19 7.4 CHS connections in compression... 7-29 7.4.1 Equation suggested for slotted CHS to gusset plate connections (under compression loading)... 7-29 7.4.2 Equation suggested for slotted gusset plate connections (under compression loading)... 7-30 7.5 Evaluation of recommended equations against experimental data... 7-31 7.5.1 Experimental program by British Steel (1992)... 7-33 7.5.2 Experimental program by Korol (1994)... 7-34 7.5.3 Experimental program by Cheng et al. (1996)... 7-34 7.5.4 Experimental program by the Authors... 7-35 7.6 Derivation of reduction (resistance) factors for the recommended equations... 7-37 7.6.1 Reduction factors for CHS connections in tension - CF failure... 7-37 7.6.1.1 Reduction factors for suggested equations for slotted CHS connections (CSA design provision format)... 7-37 7.6.1.2 Reduction factors for suggested equations for slotted gusset plate to CHS connections based on ultimate strength (CSA design provision format)... 7-38 7.6.1.3 Reduction factors for suggested equations for slotted gusset plate to CHS connections based on deformation limit (CSA design provision format)... 7-39 vi

7.6.1.4 Reduction factors for suggested equations for slotted CHS connections (AISC design provision format)... 7-39 7.6.1.5 Reduction factors for suggested equations for slotted gusset plate to CHS connections based on ultimate strength (AISC design provision format)... 7-40 7.6.1.6 Reduction factors for suggested equations for slotted gusset plate to CHS connections based on deformation limit (AISC design provision format)... 7-40 7.6.2 Reduction factors for EHS connections in tension - CF failure... 7-41 7.6.2.1 Reduction factors for suggested equations for slotted EHS connections (CSA design provision format)... 7-41 7.6.2.2 Reduction factors for suggested equations for slotted gusset plate to EHS connections based on ultimate strength (CSA design provision format)... 7-41 7.6.2.3 Reduction factors for suggested equations for slotted gusset plate to EHS connections based on deformation limit (CSA design provision format)... 7-41 7.6.2.4 Reduction factors for suggested equations for slotted EHS connections (AISC design provision format)... 7-42 7.6.2.5 Reduction factors for suggested equations for slotted gusset plate to EHS connections based on ultimate strength (AISC design provision format)... 7-42 7.6.2.6 Reduction factors for suggested equation for slotted gusset plate to EHS connections based on deformation limit (AISC design provision format)... 7-43 7.6.3 Reduction factors for CHS and EHS connection in tension - TO failure... 7-44 7.6.3.1 Reduction factors for slotted CHS connections - TO failure... 7-44 7.6.3.2 Reduction factors for slotted gusset plate to CHS connections - TO failure... 7-44 7.6.3.3 Reduction factors for slotted EHS connections - TO failure... 7-44 7.6.3.4 Reduction factors for slotted gusset plate to EHS connections - vii

TO failure... 7-45 7.6.4 Reduction factors for CHS connections in compression... 7-45 7.6.4.1 Reduction factors for slotted CHS connections in compression... 7-45 7.6.4.2 Reduction factors for slotted gusset plate to CHS connections in compression... 7-46 7.7 Summary of Chapter 7- recommended static design methods... 7-46 7.7.1 Recommended static design method for CHS connections in tension... 7-47 7.7.2 Recommended static design method for EHS connections in tension... 7-49 7.7.3 Recommended static design method for CHS connections in compression7-50 CHAPTER 8:CONCLUSIONS AND RECOMMENDATIONS FOR FURTHER RESEARCH8-1 8.1 Overview... 8-1 8.2 Recommended static design methods... 8-2 8.2.1 Recommended static design method for CHS connections in tension... 8-2 8.2.2 Recommended static design method for EHS connections in tension... 8-4 8.2.3 Recommended static design method for CHS connections in compression 8-5 8.3 Design recommendation for seismic applications... 8-6 8.4 Recommendations for further research... 8-6 CHAPTER 9:REFERENCES... 9-1 APPENDIX A: EXPERIMENTAL PROGRAM...A-1 A.1 Slotted end connections to CHS...A-1 A.2 Slotted end connection to EHS...A-3 APPENDIX B: EVALUATION OF EXPERIMENTS...B-1 B.1 Experimental program by British Steel (1992)...B-1 B.2 Experimental program by Korol el al. (1994)...B-2 B.3 Experimental program by Zhao and Hancock (1995)...B-3 B.4 Experimental program by Cheng et al. (1996)...B-5 B.5 Experimental program by Zhao et al. (1999)...B-6 B.6 Experimental program by the Authors...B-8 B.7 Experimental program by Ling (2005)...B-9 APPENDIX C: STRAIN READINGS...C-1 C.1 Connections under tension...c-1 viii

C.1.1 Slotted CHS connections - slot end not filled (A1 and A2)...C-1 C.1.2 Slotted CHS connections - slot end filled (weld return) (B1 and B2)...C-3 C.1.3 Slotted EHS connections - slot end not filled (E1 and E2)...C-5 C.1.4 Slotted EHS connections - slot end not filled (E5)...C-7 C.1.5 Slotted gusset plate to CHS connection (C1 and C2)...C-8 C.1.6 Slotted gusset plate to EHS connection (E3 and E4)...C-10 C.2 Connections under compression...c-12 C.2.1 Slotted CHS to gusset plate connection - slot end not filled (A3C)...C-12 C.2.2 Slotted gusset plate to CHS connection (C3C)...C-13 ix

NOTATION A g = gross cross-sectional area of hollow section A gt = gross area in tension for block failure A gv = gross area in shear for block failure a l = weld leg length (size) A n = net cross-sectional area of hollow section A' ne = effective net cross-sectional area of hollow section A nt = net area in tension for block failure A nv = net area in shear for block failure A w = area of effective weld throat B = width of overlapped gusset plate b = overall width of RHS and SHS, measured 90 degrees to the plane of the connection C SC = compressive strength of stub column CHS = Circular Hollow Section D = outside diameter of CHS D 1 = larger dimension of EHS D 2 = smaller dimension of EHS D avg = average between larger and smaller dimension of EHS D/t = ratio between outside diameter and wall thickness of CHS E = modulus of elasticity EHS = Elliptical Hollow Section F y = yield tensile stress F u = ultimate tensile stress h = overall height of RHS and SHS, measured in the plane of the connection HAZ = HSS = Heat Affected Zone Hollow Structural Section x

K = effective length factor LVDT = Linear Variable Differential Transformer L w = weld length L w /w = L w /D = ratio between weld length and distance between welds ratio between weld length and outside diameter of CHS l sl = length of slot in hollow section N u = calculated connection strength according to design provisions N ux = measured connection strength N ufe = N ufe-d = RHS = connection strength from FE analysis connection strength from FE analysis based on distortion limit Rectangular Hollow Section R t = tension area mean stress correction factor R v = shear area mean stress correction factor SHS = Square Hollow Section t = wall thickness of CHS t p = thickness of gusset plate T r = factored tensile resistance t sl = width of slot in CHS Tσ-Tε = uniaxial true stress - true strain curve U = reduction coefficient for shear lag in net section fracture calculation U bs = reduction factor for non-uniform stress in block shear V r = factored shear resistance V R = coefficient of variation w = distance between the welds, measured around the perimeter of the CHS w p = width of gusset plate x = eccentricity xi

x' = eccentricity reduced by half of flange-plate thickness ( = x - t p /2) x /L w = ratio between the eccentricity and weld length x' /L w = ratio between the reduced eccentricity and weld length z = longitudinal distance between strain gauges β = safety index or reliability index γ M0 = Eurocode 3 partial safety factor when neither buckling phenomena nor ultimate resistance in tension is under consideration (= 1.0) γ M2 = Eurocode 3 partial safety factor when ultimate resistance in tension is under consideration (= 1.25) ε u = ultimate strain at rupture ε ef = equivalent fracture strain ϕ m φ = mean actual-to-predicted ratio = resistance factor xii

CHAPTER 1: INTRODUCTION 1-1 Circular hollow sections (CHS) have gained in popularity in recent years, particularly for architecturally exposed structural steel. Architects appreciate the clear form of CHS as well as their excellent structural properties in compression and torsion. In order to take full advantage of these properties, the complete tube cross-section should ideally be engaged at the connection. However, the feasibility of doing this is determined by the shape of the elements merging at the connection, which may result in a complicated task for detailing and fabrication. As a result, the use of a simplified connection detail will always be desirable whenever possible. Gusset plate connections represent one of the easiest methods to connect CHS used as web members in roof trusses and brace members in buildings. During the fabrication of these connections, the gusset plate or the CHS can be slotted resulting in several possible fabrication details. The application of either detail will depend on existing tolerances during the process of fabrication and erection of the structure. Despite these connection details providing the simplest manner for connecting CHS, it is important to recognize that an incorrect understanding of their behaviour may result in their failure or an expensive conservative design. As a consequence of only part of the CHS cross-section being connected, an uneven stress distribution around the tube circumference always occurs during the load transfer at the connection. Shear lag (see Figure 1.1) leads to stress peaks at the beginning of the weld which may result in connection failure by a circumferential failure (CF) mode. Moreover, a tear-out (or block shear ) failure (TO) may also occur under tension loading. Beginning of the weld Figure 1.1 Shear lag in slotted CHS connection SLOTTED END CONNECTIONS TO HOLLOW SECTIONS, CH1: INTRODUCCION

1-2 Despite both these limit states being addressed in current North American design provisions (AISC 2005 and CSA 2001), it has been found that the predicted connection strength (in the parameter range when CF is governing failure mode) will always differ as these two design provisions use dissimilar methods to account for this phenomenon. Although it is expected that these AISC and CSA design methods will always predict conservative connection capacities when CF governs, it has been found that the number of studies (specifically in slotted end connections to hollow sections) is limited to verify the accuracy and validity limits for each method. Moreover, the model currently used in design provisions (AISC 2005, CEN 2005 and CSA 2001) to account for TO failure, which was initially developed for bolted connections, lacks studies verifying its accuracy and validity limits for these connection types. In a similar manner to tension loading, an uneven stress distribution can be expected at the connection under compression loading. However, it has been found that this phenomenon is completely disregarded by design provisions, despite the fact that it may induce tube local buckling at the beginning of the welds. 1.1 Project overview This Report is directed to clarify the behaviour of slotted end connections fabricated with CHS and Elliptical Hollow Sections (EHS), their possible failure mechanisms and the relation of these failure modes to the connection geometrical dimensions, under tension and compression loading. In order to verify the accuracy of models currently used by design provisions, these are compared against available experimental data from previous studies and data from an experimental program undertaken at the University of Toronto. Results from these comparisons revealed the deficiency of these provisions to correctly predict the connection strength and governing failure mechanisms. A further parametric analysis based on finite element models of CHS and EHS connections has provided information on the behaviour of these connections and also provided further evidence of the imprecision of current design provisions. Therefore, a new comprehensive static design method is recommended here which also illustrates the possibility of effectively diminishing the influence of shear lag in these connections. SLOTTED END CONNECTIONS TO HOLLOW SECTIONS, CH1: INTRODUCCION

CHAPTER 2: LITERATURE REVIEW 2-1 The use of slotted end connections to hollow sections is very popular nowadays. However, the design methods against the most frequent failure modes such as circumferential tensile fracture (CF) of the HSS (see Figure 2.2) and tear-out (TO) failure along the weld (see Figure 2.3), seem to still require further attention. During the load transfer from the tube to the gusset plate, a nonuniform strain distribution takes place in the tube cross-section as the unconnected material is less able to participate in the load transfer. This phenomenon, known as Shear Lag, creates a high strain concentration at the weld region which eventually can trigger the fracture of the tube material there. Moreover, the propagation of this crack (defining a typical failure mode) and the connection strength are strongly influenced by the weld length (L w ). Figure 2.2 Circumferential tensile fracture Figure 2.3 Tear-out failure 2.1 The shear lag phenomenon Since the first model to account for the shear lag phenomenon was proposed by Chesson and Munse (1963), it has been included in several design specifications. Initially it was applied to riveted and bolted connections. Afterwards, the same model was utilized for the design of welded connections. Even though this phenomenon has been studied extensively for open structural sections, studies from Easterling and Giroux (1993) and Kirkham and Miller (2000) SLOTTED END CONNECTIONS TO HOLLOW SECTIONS, CH2: LITERATURE REVIEW

2-2 have revealed that existing design approaches are overly conservative and that further research may be required. In addition, this model has been applied to tubular connections. However, the research for these connection types is relatively recent and limited in scope. To allow for shear lag on connections fabricated with Hollow Structural Sections (HSS), Packer and Henderson (1992) proposed that the distance between the welds (w) be measured along the developed perimeter of the HSS (see Figure 2.4). In addition, they also suggested an efficiency coefficient for connections with L w /w ratios less than unit. At this time, the use of small ratios was not considered for CAN/CSA-S16.1-M89 (CSA 1989) since it was estimated that the weld was critical for L w /w ratios less than one. Figure 2.4 Important dimensions in slotted end connections A specific study of shear lag-induced fracture in tubular connections started in early 1990s when British Steel (1992) studied gusset plate connections to circular hollow sections (CHS), square hollow sections (SHS) and rectangular hollow sections (RHS) under tension and compression loading. An experimental program on slotted SHS and RHS to gusset plate connections was undertaken by Korol et al. (1994). In this program, a total of 18 specimens with L w /w ratios ranging from 0.40 to 1.00 were tested. Their results confirmed that a net section failure can occur in connections with ratios L w /w < 1.00. Moreover, a ratio of L w /w = 0.60 was proposed as a lower limit for the net section failure mode. A FE analysis of these connections was made considering only their elastic response, hence the FE models could not predict the failure mode. Based on these models, a further parametric analysis determined the influence that geometrical ratios have on the shear lag phenomenon; the L w /w ratio was shown to have the major influence and tube effective depth-to-width ratio a minor influence. Finally, the results indicated the need for variable shear lag factors for slotted SHS and RHS connections. SLOTTED END CONNECTIONS TO HOLLOW SECTIONS, CH2: LITERATURE REVIEW

2-3 Girard et al. (1995) generated a FE model of a connection between a SHS and a gusset plate. Even though this FE model exhibited some limitations, their results displayed differences with the equations in CAN/CSA-S16.1-M89 (CSA 1989). Cheng et al. (1996) studied the phenomenon in CHS, undertaking an experimental program and a FE analysis of these connections. A total of nine connections were tested, these connections were fabricated with a slotted tube and, except for one, all had a weld return. The results showed the inaccuracy of the shear lag factors in CAN/CSA-S16.1-94 (CSA 1994) for this type of connection. Additionally, the results for the connection with no weld return always presented an uneven strain distribution at the slotted end. For the same CAN/CSA-S16.1-94 (CSA 1994), Korol (1996) reached a similar conclusion for slotted gusset plate connections fabricated with SHS and RHS. Cheng et al. (1998) and Cheng and Kulak (2000) suggested that the reduction in the effective net area would be eliminated for CHS connections if a minimum weld length (L w ) of 1.3 times the tube diameter is provided. Experimental programs in gusset plates slotted into RHS were also undertaken by Zhao and Hancock (1995), Zhao et al. (1999) and Wilkinson et al. (2002). Although the failure mode in the latter was not directly related to the shear lag effect, the results suggested the need to verify the factors to account for shear lag. Recently, CHS connections with very high strength tubes have been studied by Ling (2005), resulting in a design method which considers the heat affected zone. However, due to the characteristics of the tube material used during this experimental program these results may not be suitable for regular grade HSS connections. Humphries and Birkemoe (2004) studied primarly double channel to gusset plate connections but these were compared with RHS to gusset plate connections. The results showed that the channels had a better behaviour than the RHS as they were able to deform reducing the eccentricity ( x ), thus increasing the connection effiency. This study also pointed out the influence that the weld leg size (a l ) has on the connection strength, as an increase in this was associated with an enhancement of the connection efficiency. Although these research studies have contributed information related to the influence that shear lag has in tubular connections, they have also showed the need to continue with more definitive studies in order to provide design provisions with formulae that accurately reflects this phenomenon. SLOTTED END CONNECTIONS TO HOLLOW SECTIONS, CH2: LITERATURE REVIEW

2-4 2.2 Tear-out failure In general, the research on tear-out failure or block failure has been mainly aimed at bolted connections, using gusset plates, coped beams or angles. The first model for tear-out failure (based on tests of coped beam connections) was proposed by Birkemoe and Gilmor (1978) and was eventually included in the AISC specification (1978). This model calculates the connection resistance by adding the shear resistance of the shear area and the tensile resistance of the net tensile area. Since then, several investigations have been undertaken on different bolted connections types. In order to verify the accuracy of the AISC specification (1978), Yura et al. (1982) tested twelve beam web shear connections. During these tests, several parameters such as: the edge distance, standard and slotted holes, coped beams, uncoped beams and bolt arrangement were studied. The results revealed a decrease in the connection capacity (approximately 20%) when slotted holes were used, and the use of two rows of bolts clustered at the top of the web produced a lower safety factor than that expected. Finally, for a connection with a single row of bolts, a recommendation to calculate the connection capacity as the sum of the bolts single capacity rather than a group capacity was made. In a further study (Ricles and Yura 1983), a finite element analysis of these connections (considering only the connection elastic response) showed a uneven stress distribution along the vertical plane at the cope. These results disagreed with an ideal stress distribution calculated by simple beam theory. In general, fracture initially started at the tension region where an uneven stress distribution was taking place and it was combined with a substantial material yielding along the shear plane. Based on these results, a new block shear model (with a triangular stress distribution on the tension region) was proposed for double row bolted connections. Hardash and Bjorhovde (1985) evaluated the application of the block-shear concept in gusset plates connections via the testing of 28 specimens. The test specimens were fabricated with two lines of bolts with various bolt rows, pitch spacing and bolt diameters. During these tests, the dominating failure mode corresponded to the attainment of the ultimate stress along the net area in tension (at the last row of bolts) and yielding of the gross area in shear (outside of the line of the bolts). In addition to this, the data from experimental programs at the University of Illinois and the University of Alberta were combined with these results to develop a new block shear model. In general, this new model followed the original block failure model. Nevertheless, it included several new factors to calculate the ultimate resistance of the connection which made its use difficult. SLOTTED END CONNECTIONS TO HOLLOW SECTIONS, CH2: LITERATURE REVIEW

2-5 Epstein (1992) undertook an experimental program to study block shear failure in angles, with a total of 38 angle connection tests. These results showed variations with the values recommended by AISC design provisions (1986, 1989) at that time. These differences were mainly associated with the effect that the unconnected leg eccentricity had over the connection behavior, modifying the failure mechanism. Gross et al. (1995) tested 13 angle connections fabricated with a single line of bolts and steel grades A-36 and A588. In general, the results showed good correlation with AISC design provisions (1989, 1994) based on agreement with the failure load. However, an inconsistency was observed between the failure mechanism predicted by design provisions and experimental test. Based on data published in previous experimental programs, Cunningham et al. (1995) suggested a model to predict block shear failure in connections fabricated with angles and bolts. Orbison et al. (1999) tested several angles, WT and W sections which failed in block shear (a total of 17 specimens). The failure mechanism observed during the tests consisted of a fracture at the tension area which was combined with a considerable inelastic deformation along the gross shear area. Even though the predicted connection capacity by the (then-current) design provision (AISC 1994) resulted in conservative values, the expected failure mechanism disagreed with the tests results. Additionally, several factors such as: low ductility, hole fabrication (punched or drilled) and large in plane and out-of-plane eccentricities were found to have an influence on the connection capacity. Finally, a further study of these factors was suggested since they were not considered in design provisions. Swanson and Leon (2000) tested 48 T-stub specimens under monotonic and cyclic loading. From all these test specimens, only one failed by block failure (this specimen was tested under cyclic loading). For this test specimen, the predicted failure mechanism (AISC 1994) did not coincide with the failure observed during the test. Aalberg and Larsen (2000) tested splice plates, beam web connections loaded in shear and beams connections with a coped end using high strength steels. The results were compared with design provisions such as: Eurocode (CEN 1992), CSA (1989) and AISC (1994). In general, an important decrease in the connection ductility was observed as a result of the use of these steel types and the importance of limiting the deformation of these connections was addressed. For block shear failure, only the CSA (1989) method was found to be suitable for high strength steels. A review of the rules for block shear design (AISC 1999) by Kulak and Grondin (2001) suggested that these may be conservative for gusset plates, acceptable for angles and non-conservative for coped beams. This study recommended that further research of this failure mode was required. SLOTTED END CONNECTIONS TO HOLLOW SECTIONS, CH2: LITERATURE REVIEW

2-6 In addition to these experimental programs, several studies have been undertaken with the assistance of finite element models by Epstein and Chamarajanagar (1996), Epstein and McGinnis (2000), Barth et al. (2002) and Topkaya (2004). In this last study, new models to calculate the tear out failure have been suggested. As result of these research programs, the governing failure criteria defining the block shear as well as resistance factors have experienced several modifications in existing design provisions (Geschwindner, 2004). However, the initial model (suggested by Birkemoe and Gilmor) which adds the resistances in tension and shear continues in use. Nowadays, the new trend to design by block shear follows this model, but with the use of several reduction factors. As an example of this, the AISC design provision (2005) has suggested a reduction factor (U bs ) to consider the uneven stress distribution that can be found in coped beams. Finally, a unified equation suitable for all types of connections has been recently proposed by Driver et al. (2006), wherein the initial model is used but several factors are applied depending on the connection type. 2.3 International specifications When the capacity of a tension member is governed by the limit state of tensile fracture affected by shear lag, several values can be calculated from current design provisions as they do exhibit differences. In general, these provisions consider the non-uniform stress distribution caused by shear lag by including an efficiency factor (U). This factor decreases the tube net area (A n ) at the connection to an effective net area (A e or A' ne ). A e = A n U (as in AISC 2000, 2005) (2-1) A' ne = A n U (as in CSA 2001) (2-2) This effective net are is then used to calculate the connection strength. In order to calculate this efficiency factor (U), two general methods are currently most common. The first method can be found in American specifications (AISC 2000, 2005), where the connection eccentricity ( x ) is compared with the weld length (L w ), as proposed by Cheeson and Munse (1963) to allow for the shear lag phenomenon in riveted and bolted connections. Specifications using this approach are summarized in Table 2.1. By this method: SLOTTED END CONNECTIONS TO HOLLOW SECTIONS, CH2: LITERATURE REVIEW

2-7 x U = 1 ----- L w D where x = --- π for CHS; (2-3) 2 and x ------ 2 D 1 + 2D 1 D = ----------------------------- 2 for EHS (see Figure 2.5). (2-4) 3π D 1 + D 2 Equation 2-4 considers that the gusset plate is aligned with the dimension D 2 (see Figure 2.5). When the gusset plate orientation is parallel to the dimension D 1, the dimension D 1 2 should be replaced by D 2 2. The conventional interpretation of x has been the measurement from the tube centroidal axis. However, when a thick gusset plate is utilized. It may be feasible to consider a reduced x', which is the distance from the gusset plate surface to the centre of gravity of the half tube as shown in Figure 2.4. Figure 2.5 Eccentricity x of top half, for EHS. The second method compares the circumferential distance between the welds (w) with the weld length (L w ). Here the efficiency factor (U) is determined by values assigned to the ratio L w /w (see Table 2.1). This method can be found in the Canadian specification (CSA 1994, 2001) as well as in the design guide for hollow structural sections by Packer and Henderson (1997). Moreover, for slotted connections to hollow sections the distance w equals half of the HSS circumference minus the gusset plate thickness (t p ) or the slot width (t sl ). Eurocode3 (CEN 2005) only considers the effect of shear lag on bolted connections using angles connected by SLOTTED END CONNECTIONS TO HOLLOW SECTIONS, CH2: LITERATURE REVIEW

one leg and other unsymmetrically connected tension members. Eurocode3 (2005) hence is not listed in Table 2.1. 2-8 Table 2.1 Shear lag design provisions for round (and elliptical) hollow sections Specification or design guide Effective net area Shear lag coefficients Range of validity AISC (1999): LRFD Specification for Structural Steel Buildings AISC (2000): LRFD Specification for Steel Hollow Structural Sections A e = A n U x = x U = 1 ----- 0.90 with x = 2 2 ------ D 1 + 2D 1 D ----------------------------- 2 3π D 1 + D 2 L w D --- (for CHS) π (EHS, see Figure 2.5) no restrictions AISC (2005): Specification for Structural Steel Buildings U = 1- ----- x L w for 1.3D > L w D U = 1 for L w 1.3D (only CHS) L w D CSA (1994): Limit States Design of Steel Structures L w L w L w U = 1.0 for U = 0.87 for 2.0 > U = 0.75 for 1.5 > w 2.0 w 1.5 w 1.0 L w w CSA (2001): Limit States Design of Steel Structures A' ne = A n U U = 1.0 for L w w 2.0 U = 0.5 + 0.25 L w w for 2.0> L w w 1.0 U = 0.75 w for L w w < 1.0 L w no restrictions Packer and Henderson (1997): Hollow Structural Section Connections and Trusses - A Design Guide L w L w L w L w U = 1.0 for U = 0.87 for 2.0 > U = 0.75 for 1.5 > U = 0.62 for 1.0 > w 1.0 shear lag not critical for L w < 0.6 w w 1.5 w 1.0 w 0.6 T r = φ A e F u (AISC Specification, φ = 0.75) or T r =0.85 φ A' ne F u (CSA Specification, φ = 0.9). For block shear failure, the connection resistance is calculated by adding the portion of the load transferred as tension load, T r, and the portion of load transferred as shear load, V r. The different national/regional design specifications (AISC, CSA, Eurocode) either use the gross or net area for the calculation of T r and V r (see Table 2.2). In welded connections, the gross area becomes equal to the net area for the calculation or T r and V r due the absence of bolt holes. For the calculation of the shear load, the material strength is reduced to 0.60 F y or F y SLOTTED END CONNECTIONS TO HOLLOW SECTIONS, CH2: LITERATURE REVIEW

2-9 / 3. The factor U bs used in the American specification (AISC 2005) has been introduced to account for the stress distribution that can be found in coped beams, where U bs =0.5 is recommended. In gusset plate connections U bs is taken equal to unity. The Canadian specification (CSA 2001) uses a separate design formula for coped beams but it also results in the same reduction factor as the American specification. It is worthwhile noting that the latest Canadian and American specifications, while having essentially the same model for the block shear limit state, result in considerably different safety levels due to their different resistance factors ( φ ), as shown in Table 2.2 (although the Canadian value is currently under review). This is not the case for the shear lag design provisions (Table 2.1), where ( 0.9) ( 0.85) 0.75. Table 2.2 Block shear design provisions Specification or design guide AISC (1999): LRFD Specification for Structural Steel Buildings AISC (2000): LRFD Specification for Steel Hollow Structural Sections AISC (2005): Specification for Structural Steel Buildings CSA (2001): Limit States Design of Steel Structures Eurocode (CEN 2005): Design of Steel Structures - General Rules - Part 1-8: Design of Joints a) Block shear strength When A nt F u 0.6 A nv F u : T r + V r = φ [A nt F u + 0.6 A gv F y ] φ [A nt F u + 0.6 A nv F u ] When A nt F u < 0.6A nv F u : T r + V r = φ [A gt F y + 0.6 A nv F u ] φ [A nt F u + 0.6A nv F u ] with φ = 0.75 T r + V r = φ U bs A nt F u + 0.6 φ A gv F y φ U bs A nt F u + 0.6 φ A nv F u with φ = 0.75 and U bs = 1 T r + V r = φ A nt F u + 0.6 φ A gv F y φ A nt F u + 0.6 φ A nv F u with φ = 0.9 1 T r + V r = A nt F 1 1 u A nv F γ + y M 2 3 γ M 0 γ M0 γ M 2 = 1.0 and = 1.25 a) Design rule for bolted connections differs slightly. SLOTTED END CONNECTIONS TO HOLLOW SECTIONS, CH2: LITERATURE REVIEW

2-10 2.4 Summary of Chapter 2 As has been exposed throughout this chapter, research on TO failure has been mainly focused on several types of bolted connections. As a result of this, the first model suggested to predict the connection strength has experienced several modifications throughout the years. Nevertheless, the accuracy of this model still seems to need further attention or verification (especially for welded tubular connections). To account for shear lag (inducing a CF) in tubular connections, two general approaches are prevalent nowadays in current design provisions. However, the accuracy of these models has not been totally verified for slotted end connections to CHS or EHS. In order to asses the accuracy and suitability of the models recommended in current design provisions (which are suggested for the TO failure limit state and to account for shear lag phenomenon), these models are compared against the results from an experimental program carried out at the University of Toronto (Chapter 3 of this Report) and other relevant research programs undertaken on tubular connections (Chapter 4 of this Report). SLOTTED END CONNECTIONS TO HOLLOW SECTIONS, CH2: LITERATURE REVIEW

CHAPTER 3: EXPERIMENTAL PROGRAM 3-1 An experimental program has been undertaken at the University of Toronto on slotted end connections to hollow sections (CHS and EHS). The objective of this study was to identify the influence of parameters such as: the weld length (L w ), the eccentricity of the connection ( x ), the gusset plate orientation (for the EHS) and fabrication detail on the connection strength. In general, these parameters have been shown to affect the shear lag phenomenon in previous experimental programs and the calculated connection strength by current design codes is based on these parameters. As part of this program, a total of 13 connections were fabricated and tested under quasi-static tension and compression loading. A description of the connections, the material properties, the testing arrangement and results from the tests are given in this chapter. 3.1 Material properties For the fabrication of the connections, a CHS with a nominal size of 168 x 4.8mm was used and it was cold-formed Class C material with a minimum specified yield stress of 350MPa (CSA 2004). An EHS with a nominal size of 220x110x6.3mm was used and it was hot-finished with a minimum specified yield stress of 355MPa (EN 10210-1, CEN 1994). Plates with 25mm and 32mm thickness were required for the fabrication of the gusset plates; these plates had a minimum specified yield stress of 300MPa (CSA 2004). A group of test coupons was fabricated from tubes and plates in order to determine their material properties. Seven test coupons were taken from the CHS with two of them cut from the Heat Affected Zone (HAZ). A 25mm plate was used in the fabrication of the CHS connections and two test coupons were cut from this plate. Four test coupons were cut from the EHS and three 32mm plates were used in the fabrication of these connections so a total of six coupons were tested from these plates. The size and location of these coupons were made according to ASTM (2003). During testing, the engineering stress-strain relationship was acquired before the coupon test developed a neck. Afterwards, the clip gauge was removed from the test coupon. In some test coupons from the CHS, it was possible to acquire information beyond the formation of the neck but eventually the clip gauge had to be removed. In all the cases, the load and maximum elongation at rupture were determined for each coupon test. The engineering stress-strain curves from the materials are shown in Figures 3.1 to 3.4 and their measured material SLOTTED END CONNECTIONS TO HOLLOW SECTIONS, CH 3: EXPERIMENTAL PROGRAM

3-2 properties are given in Table 3.1. Additional information from the tube and gusset plate material is given in Appendix A. Figure 3.1 Coupon tests for CHS Figure 3.2 Coupon tests for 25 mm plate SLOTTED END CONNECTIONS TO HOLLOW SECTIONS, CH 3: EXPERIMENTAL PROGRAM

3-3 Figure 3.3 Coupon tests for EHS Figure 3.4 Coupon tests for 32 mm plate SLOTTED END CONNECTIONS TO HOLLOW SECTIONS, CH 3: EXPERIMENTAL PROGRAM

3-4 Table 3.1 Measured material properties E(GPa) a) F y (MPa) a) F u (MPa) a) ε u (%) a) CHS 196 498 b) 540 25.9 EHS 216 421 530 34.7 Plate (t p =25.7mm) 201 358 482 28.0 Plate (t p =32.0mm) 214 356 472 30.0 a) Properties determined by the average measurements from several tensile coupon tests. b) Using the 0.2% offset method, as material was cold-formed. 3.1.1 Stub column tests In addition to the test coupons, a stub column test was performed on both the CHS and the EHS to determine their properties under compression load. The specimen size and the testing procedure were as recommended by SSRC (Galambos 1998). Before testing, four strain gauges were placed around the tube s circumference at the mid-height (see Figure 3.5). This allowed the generation of an average σ-ε relationships for the tube materials. Results from the tests are given in Table 3.2 Figure 3.5 Strain gauges on stub columns Table 3.2 Stub column properties and test results Length (mm) Weight (Kg) Area a) (mm 2 ) c) C sc (kn) CHS 150 b) 2.91 2471-1213 EHS 104.7 b) 2.51 3053-1393 a) Measured area obtained by weighing a tube segment and using a density of 7850 kg/m 3. b) Average length measured with a caliper. c) C sc = Stub column ultimate compressive strength. SLOTTED END CONNECTIONS TO HOLLOW SECTIONS, CH 3: EXPERIMENTAL PROGRAM

3-5 Using the data acquired through the test of the CHS (see Figure 3.6), the calculation of average Young s Modulus agreed with the value previously determined from the tensile test coupons. A similar conclusion was achieved from the computation of the average yield stress. For the EHS, the average Young s Modulus (see Figure 3.7) also agreed well with the value previously determined by tensile test coupons. However, an increase of 8% was observed when the EHS stub column yield stress was compared to the tensile test coupons. This difference has been attributed to the uneven manner in which the EHS stub column changed its shape through the test, which likely resulted in a higher value. Figure 3.6 Stub column response of CHS SLOTTED END CONNECTIONS TO HOLLOW SECTIONS, CH 3: EXPERIMENTAL PROGRAM

3-6 Stress (MPa) 500 450 400 350 300 250 200 150 100 SGE SGN SGW SGS 50 Strain (mm/mm) 0 0 0.0002 0.0004 0.0006 0.0008 0.001 0.0012 0.0014 0.0016 0.0018 0.002 Figure 3.7 Stub column response of EHS 3.2 Test specimens and instrumentation A total of six connection types were examined throughout this experimental program (see Figure 3.8). Connection type A was fabricated with a slotted CHS which was connected to a 25mm thick gusset plate by longitudinal fillet welds. Connection type B was originally fabricated as connection type A, however, the slot was filled in when the weld return was included. This connection type eliminates the reduction in the gross cross-sectional area of the tube due to slotting. For connection type C, a 25mm thick gusset plate was slotted so the CHS gross crosssectional area remained unaffected. For this connection type, the tube and the gusset plate were connected by longitudinal fillet welds too. For the CHS tension tests, two specimens were fabricated for each connection type (A, B and C) and the main difference between specimens (from a similar connection type) was their weld length. Hence, they were labelled in a progressive order as the weld length increased. Additionally, specimens from the connection types A and C were fabricated and tested under compression loading. Five EHS specimens were fabricated for tensile testing. In order to avoid confusion amongst the EHS connections, these were simple labelled in a progressive order (E1 to E5) SLOTTED END CONNECTIONS TO HOLLOW SECTIONS, CH 3: EXPERIMENTAL PROGRAM

3-7 depending on their connection type and weld length. The connection types E1 and E2 were fabricated with a slotted EHS, with the gusset plates oriented to give a large eccentricity and only longitudinal weld lengths were used to transfer the load. Connection type E5 was similar to these connections, however the orientation of the gusset plate was changed to give a smaller eccentricity. In general, the connection types E3 and E4 were similar to connection type C, but the EHS was oriented to produce a large eccentricity. Figure 3.8 Connection types examined In all cases, the test specimens had a L w /w ratio within the range from 0.60 to 0.90 which guaranteed the presence of the shear lag phenomenon during the tests. All gusset plates and welds were dimensioned so as not to be critical. Fillet welds had a nominal size of 10 or 15 mm and they were fabricated using E480XX electrodes (CSA 2003). The tube lengths were 1.5 and 2.0 metres for the CHS and EHS respectively. In order to facilitate the tests, two identical connections were fabricated at each tube end, which allowed the testing of two connections with very similar weld lengths simultaneously (see Figure 3.9). The average dimensions and properties of the specimens are shown in Table 3.3. Additionally, all measured dimensions from the test specimens are given in Appendix A. SLOTTED END CONNECTIONS TO HOLLOW SECTIONS, CH 3: EXPERIMENTAL PROGRAM

3-8 Table 3.3 Measured dimensions and geometric properties of test specimens Specimen Tube a l (mm) L w (mm) w (mm) L w /w (mm) t p (mm) W p (mm) A1 10 156 0.65 197 A2 10 192 0.80 198 A3C 10 206 238 0.86 197 CHS B1 9 169 0.71 197 168.5x4.89 25.7 B2 A a) =2471 mm 2 9 208 0.87 198 C1 14 162 0.67 2 x 74.3 C2 14 195 239 0.81 2 x 75.5 C3C 14 200 0.83 2 x 74.3 E1 13 145 0.61 161 234 E2 EHS 14 182 0.77 161 E3 110.9x221.2x5.94 15 146 0.61 32.0 2 x 94.0 A a) =3054 mm 2 237 E4 15 175 0.73 2 x 93.8 E5 15 185 234 0.79 270 a) Measured area calculated by weighing a piece of HSS and using a density of 7850 kg/m 3 All the specimens were loaded in quasi-static axial tension to failure in a universal testing machine and displacement control was used throughout each test. Four LVDTs (linear variable differential transformers) were placed on each specimen to measure deformations during the test. The tube deformation reported herein corresponds to the average deformation measured by two LDVTs from the centre of the tube to the gusset plate. Each specimen was also equipped with 10 strain gauges to establish the strains in the connection region (see Figure 3.10). All this information was acquired with a computer during the tests and the use of white-washing allowed the identification of regions with high strain concentration that in most cases induced an early fracture in the tube material. For the compression tests performed on specimens A3C and C3C, a minimum free distance of 2t p was provided in the gusset plate between the machine clamps and the tube ends. In addition to the instrumentation used in the tension tests, a fifth LVDT was placed at the test specimen mid-height to measure its out-of-straightness during the test. SLOTTED END CONNECTIONS TO HOLLOW SECTIONS, CH 3: EXPERIMENTAL PROGRAM

3-9 Figure 3.9 Experiment setup for tests Figure 3.10 Location of strain gauges on test specimens SLOTTED END CONNECTIONS TO HOLLOW SECTIONS, CH 3: EXPERIMENTAL PROGRAM

3-10 3.3 Experimental test results During the tests, the connection strength was determined principally by one of the following failure mechanisms: a) a tear out failure (TO) where the crack initiated at the weld termination then propagated through the tube base material near the weld toes, b) a circumferential fracture (CF), where the crack initiated at the weld termination then propagated around the tube circumference, and c) a combination of both failure modes (TO-CF). In the latter, both failure mechanisms occurred simultaneously at the connection end. The four LVDTs installed on each test specimen measured the overall elongation from the mid-length of the test specimen to the gusset plates. Even though two connections were fabricated alike for each test specimen (one at each tube end), failure was generally concentrated at one end. This behaviour has been attributed to variations in actual weld lengths and imperfections included during fabrication. The load-deformation response shown for the test results corresponds to the failed connection. In general, all the connections exhibited an uneven strain distribution along the connection and around the tube circumference. From the data acquired during the test, the strain distribution in the connections is only presented for a stage near the end of the connection elastic response. The rest of the strain readings are given in Appendix C. 3.3.1 Slotted CHS connection - slot end not filled (type A) The use of this connection type is advantageous since the fabrication tolerance for the slot makes assembly of the parts easier. However, the presence of an open slot end can affect the overall connection behaviour, as seen by the tests. In general, the behaviour of these connections can be described in several stages. Initially, the connections showed an elastic response with an equivalent constant stiffness. Afterwards, the strain concentration in the slot region (due to the presence of the shear lag phenomenon) induced yielding of the tube material there, thus modifying the overall connection stiffness. The magnitude of the shear lag (affecting each connection), which is determined by the weld length, increases as the weld length decreases, and the weld length was the only distinction between the two test specimens. (Figure 3.11 shows a superior performance for the test specimen with the longer weld length, A2). At this yielding stage, whitewash flaking confirmed the strain concentration taking place in the tube base material near the weld start (in the slot region). The strain gauge readings from SLOTTED END CONNECTIONS TO HOLLOW SECTIONS, CH 3: EXPERIMENTAL PROGRAM

3-11 the test specimens also confirmed this, as they showed an uneven strain distribution around the tube circumference and along the connection (see Figure 3.12). Figure 3.11 Load-deformation response for connections type A As the test specimens elongated, deformation was concentrated in the slot region producing a gradual change of the tube shape (inducing the formation of a neck there). In addition, the uneven strain distribution at the slot cross-section (due to the shear lag phenomenon) stimulated a quick increase in the strains at the weld start location, where straining of the tube material continued until fracture occurred. In general, a longer weld length allowed a better load transfer over the connection which diminished the connection deformation, however tube material fracture always governed the connection behaviour. Once fracture started, the crack continued to propagate gradually from the weld heel to its toe. Then, depending of the load level and the strain distribution in the connection, the crack would continue to propagate over the weld length (TO) or around the tube circumference (CF). Specimen A1 showed both failure modes and specimen A2 only CF (see Figure 3.13). The maximum load and deformation attained by these test specimens are shown in Table 3.4. SLOTTED END CONNECTIONS TO HOLLOW SECTIONS, CH 3: EXPERIMENTAL PROGRAM

3-12 Figure 3.12 Strain distribution for test specimens A1 and A2 at 800kN Table 3.4 Ultimate capacity for connections type A Weld Length (mm) Test Load N ux (kn) Deformation @ Max Load (mm) Failure Mode N ux /A n F u Specimen A1 156 1032 8.8 TO-CF 0.87 Specimen A2 192 1154 8.8 CF 0.97 Figure 3.13 Failure in test specimens A1 and A2 SLOTTED END CONNECTIONS TO HOLLOW SECTIONS, CH 3: EXPERIMENTAL PROGRAM

3-13 3.3.2 Slotted CHS connection - slot end filled (with a weld return) (type B) The addition of a weld return to these test specimens eliminated the possibility of a failure through the tube net area. Moreover, it allowed the attainment of the maximum load with small deformations (see Figure 3.14). In both tests, the load increase produced a strain concentration that was located at the weld return region (specifically at the weld toe). This behaviour has been attributed to the difference in the ductility of the return welds, since these were loaded at 90º with respect to their longitudinal axis which creates a region of high stiffness. Figure 3.14 Load-deformation response for connections type B Whitewash flaking confirmed the strain concentration taking place at the weld return region as the tube material yielded there at an early stage of the tests. Moreover, the readings of the strain gauges always exhibited very uneven strain distributions around the connections. Figure 3.15 shows the strain distribution around the tube and along the connection length, at the end of the elastic response. The strain gauge readings around the tube circumference showed an improvement compared to the strain distribution from connections type A. However, the strains experienced an increase right at the weld return region, relative to connections type A (see Figures 3.15 and 3.12 at z=+50mm). For specimens B1 and B2, the strain distribution presented a dependency SLOTTED END CONNECTIONS TO HOLLOW SECTIONS, CH 3: EXPERIMENTAL PROGRAM

3-14 on the weld length and the specimen B1 (which had the smaller weld length) showed higher strain at z=+50mm (see Figure 3.15). Figure 3.15 Strain distribution in test specimens B1 and B2 at 800kN Once the overall connection stiffness noticeably changed, any load increment was associated with a gradual increase of the strains in the weld return region and a change in the tube cross-section shape. The maximum load was limited by the propagation of a crack in the tube material near the weld return toe. This crack spread gradually at a 45 degree angle from the gusset plate. Finally, specimen B1 showed a TO failure and specimen B2 a CF (see Figure 3.16). The maximum load and deformation attained by these test specimens are shown in Table 3.5. Table 3.5 Ultimate capacity for connections type B Weld Length (mm) Test Load N ux (kn) Deformation @ Max Load (mm) Failure Mode N ux /A n F u (A n =A g ) Specimen B1 169 1087 6.1 TO 0.91 Specimen B2 208 1211 6.1 CF 1.02 SLOTTED END CONNECTIONS TO HOLLOW SECTIONS, CH 3: EXPERIMENTAL PROGRAM

3-15 Figure 3.16 Failure in test specimens B1 and B2 3.3.3 Slotted EHS connection - slot end not filled (gusset plate oriented to give a large eccentricity) The behaviour of these connections emulated the response of specimens type A. However, some differences occurred herein which have been associated mainly to the tube geometry. During these tests, the overall connection response can be described by several stages. At first, the test specimens had a similar elastic stiffness, while strain concentrations developed at the slot region (specifically in the tube near the weld start). This eventually caused tube material yielding at that location and affected the overall connection response. In general, the magnitude of this strain concentration was directly determined by the weld length. As a consequence, the elastic response of specimen E1 had an early ending (relative to specimen E2) as it had the shorter weld length (see Figure 3.17). SLOTTED END CONNECTIONS TO HOLLOW SECTIONS, CH 3: EXPERIMENTAL PROGRAM

3-16 Figure 3.17 Load-deformation response for connections type E1 and E2 At a load of 600 kn (near the end of the elastic response), the strain gauge readings around the tube circumference showed an uneven strain distribution (as expected for this connection type). In both tests, the maximum strain along the longitudinal weld took place at the weld beginning and a much lower value was recorded at the slot open end (see Figure 3.18). At this load level, considerable differences were observed between the readings from specimens E1 and E2 in the weld region. E2 had higher local strains than E1, despite having a longer weld length, which initially represented an inconsistency with the results from other connections (where the strain concentration decayed as the weld length increased). A further examination of specimen E1 revealed that during the fabrication of specimen E1 the tube was over-slotted, with a slot length of 268 mm. This dimension far exceeded the required weld length which was only 145 mm. Moreover, the weld fabrication started near the slot end leaving a considerable portion of the slotted tube free behind the welds (see Figure 3.19). Hence, the progressive deformation of connection E1 was accompanied by a bowing outwards of the free slotted tube portion as the load increased. This may have positively affected the strain distribution in the connection since it modified the strain concentration at the slot end. The bowing in the slotted tube E1 did not eliminate the shear lag phenomenon, but was sufficient to change the connection strain distribution. SLOTTED END CONNECTIONS TO HOLLOW SECTIONS, CH 3: EXPERIMENTAL PROGRAM

3-17 Figure 3.18 Strain distribution for test specimens E1 and E2 at 600kN Once the tube material started to yield, the connection deformation began to concentrate near the open slot region (adjacent to the beginning of the welds). This local straining was combined with gradual propagation of material yielding in surrounding areas, illustrated by flaking of the whitewash along the connection. In addition, yield lines emanated from the slot into the tube. In both test specimens, these yield lines were neatly depicted on the tube surface (this contrasted with the CHS connections where material yielding was mainly exemplified by a region rather than lines). This different behaviour has been attributed to the EHS tube material properties, which exhibited a clear yield plateau unlike the CHS material. Finally, close to the attainment of the maximum load, the tube started to neck at the open slot region, slowing the load increase. Then, the connection distortion stopped as the tube material fractured (see Figure 3.19). SLOTTED END CONNECTIONS TO HOLLOW SECTIONS, CH 3: EXPERIMENTAL PROGRAM

3-18 Figure 3.19 Failure in test specimens E1 and E2 The crack continued propagating around the tube circumference (CF) in both specimens until complete tube rupture. Although the maximum load in specimen E2 nearly reached the tube gross cross-sectional area yield load (A g F y =1286 kn), the capacity was still limited by the uneven strain distribution induced by shear lag. Finally, the maximum load and deformation attained by these test specimens are shown in Table 3.6. Table 3.6 Ultimate capacity for connections type E1 and E2 Weld Length (mm) Test Load N ux (kn) Deformation @ Max Load (mm) Failure Mode N ux / A n F u Specimen E1 145 1109 9.9 CF 0.81 Specimen E2 182 1236 11.1 CF 0.90 3.3.4 Slotted EHS connection - slot end not filled (gusset plate oriented to give small eccentricity) The change in the gusset plate orientation significantly improved the behaviour of this test specimen relative to its counterpart with a large eccentricity (see Figure 3.20). SLOTTED END CONNECTIONS TO HOLLOW SECTIONS, CH 3: EXPERIMENTAL PROGRAM

3-19 Figure 3.20 Load-deformation response for connection type E5 At a load of 1040 kn (near the end of the elastic response), the strain gauge readings around the tube circumference showed a very uneven strain distribution, illustrated by Figure 3.21 (as was observed previously in specimens E1 and E2). Along the parallel welds, the strain distribution again reached its maximum value at the beginning of the weld as before. Figure 3.21 Strain distribution in test specimen E5 at 1040 kn SLOTTED END CONNECTIONS TO HOLLOW SECTIONS, CH 3: EXPERIMENTAL PROGRAM

3-20 As part of the transition from an elastic response to a distinct yield plateau, the connection deformation began to concentrate at the open slot region and shear yield lines (visible due to the whitewash flaking) also emanated from this region towards the tube mid-length. The low connection eccentricity significantly improved the load transfer from the EHS to the gusset plate, relative to its counterpart with a large eccentricity. This also decreased the strain concentration occurring at the beginning of the weld, thus allowing the attainment of the yield stress across the tube net section. At this load level, shear yield lines continued to propagate but now over the entire tube length, increasing the overall deformation from 12 to almost 27mm. In contrast with test specimens E1 and E2 (where the overall deformation was mainly concentrated at the slot region), the total deformation here was a combination of the deformation at the slot region plus the overall tube elongation due to material yielding. In order to continue increasing the load, the material at the net section started to strain harden. The uneven strain distribution taking place at the open slot, aggravated by the shear lag phenomenon, eventually caused tube fracture there (see Figure 3.22). Once material fracture began, the load decreased as a consequence of the crack propagation around the tube circumference (CF), until complete tube rupture. Even though the tube material reached strain hardening, the maximum connection efficiency (N ux /A n F u ) was restrained to only 94%. Nevertheless, this connection did allow the attainment of complete tube yielding (A g F y =1286 KN) which may represent an advantage of this structural shape over the CHS. The maximum load and deformation attained by this test specimen is shown in Table 3.7. Table 3.7 Ultimate capacity for connection type E5 Weld Length (mm) Test Load N ux (kn) Deformation @ Max Load (mm) Failure Mode N ux / A n F u Specimen E5 185 1282 31.8 CF 0.94 SLOTTED END CONNECTIONS TO HOLLOW SECTIONS, CH 3: EXPERIMENTAL PROGRAM

3-21 Figure 3.22 Failure in test specimen E5 3.3.5 Slotted gusset plate to tube connections in tension This connection type avoids any loss of the tube cross-sectional area and its potential effect on connection strength. Even though this may be considered its principal advantage over slotted tube connections, the slot in the gusset plate can negatively affect the connection stiffness, leading to excessive deformation of the gusset plate and consequently to the tube cross-section (as was observed during the tests). 3.3.5.1 Slotted gusset plate to CHS connection (type C) A strain concentration took place at the beginning of the welds (in the CHS) and interior corners of the gusset plate. Close to 600 kn, the gusset plate yielded and caused flaking of the whitewash there and a change in the overall connection stiffness (see Figure 3.23). SLOTTED END CONNECTIONS TO HOLLOW SECTIONS, CH 3: EXPERIMENTAL PROGRAM

3-22 Figure 3.23 Load-deformation response for connections type C For both test specimens, this happened at a lower load level than for the slotted tube connections. At this load stage, the strain gauge readings around the tube circumference showed an uneven strain distribution and the maximum strain value took place at the beginning of the welds (see Figure 3.24). In addition, the minimum value (near zero) was detected for the strain gauge located at 90º (see Figure 3.24), as for slotted tube connections. Moreover, close to attainment of the maximum load the readings at 90º switched to negative values (indicating compressive strains). This initially-unexpected behaviour was attributed to the excessive distortion of the tube cross-section, due to the gusset plate bowing and the necking of the tube. The readings along the parallel welds also showed typical variations, with the highest strain concentration occurring at the beginning of the weld (see Figure 3.24). Of the two tests, the higher strains were registered in specimen C1 which has the shorter weld length. SLOTTED END CONNECTIONS TO HOLLOW SECTIONS, CH 3: EXPERIMENTAL PROGRAM

3-23 Figure 3.24 Strain distribution in test specimens C1 and C2 at 595kN Beyond the elastic response, each load increment resulted in increasing distortion of the tube cross-section. Moreover, the bowing outwards of the gusset plate introduced out-of-plane strains at the tube surface which are believed to have induced a triaxial state of stress at the beginning of the weld. This behaviour continued throughout the tests until the material fractured (see Figure 3.25). Once the fracture started (at the beginning of the welds), the crack continued propagating around the tube circumference (CF) in both tests. These tests again corroborated how the presence of shear lag can affect the strain distribution in such connections. Nevertheless, the magnitude of this strain concentration (which triggers the material fracture) is a consequence of factors such as: magnitude of the shear lag, bowing of the gusset plate, tube cross-section distortions and tube necking. Based on these two tests, it seems necessary to consider the influence that the gusset plate dimension may have on the connection strength in a further parametric analysis. Moreover, the potential need to limit the maximum load based on the tube cross-section distortion has arisen herein since large distortions were observed before the attainment of the connection maximum load. Finally, the maximum load and deformation attained by these test specimens are shown in Table 3.8. SLOTTED END CONNECTIONS TO HOLLOW SECTIONS, CH 3: EXPERIMENTAL PROGRAM

Table 3.8 Ultimate capacity for connections type C Weld Length (mm) Test Load N ux (kn) Deformation @ Max Load (mm) Failure Mode N ux /A n F u (A n =A g ) Specimen C1 162 1107 13.9 CF 0.83 Specimen C2 195 1196 16.8 CF 0.90 3-24 Figure 3.25 Failure in test specimens C1 and C2 3.3.5.2 Slotted gusset plate to EHS connection (gusset plate oriented to give a large eccentricity) These two connections exhibited a strain concentration taking place in the EHS near the beginning of the welds and also in the gusset plate adjacent to the end of the welds (at the tube end). Near 850 kn, the materials in specimen E3 commenced yielding in these regions (as was confirmed by whitewash flaking) producing a change in the overall connection stiffness (see Figure 3.26). SLOTTED END CONNECTIONS TO HOLLOW SECTIONS, CH 3: EXPERIMENTAL PROGRAM

3-25 Figure 3.26 Load-deformation response for connections type E3 and E4 Once yielding started, it continued propagating from there towards the tube mid-length while the strain concentration continued increasing near the weld region. In addition, shear yield lines (visible from whitewash flaking) emanated from the connection region to the tube midlength. Even though the response of specimen E4 remained elastic at this load level (850 kn), it followed a similar behaviour afterwards. In order to transition from an elastic response to a distinct yield plateau, both connections required a large elongation (approximately of 20 mm). These slotted gusset plates exhibited smaller deformations throughout this transition, and even during the incursion of the EHS material into the plastic range, in comparison to their CHS test counterparts (connection type C). This better behaviour has been related to the higher moment of inertia of the gusset plates for these connections (E3 and E4). Gusset plates with a 32mm thickness were used herein, whereas 25.7mm gusset plate were used for connections type C (see Table 3.3). At a load of 800 kn, the strain gauge readings around the tube circumference showed the typical uneven strain distribution observed previously. An increase in the weld length (for SLOTTED END CONNECTIONS TO HOLLOW SECTIONS, CH 3: EXPERIMENTAL PROGRAM

3-26 specimen E4) only produced a marginal reduction in the strain magnitude (see Figure 3.27). Moreover, close to attainment of the maximum load the strain readings at the gusset plate corners reported a considerable increase which has been attributed to the tube ovalization in the connection region. Figure 3.27 Strain distribution for test specimens E3 and E4 at 800kN Beyond the yield load, the connections entered the strain hardening range, but this load increase was associated with a gradual increase in the bowing outwards of the gusset plate. This distorted the tube cross-section and induced a triaxial state of stress near the beginning of the weld. This gradual degradation of the connection continued until the material fractured (see Figure 3.28). SLOTTED END CONNECTIONS TO HOLLOW SECTIONS, CH 3: EXPERIMENTAL PROGRAM

3-27 Figure 3.28 Failure in test specimens E3 and E4 Once the fracture started, the crack continued propagating around the tube circumference (CF) in both tests. In a similar manner to the CHS connections, these tests illustrated the need to consider the influence that the gusset plate dimensions may have on the connection strength and the need to potentially limit the maximum load based on the tube cross-section. Finally, the maximum load and deformation attained by these test specimens are shown in Table 3.9. Table 3.9 Ultimate capacity for connections type E3 and E4 Weld Length (mm) Test Load N ux (kn) Deformation @ Max Load (mm) Failure Mode N ux /A n F u (A n =A g ) Specimen E3 146 1336 43.9 CF 0.82 Specimen E4 1175 1400 53.4 CF 0.86 3.3.6 Connections under compression load In order to comprehend the influence that the shear lag phenomenon may have on connections under compression load, two connection types were studied during this experimental program; the first connection was a slotted CHS to a gusset plate, and the second was a slotted gusset plate to a CHS. In both cases, the governing failure mechanism corresponded to a local bucking (LB) of the CHS at the connection region. Nevertheless, the SLOTTED END CONNECTIONS TO HOLLOW SECTIONS, CH 3: EXPERIMENTAL PROGRAM

3-28 occurrence of this failure mechanism in each connection was influenced by factors characteristic of each connection type. 3.3.6.1 Slotted CHS to gusset plate connection - slot end not filled The behaviour of this connection type may be explained by several stages. Initially, the connection exhibited an elastic response that allowed the attainment of almost 70% of its maximum load (see Figure 3.29). In a similar manner to the tension tests, a strain concentration developed at the slot region but, due to the difference in the load condition herein (compression loading), this strain concentration induced tube local buckling at the slot region rather than material straining to fracture. Figure 3.29 Load-deformation response for connection A3C The tube local buckling started near the beginning of the welds and then gradually extended to the entire slotted tube cross-section. As result of this, the tube deformation was concentrated at that part of the connection. With increasing loads, the slot local buckle grew and the gusset plate moved towards the CHS wall, thus reducing the slot length. From LVDT readings, it could be ascertained that the tube out-of-straightness increased slowly before the attainment of the maximum load. At the maximum load, the LVDT indicated an out-ofstraighness of 1.3 mm. After this load, full contact between the gusset plate and the tube wall SLOTTED END CONNECTIONS TO HOLLOW SECTIONS, CH 3: EXPERIMENTAL PROGRAM

3-29 occurred and this triggered an increase on the tube out-of-straightness and the rotation of the gusset plate free length (the distance left between the CHS and the machine clamps). This behaviour was associated with the start of overall bucking of the CHS. At this point, the test was stopped in order to avoid large moments being applied to the testing machine. In a similar manner to the tension tests, an uneven strain distribution existed throughout the test. Near the end of the elastic response (at 800 kn), the compressive strain around the tube circumference reached its maximum at the beginning of the welds in a comparable manner to the tensile test specimens (see Figure 3.30). The strain distribution along the connection also reproduced the behaviour seen during the tension tests (see Figure 3.30). z (see Figure 3.10), Figure 3.30 Strain distribution in test specimen A3C at 800kN. Even though the maximum load attained corresponded to 93% of A g F y, this required considerable deformation at the slot (see Figure 3.31). Besides the weld length, the slot length is likely to have a considerable influence on the connection capacity.these two parameters will hence be considered in further parametric analysis. For a long member (such as a brace in a regular building) an efficiency as high as 0.93 A g F y would never be required, as the brace capacity will be governed by the member slenderness ratio. The maximum load and deformation attained for the test specimen is shown in Table 3.10. SLOTTED END CONNECTIONS TO HOLLOW SECTIONS, CH 3: EXPERIMENTAL PROGRAM

Table 3.10 Ultimate capacity for connection type A3C Weld Length (mm) Test Load N ux (kn) Deformation @ Max Load (mm) Failure Mode N ux / A g F y Specimen A3C 206-1145 4.8 LB 0.93 3-30 Figure 3.31 Failure in test specimen A3C 3.3.6.2 Slotted gusset plate to CHS connection In a similar manner to the tension test, this connection type had an initial elastic range of almost 70% of its maximum load. Again, an uneven strain distribution formed at the connection region with a high strain concentration near the beginning of the welds, resulting in local buckling of the tube at the location. Even though an un-slotted tube would generally require a high load to reach local buckling, the bowing inwards of the gusset plate exacerbated the tube local instability. Near the end of the elastic response (600 kn), whitewash flaking was visible at interior corners of the gusset plate. Then, the bowing inwards of the gusset plate began to affect the connection deformation which, became a combination of the deformation of the tube and the gusset plate. LVDT readings showed that the tube out-of-straightness slowly increased until the attainment of the maximum load, at which time the out-of-straightness reached a value of 2.0 mm. After the maximum connection deformation progressed in a stable, ductile manner (see Figure 3.32). SLOTTED END CONNECTIONS TO HOLLOW SECTIONS, CH 3: EXPERIMENTAL PROGRAM

3-31 Figure 3.32 Load-deformation response for connection C3C In the same way as the tension connection type C, similar uneven strain distributions were present throughout the test. The connection strain distributions (at 600 kn) recorded are shown in Figure 3.33. Nevertheless, the lack of a slot reduced the strain concentration in front of the weld region (at z = 50mm in Figure 3.33). Figure 3.33 Strain distribution in test specimen C3C at 600kN. SLOTTED END CONNECTIONS TO HOLLOW SECTIONS, CH 3: EXPERIMENTAL PROGRAM

3-32 The bowing inwards of the gusset plate negatively affected the connection behaviour, causing premature the tube local bucking at the weld region and a subsequent increase in the tube distortion (see Figure 3.34). Because of this, the maximum load attained here corresponded to only 70% of A g F y. This illustrates the need to limit the ultimate connection capacity by a distortion limit, as has been suggested previously for the comparable tension test. The influence of weld length and gusset plate dimensions will thus be considered in further parametric analysis. The maximum load and deformation attained for the test specimen is shown in Table 3.11. Table 3.11 Ultimate capacity for connection type C3C Weld Length (mm) Test Load N ux (kn) Deformation @ Max Load (mm) Failure Mode N ux /A g F y A n= A g Specimen C3C 200-869 6.0 LB 0.70 Figure 3.34 Failure in test specimen C3C 3.4 Summary of this experimental program Even though all the tensile test specimens were designed to avoid failure by modes other than circumferential fracture (CF), this was only accomplished by the test specimens A2, B2, C1, C2, E1 and E5. Specimen B1 failed by block shear tear out (TO) of the base material along the weld and specimen A1 presented a combination of both failure modes (CF-TO). In addition, failure of the specimens tested under compression loading was due to local bucking at the connection end (LB). By means of this program, it was found that several factors influenced SLOTTED END CONNECTIONS TO HOLLOW SECTIONS, CH 3: EXPERIMENTAL PROGRAM