NEW DESIGN APPROACH FOR WOOD BRITTLE FAILURE MECHANISMS IN TIMBER CONNECTIONS

NEW DESIGN APPROACH FOR WOOD BRITTLE FAILURE MECHANISMS IN TIMBER CONNECTIONS Pouyan Zarnani 1, Pierre Quenneville 2 ABSTRACT: Timber construction has experienced considerable progress in recent years. In such progress, apart from the implementation of new engineered timber products, the advancement of timber joints has played a significant role. The design procedures for timber connections in most design codes are based mainly on the yielding capacity of the fasteners using the European Yield Model (EYM). While the EYM theory provides accurate predictions for connections that fail in a ductile fashion, it does not take into account the failure of the connections due to the brittle rupture of wood as the consequence of fasteners group effect. Such a significant gap in the design of connections also applies to the New Zealand (NZS 3603) and Australian (AS 1720.1) timber design standards. A new design approach is presented which allows the practitioners to predict the connection capacity associated with different brittle wood failure mechanisms. An extensive testing regime has been conducted on high load-transfer capacity joints using timber rivets under longitudinal and transverse loadings on New Zealand Radiata Pine laminated veneer lumber (LVL) and glulam. The results verify the proposal and prove its reliability. A design guide was also developed which could eventually become a design clause in the next revision of the New Zealand timber design standard NZS 3603. KEYWORDS: Connection, wood capacity, fastener capacity, timber rivet, failure mode, LVL, glulam 1 INTRODUCTION 1.1 GENERAL Timber construction has experienced considerable progress in recent years. In such progress, apart from the implementation of new engineered timber products, the advancement of timber joints has played a significant role. Structures are detailed from a combination of hinged and moment resisting joints. Therefore, connections are often the most critical components of any type of structure. Evaluation of timber buildings damaged after extreme wind and earthquake events have shown that weak connections (Figure 1) are one of the major causes of problems [1]. 1.2 BACKGROUND The design procedures for timber connections in most design codes are based mainly on the yielding capacity of the fasteners (Figure 2) using the European Yield Model (EYM). While the EYM theory [2] provides accurate predictions for connections that fail in ductile fashion, it does not take into account the failure of the connections due to the brittle rupture of wood as the consequence of fasteners group effect. Such a significant gap in the design of connections also applies to the New Zealand NZS 3603 [3] and Australian AS 1720.1 [4] timber design standards. Figure 2: Fastener failure modes in single shear for timber connections (involving fastener bending and/or member bearing) Figure 1: Connection brittle failure at Siemens Arena, Ballerup, Denmark 1 Pouyan Zarnani, Postdoctoral Fellow in Structural Timber Engineering, Department of Civil and Environmental Eng., University of Auckland. Email: pouyan.zarnani@auckland.ac.nz 2 Pierre Quenneville, Professor of Timber Design, Department of Civil and Environmental Eng., University of Auckland. Email: p.quenneville@auckland.ac.nz Based on the state-of-the-art of wood strength prediction models for parallel-to-grain failure in connections using dowel-type fasteners, different methods consider the minimum, maximum or the summation of the tensile and shear capacities of the failed wood block planes. This results in disagreements between the experimental values and the predictions. It is postulated that these methods are not appropriate since the stiffness in the wood blocks adjacent to the tensile and shear planes differs and this difference leads to uneven load distribution amongst the resisting planes [5,6]. For instance, in a block tear-out failure (Figure 3), the

contribution of the bottom or lateral shear planes to the wood resistance cannot simply be considered as a function of their respective area as the connection load is not shared uniformly among the resisting planes due to the unequal stiffness of the adjacent wood volumes loading the fasteners. Figure 5: Proposed spring model assuming linear elasticity Figure 3: Wood block tear-out failure in two joints with identical shear and tensile resisting planes and different bottom and edge distances In the case of perpendicular-to-grain loading, the available models for the prediction of the splitting failure of dowel-type connections are determined generally based on a crack growth of the entire member cross-section. These models can be appropriate for stocky or rigid fasteners installed through the full thickness of the wood member. However, for slender dowel-type fasteners such as timber rivets, particularly when the penetration depth of the fastener does not cover the whole member thickness, partial splitting combined with rolling shear can happen rather than splitting of the entire width (Figure 4). Figure 4: Cross-section view of wood splitting perpendicular-to-grain: (a) full width failure mode, (b) partial width failure mode 2 NEW DESIGN APPROACH 2.1 WOOD BLOCK TEAR-OUT RESISTANCE PARALLEL-TO-GRAIN The wood block tear-out resistance under parallelto-grain loading is predicted using the stiffnessbased model proposed by Zarnani and Quenneville [7]. The proposed analysis for wood strength is best explained using the analogy of a linear elastic spring system in which the applied load transfers from the wood member to the failure planes in conformity with the relative stiffness ratio of each resisting volume adjacent to the individual failure plane (Figure 5). The difference in the loads channelled to the tensile and shear planes is a function of the modulus of elasticity (E) and modulus of rigidity (G), the volume of wood surrounding each of the failure planes (bottom, end and edge distances - d z, d a and d e ) and also of the connection geometry (Figure 6). For details regarding the determination of the stiffness of the resisting planes, refer to Zarnani and Quenneville [7]. Figure 6: Simplified analytical model By predicting the stiffness of the wood surrounding each of the failure planes (K h, K b and K l ), one can predict the proportion of the total connection load applied to each plane, Ri Ki K. By further establishing the resistance of each of the failure planes as a function of a strength criterion, one can verify which of the failure planes governs the resistance of the entire connection due to being overloaded. It should be asserted that the strength of the shear planes cannot be higher than the tensile capacity of the adjacent wood volume where the load is channelled to these resisting planes. If the attracted load by the resisting shear planes is larger than the tensile capacity of the associated wood volume, then the wood block torn out from the member would be as wide as or as deep as the member. Therefore, as shown in Figure 7, three different failure modes are possible and the wood load carrying capacity of the connection (Equation 1) equals the minimum of the resistances corresponding to these failure modes, P w,a, P w,b and P w,c.

Figure 7: Different possible failure modes of wood block tear-out P w = n p. min (P w,a, P w,b, P w,c ) (1) where P w,a = min (p wh, p wb, p wl ), Mode (a) P w,b = min (p wh, p wb ), Mode (b) P w,c = min (p wh, p wl ), Mode (c) in which p wh = wood resistance for failure of head tensile plane K K b l ft,m A th( 1 ) Kh Kh p wb = wood resistance for failure of bottom shear plane K K h l fv,m Cab A sb( 1 ) Kb Kb p wl = wood resistance for failure of lateral shear planes K K h b fv,m Cal A sl( 1 ) Kl Kl (2) (3) (4) Due to the absence of the lateral and bottom shear planes in the failure mode (b) and (c) respectively, the following considerations are necessary for the resistance calculation of these failure modes: I. For failure mode (b), the K l /K h and K l /K b ratios should be considered as zero; and the tensile and shear areas as wide as the member. II. For failure mode (c), the K b /K h and K b /K l ratios should be considered as zero; and the tensile and shear areas as deep as the member (in the case of double-sided joint, half of the depth should be considered). In Equations (2) to (4), f t,m and f v,m are the wood mean strength in tension and in shear along the grain (MPa). A th, A sb and A sl are the areas of the head, bottom and lateral resisting planes with respect to the wood effective thickness, t ef, subjected to tension and shear stresses. Also, C ab and C al are the ratios of the average to maximum stresses on the bottom and lateral shear planes respectively [7]. n p is the number of the plates equal to 1 or 2 for one-sided or double-sided joints, respectively. It is also important to note that the connection resistance corresponding to each failure mode is a summation of the critical plane failure load, plus the load carried by the other planes. Thus, when one plane fails, then the entire connection load transfers to the remaining planes in accordance with their relative stiffness ratios. It could be possible that the occurrence of the first failure of one plane does not correspond with the maximum load of the connection. Therefore, it is recommended that after the failure of one plane, one recalculates the wood resistance for the remaining planes to verify whether the residual planes can resist a higher load by defining: K h /K b and K h /K l = 0 after the head tensile plane failure; K b /K h and K b /K l = 0 after the bottom shear plane failure; and K l /K h and K l /K b = 0 after the lateral shear planes failure. In the case of fasteners which are inserted into predrilled holes, the area corresponding to the cutting diameter is to be subtracted from the resisting plane surfaces. This affects the strength of the tensile and shear resisting planes and not their stiffnesses. 2.2 WOOD SPLITTING RESISTANCE PERPENDICULAR-TO-GRAIN The wood splitting strength in perpendicular-tograin loading is predicted using the model proposed by Zarnani and Quenneville [8]. The proposed approach is based on two different possible crack formations on the member cross-section: with partial splitting on each side of the member corresponding to the effective embedment depth, t ef (Figure 4b) or with full width splitting (Figure 4a). In fact, for connections with a large penetration depth in slender members, the governing failure mode will be the full width splitting, and as the ratio of member thickness to penetration depth increases, the transition of wood failure mode from full to partial width splitting will occur (Figure 8). Figure 8: Occurrence zone of possible failure modes of wood splitting Therefore, the ultimate splitting resistance of the connection is determined as the minimum strength corresponding to these two failure modes and is given by Equation (5):

P w = min (P s,tef, P s,b ) (5) The wood capacity for partial width splitting, P s,tef, is predicted using a stress-based analysis (Equation 6) and involves the perpendicular-to-grain tensile capacity of the splitting surface of the wood corresponding to the effective embedment depth, t ef and the crack length that propagates along the member. The crack length along the member is considered as the summation of the joint net section width, w net, and the symmetrical crack growth on the left and right sides of the joint as a factor of the effective depth, h e (Figure 9). which cause large localized stresses and force brittle ruptures in the timber. Of this family of fasteners, the timber rivet is a well-established example in timber connection technology [10]. Timber rivets are hardened nails with a rectangular cross section of 3.2 by 6.4 mm. They are available in three standard lengths; 40, 65 or 90 mm (Figure 10). Timber rivets are always driven with the long axis parallel to the wood grain. Rivets are tight-fit fasteners since the head is wedged into the steel plate hole. Timber rivets provide high load-transfer capacity, high stiffness and easy to install steeltimber connections [11]. Rivets are part of the Canadian CSA-O86 [12] and American NDS [13] wood standards. However, there is no closed-form solution for the strength prediction of this type of connection. Therefore, timber rivets were used in the current joint failure tests to verify the model. Figure 9: Effective crack length on either side of the joint In Equation (6), f tp is the tensile strength perpendicular-to-grain; d a,l and d a,r are the unloaded end distance on the left and right side of the joint, respectively; C t is a coefficient function of the unloaded edge distance and the connection length. P s,tef =n p C t f tp t ef [w net +min(βh e,d a,l )+min(βh e,d a,r )] (6) The predictive equation presented for wood splitting in the entire member cross-section, P s,b (Eq. (7)) is adopted from the fracture mechanics based model developed by Van der Put and Leijten [9]. The significant difference is the application of the η factor which accounts for the effect of unloaded end distance and the connection width. P s, b 2 in which, bc fp he he 1 h min( h, d ) min( h, d ) w 2 h e a, L e a, R net e (7) and C fp is the fracture parameter. For more details regarding the wood splitting model, refer to Zarnani and Quenneville [8]. 3 EXPERIMENTAL PROGRAM As demonstrated over numerous studies, smalldiameter fasteners have shown a significant advantage over large-diameter ones such as bolts L=90 mm L=65 mm Figure 10: Timber Rivet L=40 mm 3.1 TEST SETUP Laboratory tests were set up to prompt wood failures and maximize the amount of observations on the brittle mechanism by considering higher capacity for the fasteners compared to the wood. Specimens were manufactured from New Zealand Radiata Pine LVL grade 11 and GL8 grade glulam. An extensive testing regime was conducted including 32 test groups under parallel-to-grain loading and 24 test groups under perpendicular-tograin loading. 3 replicates were tested for each group of specimens for LVL and 4 replicates for glulam. The specimens had riveted plates on both faces of timber, resulting in a symmetric connection. The steel side plates were 8.4 mm thick of 300 grade (F y = 300 MPa) with predrilled 6.8 mm holes to ensure adequate fixity of the rivet head. The effect of geometry parameters such as connection width and length, fastener penetration depth, loaded and unloaded edge distances, end distance, and member thickness were evaluated. For more details regarding the connection configurations, refer to Zarnani and Quenneville [8,14]. The testing protocol outlined in ISO 6891 [15] was followed. The tension load was applied to the specimens using a displacement controlled loading system. The deformation of the connection was

measured continously with a pair of symmetrically placed linear variable differential transformers (LVDT). The loading rate was adjusted to 1 mm/min and kept constant until the occurence of failure in both or either side of the riveted connections. A typical specimen in the testing frame is shown in Figure 11. in Figure 12b, in brittle failure mode, the failed block thickness, t block corresponds to the elastic deformation of the rivets (t ef,e ) since no plastic deflection was observed. However, in the mixed failure mode (Figure 12c), the t block is significantly lower with observable small deformation of the rivets. In the mixed failure mode cases, the loadcarrying capacity of the wood is based on the stiffness and strength of the tensile and shear planes corresponding to the bearing length of the fastener (t ef,y ) depending on its yielding mode. Figure 11: Typical specimens in testing apparatus: (a) longitudinal loading, (b) transverse loading - mid-span, (c) transverse loading - end of member 3.2 MATERIAL PROPERTIES All specimens were conditioned to 20 C and 65% relative humidity to attain a target 12% equilibrium moisture condition (EMC). The wood had an average density of 605 and 465 kg/m 3 for LVL and glulam members respectively. For the connection capacity parallel-to-grain, the average tensile and shear strengths evaluated were 34.3 MPa (coefficient of variation (COV)=12%) and 6.8 MPa (COV=10%) for RP-LVL and 24.1 MPa (COV=24%) and 4.2 MPa (COV=15%) for RPglulam (samples from inner laminations) respectively [14]. For the stiffness properties, based on data available in the literature, an average ratio of modulus of rigidity to modulus of elasticity (G/E) is considered equal to 0.045 and 0.069 for LVL and glulam respectively in order to make the planes stiffness equations independent of G and E values. The fracture parameter value, C fp, reported in Jensen et al. [16] and tensile strength perpendicular-to-grain values, f tp, evaluated by Song [17] were used in the proposed splitting model. The average C fp and f tp were 22.7 N/mm 1.5 and 2.06 MPa (for the tangential direction with a COV=18%) for RP LVL and 18.4 N/mm 1.5 and 1.99 MPa (at 45 to the radial direction with a COV=24%) for RP glulam. 3.3 TEST OBSERVATIONS Test series with a tightly spaced rivet pattern exhibited a brittle/mixed failure mode. For parallelto-grain loading, a sudden failure happened where a block of wood bounded by the rivet group perimeter was pulled away from either one side or both sides of the specimens (Figure 12a). As shown Figure 12: Wood failure parallel-to-grain: (a) block tear-out bounded by the rivet cluster perimeter, (b) brittle failure - rivets within the elastic range, (c) mixed failure - rivets with small deformation For the tests subjected to a perpendicular-to-grain loading, splitting of the wood occurred along the row of rivets next to the unloaded edge and propagated towards the timber member ends until reaching the unstable zone (Figure 13a). The splitting crack formed either through the entire member width (for thin members with high penetration-to-thickness ratios), as shown in Figure 13b or with a depth similar to the rivet effective embedment depth (for thick members with low penetration-to- thickness ratios), as shown in Figure 13c. Figure 13: Wood failure perpendicular-to-grain: (a) crack propagation along the top row of rivets, (b) full width splitting, (c) partial width splitting 4 VERIFICATION OF PROPOSED DESIGN APPROACH 4.1 RIVETED JOINT SUBJECTED TO PARALLEL-TO-GRAIN LOADING The connection ultimate capacities were calculated using the proposed stiffness-based model. For the brittle failure mode, the elastic deformation of the rivet was considered to determine the t ef,e [7]. The

estimated t ef,e was 27.1, 45.5 and 58.9 mm for a rivet penetration (L p ) equal to 28.5, 53.5 and 78.5 mm correspondingly. In the case of mixed failure, the t ef,y was predicted based on the bearing length corresponding to the governing yielding mode of the rivets equal to 28.4, 40.4 and 24.7 mm for the different penetration lengths. The t ef,y value for the longer rivet is lower compared to the other ones. This is due to the formation of the two plastic hinges (see Figure 2f) for this length of rivet during the yielding failure, whereas for the smaller rivet sizes, there is one plastic hinge (see Figure 2d). Figure 14 shows the strength predictions of the experimental groups compared to the test results. One can note that the proposed analysis results in precise predictions with a coefficient of determination (r 2 ) of 0.89, a mean absolute error (MAE) of 7.8% and a standard deviation (STDEV) of 10.3%. (MAE) and standard deviations (STDEV). This new design approach can be extended to other small dowel-type fasteners such as nails and screws for connection design improvement and failure mode prediction [18]. Figure 15: Predictions vs. observations for the joint load-carrying capacity under perpendicular-to-grain loading Figure 14: Predictions vs. observations for the joint load-carrying capacity under parallel-to-grain loading 4.2 RIVETED JOINT SUBJECTED TO PERPENDICULAR-TO-GRAIN LOADING There was a considerable difference between the predicted wood splitting strength and the rivet yielding resistance for the tested joints, therefore, the splitting failure was predicted to occur in a brittle fashion corresponding to the rivet elastic deformation [8]. The estimated t ef,e was 24.2, 40.1 and 51.0 mm for the L p equal to 28.5, 53.5 and 78.5 mm respectively. The proposed design approach leads to a relatively good agreement between the predictions and the test results with a coefficient of determination (r 2 ) of 0.78, a mean absolute error (MAE) of 17.2% and a standard deviation (STDEV) of 15.1% (Figure 15). One can note that the proposed design approaches result in precise predictions with high coefficients of determination (r 2 ) and low mean absolute errors 5 FROM THEORY TO PRACTICE 5.1 DEVELOPED DESIGN GUIDE A design guide has been developed [19] which outlines the new method to establish capacities for timber rivet connections (Figure 16). The guide provides an aid for engineers for designing timber rivet connections in structural wood products made of Australasian Radiata Pine, including sawn timber, glulam and laminated veneer lumber (LVL). The design guide includes the design checks for both wood and fastener load-carrying capacities under different loading directions; parallel- and perpendicular-to-grain. Furthermore, it allows practitioners to predict the potential brittle (wood block tear-out/splitting), ductile (rivet yielding) and mixed failure modes of the connection. To illustrate the design process, a series of examples are also included. Figure 15: Developed design guide

The guide is accessible through EXPAN (formerly STIC) administered by the Engineered Wood Products Association of Australasia (EWPAA). 5.2 CASE STUDIES 5.2.1 Carterton Event Centre Timber rivets have been used for the first time in New Zealand in 2011, in the truss connections of the Carterton Event Centre (Figure 16). The trusses were constructed using LVL members supplied by Juken New Zealand Ltd. The auditorium trusses were up to 24.6m long and 4.8m high, so they were delivered to site in two pieces. The riveted connection uses easily man handle-able components that allowed the required mid-span splice connection to be completed without specialist lift equipment. Use of the rivets allowed the fabricator, McIntosh Timber Laminates of East Tamaki, to save over $30k on this project when compared to the detailed bolted connection option. It also allowed for adjustments on site not possible with bolted fastenings. The timber rivets were found very userfriendly both in the work shop and onsite with significantly less visual impact and material cost compared to conventional fasteners [20]. 5.2.2 Trimble Building in Christchurch In another project by TimberLab Solutions Ltd. (formerly McIntosh Timber Laminates Ltd.), using rivets in the connections of the structure and energy dissipating system of the building (Figure 17) demonstrates the advantages of this timber fastener. The Trimble Building in south-west Christchurch was damaged by the September 2010 and February 2011 earthquakes. The new building was a designbuild project, which was undertaken by Mainzeal and Opus, commencing in February 2012. Figure 16: Riveted connections of Carterton Event Centre (McIntosh Timber Laminates Ltd.) Figure 17: Riveted connections of Trimble Building in Christchurch (c/o A. Buchanan) The building holds over 6,000 m 2 of office space over two levels and utilises LVL Pres-Lam frames in one direction and Pres-Lam walls in the other to resist seismic loads. The LVL of this building was supplied by Carter Holt Harvey Ltd. The principal structural engineer from Opus, and the structural

design team leader for the Trimble project found that the compact timber rivets provide high strength and stiff connections to take significant seismic loads [21]. 5.2.3 Portal frame for URM Building seismic retrofit In this recent project, EQStruc Ltd implemented efficiently LVL timber portal frames to resist seismic loads for an unreinforced masonry structure about 45m long by 12m wide (Figure 18). To provide strong and stiff joints between the timber elements, rivet connections were designed with controlled deflections. The client, a first-time user of timber portal frames in an unreinforced masonry (URM) building seismic retrofit, found that the rivet connections offered a cost-effective solution. system ductility only through the extension of the energy dissipaters under earthquake events. Innovations Nelson Ltd. is working on another similar project related to a 2-story commercial building in which rivets will be applied to provide high capacity and high stiffness hold-down connections. Figure 19: Riveted connections of CLT shear walls (Innovations Nelson Ltd.) Figure 18: Riveted connections of portal frames for URM building seismic retrofit (EQStruc Ltd.) 5.2.4 Kaikoura District Council Building Recently, Innovations Nelson Ltd. used timber rivets in the hold-down connections of the CLT shear walls of the Kaikoura event centre (Figure 19). Elastic deformation of the rivet connections were limited to 0.5 mm to satisfy the targeted 6 CONCLUSIONS Through the emergence of advanced engineered timber products as reliable structural alternatives along with scientific developments in timber engineering, many international building standards are revising their codes to accept high-rise timber buildings. In such heavy structures where the joints need to transfer large loads either parallel- or perpendicular-to-grain, the brittle group failure of the joint is more susceptible. The design procedures for timber connections in most design codes are based mainly on the European Yield Model which is only applicable for the ductile failure of the connections but not the brittle failure modes. Such a significant gap in the design of connections also applies to the New Zealand NZS 3603 [3] and Australian AS 1720.1 [4] timber design standards. A new design approach has been presented to determine the wood resistance in connection brittle failure mechanisms. The proposed method has been verified using extensive testing regimes conducted on riveted joints under longitudinal and transverse loadings on New Zealand Radiata Pine LVL and

glulam. Based on the proposed design model, an efficient connection design can be made by decreasing the difference between the capacity of the wood and the rivets. This new design approach can be extended to other small dowel-type fasteners such as nails and screws for connection design improvement and failure mode prediction. ACKNOWLEDGEMENT The authors wish to thank the New Zealand Structural Timber Innovation Company (STIC) for funding this research work. The authors wish also to express their gratitude to University of Auckland undergraduate students Samuel Wong and Shuai Ma, who did the experimental work related to glulam splitting as part of their final year project. REFERENCES [1] Smith, I., and Foliente, G. (2002). Load and resistance factor design of timber joints: International practice and future direction. J. Struct. Eng. ASCE, 128(1), 48-59. [2] Johansen, K. W. (1949). Theory of timber connections. Publications of International Association for Bridge and Structural Engineering, 9, 249-262. [3] Standards New Zealand (1993). Timber Structures Standard. NZS 3603:1993, Wellington, New Zealand. [4] Standards Australia (2010). Timber Structures: Design Methods. AS 1720.1:2010, Sydney, Australia. [5] Johnsson, H., and Stehn, L. (2004). Plug shear failure in nailed timber connections: Load distribution and failure initiation. Holzals Roh und Werkstoff, 62, 455-464. [6] Zarnani, P., and Quenneville, P. (2013). Wood load-carrying capacity of timber connections - An extended application for nails and screws. Materials and Joints in Timber Structures, RILEM Bookseries, Springer, 9:167-179. [7] Zarnani, P., and Quenneville, P. (2013). Wood block tear-out resistance and failure modes of timber rivet connections A Stiffness-based approach. Journal of Structural Eng. ASCE, 140(2): 04013055. [8] Zarnani, P., and Quenneville, P. (2013). Splitting strength of small-dowel-type timber connections: Rivet joint loaded perpendicular to grain. Journal of Structural Eng. ASCE (Accepted/in press). [9] Van der Put, T. A. C. M., and Leijten, A. J. M. (2000). Evaluation of perpendicular to grain failure of beams caused by concentrated loads of joints. Proc., International council for research and innovation in building and construction, CIB-W18, The Netherlands, paper 33-7-7. [10] Williams, C. C. (2006). Timber rivets. NZ Timber Design Journal, 16(2), 1-5. [11] Begel, M., Wolfe, R. W., and Stahl, D. C. (2004). Timber rivet connections in US domestic species. Research Paper FPL-RP- 619, U.S. Department of Agriculture Forest Products Laboratory, Madison, Wis. [12] Canadian Standards Association (CSA). (2009). Engineering design in wood (limit states design). CAN/CSA-O86.09, Mississauga, Ontario. [13] National design specification (NDS). (2012). National design specification for wood construction. Standard ANSI/AWC NDS- 2012, American Wood Council, Washington, DC. [14] Zarnani, P., and Quenneville, P. (2012). Predictive analytical model for wood capacity of rivet connections in glulam and LVL. Proc., 12 th World Conference on Timber Eng., Auckland, New Zealand. [15] International Organization for Standardization (ISO). (1983). Timber structures-joints made with mechanical fasteners-general principles and determination of strength and deformation characteristics. ISO 6891:1983. [16] Jensen, J. L., Quenneville, P., Girhammar, U. A., and Källsner, B. (2012). Splitting of timber beams loaded perpendicular to grain by connections - Combined effect of edge and end distance. Construction and Building Materials, 35, 289-293. [17] Song, Y. K. (2010). Investigation of tension strength perpendicular to grain of New Zealand Radiata Pine. Final Year Research Project Report, Depart. of Civil and Environmental Eng., The University of Auckland, Auckland, New Zealand. [18] Zarnani, P., and Quenneville, P. (2013). Strength of timber connections under potential failure modes - An improved design procedure. Construction and Building Materials, Elsevier, 60:81-90. [19] Zarnani, P., and Quenneville, P. (2013). Timber Rivet Connections Design Guide (Australia/New Zealand). EXPAN Structural Timber Solutions (formerly STIC) Administered by Engineered Wood Products Association of Australasia (EWPAA), Christchurch, New Zealand. [20] Structural Timber Innovation Company (STIC) (2012). Waking Up to the Advantages of Timber Rivets. Retrieved December 10, 2012, from http://www.expan.co.nz/timber-rivetspresent-strong-case. [21] Brown, A., Lester, J., Pampanin, S., and Pietra, D. (2012). Pres-lam in practice a damage-limiting rebuild project. Proc., SESOC Conference, Auckland, New Zealand.