Limited-damage 3D-printed interlocking connection for timber volumetric structures: Experimental validation and computational modelling

11 Cross laminated timber volumetric construction (CLTVC) is an innovative construction technology that 12 combines the sustainability of timber and the efficiency of modular construction, as opposed to 13 conventional construction. However, the connection installing methods of CLTVC, such as fastening, 14 are laborious with limited accessibility for connection installations, thus hindering the application of 15 CLTVC in midand high-rise structures. Therefore, a new way of connecting CLT modules by sliding 16 and stacking is explored herein with a proposed damage-control interlocking connection system, aiming 17 to provide a more efficient assembly solution to CLTVC that does not require onsite screwing. Quasi18 static monotonic and cyclic test, and numerical analyses were conducted to assess the mechanical 19 performance of the proposed connections, which possessed adequate translational stiffness and strength 20 of the proposed connections. The connections’ ability to control deformation – damage is moved away 21 from timber and the embedded fasteners, was also well demonstrated in the test, as both screws and 22 timber remained mostly intact after testing. The proposed connection design showcases a new concept 23 of modules’ assembly in volumetric construction with higher efficiency and flexibility; meanwhile 24 demonstrates the potential in reducing the permanent damage to structural materials during service life 25 and enabling reuse. 26


1.
Introduction (more recent studies, new interlocking connections and 28 their limitations) 29

Multi-storey timber volumetric construction 30
Volumetric construction is an emerging modern method of construction (MMC) that has the highest 31 degree of prefabrication (~95%) [1] with the main structure and other building accessories such as 32 cladding, internal finishes and MEP services all being manufactured and assembled into flat modules 33 in factories (offsite manufacturing -OSM) before being transported to sites for the final assembly. With 34 the advent of automation and the future factories (the fourth industrial revolution), better quality and 35 higher precision of assembly work compared to the on-site manual operation can be ensured in 36 volumetric construction. Being able to combine the sustainability of timber materials and the efficiency 37 of the OSM, volumetric timber construction is considered as one of the direct solutions to the growing 38 demand in affordable urban accommodation as well as the high carbon footprint problem own to the 39 construction industry [2]. Yet, this construction technology is limited to mid-and low-rise structures 40 [2], and the current trend of building tall volumetric timber structures imposes new challenges on the 41 wind and seismic resistances of such structural systems, requiring further investigations on varied 42 aspects of VTC; for example updates on design strategy in standards, high-performing connection 43 systems and assembly tools, as well as the development of more accurate numerical modelling methods 44 [3] for the confident delivery of secured high-rise volumetric timber structures. 45

Conventional timber connection systems for multi-story VTC 46
Among all the currently available forms of timber volumetric structures, the Cross Laminated Timber 47 (CLT) volumetric structure that employs load-bearing CLT panels, has somewhat better potential over 48 others (e.g., Framed Volumetric Structures and Structural Insulated Panels, SIPs) to be used for tall 49 buildings due to the high stiffness of CLT. However, suitable connections for this structural system still 50 lack sufficient investigation, as it is a relatively new construction technology. In volumetric 51 construction, the connections between modular units, also known as inter-module connections, are 52 critical to both the project efficiency and the overall structural performance of the buildings, as they 53 link prefabricated modular units together on-site to form the entire structure and are designed to transfer 54 horizontal and uplift forces from the wind and seismic actions. Previous quasi-static tests on CLT 55 panelised systems [4] [5] and shake table tests on full-scale low-rise CLT buildings [6] [7] indicated 56 that structures made of CLT panels generally demonstrated high strength and stiffness with most of the 57 deformation and energy dissipation being processed in connections and friction between timber panels. 58 Therefore, the proposal of high-performing connections is crucial to the realisation of high-rise 59 volumetric CLT structures with improved wind and seismic performances. 60 construction, which are originally designed for post-and-beam timber structures, so they are not fully 64 applicable in CLTVC. Some well-known disadvantages of these connection systems in both 65 construction and structure aspects currently limit their application in high-rise CLT volumetric 66 structures. To name a few: 67

Incompatibility with CLTVC and low construction efficiency 68
The limited accessibility is a widely recognised challenge in volumetric construction. The installation 69 of conventional timber connections requires external access, while accessibility for inter-module 70 connections has long been a key challenge in volumetric construction [9], especially when CLT 71 modules become enclosed as panelised structures, which makes the inter-module connections in 72 CLTVC entirely inaccessible (Fig. 2). Consequently, in some CLT volumetric structures [2,10], the 73 inter-module connections are absent and the lateral resistance of the structure is provided by the friction 74 between modules or the additional structural reinforcing systems, such as steel frames and concrete 75 cores. This method is commonly adopted in low-rise CLTVC and the inaccessibility of inter-module 76 connections is treated as one of the main obstacles to achieving tall CLT  The above-mentioned features are attributed to the cavities around fasteners formed by the crushing of 91 timber and the deformation of fasteners during loading, which damage the capacity of surrounding 92 timber and are highly unfavourable to the seismic design of high-rise structures, as they make the 93 connection cyclic behaviours unpredictable and reduce the structure's capacity of energy dissipation 94 during seismic events [14]. 95 In addition, as the connection is the governing factor of the timber structures' ductility, the common 96 concept of designing ductile timber structures is achieved by increasing the number of connections 97 while using small-diameter fasteners. Large number of conventional connections with slender fasteners 98 are therefore commonly adopted in large timber structures to achieve desired structural capacity ( Fig.  99 3), which leads to time-consuming on-site fastening work that negatively impacts the project progress. 100 This also limits the maximum reclamation and the reuse of timber components at the end-of-life of 101 buildings, as the removal of nails and screws is labour-intensive and can damage the structural material 102 [15]. Another concern of using small-diameter fasteners is the high stress they introduce on timber after 103 experiencing significant deformation at the later stage of loading, which can pinch through the fibres 104 before the capacity in timber is fully developed (Fig. 4(a)), causing sudden drop in connection strength 105 and great residual displacement even after the removal of the external loading. This can be considered 106 as brittle failure and cause irreversible damage to the structural elements, which results in difficulties 107 in structural maintenance and should be avoided in connection design. 108 109 Fig. 3. Numerous angle brackets used on a single CLT shear wall to achieve sufficient capacity and ductility 110

Insufficient accuracy in analytical models for connection capacity calculation 111
Comparison studies between analytical models and experiment results were conducted in previous 112 research [8,12], which indicated that the existing analytical models for timber connections provide 113 conservative predictions on connection strength but seems to significantly overestimate the stiffness 114 (up to 9 times higher according to the study of Gavric and Fragiacomo [8]

Innovative connection systems 143
To address the limited accessibility of inter-module connections in volumetric construction, a new way 144 of assembling flat modules using interlocking techniques was explored in previous research [21,22]. 145 These conceptual studies demonstrate the potential of interlocking technique in achieving 146 replaceability, adaptability and dismantlability in volumetric structures. This technique was also 147 employed in the design of some recently proposed connection prototypes for CLT volumetric structures, 148 as summarised in Table 1. These new connections eliminate the onsite manual operation in CLTVC 149 needed for screwing, so that the access to inter-module connections is not required, and better 150 construction efficiency and accuracy can be achieved. Similar interlocking design can also be observed 151 in some new connections proposed for CLT panelised structure ( A specially designed notched CLT panel used as the side wall of modules and the inter-module connection for the easy on-site modules assembly. [2] A novel intermodule horizontal connection proposed by the University of British Columbia A replaceable inter-module connection to achieve good energy dissipation and limit plastic deformation in timber. [24] Interlocking connections for CLT panelised structures LOCK Connection from Rothoblaas Ltd.
A concealed, interlocking wall-tofloor connection for CLT structures. The Lock Connection is attached on the surfaces or edges of wall and floor CLT panels, so the floor panels can be slid into the right positions to connect with wall panels, eliminating on-site fastening work. [25] J o u r n a l P r e -p r o o f

Metal dovetail connection
An interlocking connection inspired by traditional mortise and tenon joints and designed for the quick assembly of CLT panels. [26] 160 In some novel connections [27-31] designed for CLT panelised structures, another commonly adopted 161 design strategy is to introduce an individual 'weak component' in the connections to control the plastic 162 deformation. By applying capacity-based design to strengthen timber and fasteners, this design strategy 163 can efficiently isolate timber from deformation and localise the deformation within the additional metal 164 connectors. In this way, much improved strength and ductility can be achieved in connections with 165 reduced risks of brittle failure, considering the limited ability of timber to deform plastically. A similar 166 approach is also widely adopted in the recent proposals of dissipator [32-35] for CLT structures. 167

Concept of the interlocking connection 169
In order to achieve enhanced structural performance in multi-storey CLT volumetric structures and 170 simplify the module assembly process, a novel interlocking inter-module connection system is herein 171 proposed. This connection system consists of both vertical and horizontal connections that can be pre-172 installed onto CLT modules in the factory (off-site) ( Fig. 5(a)). It is suitable for most of the module 173 specifications, as it requires no further modification on panels for the fitting of connection, which can 174 greatly simplify and standardise the structural design process and become available to the entire market. 175 Owing to its interlocking feature, the modules can be accurately assembled on-site by sliding and 176 stacking without the need for additional tooling and special operations ( female connectors of both connections are groove-like devices, which are designed for accommodating 217 the male connectors. The male connector in shear connection is formed by a cantilevering thin-walled 218 curved steel band connected to a bottom steel plate via a cubic support at the middle, which is designed 219 to deform during movement. In the unit element of the tensile male connector, two symmetric L-shaped 220 steel components are connected to the steel plate with a 12mm gap in between for their free inward 221 movement when sliding along the sloping walls of the female connector. The width of the gap was 222 chosen corresponding to the maximum horizontal movement of the L-shaped components within the 223 female connector before reaching the top. Both male connectors are designed to act as the damage 224 control devices of the system and to yield first to isolate most of the deformation (without damaging 225 other integral parts of the connections) before reaching the target displacement, so the female 226 connectors, the timber and the fasteners can remain mostly intact (undeformed). In this way, the male 227 connectors become the critical components that determine the strength and ductility in the interlocking 228 connection, which ensure more predictable behaviours due to the less scattering properties of steel than 229 timber. This can also contribute to a more reliable (finite element) simulation with only basic material 230 properties such as compressive strength and tensile strength of timber and steel being used. Moreover, 231 as the ductility in this interlocking connection is designed to be achieved in the steel connectors instead 232 of the fasteners, large diameter screws (∅7mm-∅14mm) are used in this connection system, which as 233 suggested in the literature [8] [14] can improve load distribution and help to reduce the risk of in-service 234 damage or brittle failure in timber. It is worth noting that in case of changing the damaged parts (units) 235 of the connections, only the panels with the male connectors need removing. 236

238
To assess the realistic performance of the proposed connection design and to develop reliable numerical 239 models that can accurately simulate the connection performance, monotonic and cyclic tests were 240 conducted on full-scale shear and tensile connection unit elements. 241

Test configuration and testing material preparation 242
Experimental set-ups for shear and tensile connections are shown in Fig. 8. To achieve symmetric test 243 set-up for avoiding a moment being applied to the testing apparatus that may cause instability, two shear 244 connections were tested together with three panels, while each tensile connection was tested 245 individually. Overall, 3 monotonic tests (3 specimens) were conducted on the tensile connection and 1 246 (2 specimens) on the shear connection under the rate of 0.05mm/s, and 2 cyclic tests (2 specimens) were 247 performed on tensile connection and 1 (2 specimens) on shear connection under the speed of 0.02mm/s. 248 In the quasi-static (cyclic) test, the loading protocol as prescribed in EN12512 [37] was adopted, with 249 the estimated yield point of 4mm for tensile connections and 2mm for shear connections according 250 to the preliminary FE simulations. 4 strain gauges were attached to the specimens in each test with the 251 locations of which being decided considering the preliminary numerical simulations and the 252 accessibility of the assembled connections (Fig. 8). 253 All connection specimens were connected to 5-ply GL24h 350mmx350mmx120mm CLT panels with 254 density varied between 435-470kg/m 3 provided by Stora Enso using LBS7100 for the tensile 255 connections and HBSP12120 for the shear connections. The locations of specimens on timber panels 256 were all marked before the testing to measure the relative movement between timber and the steel 257 connector. After being screwed to CLT panels, connection specimens were assembled by stacking 258 (shear connection) or sliding (tensile connection) to form the experimental set-up and were tested 259 without additional reinforcement. Mechanical and geometric properties of panels and screws are listed 260 in Table 2 and Table 3. All CLT panels were conditioned in a controlled environment of 20℃ with 65% 261 humidity for the week before testing in accordance with EN 1380 [38], and all achieved the moisture 262 content of 10%-10.8%, which were within the required range in the standard (10%-14%). 263    The proposed novel connection specimens were 3D printed. Additive manufacturing was selected for 270 its limitless fabrication capabilities, considering the new complex geometry with inclined and curved 271 surfaces and the cantilevered steel band ( Fig. 10(b)&(d)) of the proposed optimised connection designs. 272 For the size of the proposed connections (Fig. 7), a metallic alloy composed of 60% 420 Stainless Steel 273 as base material and 40% Bronze for additional strength and resistance was chosen in the printing with 274 layer thickness of 100μm using Binder Jetting in Sculpteo [40]. The connection specimens were printed 275 along the length of connection. This printing direction was chosen to achieve the most stable printing 276 process and minimise the potential deformations in the cantilevering parts. Surface polishing was not 277 an option for this size of specimens from the manufacturer, thus the 3D printed connections were all 278 finished with granular surface (Fig. 10). To eliminate the risk of unpredictable behaviour due to the 279 variation of friction during the movement of male connections within the female connections, the 280 contact surfaces were polished to obtain smooth surfaces before testing. Despite the high manufacturing 281 accuracy of 3D printing, 1mm tolerance was introduced in the printing of all connection specimens, to 282 avoid the potential fitting issues caused by the dimensional distortion, considering the cantilevered 283 feature and the scale of printing in this project. In the testing of tensile connections, the remaining gap 284 after assembly was eliminated using the built-in function in the loading machine, while it is not 285 achievable in the cyclic testing on the shear connections due to the reserve loading. Therefore, small 286 amount of sliding with no reaction force can be observed at the beginning of each loading step in the 287 cycling testing on the shear connections. 288

Coupon test of 3D printed steel material 291
After the test, tensile coupon tests were undertaken to determine the tensile engineering stress-strain 292 properties of the 3D printed SS420/BR in accordance with EN ISO 6892-1 [41] and support future 293 numerical validation. Four coupons were machined from the bottom plates of the tested tensile female 294 connectors ( Fig. 12(a)), which experienced insignificant deformation during the testing. All specimens 295 were tested in the direction that parallel to the printing direction under a constant speed of 3 mm/min 296 using an Instron testing machine, with strain being measured by an extensometer attached to the 297 coupons' surface.  The original strain and stress data of the coupon tests were converted into true stress and strain, then 303 the yield strength was calculated accordingly using 0.2% offset method. The generated data were 304 relatively consistent with acceptable variation (Fig. 13), so the averaged properties from 4 coupon tests 305 were adopted in the subsequent numerical study. In comparison with the material properties provided 306 by the manufacturer [40] and the published coupon test data of SS420 [42] with the same printing layer 307 thickness and printing direction, the adopted mean values for simulation have similar young's modulus 308 and ultimate strain, while there are variations between the yield strength and the ultimate strength. These 309 can attribute to the sensitivity of 3D printing material to the varied printing parameters such as travel 310 speed and sintering temperature from different manufacturers. It should be noted that, 3D printed 311 material has different behaviours in the directions perpendicular to and parallel to the printing direction. 312 But the variations are only around 2-4% in two printing directions. According to the published 313 experimental data [42], for the same 3D printed material in this research with the same printing layer 314 thickness, the biggest property difference between directions parallel to and perpendicular to the 315 printing direction is 2.1% in the ultimate strength. It can therefore conclude that, this difference would 316 not affect the results significantly, and the material properties obtained in the printing direction of 317 coupon specimen can well represent the material performance on different directions. 318 testing set-ups were set to be consistent with the test, as illustrated in Fig. 7, Fig. 8. and Fig. 9. 327

Materials 329
In the models of both tensile and shear connections, the 5-ply CLT panels were modelled using an 330 orthotropic elastoplastic material model with the properties as listed in Table 2. Each layer of panel was 331 considered separately with the first (20mm), third (40mm), and fifth (20mm) layers being loaded 332 parallel to grain and the second (20mm) and fourth (20mm) layers being loaded perpendicular to grain. 333 For all steel fixtures and screws, a modulus of elasticity 200GPa and a Poisson's ratio of 0.3 were 334 assumed. Grade-300PLUS steel with a yield strength of 320MPa and an ultimate strength of 440MPa 335 was used in all steel fixture devices. The Grade 10.9 carbon steel with a yield strength of 940.3MPa and 336 an ultimate strength of 940.3MPa at 0.5% ultimate strain, was adopted for HBSP12120 and LBS7100, 337 which were validated by Tomasi et al. [45]. For the modelling of the 3D printed connectors, a modulus 338 of elasticity of 130GPa, a yield strength of 304MPa, an ultimate strength of 638MPa and an ultimate 339 strain of 5.9 % that extracted from the coupon test were taken. 8-node linear brick with reduced 340 integration (C3D8R) elements were adopted in the mesh throughout the model. 341

Modelling of shear connection 342
To better simulate the experiment with sufficient accuracy, the FE model of the shear connection was 343 built according to the experiment set-up ( Fig. 14(a)) with full dimensional details of steel fixtures as 344 shown in Fig.8(a) and connections as shown in Fig.7(a)  In the models of timber connections, the interfaces between timber and steel fasteners are crucial to the 352 overall accuracy of the simulation. When loaded in shear, the screws in timber connections are mainly 353 subjected to compression from the surrounding timber and the steel plate, which is mainly bored by the 354 shank part of screws. Consequently, the screws in interlocking shear connection were modelled as 355 cylindrical elements with a diameter equals to d s in Table 3. The surface-to-surface discretisation 356 method was adopted for the contact of all interfaces in the model with the "Hard Contact" option in the 357 normal direction, and the "penalty friction formulation" option in the tangential direction, while using 358 a coefficient of friction of 0 Tie constraints were adopted in all interfaces between nuts and threaded bars to simulate the fastening 360 effect, as no relative movements was observed during the test. In the tangential direction of the screws-361 timber interfaces, a non-linear relation of contact pressure and displacement (Eq.2) was used in the 362 normal direction along with the "Tabular" option in ABAQUS to simulate the weakening effect of pre-363 drilling on the timber around screws, considering the chunky screws used in the shear connections.  Where and are the control factors of the function curvature and were taken as 3.9 and 1.1, the 368 contact deformation 0 was taken as 0.35mm, all of which were extracted from the test on ∅12 dowels 369 conducted by Dorn [47]. 0 is the maximum compressive strength of timber that was assumed to be 370 30MPa [43]. This formula represents the low timber strength around the surface of screws. Once the 371 deformed screws reach the unaffected area, the strength of timber will be recovered. 17) to represent the threaded part with the diameter equals to 1 in Table 3. The 'soft material' is 384 assumed to be perfectly elastic and has the same capacity as the wood, except for the radial modules 385 that is reduced to 50MPa to eliminate its contribution in the compressive direction, representing the 386 weakening effect of threads on the timber. Wrapped by 'soft material' is the 'core' that simulates the 387 shank of screws, the dimensional value of which is taken as of LBS7100 (Table 3). Compared to the 388 first method, the second method is more computationally efficient due to the elimination of the small 389 details of the threads and the large contacts area between the threads and timber, thus it is adopted in 390 this study. 391 where the 0 , 0 and 0 are the peak allowable stresses in the normal (n), first (s) and second (t) shear 400 direction of the bonding interface, the values of which were taken as 36MPa, 6.9MPa and 6.9MPa, 401 equalling to the compressive and shear strength of timber as listed in Table 2. 402 After the cohesive resistance is reached, the strength degradation of bonding begins and the brittle post-403 damage behaviour of the connection under tension interaction was simulated by a 'linear damage 404 evolution' law, assuming that the full residual stiffness of the cohesive surface is achieved when 4mm 405 deformation first attained, so the contact behaviour became elastic-brittle, simulating the brittle failure 406 of dowel-type timber connection under tension. The cohesive interaction was combined with "Hard 407 Contact" in the normal direction and the "Penalty Friction Formulation" with a coefficient of friction 408 of 0.4 in the tangential direction to avoid the penetration of screws into timber at the post-failure stage. 409

Mesh sensitivity 410
To study the impact of mesh size on the FE predictions and ensure the adequate adoption of mesh size, 411 the mesh sensitivity analysis was conducted. The timber panels, fasteners, metal fixtures and the female 412 connectors, which experienced insignificantly deformation in the test, were proved to be independent 413 to the mesh size. Therefore, increasingly smaller mesh size was only applied in the interested areas of 414 both connection models (Fig.18), while the mesh size in other parts of the testing set-up remained 415 consistent. The male connectors of both connections, which are designed to govern connection 416 properties, were first studied with 3mm mesh size. Then the mesh size was refined to 1.5mm and 1mm. 417 It can be observed in Fig.19(a) that, the changes in mesh size have little impacts on the numerical 418 predictions of shear connector model. In the model of tensile male connector (Fig.19(b)), the initial 419 stiffness is independent to the mesh size, while yielding strength is slightly sensitive. The 1.5 mm mesh 420 size showed a good compromise between results accuracy and the calculation efficiency in both models. 421 It was therefore chosen in the subsequent study. 422 The primary deformation mode in the tensile connection was the bending of the L-shaped elements in 442 the male connector as designed (point 1 in Fig. 20(b)). The female connector and screws on the other 443 hand, remained mostly undeformed (point 2 in Fig. 20(b)), and only very slight open-up in the bearing 444 walls (point 3 in Fig. 20(b)) and slight bending in the bottom plate (point 4 in Fig. 20(b)) were visible 445 at the end of testing, so the deformed male connector can be easily slide out. 446 The FE model of tensile connection has good agreement with the experimental output, as the maximum 447 strength of each loading cycle, the initial and reduced stiffness can all be well captured. Four 448 deformation modes in the tensile connection as discussed above were also well reproduced in terms of 449 shape and dimension, proving the feasibility of the proposed modelling method for interlocking tensile 450 connection. Though slight pulling out can be observed from the screws in male connector in the model, 451 the continuous force-displacement curve indicated that the screws-timber interfaces still remained 452 elastic, as the 'linear damage evolution' in FE model was not activated. In the cyclic testing of shear connections (Fig. 21(a)), a sudden drop in force can be observed when first 457 reaching the displacement of 4mm, which was followed by another one at the displacement of 6.5mm 458 under the similar forces (around 16kN). These indicated the buckling happened in the shear connections 459 on both sides (point 1 in Fig. 21(b)), which caused instability in the experimental set-up. After the 460 buckling, the connection strength continued to develop in the deformed steel band and the end sunken 461 (point 2 in Fig. 21(b)), until the breakage happened at the buckling point of one shear connection ( Fig.  462  21(b)), leading to the plumper hysteresis loop at large displacement. Therefore, two major deformation 463 modes can be identified in the tested shear connections after the testing: buckling at the middle of the 464 deformable band and bending at the sunken design at the end of the steel band. As shown in Fig. 21(a), 465 the buckling and plastic deformations in the shear connection specimens caused unsymmetric 466 behaviours in two loading directions, which are unfavourable considering the stability of structures and 467 limit the connection's capacity in dissipating energy. The buckling locations were relatively consistent 468 among all the specimens, while unsymmetric buckling shapes can be observed in some specimens due 469 to manufacturing imperfections and the post-buckling loading on the buckled steel elements. Eigen-buckling analysis, which is a common method for predicting the buckling strength and generating 475 'imperfection' on the model to trigger nonlinear buckling analysis, was first conducted. In the Eigen 476 buckling analysis, a unit pressure was applied at the mid-span to predict the critical load of buckling 477 and the buckling mode shapes of the ideal structure based on the 'Block Lanczos' method. As shown 478 in Fig. 22, the generated four buckling modes in the male connectors on both sides agreed qualitatively 479 with the test results, so they were extracted and imported into the nonlinear analysis as 'imperfection' 480 with a factor of 0.2. As the primary deformation form in the tensile male connector was bending instead 481 of buckling, GMNIA was not considered in the tensile connector model. 482 With the inclusion of 1mm gap in the shear connection model, sliding in the initial stage of each loading 483 step that corresponding to the experiment can be observed in the numerical output. The model with the 484 initial imperfections was able to simulate the negative stiffness of buckling. The buckling shapes on 485 both sides as well as the bending at the end sunken were also well reproduced in terms of locations and 486 dimensions, which proves the adequate accuracy of the validation model. 487

Conclusions 488
The results of the monotonic and cyclic tests on the shear and tensile connections demonstrated that, 489 two connections behaved very differently in terms of mechanical properties and deformation forms. 490 The tensile connection had lower stiffness but higher maximum strength and better ductility with a 491 ductile failure mode (due to bending), while the shear connection showed higher stiffness but lower 492 maximum strength and ductility with a brittle failure mode (due to buckling). In both connections, there 493 was no sign of pulling out in screws and no visible relative movement between the connectors and 494 timber (Fig. 23)

501
To further characterise the translational behaviours of the proposed shear and tensile connections and 502 explore their potential with more common and ductile steel material (S235), a monotonic analysis was 503 performed in the validated models at the primary and the secondary working directions (Fig. 7) of both 504 connections, with material properties of S235 (modulus of elasticity of 210GPa, yield strength of 505 235MPa and ultimate strength of 360MPa). 506

Translational behaviours of the shear connection 507
The force-displacement responses of the interlocking shear connection FE model in the primary and 508 secondary directions are reported in Fig. 22. The numerical model can also monitor the deformation 509 developing trends in the shear connections to help better understand the connection behaviours, which 510 was not available in the test due to the concealed design nature of the shear connection. Similar to the 511 tested specimens, the shear connection with S235 first buckled at the cantilevered steel band, resulting 512 in negative stiffness in the FE model. With the continuous loading, the buckled steel band continued to 513 bend and the deformation started to develop in the end sucken on the other side. Corresponding to the 514 experiments, most of the plasticity was observed in the designated deformable band (Fig. 25(a)) in the 515 male connector after a 30mm displacement in the primary direction, while litter plastic strain was 516 developed in the female connector and fasteners, which also resulted in very small damage on timber 517 panels. 518 When working in the secondary direction, the interlocking shear connection showed much higher 523 strength and ductility than that in the primary direction, but the process of plastic deformation was 524 mainly managed by the bending of the middle cubic support in the male connector and the bending of 525 the wall in the female connector at the loaded side. The increased plastic strain in the screws proved 526 that the fasteners also contributed to the connection strength when working in the secondary direction. 527

Translational behaviours of the tensile connection 528
As demonstrated in Fig. 24, the behaviour of the interlocking tensile connection with S235 in the 529 primary working direction can be classified into three stages: the elastic stage, the yield plateau stage, 530 and the densification stage. When the loading started, the male connector started to move upwards 531 within the female connector, and the sloping walls of which compressed the L-shaped elements to bend 532 inwards. With the continuous vertical movement of the male connector, the L-shaped elements yielded  533  and the walls of the female connector started to open up slowly due to the increased moment at the wall  534 base, resulting in slightly decreased reaction force of the connection at the yield plateau stage. When 535 the male connector came into contact with the top of the female connector, it stopped moving and the 536 stiffness of the connection increased significantly, indicating that the behaviour of connection enters 537 the densification stage, in which the connection becomes rigid to prevent further displacement 538 (opening). Thus, the F-curve of tensile connection in the primary working direction can be idealised 539 using three parameters: the initial stiffness, the yielding strength and the displacement at the onset of 540 densification. Fig. 27 shows that, at the displacement of 15mm in the primary working direction, the L-541 shaped elements in the male connection processed most of the plastic deformation, while little can be 542 observed in other connection components. Though the stress of the screws under tension cannot be 543 reflected directly from the stress contour due to the employment of fictitious elastic 'soft material', no 544 strength reduction can be observed in the force-displacement curve, indicating that brittle failure did 545 not appear in screws at 20mm slip and the screws were still in the elastic stage. 546  plot depicts that the  551  interlocking tensile connection developed a similar working mechanism to the conventional steel plate  552 connections, in which the bending of screws and the crushing of timber are the primary deformation 553 modes. The embedment strength of timber and the bending strength of screws therefore became the 554 main contributors of the connection strength, while the interlocking steel connectors were relatively 555 rigid. 556

6.
Discussions and Limitations

557
The FE models of both interlocking shear and tensile connections with S235 demonstrated that the 558 proposed connection designs can provide adequate strength and ductility under translational forces. 559 When working in the primary direction, the male connectors in both connections yielded before 560 reaching the full capacity of the surrounding fasteners, so most of the plastic deformation is managed 561 within the designated areas (male connectors), while the female connectors, fasteners and timber remain 562 mostly intact. This indicates that the dissipating element in the novel interlocking connection shifts 563 from the fasteners like in conventional connections to the steel elements in the new connection. With 564 the specially designed deformable steel elements, this connection system can absorb energy in 565 predictable manners, preventing plastic deformation from developing in fasteners and timber to avoid 566 brittle failure and reducing the risk of significant failure of structures under extraordinary loads such as 567 seismic load. In addition, the initial deformation of the steel male connectors can be monitored and act 568 as an early warning system since they can be relatively easy to access. This also means that only the 569 male connector would fail after a severe event and needs to be repaired or replaced, while the screws 570 connecting the unit and the panel will remain intact and thus the timber fibres will not be affected by 571 the bending of the screws as it is normally the case. It therefore helps to improve the structural integrity 572 and to reduce the time needed for maintenance. When damage happens in the connection units, the 573 damaged module can be removed following the installation sequence for the repair or replacement. Due 574 to the short unit length of connection (same to the steel specimen dimensions), rapid and cost-effective 575 supply can be achieved with 3D printing of the male connector. Also, the timber panels can be fully 576 reused since the proposed novel connection system promotes no fastener failure. 577 In the secondary working direction, the deformation in the connection is less ductile and managed by 578 the composite effect between the fasteners and the steel connector, which is similar to the working 579 mechanism of the conventional steel plate connections and is characterised by its high stiffness but 580 relative lower ductility than in the primary working direction. When comparing the deformation forms 581 in two working directions, the significantly reduced deformation in screws in the primary direction 582 indicates the successful damage control in connections through the introduction of the proposed 583 connector design. 584 Different from other novel connections for CLT volumetric structures, the one proposed in this study 585 requires no modifications for fitting, thus is applicable to different flat module specifications. The 586 equipped damage-limiting capacity also makes it superior to the conventional connection in terms of 587 life-safety performance and material reuse after the end-of-life of structures. The proposed connection 588 system holds the potential of improving the dismantlability in CLT volumetric structures, promoting a 589 more flexible volumetric construction with changeable, demountable and fully reusable structural 590 elements. 591 Despite the superiorities, the immediate practical application of the proposed connection system may 592 be limited by the existing construction tools. The sliding and stacking of modules require specially-593 designed lifting and moving machines and scaffolds frames. Also, due to the geometric complexity, the 594 mass production of the proposed connections is difficult to be achieved with the conventional 595 manufacturing methods. However, with the gradually increasing manufacturing scale of 3D printing 596 and the reducing printing cost, the mass production is expected to be achievable with 3D printing in the 597 future. 598

599
In this paper, a comprehensive review of the recent advancements in the study of connection systems 600 for CLT volumetric structures is first provided. The factors that relate to connection systems and 601 constrain the development of multi-storey CLT volumetric structures have been analysed and 602 summarised. As a result, a novel interlocking connection system is then proposed for addressing the 603 low construction efficiency, inaccessibility of inter-module connections and the insufficient ductility of 604 typical timber connections in VTC. The designed ability of managing deformations within one part of 605 the connection and eliminating damage in timber panels in the primary working direction was proved 606 via experiments and numerical analyses on the 3D printed connection specimens. Numerical models 607 with good accuracy were proposed for both connections using only nominal material properties, owing 608 to the fact that the load-carrying capacity and ductility are correlated to the proposed interlocking metal 609 connectors. Ultimately, the adequate translational strength and ductility of the proposed connections 610 were proved by the validated numerical models. 611 Future research will include a comprehensive parametric study to identify the critical parameters of the 612 proposed connection systems and their detailed mechanical properties, as well as conduct a direct 613 comparative study between typical connections and the novel connection system via global models of 614 a full-scale CLT volumetric building. The study will be centred to the contribution of the connections 615 to the overall lateral resistance of the structure. 616