STUDY OF THE STRUCTURAL BEHAVIOUR OF THE CONNECTION BETWEEN MULLIONS AND TRANSOMS IN WOOD-GLASS FACADES Philip Steinhausen Department of Civil Engineering, IST, Technical University of Lisbon 1 Abstract Keywords: Beech, Connection, Facade, Glass, Glulam, Mortise, Mullion, Structure, Tenon, Timber, Transom, Wood This study analyses the structural behaviour of the connections between mullions and transoms in beech glulam wood-glass facades. The use of beech wood in construction can be a response to possible future bottlenecks in the supply of construction wood, which might result from the forest conversion. Another advantage is that it is stiffer and stronger than the usually used coniferous wood, in particular when having in mind the weight of modern three-layered glazing. State-of-the-art CNC milling machines further open up the possibility to link the wooden elements of such a facade through multiple mortise-and-tenon joints. Based on preliminary data, and an FE model, the author comes to the conclusion that beech wood can in fact be used for these purposes. It was proven that, when subjected to dead loads, the failure of these structures occurs perpendicularly to the fibre direction. Tests to determine the transverse tension and transverse bending strength of beech glulam revealed that, compared to spruce/fir glulam, the resistance of this still unconventional construction material is considerable. The results of the FE model confirm these findings, showing the stress peaks in the range of the transverse tension and transverse bending strengths of beech glulam. Further, a design concept consisting of two equations which can be used in future investigations, was formulated. 2 Introduction This study was conducted at the Karlsruhe Institute of Technology (KIT) with the objective of investigating the load bearing capacity of a beech glulam transom (see Figure 1) with multiple mortise-and-tenon connections (see Figure 2). The focus was on the strength of the transom to tensions perpendicular to the grain and the corresponding failure mechanisms that occur due to dead loads (glass loads and own weight), providing a basis for further investigations. The expanding use of three-layer glazing to improve the energy efficiency of buildings and the resulting increasing masses of the window panes intensify the risk of failure of the connections between mullions and transoms. This could make the material wood competitive in comparison to alternative construction materials like steel, even when having high impacts on the supporting structure. So far, connections as used in traditional carpentry are relatively rare in wood-glass facades. Multiple mortise-and-tenon joints themselves are not an innovation but in the last few decades it was more economical to use industrially produced connectors than to hire specialised manpower with the skills to build complex and precise shapes out of the wood. In the last years new technology became available, particularly Computerized Numerically Controlled (CNC) milling machines, which are designed for that task. 1
3 Contextualisation In recent decades, the importance given to the forest conversion in Europe has grown. Forest conversion consists in the regeneration of forests with site-adapted tree species, mostly in a semi-natural way because of economical reasons. Nowadays approximately 73% of German forests consist of mixed stands, according to the Federal Ministry of Food, Agriculture and Consumer Protection (2011b). The trends towards a sustainable and nature-oriented forest management led to a strategy of transforming the coniferous monocultures into more stable deciduous ecosystems (de Goede cit. in Nabuurs et al., 2003, p. 26). Moreover, it seems that the limits of sustainable use of coniferous wood are gradually being reached. In central Europe deciduous wood is clearly underutilised and the forest conversion, required due to the climate change, is well underway, increasing the stocks of non-coniferous trees. These facts result in an increasing amount of available deciduous wood and a decreasing production of coniferous wood. This way traditional markets are eliminated and it is therefore necessary and important to develop new products for and adapt existing products to this natural resource. In order to tap into the potential of non-coniferous wood, the timber, pulp and paper industries are called on to develop further innovative and resource-saving areas of use. (Federal Ministry of Food, Agriculture and Consumer Protection, 2011a, p. 13) Being the largest segment of deciduous wood in Germany it is especially interesting to investigate the potentials of beech wood. In this context, one field of use could be the construction industry. 4 State-of-the-art mullion-transom connections In general there are two different solutions for the connection between mullions and transoms: plugged connections (Figure 1a) and slide-in connections (Figure 1b). For plugged connection systems the first mullion is put into place and the corresponding transoms are plugged into the mullion. After that the next mullion is plugged onto the previously connected transoms and the next set of transoms is inserted, etc.. This makes the assembling process difficult. The company Schindler Fenster & Fassaden GmbH suggests a modular concept where all connections are plugged. In that case, each pre-fabricated module consists of two half mullions and the corresponding transoms. When mounted, the two half mullions that stand side by side are connected through the vertical profiles. Apart from being easy to set up, with this solution it would even be possible to screw through the mullions in to the grain side of the transoms to secure the transoms in the direction of their own axis. These screws would be invisible to the user. For slide-in connection systems all the mullions are mounted, and then the corresponding transoms are slid in orthogonally to the transom and mullion axis. This type of connection is convenient because the transoms do not require any additional, support during the assembly and can be slid into the already set up mullions. In addition, no extra securing means are required for loads parallel to the transom axis. However, special locking means are required for wind loads, resulting in relatively complex connector configurations. Most of the connections between mullions and transoms are made through connection elements like bolts, dowels or screws. In contrast to that, multiple mortise-and-tenon joints, which are plugged connections, can work without these connection elements. Only a safety element to prevent the tenons from being pulled out of the mortises could be necessary. 5 Preliminary tests Prior to this study, preliminary tests were conducted with mullion-transom structures made of beech glulam that were connected through mortises and tenons (see Figure 2) and secured in the direction of the transom axis with the aid of steel bolts. The average carried load was Q = 11, 04 kn with a deflection less than 4 mm for all tests. The failure of the multiple mortise-and-tenon joints occurred perpendicularly to the grain, more specifically in two regions: in front the first tenon (hereinafter called zone A) and at the lower edge of the slot that supports an aluminium profile (hereinafter called zone B). Figures 3a and 3b exemplify these types 2
(a) Walter Lang mullion-transom connection system (Z-9.1-688 2008). (b) RAICO mullion-transom connection system (Z-9.1-621 2011). Figure 1: Examples of plugged-in and slide-in connection systems. of failure. It seems that the failure in zone A occurred as a transverse bending failure that came from the torsional moment which was originated by the glass load. The failure in zone B was a transverse tension failure. Glass load Holes for steel bolts Screws Figure 2: Beech wood transom cross-section with multiple mortise-and-tenon connection and aluminium profile to support the glass panes. The results of the preliminary tests can be compared to tests that were made at the KIT with stateof-the-arts connection systems from different manufacturers (see Table 1). The comparison between the tests of the state-of-the-art connection systems and the multiple mortise-and-tenon joint can only be made within certain limitations because they were not obtained with the same test configurations. The beech glulam multiple mortise-and-tenon connection is stiffer than the other connection systems. It can also be seen that the beech glulam connections had a very high strength. 3
(a) Failure of specimen V1-2 in zone A. (b) Failure of specimen V1-1 in zone B. Figure 3: Failure pattern at the connection cross section of the tested transoms. Connection system Experimental strength (dead loads) Transom size Qav in δav in (h/b/l) in mm kn mm Walter Lang 50/260/845 6,85 6,96 Seufert-Niklaus 50/160/845 7,93 5,41 Knapp 50/180/500 8,78 5,54 RAICO 50/260/850 11,43 8,82 Hoffmann 50/150/850 3,32 5,12 Table 1: Experimental loads that were carried by different connection systems. 6 Tests Having in mind the results and specially the failure modes of the preliminary tests, two different types of tests were made with beech glulam: four point transverse bending tests (see Figure 4a), to determine the transverse bending strength and transverse tension tests (see Figure 4b), to determine the transverse tensile strength. The four point transverse bending tests were made based on the directives described in DIN EN 408 (2012), section 10 and 19 and the transverse tension tests were based on section 16 and 17 of the same standard. For the transverse tension tests the loads were applied with the aid of screws, instead of metal plates. The use of screws implies a stress peak in the region of the screw tip, making this point a predetermined breaking point. Some samples had to be glued together to get specimens with sufficient length for the tests. Those specimens often failed due to problems with the glue joint. This way, the tests series with those specimens are not further referred in this paper. Within the transverse bending test series V1 it was interesting to observe that the failure always occurred in layer 5, except for the last tested specimen (V1.9.2), which failed in layer 7. For all of these tests the failure occurred in radial direction, parallel to the medullary rays (see Figure 5), which is the weakest direction of wood. The results of test series V1, especially the tests with bsp hsp = 70 13 mm2, show that within the same beam/transom, and even within the same layer, there is a great dispersion of the strength (from 6, 60 N/mm2 to 13, 59 N/mm2 ) due to the anisotropic structure of wood. Within the transverse tension tests, the failure often occurred in the region of the screw tip where the stress peak occurred (see Figure 6). This means that, in those cases, the measured strengths are lower than the ones that could have been obtained with an optimized test set-up without stress peaks. The pre-drilled holes of series V2 were not straight. This flaw had no visible negative impact on the strength of the specimens when compared to test series V3. In fact, when observed separately, the average strength of series V2 is higher: ft,90,v 2 = 7, 07 N/mm2, compared to ft,90,v 3main = 5, 67 N/mm2 for the main series of V3. 4
(a) Transverse bending. (b) Transverse tension. Figure 4: Transverse bending and tension test set-up. Figure 5: Failure of specimen V1.3. Figures 7a and 7b, show the force-deflection curves of one test of series V2 and V3, respectively. It can be observed that the effect of the not straight pre-drilled holes is a pronounced differential deflection of the front- and backside of the specimen. In fact it can be seen in Figure 7a that the backside (blue line) starts with negative deflections, which lead to compressions. The front side (red line) starts with pronounced tensions. This behaviour can be explained with the tendency of the screws to become straight, introducing a moment which causes tension on one side and compression on the other. Moreover, it can be seen in both Figures that the average force deflection curve is approximately linear and that the failure is sudden (brittle). This last observation was made in all other tests and stands for both tension, as well as bending tests. As explained before, the force-deflection curves were approximately linear for both the tension and bending tests. Figure 8a emphasizes this fact, showing the results for all the bending tests of series V1. This Figure also shows that, as can be expected, the slope varies with the cross-section dimensions (black lines: b sp = 70 36 mm 2 ; red lines: b sp = 30 36 mm 2 ;green lines: b sp = 70 13 mm 2 ). Nevertheless, as can be seen on Figure 8b, the slope of the stress-strain curves is very similar, even with a high variation of the maximum strength. 5
Figure 6: Failure of specimen V2.2 due to the stress peak at the screw tip. (a) Force-deflection curve of test V2.3. (b) Force-deflection curve of test V3.1. Figure 7: Examples of force-deflection diagrams of test series V2 and V3. (a) Force-deflection curves of tests series V1. (b) Stress-strain curves of test series V1. Figure 8: Results from test series V1 displayed graphically. The specimens of test series V7 and V5 were made of five different beams of strength class GL40h. The following average results were obtained: Transverse tension test series V5: ft,90,av = 5, 92 N/mm2 Et,90,av = 1254, 25 N/mm2 Transverse bending test series V7: fm,90,av = 10, 84 N/mm2 Em,90,av = 1305, 29 N/mm2 These results can be compared to the values given in the general technical approval for beech Glulam (Z-9.1-679 2009): ft,90,k = 0.5 N/mm2 E90,mean = 690 N/mm2 E90,05 = 550 N/mm2 It can be seen that the average traverse tension strength is ten times higher than the one that is allowed according to the general technical approval. It is also worth a notice that even the lowest transverse 6
tension strength f t,90 = 2.20 N/mm 2, obtained for test V8.5 which had a considerable imperfection, is four times higher. The average modulus of elasticity that was measured in the tests is approximately twice as high as the one given in Z-9.1-679 (2009). 7 Finite element model The model built for this investigation was a simulation of a transom with multiple tenons. It was a three-dimensional model because the focus was on a singularity, the connection, where the behaviour can not be simplified. The glass load was applied to a metal profile on two points, to simulate the effect of the two glass supports. The profile had a T -shaped cross-section and was supported by a continuous slot in the transom. The load was transmitted from the profile to the transom by screws and through contact, leading to a local compression. Since the interaction between profile and slot was similar to the behaviour of a notched beam, transverse tensions occurred. The screws through both flanges of the profile transmitted forces through shear/hole bearing and pulling out. To reduce the processing time of the simulation, the symmetry of the problem was taken into account. This way, considering the symmetry conditions, only half of the transom had to modelled. The fact that the transom to be studied was made of glulam was not explicitly taken into account in the model, neglecting that in general the glue joint should be more resistant than the wood. This was a conservative assumption. In the tests the transoms were supported on each side by tenons that were slid into mortises (the slide-in direction is parallel to the z-axis). In x and y direction the tenons were held by the mortises. As a simplification, the mullions, and consequently the mortises, were not modelled and their influence on the behaviour of the tenons was simplified by fixing two lines of each tenon in x and y direction. The model ignored the effect of the friction between mortises and tenons, putting no restrictions in z direction. To use the advantage of the symmetry of this model, only half of the transom (400 mm) was modelled and the nodes that were located on the plane of symmetry were fixed in z direction. In general, it is common practice to support glass panes only at two points/regions (glass supports), to have a defined static system. This was simulated by applying the force at a point located l a = 100 mm from the transom supports (tenons). The applied load for the main run was the average load that was obtained from the preliminary tests with beech glulam transoms was applied (Q = 11040 N). The resulting stresses of the FE modeling are shown in Figure 9a and Figure 9b. As can be seen the results are in the range of the results obtained from the transverse bending and tension tests: σ x,max = 11, 13 N/mm 2 can be compared to the obtained f m,90 = 10, 84 N/mm 2 and σ y,max = 13, 76 N/mm 2 to f t,90 = 5, 92 N/mm 2. The stresses in zone A and B (as defined in Figure 3) are the highest transverse tensions in x and y direction, as can be seen in Figures 9a and 9b respectively. This can be interpreted as a realistic result since it is confirmed by the failure modes of the preliminary tests. One reason for the difference in the values might a non-ideal element size and the fact that the the model is based on the linear elastic theory, which is not an ideal explanation for singularities. Another possibility is that the stressed volume of the tests is much higher than the one that would be admitted in practice. This way, and admitting that the size effect has to be taken into account for the occurred types of failure, the 11.04 kn that were applied to the FE model would have occurred for higher localized stress peaks. A third possibility could be that the average maximum load of the preliminary tests happened after a partial failure of the transom. This would mean that the initial failure would have happened for lower loads and that the stresses that this lower load would cause in the FE model would be lower too. 8 Design concept According to the preliminary tests and the main run of the FE model, zones A and B seem to be the ones that limit the strength of the whole transom. With the objective of formulating a design concept that enables the prediction of the stresses in zones A and B, the FE model was run to show the variation of the stresses when varying three parameters: the applied load (Q), the distance of the load application point to the support (l a ) and the slot depth (b n ). The following expressions were deduced: 7
(a) Stresses in x direction (σ x) in the connection cross-section. (b) Stresses in y direction (σ y) in the connection cross-section. Figure 9: Stress distributions in the connection cross-section. Zone A (Transverse bending) σ m,90 = 25 Q (2 b st + b t b n ) 8 h 2 k A (1) Zone B (Transverse tension) σ t,90 = Q l a k B (2) In those expressions k A = 1/(d z ) and k B = 1/(d x c z ). To adapt the calculated values to the ones obtained in the modelling, values of k A = 0, 012 mm -1 and k B = 0, 15 mm -1 were chosen. The following graphs (Figures 10a, 10b and 10c) show the results of the different runs that were made with the FE model. In the graphs σ max are the maximum stresses at zone A or B obtained from the FE model and σ calc are the stresses obtained through Equations 1 and 2. In the graphs of Figure 10, the solid lines show the stresses that were obtained from the FE model while the dashed lines show the stresses that the previously deduced expressions predict for k A = 0, 012 and k B = 0, 15. The blue lines refer to zone A and the red lines refer to zone B. It is noted that when l a = 25 mm, the maximum σ x occurs in zone B instead of zone A. Also, the the maximum σ y occures in zone C (below the first tenon) for l a = 150 mm and l a = 175 mm. For l a = 200 mm the the maximum σ y occurs again at the inner edge of the slot but in z direction it occurs at the load application point. It is mentioned that when b n = 10 mm, the maximum σ x occurs in zone B instead of zone A. This leads to the conclusion that the slot should not be to deep. It also has to be mentioned that the obtained expressions only apply to the given geometry and the range of the variations. The values that were given to k A and k B would result in d z 83, 33 mm and d z c z 6, 67 mm. It could be that these values occur due to other geometrical factors that were not taken into account. 8
(a) Variation of the stresses in zones A and B when varying Q. (b) Variation of the stresses in zones A and B when varying l a. Figure 10: Variation of the stresses in zones A and B. (c) Variation of the stresses in zones A and B when varying b n. 9 Discussion of the research results It was shown that beech wood is a material that could avoid bottlenecks in the supply of construction wood, which might result from the forest conversion. In fact, beech wood could even open new markets for wood since it is stronger and stiffer than the usually used coniferous wood. In this context it could be a solution for the challenge that the increasing use of three layered glazing posed to the design of wood-glass facades. The availability of CNC milling machines opens up the possibility to establish the connection between mullions and transoms through multiple mortise-and-tenon joints instead of relying on other more or less complex specifically designed connection means. If the milling process is relatively economical, a decrease of the global cost of the structure can be possible due to the reduction, or even elimination, of additional connection means. The results of tests that were made prior to this study show that the use of beech glulam mullion-transom structures with multiple mortise-and-tenon joints can be an alternative to state of the art wood-glass facades. These tests also reveal that, when subjected to dead loads, the failure of these structures occurs due to tension stresses that occur perpendicularly to the fibre direction (transverse tension stresses). With those results in mind there were made test to determine the transverse tension and transverse bending strength of beech glulam. The results revealed that, compared to spruce/fir glulam, considerable transverse tension and transverse bending loads can be carried by beech glulam. In fact, the transverse tension and transverse bending strengths of spruce/fir are often ignored for safety reasons. It seems that this would not be necessary for beech, especially when used in wood-glass facades, due to the high visual requirements that reduce the presence of imperfections (cracks, knotholes, etc.). The FE model showed that there were stress peaks at the zones that failed in the preliminary tests. It could also be seen that the stress peaks were in the range of the transverse tension and transverse bending strengths that were obtained in the preliminary tests. The difference in the values could be explained by a non-ideal size of the finite elements and the fact that the FE model is based on the linear elastic theory. 9
Another reason could be the size effect associated to the problem of the need to determine the stressed volume in the affected singularities. It was observed that the variation of different parameters caused variations of the stresses at the two prior identified failure zones. These variations had clear tendencies that helped with the formulation a design concept for the connection. The design concept consists of two equations (Equation 1 and 2) that were deduced through the equilibrium of the forces in the transom. However, for each of the equations a constant had to be defined to adapt the calculated values to the ones obtained from the FE model. The values that were admitted for the constants might be explained by geometrical parameters that were not taken into account in this study. Another approach to the problem through fracture mechanics would result in similar expressions to the ones obtained but in an improved explanation for the admitted values since it gives a better explanation for the fracture behaviour. Nevertheless the equations serve to estimate the stresses at zones A and B for the given geometry and the range of the variations. In further studies, the effects of the wind loads on the transom should be studied. With those results it would be interesting to study the configuration and distribution of the tenons in the connection section to optimize the resistance to the different acting loads. For an extended practical use of beech wood in the construction industry it would be necessary to amplify the number of transverse tension and bending tests to obtain more representative values. These tests could also serve to give a more precise look at the variation of the strength for varying stressed volumes (size effect). Since the failure occurs due to transverse tension and wood is weakest perpendicular to the fibres, one can think about a means to reinforce that direction, in particular in the zones where the stress peaks occur. One option could be the use of longer screws to secure the aluminium profiles. If those screws would be nearly as long as the transom is wide, and had an unthreaded shank, the axial forces in the screws would be transmitted to the less stressed part of the transom. The treadles shank would allow the profile and the front part of the transom to be pulled against the back side of the beam, resulting in compression of the front side. This way the front part could be considered as being pre-stressed. This would have to be done for both the top and bottom screws to prevent possible errors during the assembling process. Thus it would also be necessary to pay attention to the peak compression at the bottom of the transom. 10 References 10.1 Bibliography Federal Ministry of Food, Agriculture and Consumer Protection (Nov. 2011a). Forest Strategy 2020. Sustainable Forest Management: An Opportunity and a Challenge for Society. Governmental Report. Bonn (Germany): Federal Republic of Germany. Federal Ministry of Food, Agriculture and Consumer Protection (Mar. 2011b). German Forests. Nature and Economic Factor. Governmental Report. Berlin (Germany): Federal Republic of Germany. Nabuurs, G. J., M. J. Schelhaas, A. Ouwehand, A. Pussinen, J. van Brusselen, E. Pesonen, and A. Schuck (2003). Future wood supply from European forests. Implications for the pulp and paper industry. 10.2 Cited Standards and General Technical Approvals DIN EN 408 (2012). Norm DIN EN 408. Timber structures Structural timber and glued laminated timber Determination of some physical and mechanical properties. Brussels, Belgium: Comité Européen de Normalisation (CEN). Z-9.1-621 (May 2011). Allgemeine bauaufsichtliche Zulassung Z-9.1-621. RAICO Pfosten-Riegel-Verbinder für Holzfassaden. Berlin (Germany): Deutsches Institut für Bautechnik. Z-9.1-679 (Oct. 2009). Allgemeine bauaufsichtliche Zulassung Z-9.1-679. BS-Holz aus Buche und BS-Holz Buche-Hybridträger. Berlin (Germany): Deutsches Institut für Bautechnik. Z-9.1-688 (Nov. 2008). Allgemeine bauaufsichtliche Zulassung Z-9.1-688. Lang Posten-Riegel-Verbindungen Holz-Glass-Fassaden. Berlin (Germany): Deutsches Institut für Bautechnik. 10