Seismic Performance of CLT Shear Wall Infilled Hybrid Steel Frames with Concealed Steel Plates and Drift Pin Connections

Fig. 17 (in “Numerical Verification” section) summarizes the comparison between the test hysteresis result and numerical hysteresis results, together with the plotting of the yielding of steel beam, shear failure in the CLT infill, and yielding of DP shown. From the figure, it can be observed that the yielding of steel beam occurred at around $1/100\mathrm{rad}$ in all specimens. As for the shear failure of CLT, with the exception of FM-S60-32-S, the CLT-infilled specimens with FM beams experienced strength degradation at an angular deformation of around $1/50\mathrm{rad}$ due to the shear failure of the CLT infill, as shown in Fig. 8(a). It can be said that the shear capacity of the CLT infill was fully utilized. As for FS-S60B-32, strength degradation did not occur because there was no shear failure in the CLT infill, due to the lateral torsional buckling of one of the two steel beams markedly at $1/30\mathrm{rad}$ [Fig. 8(b)] and also the shear failure of HTB at the wall-to-frame connection at the final cycle [Fig. 8(c)]. It should be noted that all specimens were tested successfully, per the loading schedule, except for Specimen FM-S60-32-S, in which test had to be terminated prematurely because of the significant lateral torsional buckling of the beam during the negative direction of the first cycle of $1/30\mathrm{rad}$ [Fig. 8(d)], after which the shear failure of CLT followed almost immediately. Concerning the yielding of DP, the DP in specimens with $32+10+32$ configuration yielded at similar displacement of $1/100\mathrm{rad}$, except Specimen FS-S60B-32. In Specimen FS-S60B-32, the DP yielded at $1/50\mathrm{rad}$ because there is less effectiveness in the load transmission from the beam to the wall-to-frame connection, due to the more slender beams. On the other hand, the DP in specimens with $22+14+22$ configuration were observed to undergo early yielding and failure at the wall-to-frame connection, of a type that involved compressive buckling at the concealed steel plate [Fig. 8(e)] at a relatively smaller displacement at $1/200\mathrm{rad}$ due to the lesser number of DP, as can be observed from the plotting for Specimen FM-S60B-32. However, for Specimen FM-S60B-22-M, the DP yielded at $1/100\mathrm{rad}$ because of the confinements from the infilled nonshrink mortar at top and bottom of the CLT panel, where the CLT panel was unable to deform freely. Because of the early yielding of the DP connectors, both specimens showed particularly significant failure in the CLT shear wall, which split in half vertically in the middle, as shown in Fig. 8(f). Table 3 provides the summary of the failure mechanisms during the experiments. Generally, almost no slip behavior was observed among the hysteretic behavior of the specimens with CLT infill, an outcome that was similar to the hysteretic behavior of the steel frame specimens. This showed that the steel frame was dominant in terms of the behavior of the hybrid system.

Fig. 18 (in “Numerical Verification” section) shows comparison of the load sharing effect between the test and numerical results. As described previously, in the CLT-steel hybrid system, the shear force is jointly resisted by the CLT panel and the columns of the steel frame. In order to evaluate the effectiveness of the hybrid system’s lateral load resistance, the percentage of shear force resisted by each of the subsystems was calculated for all specimens using the values of Eqs. (10) and (11). From the figure, it can be seen that in FS-S60B-32, the CLT infill resisted more than half of the shear force until the end of the experiment. That was because, as compared with the tests based on the FM frame, this specimen used the more slender FS beams, which ultimately failed in this case by an out-of-plane buckling deformation at 1/30 rad without the CLT shear wall having shear failure, after which the experiment was terminated. As for the specimens in which FM beams were applied, the steel frame resisted more than 50% of the shear force throughout the experiment because of the more robust beams, which were more capable of transferring the lateral loads to the steel columns. At around $1/50\mathrm{rad}$, the percentage of force resisted by the CLT infill started to decrease gradually due to the accumulation of damage, except Specimen FM-S60-32-S. This was because in this specimen in which the CLT infill is positioned at the side instead of in the center of the steel frame, the shear properties of CLT infill was not effectively taken advantage of at small displacements. This led to the CLT infill having adequate shear capacity for resisting shear load until large deformation (i.e., $1/30\mathrm{rad}$), after which it only started to show signs of shear failure.

The hysteretic response of each of the steel member (e.g., beams or columns) was reproduced using the modified Ramberg-Osgood model proposed by Meng et al. (1992). The force ($f$)-displacement ($d$) relationship that defines the hysteretic loop of the model is given in Eq. (12)where $f_1$ and $d_1$ = both the unloading force and displacement; $K_u$ = unloading stiffness at $(f_1,d_1)$; and $\gamma$ = parameter used to define the curvature of the hysteresis loop for calibration purposes. The coefficient $b$ defines the unloading path of the loop targeting for point $(f_2,d_2)$ and is given in Eq. (13)where $(f_2,d_2)$ = target point that the unloading path will arrive at. This target point can be shifted along the horizontal axis by multiplying $d_2$ by parameter $Ø$, when the unloading point $(f_1,d_1)$ occurs after the yielding point $(f_y,d_y)$. When $Ø<1$, the stiffness of the unloading path will be less than the initial stiffness $K_0$ which enables consideration of the strain-softening phenomenon to take place.

A model of the bare steel frame whose sectional behavior follows the previously mentioned modified Ramberg-Osgood equations was developed for calibration to fit the cyclic experiment results of the bare steel frame. The steel bare frame model is conceptually the same as the example shown in Fig. 9, minus the CLT infill elements. The steel frame’s columns and beams were modeled using software-defined column elements and beam elements, which can resist bending, shear, and axial forces. The elastic and yield values used for these elements are defined using a previously conducted steel bar tensile element test, illustrated in Fig. 10(a). The bending strength $F_b$ can be obtained directly from the test, while the shear strength $F_s$ was obtained by dividing the test result by $\sqrt{3}$. $F_b$ and $F_s$ were used to calculate the ultimate moment force $M_u$ and ultimate shear force $Q_u$, respectively, forces that defined the yield properties of the steel frames in the model. The values obtained for $M_u$ and $Q_u$ for the FM steel beams were around 2.8 times and 2 times larger than the FS steel beams, respectively. The material properties of the steel are provided in Table 2.

Pin supports were inserted at the bottom of the steel frame to allow flexibility in rotation while constraining the vertical and horizontal displacements. In order not to overevaluate the rigidity at the ends of each steel member, the steel column-to-beam panel zone was not set to be rigid. Previous literature has already reported that steel frame might show semirigid behavior at the welded column to frame connection (Vertes and Ivanyi 2004). The postyield behavior was based on the formation of plastic hinges, which were set to occur outside the panel zone of the steel column and beam connections.

During the calibration process, the parameters $\gamma$ and $Ø$ were changed accordingly through trial and error. The definable range in the analysis software for $\gamma$ and $Ø$ are 1–20 and 0–1, respectively. Therefore, increment steps of 1 and 0.1 in changing the values of $\gamma$ and $Ø$ was conducted until the numerical result fit the experiment result. As the result of the calibration process, the hysteresis loop parameters, $\gamma =5$ and $Ø=0.8$ were derived. Similar parameter values have also been reported in earlier literature (Ohi et al. 1992), which corroborates the values’ validity. A good agreement between the experimental and analytical results can also be seen in Fig. 11.

The infilled CLT shear wall was modeled as an I shape with a vertical beam element in the middle and horizontal beam elements on the top and at the bottom, and which is defined to have the same dimensions as CLT shear wall in the experiment. In most of the previous literature concerning CLT-hybrid structures, the CLT infill shear wall is defined as an elastic element. However, several CLT panels experienced severe shear failure during the experiments because the utilization of the CLT’s shear capacity was maximized. Therefore, the vertical beam element is defined as nonlinear to reproduce the nonlinear strength and stiffness degradation of the CLT panel near its point of ultimate failure in shear. Fig. 12(a) shows the stress–strain curve obtained from the shear element test and the model for specimens infilled with S60 grade CLT. The shear strain was obtained from the diagonal LVDTs [#3 and #4 in Fig. 10(b)] by Eq. (14) with reference to Fig. 12(b)

As shown in Fig. 12(a), much as with the failure mechanism observed during the experiments on the hybrid test specimens, there was no sudden strength degradation during the CLT shear element test. The backbone parameters for the 4-linear stiffness degradation model shown in Fig. 13 were decided by tracing the averages of the test results. The reason for adopting the 4-linear stiffness degradation model is that this model is one of the very few models incorporated within the SNAP analysis software that allows for strength degradation modeling. Second, this model is relatively easy to define because it does not require many parameters, which may also facilitate a simpler design for such a hybrid structure. Negative values can be inputted for the stiffness ratio $\beta$ and $\eta$, so that the path of the hysteresis loop can be adjusted according to the displacement at which the strength degradation occurs. The CLT material properties used for the beam element (Table 4), except the elastic modulus $E$ obtained from CLT compression test [Fig. 10(c)], were obtained from the previously conducted CLT shear element test illustrated in Fig. 10(b).

The horizontal beam elements, located at the top and bottom of the CLT beam element, were defined as rigid to allow effective load transfer between the CLT and the steel frame through the DP connections in the numerical model.

As for the compressive resistance incurred at the contact surface between the CLT panel and the nonshrink mortar, it was modeled using spring-bearing elements placed 100 mm apart. In order to reproduce the uplifting deformation of the CLT, the bearing springs were adjusted to resist only compression. Each of the bearing springs was modeled as an elastic-plastic element, and the parameter input was obtained from the CLT elemental test in compression, illustrated in Fig. 10(c), which yielded the Young’s modulus $E_w$ and compression strength $F_c$ for the CLT. It should be noted that $E_w$ was obtained using the local strain-measurement LVDTs (#3 and #4). Eqs. (15) and (16) were used for the calculation of the stiffness $K_c$ and the yield strength $F_y$ properties for the bearing spring elements. In the analysis model, each of the bearing springs is assumed to bear the compression load partially by a certain compression area. The compression springs at the ends were considered to only have half the stiffness and yield strength. The input values are provided in Table 5where $t$ = thickness of the CLT (105 mm); $d$ = distance between the bearing springs (100 mm); and $F_c$ = compression strength of the CLT. Ec/290 ($\mathrm{kN}/{\mathrm{mm}}^3$) is an equation for calculating the bearing stiffness of the CLT, proposed by Fukumoto et al. (2021). This equation was defined focusing on the compressive stress transfer from the CLT panel to the steel beam. It also takes into account of the curvature deformation of the steel beam under compression by the infill.

The overall frame-to-wall connection is the key component that ensures the two subsystems (CLT infill and steel frame) work well together to resist lateral loads in a steel-timber hybrid structure (Li et al. 2015). Modeling of this overall connection must be carefully conducted if a satisfactory result is to be obtained from the analysis. Under lateral loading, the DPs group located at the four corners of the panel may experience rotation. Due to the rotation, it may induce each DP to move at different direction and such behavior should be accounted in the model. However, from calculations of the angle deformation of the infill using the recordings of LVDTs (SPH-1–4) at the corners of the infill (refer to Fig. 6), the results showed that the infill was only subjected to less than 1° of angle deformation. Furthermore, it has been reported in previous literature that the resisting directions of the DPs in the CLT panel do not have much effect on the capacity and stiffness of the connection at least for the case of fewer number of DPs in a row, even when the DPs are in different arrangements (Nakashima et al. 2016). With the aforementioned reasoning, it was considered that the rotation behavior of the DP groups be eliminated from the analysis model, since the rotation effect is insignificant. On the other hand, during the tests, the LVDTs generally recorded only small (effectively negligible) displacements between the steel frame and the bolted connections, except for Specimen FS-S60B-32, whose ultimate failure mode of beam buckling included shear failure at the bolt connection [refer to Fig. 8(c)]. In the other specimens, no severe damage was observed at the bolted connections. Therefore, the bolted connections were considered rigid, and only the DP connection was considered when modeling the frame-to-wall connection. However, for Specimen FS-S60B-32, the stiffness of the connection was reduced by modeling the bolted and DP connections as series spring connections, in the manner shown in Fig. 14. The equation for the calculation of the stiffness in that case is given in Eq. (17)where $K_{HTB}$ and $K_{DP}$ = stiffness of the HTBs and DPs, respectively; and $K_{DP}$ was obtained from the DP cyclic elemental test, which will be describe subsequently. On the other hand, $K_{HTB}$ was obtained using Eq. (18)where $F_{shear}$ = design shear force of an F10T HTB as stated in the AIJ Standard (AIJ 2019) (equal to 33.9 kN); $n_f$ = number of surfaces for shear force transmission; $n_b$ = number of bolts in the connection; and ${\delta }_{confined}$ = displacement suggested by Shi et al. (2011). In the case of the bolt shear failure mode, which was a feature of the observed failure mode in the experiment performed on Specimen FS-S60B-32, ${\delta }_{\mathrm{confined}}$ was proposed to be 1.35 mm.

The connections of CLT panels to the steel frames were modeled using a series of axial-shear spring elements, which also have tensile and compression resisting capabilities.

The axial-shear spring elements were represented using the modified normal characteristic loop (NCL) model for wooden structures (Fig. 15), as proposed by Matsunaga et al. (2008). The original NCL model consists of a skeleton curve and a normalized characteristic loop, which can trace the slip effect, enlarging hysteresis loop area phenomenon, and stiffness degradation, which are all observed in the hysteretic behavior of RC structures. Here, the modified NCL model takes advantage of these capabilities to similar model behaviors, which can be observed in the wooden structure, while also being able to capture the complexity of varying slip lengths in the hysteresis curve obtained from the wooden shear wall tests by introducing a new exponent parameter $n$ in the characteristic loop function given in Eq. (19)where $A$ = loop area coefficient; and $B$ = slip pattern coefficient. The exponent $n$ serves to adjust the slip length of the NCL, where two different values of $n$, for example $n_{\min }$ and $n_{\max }$, are required to define the loading curve (curve showing load increment ) and unloading curve (curve showing load decrement) in each direction of the loop, as shown in Fig. 15.

Cyclic loading tests were conducted to obtain the load-deformation relationships for the frame-to-wall connection. The test setup is illustrated in Fig. 10(d). The values from the test results given in Table 6 were used to calibrate the model parameters, and the calibration results for the NCL model are listed in Table 7. It should be noted that the calibration was conducted accordingly to a specific modeling procedure proposed by Matsunaga et al. (2008). In order to define the yield strength and stiffness ratios for the backbone curve of the model, the strength properties and stiffness were adjusted by a ratio obtained by dividing the number of DPs used in the timber-steel hybrid systems by the number of the DPs in the cyclic test setup. Fig. 16 compares the connection test results and the modified NCL model as calibrated. It can be seen from the comparison that the analysis result can accurately reproduce the observed slip and pinching behavior of the DP connection under its cyclic experimental test loading.

The numerical model was subjected to the same loading protocol as the structural experiment. The model’s predictions of hysteresis loops, shear force sharing effect, and the cumulative energy dissipation of the hybrid systems were compared with the test results in this section, with outcomes shown in Figs. 17–19.

From Fig. 17, it can be observed that the plotting of the yielding of steel beam and shear failure of CLT from the test and numerical results are in good agreement. As mentioned in the section “Test Results and Discussions,” the test had to be terminated prematurely for Specimen FM-S60-32-S, due to the significant lateral torsional buckling of the beam. This explains the reason why the shear failure of CLT in the model occurred later than in the experiment for this particular specimen, because lateral torsional buckling was not considered in the numerical model. As for the yielding of DP, the DP in specimens with fewer number of DP tend to yield at relatively smaller angular displacements. The analysis model is also able to simulate such behavior and the occurrence of the yielding of DP in the experiment and analysis show good agreement, as shown in the figure. Moreover, the hysteretic behavior of the hybrid walls was similar to that of the steel bare frame, indicating that the steel frame governs the behavior of the hybrid system. The numerical results were also able to reproduce such behavior of the hybrid system.

From Fig. 18, it can be seen that the CLT infill resisted more than half of the shear force in Specimen FS-S60B-32, as described previously. As for the specimens with the more robust beams, different results were observed, in which the steel frame resisted more shear force than the CLT infill. However, when the hybrid system started to show signs of failure as the displacement increases, the prediction accuracy of the model starts to drop by a little, due to the slight difference in the timing when the failures occurred. Nonetheless, the model predictions were capable of capturing the trend of the load sharing effect between the two subsystems.

The accumulative energy dissipation, obtained by calculating the area under the hysteresis loops, between the numerical and test results is compared in Fig. 19, which also indicates good agreement at all test cycles.

As described previously, overall, the numerical model provided reliable predictions for the behavior of all the specimens. Since the model’s predictions can reproduce the empirical test results, including the unloading paths, the model could be applied to the seismic analysis of a structure that consists of such a hybrid system in future studies.