Validation and verification of automated non-linear modelling for the seismic response of cross laminated timber structures

I. P. Christovasilis, L. Riparbelli (Aether Engineering s.a.s.)


The scope of this study is to validate a structured methodology for the modelling of timber structures under seismic excitation. It relies on an advanced numerical simulation model which performs design codes verifications based on the results of linear analyses and allows considerations for performance-based and resilience-based design from the results of non-linear analyses.

Cross Laminated Timber (CLT) panels constitute a relatively new type of component that has played an important role in the significant growth of timber construction in the building sector over the last years. This is a result of enhanced structural performance compared to traditional timber systems, other than the positive impact in the carbon footprint of the building due to the use of a renewable material. Moreover, CLT structures are highly suitable for off-site modular construction with accelerated assembling on-site that is currently revolutionizing the practices of the construction industry.

While research conducted on the structural performance of CLT buildings has been increasing over the last years with considerable knowledge gain, the prediction of the response of these structures under seismic excitation still remains a challenging topic, especially when considering a realistic numerical model that takes into account non-linear effects.

A CLT structure is composed of bi-dimensional wall and floor panels that are connected along the adjacent common edges with various types of metal connectors and connection devices. Compressive loads between panels are transferred through contact forces ; tensile loads are transmitted solely through specific uplift-restraint connections ; while shear loads are transferred by both friction forces in contact areas and shear-restraint connections. All of these mechanisms coexist in the interaction among all the wall and floor CLT panels of the structure.

Modelling approach

The modelling approach builds upon an existing framework that was specifically designed for the practical creation of three-dimensional models of CLT buildings with the Finite Element Method (FEM). To facilitate the entire workflow, a dedicated tool was developed for managing all information related to each structural component at each point of the workflow and for performing specific script-driven operations. These allow the automatic creation of the input data needed for creating the geometry and mesh of a building model and for performing the analysis with code_aster.

The 2nd generation approach presented in this study extends the existing one by introducing two modelling features. On one hand, the algorithm that computes and classifies the geometric intersections between the CLT panels has been upgraded to allow for the definition of all the common boundaries needed to assign all pairs of master-slave edge groups for mono-lateral contact and friction phenomena between panels. Figure 1 shows a deformed configuration of a CLT assembly with mono-lateral contact between panels. On the other hand, user-defined 2-noded discrete elements can be introduced to explicitly model each and every metal connector and connection device in the structure. A combination of two discrete elements in series – a cable and a bilinear element - is used to model uplift-restraint devices that have no resistance in compression and exhibit a pinching response in tension, as shown in Figure 2.

All existing script-driven operations have been upgraded to handle the additional data of the new approach and to maintain an automated level of implementation of the geometry and mesh construction as well as the related commands for performing non-linear analyses with code_aster.

Figure 1 : Deformed configuration of a CLT assembly under lateral load with non-linear mono-lateral contact
Figure 2 : Force-Slip response of an uplift-restrain device

Test structure and numerical model

The validation and verification of the modelling approach is performed using the experimental data from shake-table tests of a full-scale 3-storey CLT structure that took place in Japan as part of the CNR-IVALSA SOFIE project. The test structure is rectangular in plane with a double-pitched roof.

The building consists of 61 CLT panels (40 wall panels, 5 beam panels and 16 floor panels) that are composed of 449 faces. Figure 3a shows the geometry of the numerical model, identifying with the same face colour each group of faces that constitute a CLT panel. These panels share 204 common edges, where contact and friction conditions and master-slave groups are automatically defined. Moreover, the numerical model includes 1543 discrete elements, as shown in Figure 3b, that represent 9 different types of connectors and connection devices. In this case, the CLT panels are connected with 211 different connections, different in terms of panels connected and type of connections.

Figure 3 : Geometry of the numerical model, with different face colours for each CLT panel in the structure (a) and different line colours for each type of discrete connections (b)

Numerical studies

A series of validation and verification studies are presented based on some of the various unidirectional shake-table tests that were performed with scaled and unscaled ground motions.

We first present numerical predictions from a code-based linear modelling approach and evaluate its accuracy with the experimental data (Figure 4).

Consequently and based on the results from the linear modelling approach, we explore the potential of non-linear modelling, quantifying the differences in the numerical response for different cases regarding the consideration of contact and friction phenomena, as well as material non-linearity in the connectors.

Furthermore, an evolutionary verification index according to the Eurocodes is introduced that takes into account the various failure mechanisms during the transient history of the response.

Figure 4 : Comparison between test results (grey) and numerical results for 5% (red), 10% (black), 16% (red-dashed) and 21% (black-dashed) damping values in terms of base shear for the Kobe JMA ground motion (N-S component) scaled to a PGA of 0.15g