Experimental testing and analysis of the axial behaviour of intermeshed steel connections

A new, universally applicable, structural steel connection mechanism has not been introduced in more than a century. Bolting and welding remain the two primary options for connecting structural steel elements on site, despite both being expensive and time-consuming operations. In addition, with these connection options, easy disassembly cannot always be facilitated for material reuse (BCSA et al., 2018a; Sansom and Avery, 2014), particularly in the case of welding. It is notable that this remains the case even though steel is the most recycled construction material (SRI, 2017).

However, recent advances in manufacturing technologies have created a platform that enables the development of a new universal connection in steel. To date, advanced cutting technologies have only been used to accelerate traditional processes by cutting sheet metal or providing faster conventional fabrication (e.g. cutting instead of drilling bolt holes). This paper presents an overview of a collaborative tripartite research project, entitled Advanced manufacturing and assembly of steel structures (AMASS), which is the first of its kind to exploit fully automated, precise, advanced manufacturing cutting technologies (e.g. laser, waterjet and high-definition (HD) plasma cutting) in the creation of a new class of efficient intermeshed steel connections (ISCs) (Al-Sabah and Laefer, 2017a, 2017b; Al-Sabah et al., 2020) that eliminate the need for on-site welding and most on-site bolting.

The potential impact of an alternative steel connection of this type that targets reductions in time and cost and simplifies disassembly for reuse is further highlighted by the dominance of steel in the construction industry. Considering the UK alone, structural steelwork currently holds a dominant share of 67% in the multi-storey buildings market and similar shares (∼60%) are reflected in corresponding EU markets (Muller, 2008). In the USA, the structural steel market size for both residential and non-residential construction has grown steadily since 2014 (GVR, 2019). For a typical multi-storey commercial building, fabrication costs can rival the material costs, reaching 30–40% of the overall cost (BCSA et al., 2018b). Furthermore, a major part of the steel frame cost (10–15%) is related to on-site assembly and erection, with longer schedules and higher associated costs expected for more complex connections. The AMASS project aims to deliver an efficient new connection that improves and refines these features of steel buildings and structures, including bridges, leading to savings in time and hence cost across the UK and EU construction industries, in addition to fast-growing markets in the USA, China and India.

Recent decades have seen the proposal of alternatives to existing steel connections, a number of which were motivated by the brittle failure mechanism of steel moment-resisting connections during the 1994 Northridge earthquake (AISC, 2016a; FEMA, 2000; Hamburger et al., 2009; Han and Moon, 2007; Naimi et al., 2013; SAC JV, 1995; Stojadinovic et al., 2000). While proprietary connection solutions have resulted, these have limited use in wider practice and thus have had limited success (Cordova and Hamburger, 2011). Among the most successful alternative connections developed in the USA are the Kaiser bolted bracket moment connection (Adan and Gibb, 2008, 2009), the ConXL moment connection (ConXtech, 2020; Hamburger, 2006), the ATLSS connection (Fleischman et al., 1990; Perreira et al., 1993), the SidePlate connection (Rafezy et al., 2015, 2018), the reduced beam section moment connection (Engelhardt et al., 1998; Jin and El-Tawil, 2005) and the pin–fuse connection (Cordova and Hamburger, 2011). Alternative connections have also been developed in the UK, such as the Lindsay adapter (Lindapter, 2018) and the Quicon connection (Burgan, 2002; Heywood, 2004), designed by the UK Steel Construction Institute. However, all the above continue to rely predominantly on conventional bolted or welded methods or require more expensive casting techniques.

The ISC method developed by the AMASS project team eliminates the need for these on-site operations by means of a ‘snap together’ connection mechanism. This mechanism is summarised in Section 2 for two types of ISC – the front ISC and the side ISC. Section 3 details the fabrication, design and scaled experimental testing of the flange connection component of the side ISC, with a focus on its axial behaviour. Section 4 presents the results of finite-element analysis (FEA) of the tested connections while Section 5 provides a brief design review and recommendations for future design.

The ISC method employs advanced manufacturing technologies to create precise geometries that snap together, with load transfer achieved through common bearing surfaces at multiple contact points. The fabrication of such geometries from both two- and three-dimensional steel sections is now possible due to the introduction of computer numerical control technology and robotic arms in conjunction with advanced cutting technologies such as laser, waterjet and HD plasma cutting and wire electrical discharge machining (WEDM) (Krar and Gill, 2003; Ramakrishnan and Rogozinski, 1997). A comparison of these cutting technologies is provided in Table 1. For the experimental work described in this paper, the typical cutting tolerance achieved for connection samples was ± 0.2 mm using a flatbed laser cutter.

Figure 1 shows the first type of ISC developed – the front ISC (Al-Sabah and Laefer, 2017a). Bending moment is transferred across the connection by way of tension and compression in the two flanges; this is achieved by bearing and friction by the intermeshed ‘teeth’ of the flanges, the dovetail configuration of which can also be seen in Figure 1. Shear load transfer is achieved by direct contact bearing at multiple points along the stepped web connection, while a small locking piece can be inserted to provide resistance to uplift or unexpected lateral loading. The arrangement is ideally suited for connecting beams at or near-ideal inflection points to create gravity load framing, as illustrated in Figure 2, where beam stubs can be welded to columns off-site. The connection geometry also lends itself to a straightforward assembly and disassembly on site, which is desirable – a beam can be dropped into place prior to inserting the locking element. However, it is worth noting that this also means the connection is primarily suited to scenarios where cutting and site erection tolerances are expected to be small and highly controlled. Furthermore, as a new connection type with faying surfaces, careful attention needs to be paid to the type of coating system used if a spray or paint is to be applied for fire or corrosion protection (NSAI, 2018). As the front ISC relies on bearing and friction for load transfer, it is preferable that any applied coating provides a specified slip factor (friction coefficient) of at least a value of 0.3 for the surfaces in contact, verified by testing in accordance with the relevant standard (BSI, 2011). In the case of possible damage to the coating, the connection can be inspected after installation and any damage to the paint repaired as necessary.

The related load-transfer behaviour and performance of the front ISC connection type have already been studied numerically under simulated mixed-mode loading scenarios by the authors (Shemshadian et al., 2019) for a 254 × 102 UB 28 beam section, with the connection exhibiting excellent shear resistance even in the presence of flexural and axial loading. The non-linear FEA carried out by the authors enabled the interaction diagrams of axial, shear and moment capacities of this ISC to be obtained; interaction diagrams currently available in design codes were found to be inadequate in capturing the behaviour of the connection.

However, as mentioned earlier, this connection type is sensitive to assembly tolerance and it was found that its axial and flexural behaviour were significantly affected by the alignment of the flanges, which can be altered in response to loading, even for ‘perfect’ initial configurations, and this alignment can also potentially differ between fabrication and erection due to thermal effects, particularly for long members. Therefore, while it was established that the connection can be classified as a simple, partially restrained (or semi-rigid) connection according to ANSI/AISC 360-16 (AISC, 2016b) and BS EN 1993-1-8:2005 (BSI, 2005) and thus used for gravity load bearing structures, further study of the alignment of the flanges is required in order to better understand how to avoid axial and/or flexural failures. A final recommendation on modelling of this connection is provided, suggesting that the intermeshed connection can be represented as a pin connection, but limited additional flexural stiffness may be required for a fully accurate numerical representation. The authors have proposed alternative flange geometries in order to improve the axial and flexural capacities of this connection type and work in this area is ongoing (Matis et al., 2018a, 2018b).

Figure 3 shows the second type of ISC developed – the side ISC (Al-Sabah and Laefer, 2017b). This connection is an improvement on the front ISC configuration and can accommodate greater tolerances in assembly and fabrication through the introduction of side plates at the top and bottom flanges, which transfer tension and compression forces resulting from bending moments. For this purpose, teeth are cut along the flange edges, which allow side plates with matching holes to be slotted into position. The side plates can then be fixed in place using a variety of methods, including T-head bolts (configuration shown in the figure). As per the front ISC, shear load transfer can also be achieved by direct contact bearing at multiple points along the stepped web connection, as shown in Figure 1. However, an alternative version is illustrated in Figure 3, which targets greater acceptance in the construction industry using bolted shear web plates to achieve shear load transfer. The configuration shown in Figure 3 is for a 254 × 146 UB 31 beam section; the dimensions of the four side plates are 557 × 65 × 10 mm while those of the two web plates are 200 × 201 × 5 mm. Slots are provided in the flange teeth to allow eight T-head M8 grade 8.8 bolts to fix the side plates in place while eight M12 grade 8.8 bolts are used for the web plate (not shown). Due to the predominance of flatbed cutting machines used in industry at present, the fabrication of built-up sections to form the side ISC is also possible, offering potential applications for longer span beams and bridge structures.

The behaviour and performance of this connection type are currently under ongoing investigations by the AMASS team. The experimental testing and FEA of scaled flange connections under tensile loading are presented in Sections 3 and 4 of this paper respectively, investigating the local axial behaviour of the side ISC in flanges and side plates. A description of the full-scale beam testing programme is outside the scope of this paper, but the authors have recently reported the results of an experimental testing programme (Shemshadian et al., 2020) in which simply-supported beams utilising the side ISC with side angles were designed to resist gravity loading in frames, and were tested under pure bending and combined bending and shear. In all cases, the beam connections performed as intended but were found to provide a greater moment capacity than designed for; this was attributed to material overstrength and a conservative design procedure. These factors are also addressed in this paper regarding the axial performance of the scaled connection with side plates.

To investigate the constructability and erection time requirements of the side ISC, a trial erection test of the connection in a two-storey assembly has been completed recently (Al-Sabah et al., 2020) and a video report on this can be viewed online (UMG, 2019). The trial assembly included geometric deviations in connection fabrication within the tolerances accepted by the steel industry. It also included the use of shim plates to introduce the maximum allowable site tolerances in the column (10 mm shim under the base plate) and beam stub (1 mm shims) alignments for erection. It is notable that these deviations did not cause any delay in the frame erection. As a primary cost benefit for this connection type relates to the efficiency of assembly and disassembly, this was also analysed during the trial erection test. It was found that the side ISC assembly provided a fast learning curve, with completion time decreasing from 12 min 58 s initially per beam connection to an average completion time of 4 min 32 s for the last two connections. Subsequent disassembly was also straightforward and much faster, averaging around 2 min 30 s per beam connection.

The average web splice plate installation time was 2 min 27 s. This involved installing the two plates, inserting and tightening eight M12 bolts, giving an average time per individual M12 bolt of around 18 s. It is expected that this time would increase with bolt diameter. For the equivalent bolted beam splice connection, 40 M12 bolts would be required, with ten on each side of the two splice plates for the flanges. The time required to assemble this connection, based on the average 18 s per bolt in the frame test, would be 18 s × 40 bolts = 720 s = 12 min. Note that these times ignore the fitting time for the heavier flange splice plate as it should not add much to the total time. Based on these average assembly times, the ISC could potentially offer a time saving of 62.5% over conventional bolted connections – which is significant. Further work is required to establish definitively the fabrication costs and erection times relative to equivalent bolted and welded connections, including on-site erections. To this end, an equivalent bolted frame assembly test is planned.

This section presents the results of uniaxial testing of scaled flange geometries for the side ISC. Due to the potential issues encountered with the misalignment of beam flanges for the front ISC, it is important to study the axial behaviour of this connection type experimentally to verify the design procedure and identify any factors that could potentially influence its behaviour at full scale.

Tensile tests were carried out with a displacement rate of 1 mm/min using a Zwick/Roell RetroLine tC II universal testing machine (UTM) (Figure 4(a)), which was calibrated in accordance with ISO 7500-1 (BSI, 2004a). This had a tensile capacity of 100 kN with an accuracy of ± 2% and cross-head speed range of 0.001–1000 mm/min. Tinius Olsen mechanical wedge action grips were used to hold samples in place.

Load measurements were directly output by the UTM control system, while sample deformations were measured using a LaVision StrainMaster portable digital image correlation (DIC) system (LaVision, 2018) (Figure 4(b)). DIC is a non-invasive technique that can be used to measure deformations by analysing a series of images (Sutton et al., 2009) and, in these tests, allowed deformations to be observed across full samples rather than at one localised strain gauge location. For this purpose, test samples were spray painted with a matt black and white speckle pattern that the DIC system could track and analyse. The system comprised a dedicated laptop computer running LaVision StrainMaster analysis software along with the following components mounted on an adjustable lightweight bar and tripod, as shown in Figure 4(b) 

• 2 × VC-Imager E-lite cameras (5 megapixels, USB3)

• a compact controller unit

• 2 × 20 W LED linear illumination units producing focused white light.

The scaled flange geometry tested in this study is shown in Figure 5; this geometry ensured that connection behaviour up to failure could be observed; that is, the failure load would not exceed the capacity of the UTM (100 kN). In total, six samples of this arrangement were laser cut from 6 mm S275JR steel plate; thus all components – flanges and side plates – were 6 mm thick. An example of the fabricated geometry is shown in Figure 6(a). Laser cutting is suitable for material thicknesses up to 20 mm. Sample fabrication using a flatbed laser cutter (a Trumpf TruLaser 4050 laser cutting machine) was thus appropriate, with a minimum geometric tolerance of ± 0.2 mm specified by the fabricator. As can be seen in Figure 5, the connection design has an allowable tolerance of 1 mm for each slotted hole – that is 10 mm wide × 6 mm thick teeth in 11 mm × 7 mm holes.

In the full-scale connection (Figure 3), M8 bolts are used to hold the side plates in place. However, as the testing scale did not allow for these to be included, for safety purposes during testing, the flange teeth were extended by 4 mm and plastic ties were used to provide a limited horizontal restraint, keeping the side plates in place during the test setup (Figure 6(b)). Figure 6(b) shows extended flange lengths – all the test samples required an extra length of 70 mm at either end to fit the grips of the UTM.

Material coupon tests carried out on the source 6 mm S275JR steel plate according to BS EN ISO 6892-1:2016 (BSI, 2016) established an average yield strength of 283 MPa (±4.7% coefficient of variation) and an average ultimate strength of 476 MPa (±1.8% coefficient of variation) for the material, compared with a minimum yield strength of f_y = 275 MPa and a minimum ultimate tensile strength of f_u = 410 MPa specified by BS EN 10025-2:2004 (BSI, 2004b). Taking the minimum values for strength as the characteristic values for design, the test and geometry were designed such that the initial yield and failure would occur in the side plates rather than the flange, with yielding expected in the side plates at a load of 42.9 kN.

Table 2 shows the relationships between key design parameters for the scaled flange connection. It was assumed that flexural theory applies to the flange teeth. The simplified plastic design resistance procedure for the connection was carried out in accordance with BS EN 1993-1-1:2005 (BSI, 2014; Lam et al., 2014) and takes account of axial forces in the flange and the side plates, in addition to combined bending and shear in the flange teeth. Therefore, the number of teeth required on each side of the flange, $nt$⁠, depends on the applied axial load on each side of the flange, $Fti$⁠, and the calculated plastic shear resistance of each flange tooth, $Vpl,Rd$⁠, limited to 75%. In this study, a flange tooth width of b_t = 10 mm was initially assumed, which allowed calculation of $Vpl,Rd$⁠. Assuming an equal distribution of load between the teeth, $nt$ can be calculated using Equation 1. A simple expression allowing the length of the side plate, $lsp$⁠, to be calculated is shown in Equation 2. A tolerance g₀ = 1 mm was (Table 2); this defines the lateral gap between the side plate and the flange between the teeth. As a result, $g0$ is also dependent on the flange tooth corner radius, r_t. Here, r_t = g₀ = 1 mm.

Bending deformation of the side plate about its centroidal axis was not considered in the design approach presented here and thus is not considered in Equation 2. Although its effect is presumed to be very small based on flexural testing (Shemshadian et al., 2020), further analysis is needed to quantify the effect of bending deformation on the side plate and hence $lsp$⁠; this analysis is outside the scope of this paper. The design parameters in Table 2 are reviewed in Section 5 based on the results of experimental testing of the axial behaviour of the connection.

Following Table 2, the design procedure allowed the resistance of the connection to be calculated and compared for the three component parts at yield – (a) the flange section resistance in tension (44.6 kN), (b) the flange teeth resistance in shear (57.2 kN) and bending (46.3 kN) and (c) the tension resistance of the side plates (42.9 kN). This confirmed that resistance is governed by the side plate capacity and thus yielding should occur in the side plates first.

Figure 7 shows the load–displacement behaviour of the six side ISC flanges tested (samples A–F), obtained from the DIC measurement system. Consistent ductile behaviour across all tests was observed. Differences at the start of the figure at low load and displacements are caused by the initial ‘bedding in’ or settlement of the connection as teeth come into contact with the side plates. This region only occurred in approximately the first 1.2 mm of displacement, indicating that it was a direct result of the design and fabrication tolerances (1 ± 0.2 mm) and did not significantly affect load-carrying capacity. Figure 8 shows the corresponding load–displacement behaviour with the initial settlements removed, highlighting the consistency between tests. Each sample exhibited some strain-softening behaviour immediately prior to tensile failure, which occurred in the flange rather than the side plates. The failure mechanism is shown in Figure 9 – the failure surface occurred at an incline from corner to corner rather than as a failure surface perpendicular to the direction of loading.

This unexpected failure mechanism was initially investigated by reviewing the load performance of the connection. The average yield load from testing was 46.7 kN while the average ultimate load was 74.8 kN, 1.6 times the yield load. The average yield load exceeded the design value of 42.9 kN. Referring to Section 3.1.2 and Figure 5, this can be partially explained by considering the flange cross-section with an effective area of 162 mm²; the 46.7 kN yield load corresponds to a stress of 288 MPa, which is closer to the yield strength obtained from material testing (283 MPa) than the value used for design (275 MPa). The ultimate load corresponds to a stress of 462 MPa, clearly higher than the design value but lower than the ultimate strength obtained from material testing. For the side plates, the yield and ultimate loads corresponded to stresses of 299 MPa and 479 MPa, respectively. Although this yield stress exceeded the value obtained from material testing, very limited yielding in tension was observed in the side plates, suggesting that this stress concentration did not materialise, potentially due to the load distribution. The flange teeth were observed to deform permanently due to bending in-plane, as shown by the curved edges of the teeth in Figure 9, with the largest deformations in the teeth furthest from the centre of the connection. This also indicates that the teeth at this location potentially carry more load share than the others. Furthermore, due to the load-transfer mechanism, this deformation also caused some slight lateral bending in-plane in the side plates, perpendicular to the direction of loading. This had the effect of increasing the lever arm of moment of the force applied to the end teeth as the ends of the side plates tended to move slightly outwards, away from the flange. Overall, as the same material was used for both the flanges and side plates, the failure shown in Figure 9 is thus most likely caused by the connection geometry as fabricated, or the loading arrangement, or a combination of both.

The failure mechanism was further investigated by inspecting the DIC strain contours obtained during testing (Figure 10). Prior to failure, strains started to increase in the flanges at the point where the cross-section narrows, as expected, although there was an asymmetrical concentration in the top half focused on the right-hand side. Inspecting Figures 10(a)–10(c) in order, the sample can also be seen to shift to the left slightly during testing, which may have caused an imbalance in loading between the two side plates and induced some rotation in the flange, leading to a stress concentration on one side and hence the inclined failure surface between the corners shown in Figure 9. This is investigated further in Section 4.2.2. To prevent the occurrence of this type of failure, the introduction of a larger radius at re-entrant corners (i.e. increasing the width at the base of the flange teeth) may be needed, which could reduce stress concentrations (Schultz et al., 2019; Shemshadian et al., 2020). Overall, the consistency of the behaviour in the scaled testing highlights that it is important to be aware of this in future analysis and design of the full-scale connection, particularly if fabrication and assembly tolerances are relatively large or if a design restricting failure to the side plates is preferred. If the latter is preferred, with the side plates acting as sacrificial elements that can be replaced, a capacity design approach could be used such that the ultimate capacity of the side plates is used to design and detail the flange teeth, although this may result in an overly conservative design with a requirement for significantly more teeth that will increase the length of the side plate. Alternatively, a reduction in the effective area of the side plate would be required. To this end, a revision of the design approach presented in Section 3.1.2 is proposed in Section 5.

To obtain a more comprehensive understanding of the axial behaviour of the side ISC flange connection observed during testing, non-linear FEA of the experimentally tested connection geometry was undertaken using Abaqus/CAE (DSSC, 2019). As highlighted earlier in the paper, the fabricated geometry and loading arrangement were two influences on the behaviour observed in testing that needed to be considered. Therefore, the effects of fabrication tolerance and lateral movement during testing on the finite-element model (FEM) were also investigated.

For the purposes of the non-linear FEA in Abaqus/Standard, a three-dimensional model of the tested connection geometry was assembled, as shown in Figure 11(a). This model was based on the geometry shown in Figure 5 and is denoted SP1. Geometric and material non-linearity effects were incorporated. All parts were meshed using C3D8R brick elements, as shown in Figures 11(b) and 11(c). A mesh convergence study was carried out that identified 1.5 mm as a suitable element size for mesh refinement, resulting in a total of 25 592 elements and 35 290 nodes for the whole model. The base of the bottom flange was fixed, while the top of the upper flange was restrained in all but the vertical direction to enable a displacement-controlled analysis up to a translation of 14 mm, replicating the experimental test scenario.

Two non-linear material models were considered in the analysis. The first was a non-linear, experimentally calibrated, standardised constitutive model for hot-rolled carbon steels (Yun and Gardner, 2017), which can be defined based on only the yield stress $fy$⁠, ultimate stress $fu$ and Young's modulus E of the material. Herein, this model is referred to as the YG model; it was used to represent predicted connection behaviour based on the characteristic design values for S275 steel (f_y = 275 MPa, f_u = 410 MPa and E = 210 000 MPa). The second material model was based on the experimentally determined stress–strain curve and this was used to investigate the impact of material overstrength on the performance of the connection. The true stress–true strain curves for both material models are shown in Figure 12, with the difference between them clearly illustrating the material overstrength. Poisson's ratio of $ν=0.3$ was adopted for both material models and damage (i.e. fracture) was not incorporated.

To model the contact behaviour and transfer of forces between the flange teeth and the side plates, surface-to-surface contact with finite relative sliding was implemented using the general contact algorithm in Abaqus. It was assumed that all teeth were initially in contact. For the normal contact behaviour, the ‘penalty’ constraint enforcement method was employed using the default ‘hard’ contact pressure–overclosure relationship. Tangential behaviour was modelled using the default penalty friction formulation and the isotropic Coulomb friction model. A friction coefficient of 0.3 was selected based on early parametric studies by the authors (Matis et al., 2018a) and also in agreement with the value of 0.3 recommended for bolted connections in international standards (BSI, 2005, 2011).

The load–displacement responses obtained for each material model from the FEM for SP1 are shown in Figure 13, along with the experimental results for comparison. The general load–displacement behaviour trend was captured well by the FEA. Samples A–F exhibited a slightly lower stiffness in the elastic range during testing, which – combined with the effects of initial bedding in not being accounted for in the analysis – resulted in yield at a slightly higher displacement and load than in the FEA. However, inspecting the YG model curve, it is clear that the connection outperformed the predicted behaviour based on characteristic design values, verifying a conservative design procedure. The material test model underestimated the average ultimate capacity and corresponding displacement by 2.8% and 9.3%, respectively, but captured the ductile behaviour of the connection very well up to that point, illustrating that the axial behaviour of the connection can be predicted accurately by FEA with material testing. The underestimation highlights that additional capacity is potentially provided by the tested samples that is not explained by material overstrength alone; this is investigated in the next section.

To investigate the unexpected failure location in the flange in testing, the von Mises stress contours at yield strength were inspected. These are shown in Figure 14, with arrows indicating the location of the maximum stresses. Only the contours for the material test model are shown here as the YG model provided similar observations. It can be seen in Figure 14(b) that, aside from the local stress concentrations at the base of the teeth, the teeth furthest from the middle of the connection experienced higher stresses than those the teeth that were closer. It was also found that the yield strength (283 MPa) was first exceeded locally in the flange at the base of the teeth due to the stress concentration, despite the curved corner radius of r_t = 1 mm. This is a clear indication of the stress concentration that caused failure in the flange during testing, rather than the side plate, and this relates to tooth bending; this bending can be seen more clearly in Figure 15(a) and agrees well with the experimental observations .

However, the full side plate cross-section yielded first, as per the design, while the ultimate strength (476 MPa) was also exceeded locally in the side plate section first, followed shortly by necking as the full side plate cross-section exceeded the ultimate strength, illustrated by the grey areas in Figure 15(b), which led to complete section failure (i.e. fracture at this location). At this stage, the flange section had fully yielded but only exceeded the ultimate strength locally at the base of the teeth (Figure 15(a)). Irrespective of the failure location, a design revision, including an increased tooth corner radius, may be necessary to reduce the stress concentration in the flange that influenced the failure observed in testing. Note that the influence of the stress concentration is slightly exaggerated at this scale due to the 4 mm extra tooth length on each side of the flange, which reduced the effective flange cross-sectional area by 23%; this extra length would not normally be required at full scale.

The laser cutting fabrication tolerance was small for the test samples (±0.2 mm) but, at this scale, could influence the connection capacity. Noting the underestimation of capacities in Figure 13 and the failure location in the flange in testing, the original model was modified to investigate this effect. This modification involved increasing the side plate capacity at the holes, initially according to the specified tolerance of 0.2 mm per edge. However, on inspection of untested samples, the average hole width was found to be 6.3 mm; thus, in the modified model (called SP3), the side plate hole dimensions were reduced to 6.3 mm × 11 mm, from 7 mm × 11 mm. It is important to note that based on the material test model with f_y = 283 MPa, this modification increased the axial load capacity of the side plates at yield by 5.4% to 46.5 kN, which was greater than the yield capacity of the flange at 45.9 kN. This is a scenario unique to this test due to the extra 4 mm tooth length either side of the flange.

As shown in Figure 16, the load–displacement curve obtained using the modified model (SP3) provided an excellent match to the experimental results. The average ultimate capacity and corresponding displacement were overestimated by 2.8% and 0.5%, respectively, reflecting some additional capacity that was not present in testing. Based on this figure, it is clear that the fabrication tolerance influenced the behaviour, but the true behaviour lies somewhere between the tolerance used for the SP1 and SP3 models.

The von Mises stress contours at yield are presented in Figure 17. Similar behaviour to Figure 14 for SP1 was observed, with local yield occurring in the flange initially due to the stress concentrations at the base of the teeth, rather than its lower relative yield capacity alone. However, it was found that the ultimate strength was exceeded locally in the flange first for the SP3 model, reflecting the increased side plate capacity. Subsequently, in Figure 18, similar to Figure 15, the ultimate connection capacity was reached when the full side plate cross-section exceeded the ultimate strength and necking begins. Figure 18 also shows that a much larger area of the flange exceeded the ultimate strength compared with Figure 15, because the side plate could transfer more load; this indicates that fabrication tolerance should be considered as an influence on increased stresses in the flange for the scaled connection design that was tested. At full scale, this influence will become relatively insignificant.

As some lateral movement was observed during testing, a lateral in-plane displacement of up to 6 mm applied at the loaded edge of SP3 was investigated. However, it was found to have a negligible effect on load–displacement behaviour of the connection, so is not discussed further. In terms of the stress distribution, the trend between the flanges and side plates followed the original SP3 model, with local yield in the flange first followed by full yield of the side plate cross-section and subsequent failure occurring in the side plates. This suggests some further adjustment in the model boundary conditions may be required to fully reflect the local stress concentration causing the fracture observed in the flange in testing. However, an asymmetrical stress distribution can be seen in the flanges in Figure 19, which reflects the stress concentration at the failure location in testing, shown in Figure 10(b). The ultimate capacity of the connection corresponded to an in-plane lateral displacement of 3 mm at the loaded edge, in the left-hand direction in Figure 19(a).

Further to the discussion in Section 3.2, based on the observations of the connection behaviour in experimental testing and the non-linear FEA, the design parameters presented in Table 2 and Equation 1 were reviewed. The original design was based on characteristic design strengths, with the side plate at its full yield capacity (42.9 kN) and flange tooth shear and bending utilisation ratios of 75% and 93%, respectively.

Updating this design considering the material yield strength (f_y = 283 MPa) and the average yield capacity of the connections (F_t = 46.7 kN) obtained from experimental testing, and using the expressions in Table 2, highlights the extent by which the design resistances were exceeded in testing. Firstly, the tooth shear force exceeded the adopted 75% plastic shear resistance limit $(0.75Vpl,Rd)$ by 4%. Secondly, the tooth bending moment $MEd$ exceeded the reduced tooth plastic moment resistance $Mpl,V,Rd$ by 12%, while the side plate axial load capacity was exceeded by only 6%; this indicates that the flange experienced higher stress concentrations due to tooth bending, as observed in the FEA and reflected by the flange fracture in testing.

Overall, any updated design should

• (a)

  guarantee yielding of the full side plate cross-section while the flange section's global behaviour is elastic

• (b)

  account for fabrication and construction tolerances

• (c)

  account for potential material overstrength influencing side ISC connection behaviour

• (d)

  maintain all design resistances within their acceptable limits.

Regarding (a) and (d), the connection's ultimate capacity should also be controlled by the side plate cross-section reaching the ultimate strength. Considering the original design without the longer flange teeth required for scaled testing (i.e. with $lt$ revised to 7.5 mm), the axial load capacity of the flange cross-section alone increases by 30% from 44.6 kN to 57.8 kN due to the corresponding increase in cross-sectional area. The results of the FEA for this revision are shown in Figure 20 for the YG design material model; the SP3 side plate was used here to account for fabrication tolerances. It was found that the full side plate cross-section yielded first, as desired, with yielding in the flange restricted to local regions at the base of the teeth corresponding to the stress concentrations observed in the original analyses. The side plate also controlled the ultimate capacity, as desired, with stresses in the flange reaching 402 MPa but not exceeding the ultimate strength at the same localised regions. The FEM of this connection reached an ultimate capacity of 66.7 kN.

In addressing item (b) of the above list, two cases can be considered for yielding of the side plates based on this study – one each for the maximum and minimum potential yield resistances (⁠$Fy,sp,max$ and $Fy,sp,min$⁠, respectively). The connection's nominal design resistance can be based on $Fy,sp,min$ as per the original design but the teeth and flange dimensions should be based on $Fy,sp,max$⁠, which takes account of potentially increased side plate capacity due to manufacturing tolerances (e.g. the modified design described in Section 4.2.1). Maintaining the same nominal design resistance for the scaled connection design presented in Section 3.1.2, $Fy,sp,min=42.9kN$⁠, which will result in a requirement for the same side plate effective cross-section; however, $Fy,sp,max=45.2kN$⁠, which requires an increased flange cross-section. Alternatively, maintaining the original flange cross-section and tooth dimensions, the side plate effective cross-sectional area can be reduced by taking $Fy,sp,max=42.9kN$ to provide the desired behaviour. This was achieved by reducing the side plate width from 20 mm to 19 mm, with this reducing the nominal design yield resistance of the connection to $Fy,sp,min=39.6kN$ – a reduction of 7.7%. This reduction was applied to the SP3 FEM for analysis of the design solution. The resulting stress contours at yielding of the side plate cross-section and ultimate connection capacity are presented in Figure 21, which shows the desired behaviour. The FEM for this connection reached an ultimate capacity of 61.8 kN, reflecting the reduced side plate cross-section capacity. Local stress concentrations in the flange were again observed to occur, but the maximum stress in the flange never exceeded 339 MPa.

Additional factors can be introduced to promote the desired connection yield and failure behaviour, which also address points (a)–(d) listed earlier in Section 5, such as increasing the width or the number of flange teeth, which will reduce the influence of stress concentrations at the base of the teeth. For this purpose, an alternative recommendation based on this study is to use Equation 3 with the expressions in Table 2 to calculate a revised number of flange teeth required for a beam section using characteristic design values. Using the scaled geometry tested as part of this study as an example, with the same side plate cross-section and assuming no change in flange tooth dimensions, $nt,rev=4$ rather than the $nt=3$ used in the original design. This straightforward revision increases the length of the side plates but reduces the individual flange tooth shear and bending utilisation ratios to 56% and 53%, respectively, and was found to reduce the associated stress concentrations observed in this study. It can be seen in Table 3 that the flange stress concentration factor reduces from 2.7 to 2.1, a reduction of 20%. Hence, this solution has residual capacity, which allows for material overstrength. The associated von Mises stress contours at both full yield of the side plate cross-sections and ultimate connection capacity are shown in Figure 22. Again, the desired behaviour was achieved, with the connection yield and ultimate capacities controlled by the side plate. Similar to the design with $lt=7.5mm$⁠, an ultimate capacity of 66.7 kN was achieved in the FEA. Also, the maximum stress reached locally in the flange was 361 MPa, well below the ultimate strength.

In Section 4.2, it was proposed to increase the flange tooth corner radius $rt$ to reduce the stress concentration in the flange. However, any increase in this radius is equivalent to both a reduction in the flange width (i.e. axial resistance) and an increase in $g0$ (i.e. the lever arm increases by the same amount for the resultant force of the side plate applied to the flange tooth). Therefore, for the original design presented in this paper, the radius cannot exceed $rt=1.3mm$ without reducing the design capacity of the flange. This can be partially addressed by increasing the flange tooth width $bt$⁠. To investigate this, a limited parametric study was carried out based on the expressions in Table 2 to identify a pair of values of $bt$ and $rt$ that can maintain the original bending capacity of the flange teeth at 46.3 kN without reducing the flange axial capacity below that of the side plates. This resulted in values of $bt=11mm$ and $rt=2.6mm$⁠, as shown in Figure 23, with the resulting stress contours for this FEM shown in Figure 24. Again, the desired behaviour was achieved, although the flange stress concentration factor was only reduced from 2.7 (for the reduced $lt$ model) to 2.5 in this case. The ultimate connection capacity observed from the FEA was 61.9 kN while the maximum stress reached locally in the flange was 382 MPa. Further increases to the corner radius are not possible without a reduction in the capacity of the flange cross-section, thus the use of extra teeth as per Equation 3 may be preferrable to achieve higher resistances and lower stress concentrations for the same flange section width.

Further increases in the tooth width $bt$ would also increase tooth bending and shear resistances but also cause the length of the side plate to increase, potentially influencing bending deformation of the side plate, which is not currently considered in the design. Future work will need to analyse this behaviour in greater detail to establish its significance for the flexural behaviour of the connection at full scale.

Table 3 provides a summary of the design resistances for all of the above design revisions. The calculated values given for the side plate are based on the design geometry (i.e. SP1 type side plate), while the SP3 type with higher resistance was used in the FEA to account for fabrication tolerance.

This paper has outlined the work carried out in a collaborative tripartite project between the USA, Republic of Ireland and Northern Ireland to create and investigate the design, development and testing of a new class of ISCs that rely on neither welding nor bolting, targeting the facilitation of fast disassembly of steel structures and material reuse. This connection type could potentially offer a time saving of 62.5% over conventional bolted connections. The front ISC was introduced with a summary of a numerical study of its behaviour under mixed-mode loading scenarios, highlighting its excellent shear resistance and noting its limitations in terms of axial and flexural capacity with respect to erection tolerances.

These limitations motivated the development of the side ISC, for which the results of scaled experimental testing and non-linear FEA of the axial behaviour of the flange connection have been presented here. The connection exhibited ductile behaviour, and the measured yield and ultimate load capacities were in close agreement with experimentally determined material properties, highlighting the extent of material overstrength over the characteristic values used in design. An unexpected but consistent tensile failure pattern was observed in the flange rather than in the side plates where it was expected; this was influenced by a combination of the material overstrength, the loading arrangement during testing and the fabrication tolerance of the sample geometry, leading to an asymmetrical stress distribution, as confirmed by FEA. This has implications for future design and testing of this connection type at full scale, particularly if side plates are to be designed to fail by yielding first or if fabrication and assembly tolerances are relatively large. As a result, the calculations of the connection design resistance were revised in the design procedure, supported by FEA of the effect of varying geometric parameters including the flange tooth radius and length and the number of flange teeth required. Increasing the number of flange teeth was found to reduce the flange stress concentrations while maintaining the connection capacity.

Work is ongoing in the AMASS project, with upcoming full-scale experimental testing of the behaviour and performance of the side ISC in a structural frame. This is expected to support the establishment of a straightforward design method to enable adoption by designers in industry.

The research presented in this paper is part of a collaboration and partnership between Queen's University Belfast, University College Dublin and the University of Minnesota. The authors wish to express their gratitude for the funding received from the US–Ireland Research and Development Partnership Programme, supported by the Department for Education (DfE) and Invest Northern Ireland (grant USI-096), the National Science Foundation (grant CMMI-1563115), Science Foundation Ireland (grant SFI/15/US/B3234) and Enterprise Ireland (grant CF20160454).

$Af$

flange effective area

$Asp$

side plate net effective area provided

$Asp,req$

side plate required effective area

$Av,tooth$

tooth shear area

$bt$

flange tooth width

E

Young's modulus

$Ft$

total force in flange

$Fti$

force per flange side

$Fy,sp,max$

maximum potential yield resistance of side plate

$Fy,sp,min$

minimum yield resistance of side plate

$fu$

material ultimate tensile strength

$fy$

material yield strength

$fyr$

reduced material yield strength for shear area

$g0$

tolerance between side plate and flange

$lsp$

side plate length

$lt$

flange tooth length

$MEd$

tooth design bending moment

$Mpl,Rd$

tooth plastic moment resistance

$Mpl,V,Rd$

reduced tooth plastic moment resistance

$nt$

number of teeth per flange side

$nt,rev$

revised number of teeth per flange side

$rt$

flange tooth corner radius

$tf$

flange thickness

$tsp$

side plate thickness

$VEd$

design shear force in tooth

$Vpl,Rd$

plastic shear resistance of each flange tooth

$Wpl,z,tooth$

tooth plastic section modulus

$γM0$

partial safety factor for resistance of cross-section

$ν$

Poisson's ratio

$ρ$

factor for bending stress reduction

$τy$

shear yield strength