Professional Documents
Culture Documents
Figure 1. Configuration of panel-sheathed shear wall without opening and loading configuration for experimental tests (left), and detail of the additional metallic connector in lower corners of the frame (right).
To model these configurations, a refined FE model is first derived. The frame is modeled using elastic timber beams and the panels with plate elements. All metallic connections, including each individual nail for the panels, are modeled using a suited nonlinear semi-rigid behavior law with cumulative damage [2]. The behavior of each connection (nails, angles or HDs) are derived by fitting results of experimental tests under monotonic and reversed cyclic loading. Those tests present the advantage to be much less costly than full-scale tests on shear walls. Lumped masses and dampers are added on the top plate in order to account for the vertical loading on the wall and some minor dynamic effects. The second FE model adopts a macro-scale approach to model the same wall element. It is composed of a rigid frame of bar elements, allowing for a parallelogram-like deformation only. A single finite element which uses the same nonlinear semi-rigid behavior model as above connects two opposite corners of the frame to confer its final behavior to the wall model. Additionally, lumped masses and dampers are added on the top plate to model the overall mass of the wall and its vertical loading, as well as a limited damping. The dampers are exactly the same in both FE models, with a small damping ratio of a few percents at most since the structural damping is already accounted for in the nonlinear semi-rigid behavior laws of both models. These dampers mainly prevent numerical oscillations, and model minor dynamic effects not taken into account otherwise. In the end, the refined FE model has more than 500 elements, whereas the macro-scale model has only about 10 elements.
- 24 -
III. Results and Discussion These two models of shear wall were already validated under quasi-static monotonic and cyclic loading in a prior research [3]. Here, we complement this validation by checking their response under a dynamic seismic loading. First, we use the input motion from the Landers earthquake (California, 1992). Figure 2 presents the comparison between the experimental result of a shaking table test and the numerical prediction of the refined FE model of the wall configuration with opening [1]. The graph shows the temporal evolution of the drift of the top plate. The refined FE model is able to accurately predict the phase, frequency and amplitude of the response in the initial part of the earthquake. The peak displacement error is less than 10 %. In the second part, from 17 seconds, the amplitude is slightly underestimated by the FE model, with an error about 21 %, yet no phase shift or frequency mismatch occur between the experimental and numerical responses. Moreover, the error increases with time, as it is often the case in history-dependent analyses, and the model proved very accurate before the peak force is reached. Other tests with different wall configuration and input motion report similar conclusions, with some tests as low as 1 % of error on the peak displacement. Therefore the refined FE model is considered validated for use under dynamic seismic loading.
Figure 2. Experimental vs. numerical results on the refined FE model undergoing the input motion of the Landers earthquake (California, 1992).
Figure 3. Refined FE model vs. macro-scale model undergoing the Loma Prieta input motion (California, 1989).
The response of the macro-scale model is then compared to the predictive response of the validated refined FE model (Figure 3). In this example, we conducted numerical tests using the Loma Prieta input motion (California, 1989) scaled to a peak ground acceleration (PGA) of 0.32 g (3.14 m/s2) (first graph on the right). The macro-scale model is able to predict a similar response to those of the refined FE model, with at most a few percents of error on the wall drift (second graph on the right). Moreover, the largest errors are observed past the peak force when the wall is heavily damaged and the overall resistance of the wall has much decreased. The forcedisplacement responses are shown on the left graph of Figure 3. As a concluding remark, it should be noted that while both models predict a similar response, the macro-scale model drastically decreases simulation time down to a 1:10 ratio as compared to the refined FE model. Therefore, the macro-scale approach is of particular interest for use in time-consuming simulations such as time-history seismic analyses.
IV. References [1] N. Richard, M. Yasumura, L. Davenne (2003) Prediction of seismic behavior of wood-framed shear walls with openings by pseudodynamic test and FE model, J. Wood Sci. 49:145-151. [2] J. Humbert (2010) Characterization of the Behavior of Timber Structures with Metal Fasteners Undergoing Seismic Loadings, Ph.D. thesis, Grenoble University, France. [3] J. Humbert, C. Boudaud, J. Baroth, L. Daudeville (2012) Modeling timber joints with metallic fasteners undergoing reversed cyclic loading, J. Struct. Engrg ASCE, under review.
- 25 -