Thermo-mechanical behavior of CLT beam-to-girder assemblies connected with T-shaped dowelled connections before, during and after fire exposure

Purpose – 3-ply cross-laminated timber (CLT) is used toinvestigate the thermo-mechanical performance of intermediate-size assemblies comprised of T-shaped welded slotted-in steel doweled connections and CLT beams at ambient temperature (AT), after and during non-standard fire exposure. Design/methodology/approach – The first set of experiments was performed as a benchmark to find the load-carrying capacity of the assembly and investigate the failure modes at AT. The post-fire performance (PFP) test was performed to investigate the residual strength of the assembly after 30-min exposure to a non-standard fire. The fire-performance (FP) test was conducted to investigate the thermo- mechanical behavior of the loaded assembly during non-standard fire exposure. In this case, the assembly was loaded to 67% of AT load-carrying capacity and partially exposed to a non-standard fire for 75 min. Findings – Embedmentfailureandplasticdeformationofthedowelsinthe beamwerethedominantfailure modes at AT. The load-carrying capacity of the assembly was reduced to 45% of the ambient capacity after 30 min of fire exposure. Plastic bending of the dowels was the principal failure mode, with row shear in the mid-layer of the CLT beam and tear-out failure of the header sides also observed. During the FP test, ductile embedment failure of the timber in contact with the dowels was the major failure mode at elevated temperature. Originality/value – This paper presents for the first time the thermo-mechanical performance of CLT beam- to-girder connections at three different thermal conditions. For this purpose, the outside layers of the CLT beams were aligned horizontally.


Introduction
In comparison to other types of construction materials, cross-laminated timber (CLT) has advantages in architectural appearance, constructability, cost and sustainability (Barber, 2018). As such, the use of CLT has been on the rise in Europe and more recently in North America (Smith and Frangi, 2014;Grasser, 2015;Pei et al., 2016). Although CLT has been employed mainly as large dimensional wall and floor panels in contemporary heavy timber structures, it is also possible to bring CLT into service as a 2D beam element (Tuhkanen et al., 2018). It has been assumed that using the CLT beam will improve the performance of the beam member against brittle splitting failure. There are different retrofitting methods such as FRP strengthening (Gilfillan et al., 2003;Dempsey and Scott, 2006;Li et al., 2014;Solarov and Gli si c, 2014), utilizing steel strips (Franke et al., 2015) and self-drilling screws (Blaß and Schmid, 2001;Palma et al., 2013) to overcome this undesirable failure; however, these solutions may cause a negative influence on the performance of the member at the dowel locations during the fire (Martin and Tingley, 2000;Blaß and Schmid, 2001;Palma et al., 2013). Additionally, steel connections play a crucial role in both seismic and fire performance (FP) of contemporary timber structures, increasing the ductility and thereby improving the seismic performance of the structure against lateral forces.
In contrast, steel connections may be seriously affected by fire, losing strength and stiffness, leading to large plastic deformations and contributing to progressive collapse. In general, FP of steel connections in timber structures is influenced by parameters such as material thermal properties, temperature dependent mechanical properties and geometry of the assembly. This is a complex thermo-mechanical problem and the research conducted on the FP of the doweled steel connections is limited to the fire tests performed in accordance with the standard fire curves (Erchinger et al., 2010;Moss et al., 2010;Peng et al., 2012;Khelifa et al., 2014;Audebert et al., 2014;Palma, 2016;Hofmann et al., 2016). These standard fire curves are mostly adopted by prescriptive codes, while most of the contemporary timber structures with different amount of exposed combustible timber material have been designed through performance-based design (PBD) guidelines. The design fires determined for the PBD method include but are not limited to the standard fire curves. The heat flux of these design fires plays a significant role in charring depth, and as a result, load-carrying capacity of the structural timber elements.

JSFE
Additionally, most of the standard fire curves do not capture the decay phase of the realfire and they are more suitable for comparing the performance of different type of assemblies with each other. Furthermore, performing the standard fire test in the furnace restricts the loading condition, measurements and visual access (Gales et al., 2021). To overcome some of these restrictions, a set of experiments was designed to investigate the thermo-mechanical behavior of CLT beam-to-girder assemblies connected by T-shaped slotted-in doweled type connections before, during and after non-standard fire.
In a previous study, thermo-mechanical performance of glulam beam-to-column connections were investigated at ambient and fire conditions (Palma, 2016). Results showed reducing the gap between connecting members improved the FP of the connection. A recent study, however, showed that removing the gap between connecting members led to undesirable brittle splitting failure in the glulam members (Shabanian and Braxtan, 2022a). This study presented ambient, post-fire performance (PFP) and FP tests of glulam beam-to-girder connections with full contact (no gap) between connecting elements. Results showed that ductile dowels were critical in connection performance where brittle splitting failure in the glulam beam members was delayed or eliminated. Splitting failure in the glulam beam occurred at ambient temperature (AT), but only after large deformations in the connection; splitting failure was not observed in the PFP and FP tests on the fire exposed connection as charring during the fire degraded the wood properties and decreased the contact between connected elements. Embedment failure in the glulam beam was the controlling failure mechanism during the fire and post-fire tests.
Another study presented ambient, PFP and FP tests of glulam beam to CLT wall connections with full contact (no gap) between connecting elements to consider the effect of changing the geometry and timber lamination orientation of the header (Shabanian and Braxtan, 2022b). Results showed that the controlling failure mechanism remained to be splitting failure at AT and embedment failure in the beam during PFP and FP tests despite this change in geometry.
The current study presents a CLT beam-to-girder connection in an effort to reduce the splitting failure in the beam members. AT, PFP and FP tests were considered in this research. The CLT beam-to-girder connection is a more unique configuration and research is limited on the subject.

Experimental program 2.1 Assembly description
Experimental tests were conducted on symmetric, intermediate-size CLT beam-to-girder assemblies. Figure 1 illustrates the CLT beam-to-girder assembly and its geometry. Each assembly was comprised of two 3-ply CLT girders connected to a CLT beam with steel connections. Each steel connection included 4 full-length steel dowels to the joist member and 20 Type 316 stainless steel screws to each girder. Table 1 provides more detailed information on the dimensions and material type of each component.
2.2 Material description 2.2.1 Cross-laminated timber (CLT). Both the beam and girders in the assembly studied were 457 mm long by 254 mm deep, 3-ply spruce CLT with total thickness of 120 mm. A VOC-free, formaldehyde-free, PUR adhesive was used over the entire surface of joints. The adhesive was classified as a TYPE 1 adhesive and has been approved to produce load-bearing timber components. The CLT panels had an average moisture content (MC) of 12% ( ± 2%) on delivery date and density of 5.5 kN/m 3 . According to the CLT manufacturer, the CLT panels had nominal char rate of 38 mm/h and flame spread index of classification B. All the samples were loaded in the gravity direction at the mid-span of the beam, in the direction perpendicular to the grain of the outside layers and parallel to the grain of the mid-layer.   Figure 2 displays the welded steel doweled connection and its geometry.
2.2.3 Steel dowels and fasteners. The full-length dowels (120 mm) are cut from 12.7 mm diameter steel rods with similar material properties as the steel connection (ASTM A572 Gr 50). The 6.35 mm diameter heavy-duty hexagonal connector screws with 76.2 mm length and 50.8 mm thread are manufactured from low-carbon steel wire grade 1022. This fastener has 1,130 MPa bending yield strength, 6.36 kN allowable tensile capacity and 3.5 kN allowable shear capacity. Plate 1 shows the steel dowels and screws placed on top of the CLT beam.

Experimental setup
Six CLT beam-to-girder assemblies were tested in totalfour at AT, one post-fire and one during fire. Plate 2, Figures 3 and 4 show the AT, PFP and FP test setup, respectively.
To approximate the real boundary conditions of the connections in all the fire tests, the CLT beams were exposed on 3-sides and only the top surface of the assembly was covered by insulation board. Plate 3 shows photos of the two-step PFP test, including 30 min of non-standard fire exposure of approximately 480 8C and subsequent cooling, followed by mechanical loading to failure.
Plate 4 shows a photo of the FP test, including the application of constant mechanical load equivalent to 66% of the AT assembly capacity along with non-standard fire exposure of approximately 480 8C until failure. This higher than usual mechanical load was prescribed to Plate 1. 3-Ply CLT beam, fulllength steel dowels and heavy-duty screws Plate 2. Test set-up and instrumentation utilized for testing at ambient temperature JSFE accelerate the failure time of the assembly and to study the thermo-mechanical behavior of the steel connection within the assembly under a greater state of stress. Figures 5 and 6 show the locations of the k-type thermocouples used in the fire tests. Detailed information on the test set-up is provided in related research (Shabanian, 2020).

Ambient temperature results
Four replicas of the CLT beam-to-girder assembly were tested at AT. The first two tests were stopped after initial failure occurred and investigated to establish the primary failure modes. Assemblies 3 and 4 were loaded until failure and the load-carrying capacity and deflection at the mid-span of the assemblies were recorded. Figure 7 shows the load-displacement of the CLT beam-to-girder assemblies tested at AT for replicas 3 and 4. The average load-carrying capacity of this assembly was 146.7 kN. The elastic region of the curve has an average slope of 10.5 kN/mm.    Fire-performance test of the CLT beam-togirder assembly JSFE Plate 5 shows the failure modes that occurred during the AT test in the order observed. The failure in this assembly was triggered by the plastic embedment failure and bending of the dowels. Then row shear failure occurred in the mid-vertical layer of the CLT beam loaded in shear parallel to the grain at the dowel location (shown as sudden drops in loaddisplacement curves between 10 and 15 mm displacement) and finally the test was stopped after the partial splitting failure occurred at the outside layer of the CLT beam and girders.

Thermomechanical behavior of CLT beam
In the outside layers where the dimensional lumbers were laid out horizontally, the transferred shear load produced a tension perpendicular to the grain direction and a brittle splitting failure occurred.

Post-fire performance test results
One sample was tested to study the PFP, stiffness degradation and residual loadcarrying capacity of the CLT beam-to-girder assembly after heating and subsequent cooling. The results of the 30-min PFP (PFP30) test include the heat distribution during the 30 min of fire exposure and the load-displacement behavior of the burnt sample at the mid-span. Figure 8 shows heat distribution along the sample during the 30 min fire exposure. The burner temperature rapidly increased for the first 3-4 min to approximately 480 8C and then ranged from approximately 340 8C-560 8C for the duration of the fire, with the slight fluctuations in temperature due to the manually control of the gas flow. The figure also shows the ASTM E119 standard fire curve for comparison.
The beam temperatures at the location of TCs 7, 8 and 9 (bottom, middle and top of the beam) reached the wood ignition temperature (300 8C) and then began to increase in comparison to the burner temperature after one minute. Plate 6 shows visual confirmation of ignition over the height of the beam 1 min into the fire test.
Temperature of the three lower dowels at their tips was recorded by TCs 4, 5 and 6. The temperature of the dowel tips followed the same fluctuation pattern as the beam and burner Thermomechanical behavior of CLT beam 50 remain unaffected after exposure to temperatures up to 600 8C. Therefore, no reduction in residual strength of the steel connection is expected.
In the next step, the burned sample was cleaned up and loaded at the mid-span, subject to similar boundary conditions to the AT tests. The PFP30 loading test terminated when the sample experienced major failures similar to what was observed in the AT tests and when the extension at the mid-span was around 15 mm. According to Figure 9, the load-carrying capacity of the assembly after 30 min of fire exposure was reduced to 82.5 kN, 55.8% of the AT capacity. Since the steel connection itself is expected to maintain its AT strength, this reduction in capacity is attributed to a gross loss of wood cross-section due to charring and loss of strength in the thermal penetration zone behind the char layer.
Plate 7 shows the failure modes in the PFP test replica. The embedment failure around the dowels was the dominant failure mode followed by row shear of the mid-layer of the CLT beam and plug shear of the CLT girder. In the PFP30 loading test, the steel dowels did not experience a similar level of plastic deformation and bending in comparison to the AT tests. This observation is predictable based on Figure 9 and considering that the PFP30 sample did not experience the same level of plastic deformation that AT samples went through. It is interesting to notice that the maximum plastic deformation in the dowels formed where the most degradation and charring occurred in the CLT beam. In this case, maximum charring depth in the joist member occurred at the lower end of the joist, and it was approximately 70 mm while charring depth at the dowel location section was approximately 30 mm.

Fire-performance test results
The fire resistance (FR) of the CLT beam-to-girder assembly was studied through the coupled application of thermal and mechanical loading. A constant load equal to 98.6 kN (approximately 67% of the expected load-carrying capacity of the assembly at AT) was imposed while a non-standard fire of approximately 480 8C was applied until failure. Figure 10 shows the heat distribution along the exposed area during the FP test. According to the recorded heat distribution, the bottom of the CLT beam ignited after 2 min.  Heat distribution along the exposed area during the fireperformance test Thermomechanical behavior of CLT beam Figure 11 shows the deflection of the CLT beam at the mid-span recorded during the FP test. According to this figure, the sample failed after 55 min of non-standard fire exposure. The maximum displacement ratio occurred at this point, and it was 0.32 mm/min. Plate 8 shows a series of video captures taken from recordings during the FP test. Onset of ignition at the beam bottom can be seen at 2 min. The charring development around the connection is obvious in these pictures. The CLT beam started charring from the bottom corner and the lowest dowel was exposed after 30 min to the fire. This was where the embedment failure of the CLT beam was captured for the first time. By the end of the test, the lowest dowel was only surrounded with char.
Plate 9 shows the failure modes captured during the FP test. The failure initiated by embedment failure around the lower dowels (Plate 9a) and followed by bending in the top dowel (Plate 9b).   Figure 10, the temperature of the connection at the time of failure at 55 min was nearly 650 8C at the bottom of the connection. At 650 8C the steel is expected to maintain only 35% of its AT yield strength. Failure, however, was not observed in the steel connection itself, but instead in the timber elements. This is attributed to the extensive charring that occurred in the timber which in turn weakened the timber and led to the embedment failure in the beam. Plate 10 shows the residual section of the CLT beam-to-girder assembly after the FP test. Extensive charring occurred around the steel connection. The maximum char depth on the header side occurred at the bottom and was equal to 40 mm. The charring progressed faster in the mid-layer of the CLT-beam in comparison with the exterior layers. In the assembly, changes in the geometry and boundary condition caused a rotation in the steel connection. However, as there was no contact between beam and girder, this rotation followed by a plastic deformation in the lower edge of the steel connection.

Discussion and conclusions
This experimental study demonstrated the performance of a CLT beam-to-girder assembly connected with T-shaped connections subject to vertical load before, during and after nonstandard fire.
For the assemblies tested at AT, ductile embedment failure of the CLT-beam around the dowels and plastic bending of dowels were the dominant failure modes. This is significant inasmuch that brittle, splitting failure of the beam was not the controlling failure mechanism as is often seen in similar connections with timber elements loaded perpendicular to the grain. The samples also experienced secondary failures due to row shear failure in the mid-layer of the CLT-beam and splitting failure between the screws and finally at the outside layers of the CLT-beam.
For the assembly loaded after 30 min non-standard fire exposure (PFP test) embedment failure, bending of the dowels, row shear failure of the beam, and plug shear of the headers occurred, but importantly, no splitting occurred in the CLT members. The load-carrying capacity of the assembly was reduced by 44% compared to the ambient capacity. The maximum recorded temperature of the steel connection during the PFP test was 525 8C. Therefore, the reduction in load-carrying capacity of the assembly is attributed to the gross loss of wood cross-section due to charring and loss of strength in the thermal penetration zone behind the char layer.
For the loaded assembly tested during the non-standard fire (FP test), ductile embedment failure and bending of the top dowel were the only failure modes observed when the specimen failed after 55 min of coupled fire and mechanical loading, with an imposed load of 67% of ambient capacity. In this case, the steel connection deformed slightly in the lower edge.
All the deformed dowels in this research experienced second mode of failure (Johansen, 1949) and only one plastic hinge formed in the middle of the dowels. Increasing the side member thickness (joist width) with a larger embedment length and using longer dowels could increase the ductility and load-bearing capacity of the connection for this research. This could also improve the FP of the connection.
In both the PFP and FP tests, extensive charring in the assemblies caused a degradation of timber strength and gross loss of section. The presence of the steel connection promoted the charring process and facilitated the failure of the wooden beam at the connection area.