Engineering Structures 33 (2011) 30433053

Contents lists available at SciVerse ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Elastic and ductile design of multi-storey crosslam massive wooden buildings under seismic actions
M. Fragiacomo a, , B. Dujic b , I. Sustersic b
a b

Department of Architecture, Design and Urban Planning, University of Sassari, Palazzo del Pou Salit, Piazza Duomo 6, 07041 Alghero, Italy CBD d.o.o. - Contemporary Building Design Company, Lopata 19g, 3000 Celje, Slovenia

article

info

abstract
The paper discusses the seismic design of multi-storey buildings made from cross-laminated timber panels (crosslam). The use of seismic analysis methods such as the modal response spectrum and the non-linear static (push-over) analysis is discussed at length, including issues such as the modelling of crosslam walls and connections, the evaluation of the connection stiffness, and the schematization of floor panels. It was found that it is crucial to account for the flexibility of the connections (hold-downs and angle brackets) between upper and lower walls, since otherwise the vibration periods of the building would be underestimated. The basics of capacity design to ensure the attainment of ductile mechanisms in crosslam timber structures under seismic actions are presented. The ductile failure mechanism is characterized by plasticization of connectors (hold-downs, angle brackets and screws) between adjacent wall panels and between panels and foundations. The crosslam panels and the connections between adjacent floor panels must be designed for the overstrength of the connectors to ensure that they remain elastic during the earthquake and the ductile failure mechanism is attained. Based on the results of preliminary quasi-static cyclic tests, a value of 1.3 was found for the overstrength factors of hold-downs and angle brackets. A case study multi-storey crosslam massive wooden building was then analysed using the non-linear push-over analysis as implemented in the N2 method recommended by the Eurocode 8. The building was modelled using shell elements and non-linear links to schematize the hold-downs and angle brackets. The building ductility, calculated from the bilinear curve equivalent to the actual non-linear push-over curve, was then investigated. Such a quantity, defined as the ratio of the displacement at the near collapse state and the maximum elastic displacement of the top floor, was found to rise from 1.7 to 2.5 when ductile instead of brittle hold-downs and angle brackets are used. Furthermore, the maximum peak ground acceleration the building can resist raised from 0.2g to 0.4g , demonstrating the importance of using ductile connectors in seismic design. 2011 Elsevier Ltd. All rights reserved.

Article history: Available online 29 June 2011 Keywords: Cyclic tests Cross-laminated panels Ductility Earthquake design Multi-storey buildings Push-over analysis Seismic performance Timber Wood

1. Introduction Different methods can be used for the design of seismicresistant buildings. The most basic approach would be to evaluate the forces induced on a building by an earthquake with a high return period and design the structure in elastic phase. Since statistically the chance of a high intensity earthquake occurring during the lifetime of a building (in most cases 50 years) is not particularly high (about 10%), the elastic design leads to significant overdesign of the building elements. For this reason the elastic approach is generally used only in low to moderate seismicity regions. The alternative design approach is based on the principles of ductile design. A ductile structure is able to dissipate energy during the seismic event by undergoing through plastic

Corresponding author. Tel.: +39 079 9720418; fax: +39 079 9720420. E-mail addresses: fragiacomo@uniss.it (M. Fragiacomo), bruno.dujic@cbd.si (B. Dujic), iztok.sustersic@cbd.si (I. Sustersic). 0141-0296/$ see front matter 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.engstruct.2011.05.020

deformation. One of the advantages is the possibility to survive high intensity earthquakes as long as the displacement demand in the ductile parts of the structure does not exceed the displacement capacity. The ductility also allows more economical structures to be built as the design seismic actions can be reduced depending upon the ductility ratio [1]. Such an approach is generally followed for building design in medium to high seismicity regions. Current codes of practice [2] suggest two different approaches for design of ductile structures in earthquake-prone regions. The first approach, well known and widely used, is referred to as the Force-Based Design (FBD) method since it mainly focuses on designing the strength of the structure [1]. The objective is the evaluation of the behaviour factor q, which is employed to transform the elastic response spectrum into a design spectrum. In this way a non-linear structure can be designed using a linear-elastic static or dynamic (modal response spectrum) analysis under seismic action, with the structural ductility only implicitly considered when evaluating the behaviour factor q.

3044

M. Fragiacomo et al. / Engineering Structures 33 (2011) 30433053

Fig. 1. Failure modes for steeltimber (left) and timbertimber (right) connections, notations according to EC5-1-1 [15].

The second approach, which explicitly refers to the structural ductility in addition to the strength, is based on a Non-linear Static Analysis (NSA) procedure [1]. The purpose of this approach is the evaluation of the actual structural response mainly in terms of ductility demand and, hence, ultimate displacement induced in the structure by the earthquake ground motion [3]. A number of different methods have been proposed, including a modified version of the N2 method [4], which has been adopted by the new Italian regulation [5] and by the Eurocode 8 (EC8) [2] and is discussed more in detail in the following sections. An important issue in seismic design is the identification of suitable, ductile failure mechanisms. Capacity-based design must then be used [6] to ensure that brittle failure mechanisms will not occur. For multi-storey timber buildings with lightframe construction, the most ductile failure mechanism is shear in the nailed/screwed connections between the sheathing and the frame [7]. All other failure mechanisms are fairly brittle and, therefore, should be avoided by designing the corresponding members (hold-downs, bolts, timber studs, timber plates, plywood sheathing) for the overstrength of the nailed/screwed connection. New structural systems for multi-storey timber buildings have recently been proposed in Europe and in Australasia. Unlike lightframe construction, these innovative systems can also comply with the new philosophy of the Damage Avoidance Design, according to which a building should not only survive an earthquake at ultimate limit state, but be easily repairable and useable in a short time so as to reduce the disruption and the associated cost to a minimum [8]. In Australasia, hybrid systems made of laminated veneer lumber (LVL) walls and frames prestressed with unbonded tendons used together with energy dissipaters were developed [9,10]. In such systems the prestressed tendons keep the timber elements connected together and ensure that after an earthquake event, due to the restoring force, the structure returns to its initial position with little, if not, residual deformation. The energy dissipaters ensure proper dissipation with reduction in displacement demand to the structure. In Japan, multi-storey buildings constructed from prefabricated walls and slabs made of cross-laminated timber (crosslam) produced in Europe were subjected to full-scale shaking table tests. The buildings survived high intensity earthquakes with limited structural damage [11,12]. Furthermore, it was found that such systems, with a proper choice of connection details and panel sizes, can dissipate a significant amount of energy, mostly in the connections between wall panels, and between panels and foundation, leading to the possibility of carrying out static and modal response spectrum analysis assuming a behaviour factor q = 3 [13]. Despite the extensive use of the crosslam technology, there are few provisions for the seismic design of this system in current codes of practice such as the Eurocode 8 [2]. In addition to the lack of any value for the overstrength factor, there is no suggestion on the choice of the ductile failure mechanism, nor indication on the way the NSA can be carried out according for example to the N2 method. This paper provides some answers to the aforementioned queries, namely it proposes some values of the overstrength factors for typical connections used in crosslam construction based on the results of experimental cyclic tests, and presents the use of the

NSA in the design of a simple crosslam building. Further important issues such as the numerical modelling of crosslam panels and connections, the influence of the connection ductility on the global ductility of the building, and the seismic performance of the whole building are critically discussed in the paper. 2. Choice of an appropriate ductile failure mechanism for multi-storey crosslam buildings 2.1. Provisions for a ductile connection between crosslam panels Crosslam panels are solid slabs made from layers of timber boards with the adjacent layers glued at a right angle. Advantages over glued-laminated elements include improved stability in both directions, which is of particular importance for 2D elements, and the possibility to use medium to low quality timber. This technique was developed in Europe about 15 years ago and is nowadays extensively used. Crosslam buildings are erected by connecting crosslam walls and slabs together using angular metal bracket connectors, usually nailed or screwed to the timber, and self-tapping screws. The connections between adjacent panels are usually realized using 8 mm diameter screws placed at a distance of 300 mm centre to centre [14], where the penetration of the screw in the second wall is usually equal or longer than the penetration in the first wall. The foundationpanel and upperlower panel connections are made from nailed or screwed angular metal brackets, and in some cases also by hold-down connectors, nailed and bolted to the timber and reinforced concrete foundation at the first storey. The wall-to-floor panel connections are often made of 8 mm diameter screws placed at a distance of 300 mm centre to centre. Different failure mechanisms can occur in a crosslam building, and only few of them are ductile. Failure of the wall panel due to in-plane loading (shear, bending and axial force) is mostly brittle and should be avoided by designing the panel for the overstrength of the ductile elements (the connectors). The connectors between adjacent panels and between panels and foundation, however, may behave ductile or brittle under shear deformation depending on whether plasticization of the steel fastener (screws and nails) is attained or not. Therefore, according to the notation of the Eurocode 5 Part 1-1 [15], see Fig. 1(left), failure mode b is regarded as ductile (one plastic hinge formation), whilst mode a that occurs with shorter and/or thicker fasteners and has no plastic hinge formation in the steel fastener is regarded as brittle. The most desirable failure mechanism would be mechanism e, where two plastic hinges are formed in the fastener. However, this would only be possible if thicker steel plates were used, which does not apply to typical brackets and hold-downs. For screwed connections between adjacent panels, mode f is the most ductile and therefore the most desirable. Modes d and e are also ductile though only one plastic hinge is formed (Fig. 1(right)). 2.2. Definition of the overstrength factor An important issue in seismic design is the identification of suitable, ductile failure mechanisms. Capacity-based design must then be used [6] to ensure that brittle failure mechanisms will

M. Fragiacomo et al. / Engineering Structures 33 (2011) 30433053

3045

Fig. 2. Backbone curves of a crosslam wallbracket (BMF105) connection tested under cyclic loading of tension (left) and shear (right) according to EN 12512 [26] for 40 and 60 mm long nails with 4 mm diameter. Solid/dashed lines refer to the response during the first/third cycle, respectively.

not occur. This is achieved by designing the structural elements which may fail in a brittle manner under an increased seismic demand Ed given by
Ed = Rd od Ed

(1)

where Ed is the seismic demand on the structural element calculated using the design spectrum as reduced to allow for the ductile behaviour of the structure, od is the overdesign factor, given by

od =

(2) Ed where Fd is the design strength capacity of the ductile structural element, and Rd is the overstrength factor, given by

Fd

Rd =

(3) Fd where F0.95 is the 95th percentile of the actual peak strength capacity of the ductile structural element [16]. Whilst the overdesign factor depends on the rounding carried out in the actual design, the overstrength factor depends on the type of material and structural detail. The Eurocode 8 [2] provides the values of the overstrength factors for steel and reinforced concrete structures, which are in the range 11.3, however there is no provision for timber structures. Some indications can be found in the New Zealand timber standard [17], where a value of 2 is suggested for the overstrength factor. 2.3. Evaluation of the overstrength factors for timbertimber connections This section discusses the results of experimental cyclic tests performed on timber to timber connections made of angular brackets and screws. The connection behaviour governs the seismic response of timber buildings and therefore plays a major role in a seismic analysis. Due to the lack of information in standards and codes of practice, reference to only experimental tests has been made in this section. In earthquake design of crosslam buildings, a suggestion is given to design the nailed and screwed connections for the ductile failure modes discussed above. Fig. 2 shows the results of cyclic tests carried out by Dujic and Zarnic [18] on cross-laminated wall connections with metal bracket loaded in tension (left) and shear (right). Ten 4 mm diameter nails were used to connect the angle bracket to the crosslam panel, which had three layers of timber boards with 30, 34 and 30 mm thickness. The connection with 60 mm long nails was designed so as to ensure failure mode b is achieved. The ductile behaviour can be clearly appreciated an average ductility ratio of about 8 for tension and 6.5 for shear is obtained. In this evaluation, the yield point was computed by following the so-called Yasumura and Kawai procedure [19], which will be discussed in more detail in an ensuing section. Conversely the use of the 40 mm long nails leads to a rather brittle behaviour, with failure occurring in the bottom row of nails (see Fig. 3) for plug

F0.95

Fig. 3. Plug shear failure in the bottom row of nails of the bracketpanel connection loaded in tension.

shear. This plug shear failure was a consequence of the nails not penetrating deep enough in the middle layer of the wall. Hence a recommendation is given to use nails or screws at least 60 mm long to achieve a ductile failure mechanism and avoid the brittle shear plug failure. For the same specimens, the overstrength factors were computed using Eq. (3). The design strength capacity Fd was calculated by dividing the characteristic experimental strength F0.05 by the strength partial factor M , assumed to be equal to one according to the Eurocode 8 [2] for dissipative timber structures. The 5th percentile F0.05 was estimated by assuming a students t -distribution on the basis of the maximum experimental shear and uplift strength Fmax among the three specimens tested. The same procedure was used for the calculation of the 95th percentile, F0.95 . Since several different types of connections were tested, the number of specimens per setup was rather small but so was the scatter of results and also the standard deviation . The overstrength factors from shear tests are 1.3 for 60 mm and 2.2 for 40 mm nails (see Table 1). The results from uplift give an overstrength factor Rd of about 1.9 for 40 mm nails and 1.2 for 60 mm nails, the latter value being based on the results of only two specimens (see Table 1). Although too few specimens were tested to derive a final value, the results of this preliminary investigation provide a first approximation for the overstrength factor to be used for the design of the components with brittle failure. An experimental investigation performed by the same authors [18] on screwed connections between perpendicular panels involved 5 specimens with 8 screws 160 mm long with an 80 mm long threaded part that were cyclically tested under shear loading. Fig. 4 shows that the screw exhibited a ductile behaviour since either ductile mechanism d or e according to Fig. 1 had formed. If calculating the failure modes according to the Johansen equations and using the method proposed by Blass and Uibel [20] to account for the layered structure of the crosslam panels, mode f failure should occur. The difference among the predicted strengths for modes d, e and f is, however, very small. The overstrength factor calculated according to Eq. (3) using a students

3046

M. Fragiacomo et al. / Engineering Structures 33 (2011) 30433053

Table 1 Experimental results for BMF 105 angular brackets with ten 4 mm diameter nails 40 and 60 mm long and Wuerth Assy II 8 160/80 self-tapping screws. Number of specimens tested BMF 105 angular brackets with ten 40 mm nails BMF 105 angular brackets with ten 60 mm nails Wuerth Assy II 8 160/80 self-tapping screws Shear Uplift Shear Uplift Shear 3 3 3 2 5 Mean value of peak force (Fmax ) 13.5 14.8 15.0 23.1 4.7 Standard deviation of peak force 1.73 1.51 0.59 0.31 0.52 5th percentile F0.05 8.4 10.4 13.3 21.2 3.6 95th percentile F0.95 18.5 19.3 16.8 25.0 5.8 Overstrength factor Rd (Eq. (1)) 2.12 1.85 1.26 1.18 1.63

Fig. 4. Typical failure of Wuerth Assy II 8 160/80 self-tapping screws due to low cycle fatigue (top left) and cyclic response of a specimen with 1 screw loaded in shear, including backbone curves of the first 3 cycles.

t-distribution on the basis of the maximum strength Fmax among the five specimens tested is Rd = 1.6 (see Table 1), which is recommended for the overstrength factor of screwed connections between perpendicular crosslam panels. 2.4. Additional considerations on the failure mechanism of multistorey crosslam buildings under seismic actions Additional issues are whether the plasticization should be allowed also at the upper storeys and within the floor diaphragm. Since the damage observed in shaking table tests of full-scale multi-storey crosslam buildings was little [11], unlike lightframe construction it is suggested that plasticization should be allowed in the walls at all of the storeys. However, since the plasticization of the floor diaphragm may lead to a number of problems including torsional rotation of the building, influence of higher modes of vibrations, and significant increase in structural damage, it is suggested that the connections between adjacent floor panels are considered as non-ductile and designed for the overstrength of the bracket and wall-to-wall connections. A typical connection between adjacent floor panels is obtained with 50 mm overlap of panels (step joint) where each panel is cut in the middle of its depth (Fig. 5a). 6 screws are inserted usually perpendicularly at a distance of 300 mm centre to centre. Another option is the use of a plywood strip overlapping on the two panels (Fig. 5b) [11,21], which leads to increased ductility but also to reduced stiffness. A

desirable connection, with high in-plane strength and stiffness, is displayed in Fig. 5c, where screws take most of the load in the favourable axial (tensile) direction. If gluing is applied or screws are drilled at 45 angle also parallel to the panel length, a very stiff joint could be achieved for in-plane axial and shear forces. The current version of the Eurocode 8 Part 1 [2] does not provide any value for the overstrength factors, although it requires the use of capacity design and, in particular, that the overstrength requirement applies especially to: (i) anchor ties and any connection to massive sub-elements; (ii) connections between horizontal diaphragms and lateral load resisting vertical elements. It should be pointed out that the last provision is questionable for crosslam construction since the connections between inter-storey plates and walls are ductile and practically without any possibility of a brittle failure. However, to ensure the ductile mechanism, the use of crosslam elements (walls and plates) with a thickness of at least 90 mm is recommended, together with a screw length adjusted to the thickness of the floor plate so that either failure mechanism d or e (according to Fig. 1) can form. There is also a possibility of a more ductile mechanism f and the formation of two plastic hinges with even greater energy dissipation. Future studies should investigate the possibility to achieve this very ductile failure mechanism (also for angular brackets), which could enable designs with even higher behaviour factors q. Regarding the design of the panels for the overstrength of the connections, this is of particular importance when the panel is weakened by an opening. In this case, the failure may occur for stress concentration due to bending moment around the corners of a large opening (frame effect). For example, in an experimental study carried out by Dujic et al. [22], strength and stiffness were observed to depend upon the opening factor defined by Yasumura [23] for lightframe walls: r = Li . = 1+ H Li + Ai 1 H

(4)

Variables in Eq. (4) and Fig. 6 are the panel area ratio (r ), the height of the wall element (H ), the length of the wall element (L), the length of full height wall segments ( Li ), the summation of opening areas of the wall segment ( Ai ), the ratio of openings in the wall element Ai /HL), and the ratio of full wall ( = segments ( = Li /L). Based on the results of recent investigations, Dujic et al. [22] suggested that for panel area ratios greater than r = 0.4, the openings do not affect the strength and stiffness of the timber panels, and mostly brittle failures cannot occur in crosslam panels before the ductile mechanism of the steel brackets activates.

Fig. 5. Different types of connections between adjacent floor panels.

M. Fragiacomo et al. / Engineering Structures 33 (2011) 30433053

3047

Table 2 Blass and Fellmoser coefficients for in-plane loaded crosslam panels [24] and effective strength and stiffness for a homogenized panel (0/90 subscripts denote parallel/perpendicular to the grain orientation of the outer layer of the panel).

k3 = 1 1

E90 E0

am2 am4 +a1 am

k4 =

E90 E0

+ 1

E90 E0

am2 am4 +a1 am

a1 am E0 E90 ft ,90,ef = k4 ft ,0 Et ,90,ef = k4 E0

thickness of the middle layer, panel total thickness, modulus of elasticity parallel to grain modulus of elasticity perpendicular to grain fc ,0,ef = k3 fc ,0 Ec ,0,ef = k3 E0 fc ,90,ef = k4 fc ,0 Ec ,90,ef = k4 E0

Eff. strength Eff. stiffness

fm,0,ef = k3 fm,0 Em,0,ef = k3 E0

fm,90,ef = k4 fm,0 Em,90,ef = k4 E0

ft ,0,ef = k3 ft ,0 Et ,0,ef = k3 E0

The aforementioned limit value of r is generally acceptable for buildings with gravity load in seismic condition not exceeding 15 kN/m and if the walls are held down using BMF105 bracket connections at a distance of about 50 cm centre to centre. 3. Seismic modelling of crosslam timber buildings Crosslam structures are currently not explicitly mentioned either in Eurocode 8 [2] or in Eurocode 5 [15]. Due to the lack of comprehensive design guidelines, some issues that need special consideration when modelling crosslam buildings under seismic actions are discussed in the ensuing sections. 3.1. Modelling of timber panels The complex panel layout can be modelled using an orthotropic, homogenized orthotropic or homogenized isotropic material, depending on the possibilities offered by the FEM software. Blass and Fellmoser [24] proposed the Homogenised, Orthotropic plane stress Blass reduced cross Section (HOBS) method, which is based on the reduction of a multilayer to a single layer section using the coefficients k3 and k4 provided in Table 2 to modify the stiffnesses and strengths. By assuming a plane stress state, only two moduli of elasticity (E0 and E90 ), one shear modulus (G12 ) and one Poissons coefficient (12 ) need to be defined. The thickness of the panels remains the same as does the shear modulus. If the adjacent boards of individual layers are not glued along their thickness, a 10% reduction in the shear modulus is suggested. In spite of its simplicity, recent investigations demonstrated that this model provides results accurate enough for seismic design. 3.2. Modelling of floor diaphragms When performing time-history analyses of crosslam buildings, some authors [11,12] have assumed the hypothesis of rigid floor diaphragms. It should be pointed out that this assumption is not fully true. If such a simplification is made, it is advisable to prescribe gluing of adjacent floor panels or using stiffer connection with screws running at 45 angle (see Fig. 4c) so as to achieve a stiff floor diaphragm. Appropriate overstrength factors should be applied in the design of the floor panels and their connection. The modelling of the floor diaphragm is an issue that deserves further investigation through non-linear dynamic analyses to evaluate the in-plane flexibility of a floor and its influence on the building behaviour.

Fig. 6. Definition of panel area ratio.

3.3. Evaluation of the connection stiffness This is a fundamental issue in any type of seismic analysis. The modal response spectrum analysis depends on the structure vibration period and therefore on its stiffness. The connections between crosslam panels play a crucial role and therefore must be modelled accurately. Care has to be taken when modelling angular brackets and hold-downs loaded in tension with finite element links. Their stiffness must be very high in compression due to the contact, and equal to the connection stiffness in tension. Unfortunately such non-linear behaviour cannot be directly modelled in linear static and dynamic (modal response spectrum) analyses. Neglecting connections between crosslam panels, namely modelling walls with rigid connections as in concrete construction leads to large errors as it will be shown in the case study presented in the following section. Some guidance on how to evaluate the correct connection stiffness will also be given.

4. Case study: elastic seismic analysis of a multi-storey crosslam building 4.1. Geometrical properties of the case study building and types of modelling A four-storey crosslam building (Fig. 7) was modelled according to the recommendations given in the previous sections. The building has 140 mm thick 5-layer crosslam walls along its perimeter. Inside the building there are only two posts and a beam that support three adjacent 140 mm thick crosslam slabs running from wall A to wall C. Wall A is made from two separate panels, which are connected only with a beam element pinned onto the walls. The wall panels are connected at the bottom and at the top

3048

M. Fragiacomo et al. / Engineering Structures 33 (2011) 30433053

with BMF105 brackets (with ten 60 mm long, 4 mm diameter nails per bracket), which are placed at 50 cm spacing so as to resist the base shear force calculated according to the lateral force method of analysis. The building was modelled in SAP 2000 [25] using shell elements (10 10 cm mesh) in accordance with the HOBS model. The floor diaphragms were modelled as rigid. The connections among adjacent panels were schematized in different ways: (i) with rigid links (full 3D model with rigid connections), (ii) with linear-elastic springs for the top and bottom connections of walls, and without any connection between perpendicular walls at the same level (pseudo-3D model), and (iii) with linear-elastic springs for the top and bottom connections of walls, as well as connections between perpendicular walls at the same level (full 3D model with elastic connections). 4.2. Calculation of the connection stiffness An important issue is how to model the stiffness of the screwed and bracket connections. In this paper, it is suggested that stiffness, strength and ductility of the steel brackets and screwed connections are determined according to the Yasumura and Kawai procedure [19] (see Fig. 8), sometimes also referred to as the modified CEN procedure or the 10-40-90 procedure. Such a procedure was proposed for the evaluation of wood framed shear walls, which is similar to the one suggested by EN 12512 [26]. The calculation of the ultimate strength is based on the equivalence of the deformation energy, but the calculation of the elastic stiffness is slightly different from EN 12512. The yielding load is determined by assessing the intersection of two lines. The first line is drawn through the points on the loading curve corresponding to 10% and

11.2 m

6.

8.

Fig. 7. 3D view, plan and wall elevations of the crosslam building analysed (dimensions in mm).

Fig. 8. Yasumura and Kawai procedure [19] for the evaluation of strength, stiffness and ductility of a timber shear wall.

40% of the peak load Fmax . Unlike the EN 12512 procedure, the second line is drawn through the points corresponding to 40% and 90% of Fmax . The line is then moved so that it becomes tangent to the actual loading curve. The intersection of this and the former line gives the yield load and the corresponding displacement. The ratio between the yield load and its corresponding displacement gives

M. Fragiacomo et al. / Engineering Structures 33 (2011) 30433053

3049

Fig. 9. Four-storey crosslam building Comparison among different models (i, ii and iii) in terms of vibration periods (T1, T2 and T3), base shear along x (Rx) and y (Ry), and top floor displacement along x (Ux) and y (Uy).

the elastic stiffness Ky . The ultimate displacement corresponds to 80% of Fmax on the decreasing part of the loading curve. The ultimate strength Fu is calculated so that the equivalence of the deformation energies is achieved by assuming an elasto-plastic loaddisplacement curve, the area under which is marked grey in Fig. 8. When determining the elastic stiffness of brackets for a use in the 3D modal response spectrum analysis, the values of the stiffness Ky for shear and tension were calculated as described above. Each wall was then non-linearly modelled using gap elements (namely non-linear elements almost rigid in compression with null tensile stiffness) and elastic links for brackets to simulate the exact boundary conditions of contact in compression and elastic behaviour in tension and shear. The wall models were then recalibrated so that only the elastic links and no gap elements were used, and the target displacements for the non-linear and linear cases were the same under the same horizontal load. With the latter model, a modal response spectrum analysis was carried out. It must be noted that the analysis was simplified as neither the influence of vertical load on the rotational stiffness of walls nor the friction in shear were considered. Both of these simplifications result in a lower stiffness, longer vibration periods and, therefore, in potentially lower design seismic actions. It should also be noted that the elastic stiffness of the bracket evaluated according to the Yasumura and Kawai procedure should not be reduced when performing a modal response spectrum analysis. Experimental tests have shown [18] that the accumulating damage during cyclic loading does not reduce the connection stiffness in crosslam wall setups, even if it does reduce the peak load. This behaviour can be clearly observed in Fig. 2 the load curves of the first and the third load cycle are almost identical for both 40 and 60 mm long nails in the initial part of the diagram where the stiffness is defined, although the peak load differs by 15%20%. Furthermore, even the secant stiffness of the connection calculated as the ratio between the peak strength and the corresponding displacement does not markedly reduce during the cyclic test. Experimental results on full-scale buildings [11] have shown that after a seismic event only a limited amount of damage occurred, meaning that the main vibration period of the building could not have changed significantly. Since in most cases of multistorey buildings longer vibration periods result in lower seismic forces, it is proposed to conservatively refer to the initial stiffness of the connection. Even though this assumption may lead to nonconservative estimation of the displacements, this is less critical than a reduced force estimation. 4.3. Modal response spectrum analysis The following data was used for the modal response spectrum analysis of the building according to Eurocode 8: type 1 elastic

Table 3 Mass and mass radius of gyration for each floor. Mass (kg) 4th storey 3rd storey 2nd storey 1st storey 19 405 31 489 31 489 31 489 Mass gyration radius (kg m2 ) 185 156 300 461 300 461 300 461

response spectra and a rock foundation (type A soil according to EN 1998-1, corresponding to S = 1.0, TB = 0.15 s, TC = 0.40 s, TD = 2.00 s), behaviour factor q = 2.0 and lower bound factor for the design spectrum = 0.20. Ground acceleration was assumed to be 0.25g , with a building importance factor I = 1.0. The permanent load of the floor and roof was 3.5 kN/m2 and 2.0 kN/m2 , respectively. The permanent load for floors would suite a situation with medium thickness crosslam plates and a floating concrete floor atop, and includes allowance for partitions. The imposed load on the floor and roof was 2.5 kN/m2 and 2.0 kN/m2 , respectively, the former being the mean value between the imposed load for residential and office building, and the latter corresponding to the snow load on a moderately high altitude or to imposed load for residential buildings. The self-weight of the outer walls was 1.2 kN/m2 . The building was assumed to be category of use A (areas for domestic and residential activities) according to EN 1991-1-1 [27], so the value of 2i for quasi-permanent load was 0.3 and the factor was 0.5 for all floors except for the roof where it was 1.0 assuming the roof is accessible. The mass was modelled as lumped in the centroids of the floors. Mass and radius of gyration of floor mass in plan are summarized in Table 3. 4.4. Results of the linear analysis Fig. 9 compares vibration periods, base shears and top floor lateral displacements of the building for the three different types of models. The stiffness of the building is very high for model i, where all connections are assumed as rigid. This results in higher base shears and is conservative. However if the building was lower and hence even stiffer, the same hypothesis of rigid connections could yield to non-conservative results due to vibration periods being in the range of increasing spectrum outside the plateau region. On the other hand, the top displacements are only 20% of those obtained using models ii and iii where the connectors are schematized as flexible. Although this significant difference does not compromise the general stability of the buildings (unless second order effects have a remarkable impact), it can lead to underrated damage estimation. From Fig. 9 it can also be observed that the difference between considering and ignoring vertical connections between perpendicular walls at the same level (models iii and ii, respectively) is less than 4% when long wall

3050

M. Fragiacomo et al. / Engineering Structures 33 (2011) 30433053

Fig. 10. Calibration of the non-linear FEM link on the backbone curve of the 3rd cycle of the experimental results for BMF105 brackets with ten 4 mm diameter, 60 mm long nails subjected to axial (left) and shear force (right).

segments are used as in the case under study. This result may be different when shorter wall panels are used as in the buildings developed in the SOFIE research project [1113]. It should be noted, however, that the vibration period values for models ii and iii are unexpectedly high on the basis of a qualitative comparison with the first vibration period of the 3-storey SOFIE building, one would expect values of up to 0.4 s. It is believed that the main reason for such a difference is that the influence of friction at the panelpanel and panelfoundation was neglected in the numerical modelling. Friction affects the stiffness and strength but not ductility, however it is difficult to take it into account and further investigations will be necessary. 5. Case study: non-linear static analysis of a multi-storey crosslam building 5.1. Numerical modelling For a non-linear static analysis of the case study building, a 3D shell model with non-linear springs was used in SAP2000 [25]. Shear, tension and compression properties were assigned to each link element and their mechanical properties were calibrated on the experimental curves of BMF105 brackets with ten 4 mm diameter, 40 or 60 mm long nails (see Fig. 2). The backbone curves of the 3rd cycle test were used so as to account for the effects of cumulative damage of a seismic (cyclic) load. The curves were calibrated on the specimens with the lowest capacity (see Fig. 10). The comparison in terms of push-over curves (base shear vs. top floor horizontal displacement) using different nail lengths (40 mm and 60 mm, corresponding to brittle and ductile behaviour of the metal bracket, see Fig. 2) is presented in Fig. 11 for the case study building. In this figure, the assumption of lateral forces applied from right (wall B) to left (wall D) distributed along the building height according to the first mode shape was made. In the 3D FE model, only walls A and C were considered, whereas walls B and D were neglected as were torsion effects (i.e. degrees of freedom were limited to plane deformation). Furthermore, to investigate the effect of the bracket ductility on the whole building ductility, also a bracket with the same stiffness and peak strength as the BMF with 60 mm long nails, but with twice the ductility was analysed. 5.2. The N2 method and its limitations The N2 method [4] was used to carry out the NSA and identify the performance point of the building. The aim of the method is the evaluation of the seismic displacement, which is linked to the damage control of the structure and needs to be kept below some reference values. The N2 method considers a performance point defined in terms of both strength and displacement, where
Fig. 11. Comparison among different push-over curves for the case study building using metal bracket connections with different ductility, and elasto-plastic approximation according to Yasumura and Kawai procedure [19].

the structural capacity is compared with the demand in terms of seismic ground motion. The base shear force and the top displacement of a Multi-Degree-of-Freedom (MDOF) system are first computed by means of a non-linear Push-Over Analysis (POA) and then converted respectively to the spectral acceleration and displacement of an equivalent Single-Degree-Of-Freedom (SDOF) system. The demand of the seismic ground motion is represented through the response spectrum in terms of pseudo-acceleration and displacement. Such an inelastic spectrum depends upon the cyclic behaviour of the SDOF system and the characteristics of the ground motion (peak ground acceleration and shape), and can be obtained from the elastic spectrum using suitable reduction factors. Such a method was found to provide the best approximation among various NSA methods for SDOF systems with different hysteretic models and for MDOF systems [28]. The method as it is currently presented in Eurocode 8, however, is mostly suitable for reinforced concrete and steel structures, due to some limitations on specific hysteretic behaviour incorporated in the formulae for the reduction factor (R), which is needed to derive the inelastic from the elastic spectrum. Current formulae do not include the case of systems with a pinching hysteretic behaviour or larger stiffness degradation, which are typical for crosslam panels (see Fig. 12). The so-called flag shape hysteresis loops is due to rocking of the crosslam panels, friction with the foundation, and the influence of the vertical load in combination with the holddown capacity which increases the elastic stiffness of the wall. If there is no vertical load, the intermediate linear part of the hysteresis loop disappears and a loop with only pinching is left. An extensive parametric study on SDOF systems with this hysteretic behaviour is needed to derive new formulae for the reduction

M. Fragiacomo et al. / Engineering Structures 33 (2011) 30433053

3051

Fig. 12. Typical crosslam wall hysteresis loops under cyclic lateral loading when 15 kN/m of vertical load is applied.

pinching dissipate less energy than bilinear plastic loops with the same ductility. However, it should be also pointed out that for all analysed bracket setups, the SDOF systems equivalent to the multi-storey building have vibration periods longer than Tc which is usually the value from where the reduction factor (R) and ductility factor () are considered to be the same, regardless the type of hysteresis loop. The results of these analyses should therefore be considered as a preliminary study aimed to investigate the effect of different input strengths and ductility characteristics of the brackets on the strength and ductility of the building as a whole. The limit displacement in the Near Collapse (NC) limit state is defined by the attainment of 80% of the peak force on the descending base shear-top floor displacement curve when the failure mechanism is ductile. The ultimate force is derived from an elasto-plastic equivalent curve derived in accordance with the Yasumura and Kawai procedure [19] (see Fig. 11). Such a curve is then used to derive the capacity curve of the SDOF system equivalent to the case study building (see Fig. 13) and used in the N2 method to derive the performance point as an intersection with the inelastic spectrum (demand curve). In the case with 40 mm nails, due to the brittle failure, there is no decline in force, and the target displacement is set as the displacement at the failure point. It should be pointed out that the original procedure of the N2 method for the determination of the equivalent elasto-plastic diagram leads to too low vibration period values, too high ductility factors and, hence, too high allowable ground accelerations when the Near Collapse point is identified by 80% of the peak strength and by the corresponding displacement (see Fig. 14(left)). Similarly, also the use of the peak strength and the corresponding displacement lead to unrealistic results characterized by too low ductility ratios (Fig. 14(right)). Therefore it is recommended that the Yasumura and Kawai procedure as suggested in this paper is used to evaluate the equivalent elasto-plastic diagram. 5.3. Results of the non-linear analysis The results of the N2 method are displayed in Fig. 13 and summarized in Table 4 for the different types of metal bracket connections in terms of ductility ratios, seismic demand, seismic capacity and peak ground acceleration (for type A ground) of the whole case study building. The significant role played by the ductility of the connection can be clearly recognized: using a ductile ( = 8) metal bracket connection with 60 mm long nails, it is possible to raise the building ductility from 1.7 to 2.5, increase

Fig. 13. Inelastic spectrum and capacity curves of the SDOF system equivalent to the multi-storey crosslam building for three different types of metal bracket connectors. The square points denote the maximum allowable displacements and the circles the target displacements. Vibration periods of the SDOF systems are also marked.

factor R which could be then incorporated into the N2 method. The analysis is rather complex due to the influence of the vertical load which should be considered in the formulae. Therefore it must be noted that the results derived in this study could be non-conservative, because hysteresis loops with

Fig. 14. Comparison among different push-over curves for the building response and elasto-plastic approximations according to the N2 method with maximum displacement at 0.8 Fmax (left) and at F max (right).

3052

M. Fragiacomo et al. / Engineering Structures 33 (2011) 30433053

Table 4 Ductility ratios (calculated as the ratios between maximum and elastic displacements), seismic demand and capacity of the case study building in terms of maximum acceleration depending upon the metal bracket connection ductility. Bracket uplift ductility double bracket ductility BMF 105 bracket with 10 60 mm long, 4 mm diameter nails BMF 105 bracket with 10 40 mm long, 4 mm diameter nails 16 8 Bracket shear ductility 13 6.5 Building ductility 3.67 2.46 Target displacement (mm) 54 56 Maximum displacement (mm) 101 67 Max. ground acceleration (g) 0.64 0.41

6.4

1.69

66

42

0.22

the max peak ground acceleration the building can resist from 0.22g to 0.41g , and the maximum failure displacement from 42 to 67 mm. In performance-based design, different limit states are defined, such as damage limitation (DL), significant damage (SD), near collapse (NC) and total collapse (TC). For timber construction, the performance indicator usually assumed is either the interstorey drift, or the top floor horizontal displacement. Different values are suggested by design codes, however an inter-storey drift of about 3% of the wall height is commonly set as a limit. In this paper the proposal is to define the NC limit state at the attainment of 80% of the peak strength on the descending part of the push-over curve, as recommended in the Yasumura and Kawai procedure. A more exact procedure would be to define a limit state for each wall and consider the NC limit state of the whole building to be attained when the first wall reaches its NC state. The SD state could be defined at 75% of the ultimate displacement du at NC state as suggested for other structural types (e.g. reinforced concrete structures or masonry walls); however this limit state is questionable as crosslam buildings exhibit small residual damage at the end of a seismic event [1113]. The DL state could be set at the elastic inter-storey (or wall) displacement in accordance with the Yasumura and Kawai procedure or, in a less conservative way, at the displacement where the ultimate force (Fu ) is reached for the first time (in the elasto-plastic approximation this is the yielding point). 6. Summary and conclusions This paper deals with the seismic analysis of multi-storey crosslam buildings. Such systems are made from massive timber panels connected to each other and to the foundations using metal brackets and nails (or screws), and self-tapping screws. During an earthquake, energy is dissipated in all the connections as well as in friction between timber panels, although it is suggested that the beneficial contribution of the latter is conservatively neglected until further investigations are performed. Capacity design principles are provided in the paper, where the ductile failure mechanism is characterized by plasticization of metal bracket connections throughout the height of the building, and brittle failures of the crosslam panels are avoided by designing the corresponding members for the overstrength of the connections. Floor panels and their connections between adjacent panels are also designed for the overstrength of the connections. The overstrength factors for BMF105 steel brackets, for which there is currently no information provided in the Eurocode 8, were derived based on some experimental results and a value of 1.3 is proposed for shear and uplift based on some preliminary tests. Simple detail rules on the nail length which should not be shorter than 60 mm to avoid brittle failure of connections were provided. An overstrength ratio of 1.6 was derived for self-tapping screws holding together adjacent perpendicular walls, although the contribution of such connections to the building stiffness was proved to be minimal in a 3D elastic modal response spectrum analysis of a multi-storey

building when long wall segments are used. The same analysis also revealed the importance of modelling the flexibility of the connections between the upper and lower panels, which is often neglected by designers and would lead to underestimation of the building vibration periods. The paper also discusses the possibility of using a non-linear push-over analysis for seismic design as implemented in the N2 procedure recommended by the Eurocode 8. The N2 method with a modified bilinearisation procedure was used for the calculation of the maximum ground acceleration a building can withstand. The influence of the connection ductility on the global building ductility and seismic performance was analysed. The use of ductile hold-downs with 60 mm long nails was found to raise the building ductility from of 1.7 to 2.5 and the maximum peak ground acceleration the building can resist from 0.2g to 0.4g compared to the case when a brittle hold-down is used, demonstrating the importance of using ductile connectors. However, due to limited experimental data available, further investigations are needed to confirm the conclusions on some of the issues highlighted in this paper. Cyclic tests of metal bracket and hold-down connectors should be carried out on at least 710 specimens per configuration to derive more accurate values of the overstrength factors. Also non-linear time-history analyses (in addition to less time consuming non-linear static analyses) should be carried out on crosslam buildings where the actual hysteretic behaviour of the connections is modelled in a rigorous way to investigate the influence of the connection ductility on the global building ductility and seismic resistance. Acknowledgements The facilitating network provided by the COST Action E55 Modelling of the performance of timber structures (http://www.coste55.ethz.ch/) is gratefully acknowledged. The research support provided to the second and third authors by the EU through the European Social Fund Investing in your future is also acknowledged.

References
[1] Chopra AK. Dynamics of structurestheory and applications to earthquake engineering. Upper Saddle River: NJ: Prentice Hall; 1995. [2] European Committee for Standardization (CEN). Eurocode 8design of structures for earthquake resistance, part 1: General rules, seismic actions and rules for buildings. 2004. [3] Priestley MJN. Performance based seismic design. 12th European Conference on Earthquake Engineering. CD. 2000. [4] Fajfar P. A nonlinear analysis method for performance-based seismic design. Earthq Spectra 2000;16:57392. [5] Italian Ministry for the Infrastructures. New technical regulation for construction. Decree of the Ministry for the Infrastructures, Ministry of Interior, and Department of the Civil Defence. 2008. [6] Pauley T, Priestley MJN. Seismic design of reinforced concrete and masonry buildings. Wiley Ed.; 1992. [7] Beattie G. Multistorey timber buildings manual, Carter Holt Harvey (Ed.), Origin, James Hardie and GIB, 2001.

M. Fragiacomo et al. / Engineering Structures 33 (2011) 30433053 [8] Mander JB, Cheng CT. Seismic resistance of bridge piers based on damage avoidance design. Report NCEER-97-0014. 1997. [9] Smith T, Fragiacomo M, Pampanin S, Buchanan A. Construction time and cost estimates for post-tensioned multi-storey timber buildings. Proc Inst Civil Eng Construction Mater 2009;162:1419. Timber Structures (special issue). [10] Buchanan A, Deam B, Fragiacomo M, Pampanin S, Palermo A. Multi-storey prestressed timber buildings in New Zealand. IABSE Struct Eng Internat 2008; 18:16673. Tall Timber Buildings (special ed.). [11] Ceccotti A. New technologies for construction of medium-rise buildings in seismic regions: the XLAM case. IABSE Struct Eng Internat 2008;18:15665. Tall Timber Buildings (special ed.). [12] Dujic B, Strus K, Zarnic R, Ceccotti A. Prediction of dynamic response af a 7-storey massive XLam wooden building tested on a shaking table. World Conference on Timber Engineering WCTE 2010. Riva del Garda, Italy, June 2024, 2010, CD. [13] Ceccotti A, Follesa M, Lauriola MP, Sandhaas C, Minowa C, Kawai N. et al. Which seismic behaviour factor for multi-storey buildings made of crosslaminated wooden panels? Meeting 39 of the Working Commission W18Timber Structures, CIB. Florence, Italy, 2006, paper CIB-W18/39-15-4. [14] Yates M, Linegar M, Dujic B. Design of an 8 storey residential tower from KLH Cross laminated solid timber panels. World Conference on Timber Engineering WCTE 2008. Miyazaki, Japan, June 25, 2008. [15] European committee for standardization (CEN). Eurocode 5 design of timber structures part 1-1: general rules and rules for buildings. 2004. [16] Jorissen A, Fragiacomo M. General notes on ductility in timber structures. Eng Struct, (submitted for publication). This special issue. [17] New Zealand StandardTM . Timber structures standard, NZS3603:1993. Published by Standards New Zealand, Private Bag 2439, Wellington 6020, New Zealand. [18] Dujic B, Zarnic R. Report on evaluation of racking strength of KLH system. University of Ljubljana, Faculty of civil and geodetical engineering, Slovenia, 2005.

3053

[19] Yasumura M, Kawai N. Evaluation of wood framed shear walls subjected to lateral load. Meeting 30 of the Working Commission W18-Timber Structures, CIB. Vancouver, Canada, 1997, paper CIB-W18/30-15-4. [20] Blass HJ, Uibel T. Tragfhigkeit von stiftfrmingen Verbindungsmitteln in Brettsperrholz. Lehrstuhl fr Ingenieurholzbau und Baukonstruktionen 2007, Universittsverlag Karlsruhe. [21] Sandhaas C, Boukes J, Kuilen JWG, Ceccotti A. Analysis of X-lam panel-topanel connections under monotonic and cyclic loading. Meeting 42 of the Working Commission W18-Timber Structures, CIB. Dbendorf, Switzerland, 2009, paper CIB-W18/42-12-2. [22] Dujic B, Klobcar S, Zarnic R. Influence of openings on shear capacity of wooden walls. Research report, 2005, University of Ljubljana and CBD Contemporary Building Design Ltd., Slovenia. [23] Yasumura M. Racking resistance of wooden frame walls with various openings. Joint meeting 39 of the Working Commission W18-Timber Structures, CIB IUFRO-S5.02. Florence, Italy. 1986. [24] Blass HJ, Fellmoser P. Design of solid wood panels with cross layers. 8th World Conference on Timber Engineering WCTE 2004. Lahti, Finland, June 1417, 2004, p. 5438. [25] Computers & structures inc. SAP2000integrated finite element analysis and design of structures. Computers & structures inc.: Berkeley, CA, 2000. [26] European Committee for Standardization (CEN). EN 12512 Timber structures Test methods Cyclic testing of joints made with mechanical fasteners. Brussels, Belgium, 2001. [27] European Committee for Standardization (CEN). Eurocode 1: actions on structures part 1-1: general actions densities, self-weight, imposed loads for buildings. 2002. [28] Fragiacomo M, Amadio C, Rajgelj S. Evaluation of the structural response under seismic actions using non-linear static methods. Earthq Eng Struct Dyn 2006; 35(12):151131.