1. Introduction
In 2050, the urban population is expected to be close to 7 billion, an increase of 50% compared to the actual situation [
1]. To accommodate more than 2 billion new people without increasing too much urban sprawl and its environmental consequences, the densification of cities is mandatory [
2]. In compact cities, the buildings, having higher volume-to-surface ratios, are more energy efficient [
3]. Constructing in cities poses its challenges. Space for the construction site is very limited, neighbors are close and can be bothered by the noise and dust, and arteries closure can have increased consequences the longer the construction lasts for locality and nearby businesses [
4]. The construction of tall buildings in cities must therefore be accelerated, less disruptive and less costly with prefabrication [
5]. Light-frame modular construction allows for a high level of prefabrication. Insulation, mechanic and electric elements; cladding; doors and windows; wall and floor finishes; and even furniture and appliances can be installed in factories, diminishing the work to be performed onsite.
In the province of Quebec, Canada, and throughout North America, light-frame construction is limited at six stories by construction codes due to resistance and serviceability limitations. Systems designed to receive internal substructures and to act as the main gravity and lateral load resisting system (LLRS) could enable the use of light-frame construction in high-rise buildings and need to be more explored so that the benefits of prefabrication can be further exploited [
6].
Such design was explored with the building Treet, completed in Bergen, Norway, in 2015. Treet was the world’s tallest timber building in 2015, and its innovative structural system was studied by many authors [
7,
8,
9,
10]. A similar structural system, but with the main structure made of concrete, was studied in recent years [
6,
11,
12,
13]. This structural concept is named FaB and was first proposed by Xiong et al. in 2016 [
14]. Many numerical studies were also performed on innovative structural systems for high-rise timber buildings [
3,
9,
15,
16,
17,
18]. The studied systems were part of research concepts, but also real buildings, such as Mjøstårnet and Brock Commons. In the last decade, high-rise buildings using different modular structural systems were also constructed around the world [
19,
20,
21,
22].
Several numerical studies dealing with light wood-frame structures have been presented by many authors [
23,
24,
25,
26,
27,
28,
29]. Models developed vary in regard to numerical complexity and precision compared to the tested structures and elements considered (walls, diaphragms, full house, etc.) [
30,
31,
32,
33]. Some researchers model light-frame behavior through shell elements, while others model it with linear elements, springs or a mix of the three latter. An interesting approach was suggested by Chen et al. [
23], who worked on a pinned frame with diagonal members and vertical connectors.
The main aim of this study was to design a hyperstructure system that achieves the structural performance required for the use of modular light-frame in high-rise buildings. To achieve this, two different 20-story hyperstructure systems were designed, a method to model modular light-frame construction, using finite elements, was developed, and the interaction between both systems and the impact on the substructures were analyzed.
2. Materials and Methods
The structural concept to enable the use of light-frame construction in high-rise buildings can be easily pictured as a cabinet and its drawers [
8]. With this analogy, the cabinet is the hyperstructure and the drawers are the prefabricated light-frame modular substructures. That is, a hyperstructure is a structural system that carries the loads of other complete structural systems. Hyperstructure and substructure terms are used in this article.
Two different 20-story hyperstructures were designed and modeled by using finite elements: one using glulam with external and internal bracings and another one using reinforced concrete with two internal cores. The two systems were designed in such a way that five 4-story light-frame modular substructures could be inserted. For the timber hyperstructure, glulam was preferred to CLT as an initial choice, because using the latter would have resulted in a duplication of the walls already present with the light-frame modules. The software used was RFEM [
34].
The 4-story light-frame modular substructures were also modeled by using finite elements. Without a universal method to model this type of structure using finite elements, a method was developed to enable the representation of different shearwall assemblies and to obtain deflections corresponding to the empirical equations given in CSA O86-19 [
35].
The structural system is completed when the substructures are inserted into the hyperstructures and then connected to the hyperfloors only or also to the columns and cores.
The studied design is a 20-story residential tower. The footprint of the building is 22 m × 28 m. The building is 60 m high, with 3 m–high stories. It has two internal shafts of 6 m × 6 m. The hyperstructure’s floors (hyperfloors) were designed to be able to minimize the need for internal columns. There are two types of beams, namely the primary beams and the secondary beams. The light-frame substructures are laid on the secondary beams. The hyperfloor plan is shown in
Figure 1.
The vertical gravity-load-resisting system (GLRS) differs depending on the hyperstructure concept. For the glulam concept, it consists of glulam columns. For the concrete concept, it consists of concrete columns on the exterior and two concrete cores. The horizontal GLRS consists of the hyperfloors, which transfer the loads from the substructures to the vertical GLRS. For the glulam concept, they consist of glulam beams in order to keep a post-and-beam glulam system. For the concrete concept, they consist of prefabricated unidirectional concrete slabs on bidirectional concrete beams.
The LLRS differs between the concepts. For the glulam concept, the LLRS is achieved by the glulam diagonals acting as vertical trusses on the exterior walls and the service shafts. For the concrete concept, the LLRS is achieved by the two concrete cores acting as ductile shearwalls. The lateral loads are transferred through the hyperfloors by diaphragm action. For the glulam concept, the diaphragm action is achieved by steel diagonals between the glulam beams creating a horizontal truss. For the concrete concept, the diaphragm action is achieved by the concrete slabs.
For the glulam concept, the beams are assumed to be of grade SPF (spruce–pine–fir) 20f-EX and the columns and diagonals of grade SPF 14t-E. For the concrete concept, concrete with a compression strength of 30 MPa is assumed with 400 R steel rebars.
The building is assumed to be in Quebec City in the province of Quebec in Canada. Snow, wind and seismic loads are determined accordingly [
36]. The design data for gravitational loads are given in
Table 1, based on the NBCC.
For the hyperstructures design, modular substructures were considered as loads acting directly on the hyperfloors. Dead and live loads coming from the substructures were transformed into uniform linear loads acting on the hyperfloors’ secondary beams to represent the modules’ walls placed directly over them.
The design data for wind loads are given in
Table 2 based on NBCC [
36].
The wind loads acting on the building were calculated with the dynamic procedure following Article 4.1.7.8. of NBCC. Figure A-4.1.7.5.(2) was followed for the external pressure coefficient and the exposure factor. The effects of total and partial wind loads were considered following Figure A-4.1.7.9.(1) Apart from stress and strain caused by the wind loads, vibrations due to wind were also verified. Across- and along-wind accelerations at the top of the buildings were calculated with Equations (8) and (9) of the NBCC 2015 Commentary I [
38].
Since the hyperstructure system is not well-studied and because the conditions of Article 4.1.8.7.1) of NBCC [
36] are not all respected, a dynamic analysis method must be used to evaluate seismic loads and their effects. The response spectrum analysis method was used in this study. For the modal analysis, eigenvalues were extracted by the Lanczos method. To represent an adequate dynamic response, eigenvalues were extracted to have a minimum of 90% modal participating mass ratio (MPMR) in modal analysis for each orthogonal direction [
38].
Once the vibration modes were extracted, the response spectrum analysis was performed. The response spectrum used is based on the 5% damped spectral response acceleration values and peak ground acceleration of Quebec City provided in table C-3 of NBCC [
36]. The soil type assumed for the building was class D for stiff soil. A Lehr damping ratio of 1.5% was considered. The glulam hyperstructure’s LLRS was considered as braced frames with ductile connections of limited ductility with R
d = 1.5 and R
o = 1.5. The concrete hyperstructure’s LLRS was considered as ductile shear walls with R
d = 3.5 and R
o = 1.6 [
36]. Respecting Article 4.1.8.12.4)b) of NBCC 2015, the effects of accidental torsional moment were accounted for. A load case was obtained for each mode of vibration, and a combination of results for all participating modes was obtained by the complete quadratic combination (CQC) method.
All elements were designed to fulfill ultimate limit state (ULS) requirements. Load effects were obtained directly from RFEM structural analysis for all load cases prescribed by NBCC. Connections were not designed in this study. All elements were also designed to fulfill serviceability limit state (SLS) requirements. A maximal deflection criterion of L/360 was considered as suggested in Article 9.4.3. of NBCC [
36]. Apart from individual elements’ deflection limits, LLRS were designed to respect interstory drift criteria:
for wind loads and
for seismic loads. For seismic load cases, lateral deflections were multiplied by
, as prescribed in Article 4.1.8.13.2) of NBCC 2015. LLRSs were also designed so that accelerations due to wind, both along- and across-wind, were less than 15 milli-g, as prescribed by NBCC for residential buildings [
37].
For the glulam hyperstructure, all connections between members were considered pinned. Reduced section properties were used for all concrete surfaces and elements according to Article 21.2.5.2 of CSA A23.3-19 [
39]. The core surfaces were connected between themselves rigidly to represent traditional cast-in-place cores. Except for the cores, the concrete elements were assumed to be prefabricated. The columns were modeled with pinned connections at every four floors, resulting in prefabricated columns of 12 m. The beams and slabs in the concrete floor cassettes were assumed to be factory cast together, and rigid monolithic connections were modeled. At the intersection of the cassettes’ beams with the columns and core walls, pinned connections were considered. At the interface between floor cassettes’ slabs and at the interface between these slabs and the core walls, linear articulations were modeled to allow rotation along length and translation perpendicular to the intersection line. The same was performed for the roof cassettes.
The hyperstructures are shown in
Figure 2.
The 5 substructures are found at levels 1–4, 5–8, 9–12, 13–16 and 17–20. The lower substructures (levels 1–4) rest directly on the foundations. The upper substructures were at first only connected to the hyperstructure at the hyperfloors. Models were then also analyzed by connecting the modules to the columns and cores to make them participate in the LLRS.
The chosen model needed to be relatively simple and able to represent shearwalls deflections in multi-story buildings, as calculated in Article A.11.7.1 of CSA O86-19 [
35]. Hence, only the substructures shearwalls and floors are modeled (
Figure 3). All other elements normally composing prefabricated light-frame modules (studs, sill plates, headers, etc.) are not modeled, and the gravitational loads pass through the columns. A new model was developed based on the modified macro-element model of Chen et al. [
23] in which the diagonals had hysteretic stiffnesses based on the Bouc–Wen–Baber–Noori model. For this study, the diagonals were modified as non-linear members to link with the equations of Article A.11.7.1. The concept, illustrated in
Figure 4, consists of pinned frames composed of rigid beams and columns (in black). The lateral rigidity of shearwall segments is represented by diagonals with non-linear stiffnesses. Elongation of the wall anchorage system is represented with rigid connectors having linear stiffnesses in compression and tension (in blue).
To determine the non-linear stiffness of each shearwall segment, interstory deflection due to bending, panel shear and nail slip are calculated for incremental horizontal loads of 1 kN up to the maximal shear capacity of the wall configuration. To represent the nonlinear behavior of light-frame shearwalls in the software, the deflections are calculated for 6 or 7 shear loads (e.g., for a wall with a maximal capacity of 64 kN, deflections were calculated for 1, 2, 4, 8, 20, 40 and 64 kN). Although no nonlinear analysis is conducted, modeling the substructure with non-linear diagonals allows more accurate deformations under given lateral forces.
The deflection due to vertical elongation of the wall anchorage system is taken into account by attributing axial stiffnesses in compression and in tension (using equations from Reference [
40]) to the vertical connectors between modules.
Substructures’ shearwalls compositions are provided in
Table 3.
The developed methodology was used to model the 3D substructures. In
Figure 3, sections with springs are shearwall segments. The sections without springs are sections where there could be openings and structural or non-structural partitions. Substructure loads are presented in
Table 4.
In RFEM, modeling hypotheses were made. Beams and columns forming the pinned frames were modeled as highly rigid bars so that they do not deform when transferring the normal forces. All of these sections were modeled with no self-weight and with end articulations allowing rotation in the
y-axis and
z-axis. The frames’ stability was in this way only assured by the nonlinear diagonals. Floors and ceilings were modeled as infinitely rigid surfaces hinged at each section, allowing rotation along the intersection. Vertical connections between modules were modeled as rigid bars. The top node was fixed to the floor surface. The bottom node was free to rotate along the
y- and
z-axes and had different axial stiffnesses in tension and compression (
Table 5). Horizontal connections between modules were modeled as pinned rigid bars free to rotate along the
y- and
z-axes.
In the models with the modules participating in the LLRS, the connections between the modules and the columns were modeled with rigid bars pinned at both ends. At the column or core side, the articulation was free along the y-axis and z-axis. At the module’s side, the articulation was only free along the y-axis to prevent differential lateral displacement between the modules and the hyperstructure but to allow vertical deflection of the modules. For those models, some connections were modified to stiffen the modules and prevent substructures’ modes of vibration with higher periods than the whole system’s modes of vibration. The rotation along the z-axis was blocked for the horizontal connections between the modules, and pinned rigid bars were added to connect the end walls of the long modules together.
For all cases, surface loads were applied directly to the floors and roof of the modules to replace the linear loads applied to the secondary beams. Gravity-, wind- and seismic-load cases were analyzed as for the hyperstructure.
Modules’ lengths are chosen so that there is a 500 mm spacing between the modules’ end walls and the exterior columns. That spacing is reduced to 250 mm between the short modules’ end walls and the concrete core or internal columns. A spacing of 50 mm is considered between the end walls of the long substructures.