Experimental Cyclic Response of a Novel Friction Connection for Seismic Retrofitting of RC Buildings with CLT Panels

The seismic retrofitting of existing buildings is a comprehensive and contemporary research topic with diverse economic, social, and environmental entanglements. This paper is part of the European Energy and Seismic Affordable Renovation Solutions (e-SAFE) project (e-SAFE n.d.), which presents a multidisciplinary approach on renovation solutions for existing buildings, in the framework of the Horizon (2020) European goal. The authors focus on seismic retrofitting, which represents an all-embracing challenge, especially in earthquake-prone areas in southern Europe. As an example, 60% of Italian residential buildings were built between 1946 and 1990, while 25% were built before 1946 (Istat 2015). The Italian seismic hazard and the inadequacy of seismic provisions before the 1980s confirm the Italian territory’s significant seismic risk. There are several solutions for seismic retrofitting of RC buildings. Specifically, there are two types of interventions, those aiming to reduce the seismic demand and those aiming to increase the structural capacity (Pantazopoulou et al. 2016; Di Ludovico et al. 2017; Caterino et al. 2008). The structural capacity may be increased by adopting strengthening interventions or by installing stiffening elements. So far the strengthening interventions are the most widespread. They can be traditional [e.g., steel jacketing of beams and columns (Dubey and Kumar 2016)] or based on the use of advanced materials [wrapping with fiber-reinforced polymers (FRPs) or textile-reinforced mortar (Del Vecchio et al. 2014; Bournas et al. 2009)]. Properly designed stiffening elements [e.g., RC shear walls, X-bracing (Rahimi and Maheri 2018, 2020)] could partially adsorb the seismic load. However, the seismic vulnerability could also be reduced by lessening the seismic demand. These interventions include the use of dissipation devices (Javadi et al. 2020; Di Sarno and Manfredi 2010; Barbagallo et al. 2017a) or base isolation (Clemente and Buffarini 2010; TahamouliRoudsari et al. 2018; Barbagallo et al. 2017b). The solutions aiming at increasing the structural capacity are invasive and generally expensive. The interventions that reduce the seismic demand could be advantageous, being less invasive and reducing installation time. In this field, friction connections (FCs) could be considered possible candidates for increasing the structural dissipation capacity (Khoo et al. 2015; Borzouie et al. 2016). Still, these devices are not common due to difficulties in their practical installation and the uncertainty of the RC building’s coupled response. So far, full-scale experimental tests on the effect of asymmetric friction connections (AFCs) on RC frames are lacking (Rad et al. 2019; Khoo et al. 2013). However, many scholars started investigating the possibility of using timber as a strengthening solution for existing buildings, particularly cross-laminated timber (CLT) panels (Björnfot et al. 2017). In RC buildings with masonry infill, CLT panels can be used in addition or as substitution (Sustersic and Dujic 2013; Stazi et al. 2019) of existing masonry infill. The main issue related to these interventions is implementing an appropriate connection system between the CLT panel and the existing structure. The system proposed by the authors is called e-CLT (Fig. 1) and consists of installing a CLT panel from the outside of the building while leaving the masonry infills unchanged, without discomfort for the people living inside the building (Tardo et al. 2020). CLT panels have light weight (around $470\mathrm{kg}/{\mathrm{m}}^3$), thus not increasing too much the mass of existing buildings. Other advantages that make CLT attractive for retrofitting uses are the high level of prefabrication and the benefits of dry interventions, such as quick and easy installation, material recyclability, and reversibility of the retrofit intervention. The central innovation of the e-CLT retrofitting intervention stands in the connection system between CLT and the existing building: a friction connection. The FC is constituted by a couple of steel profiles, one connected to the existing beam and the other to the CLT panel, clamped together by preloaded bolts. One of the two profiles presents a slotted hole, which enables their mutual sliding. Single CLT panels are connected to the RC beams of the structure by at least two FC. The size of the CLT panels is related to that of the bays without openings where they are applied. In common RC framed buildings, the story height is generally equal to 3 m, while the width of each bay does not exceed 5–6 m. The e-CLT technology can be combined with nonstructural framed panels, which may be applied to the walls with openings and are equipped with high-performing windows that replace the existing ones. Both panels integrate insulation materials to improve the energy efficiency of the building and a finishing layer. The retrofit system also provides technological solutions to cover the FC devices after the panels installation and to ensure their inspection and maintenance (Margani et al. 2020).

This paper reports a testing campaign on the novel FC developed within the e-CLT retrofitting technology, but is also applicable to other types of structural panels or retrofit systems. The authors investigated the cyclic response of four different prototypes by focusing on their dissipative performance. The goal is to understand the prototypes’ behavior for a mindful assessment of their potential for future developments. The main novelty of this paper lies in the development and testing of a new AFC for seismic retrofitting intervention coupled with CLT panels; both design choices regarding the shape and mechanical behavior of the system will be discussed. The investigation of the connection between FC and the CLT panel and between FC and the concrete beam is outside the scope of the current paper. The experimental activity was directed at isolating and studying the friction behavior of the FC, representing the innovative part of the system. The connections to the CLT panel and to the concrete are known from the literature and standards. Their design will follow the capacity design rules with respect to the sliding force of the FC. The structure of the paper is as follows: after an overview of previous investigations on friction connections, tested specimens and setup are described; results are presented in graphical and tabulated form, along with a brief explanation of the data elaboration; a discussion completes the presentation of results, along with a possible modeling approach; and in the end conclusions are drawn.

Several researchers investigated the performance of friction connections in their basic form, usually divided into two classes: symmetric friction connections (SFCs) and AFCs (Latour et al. 2019; Loo et al. 2014c). Symmetric friction connections present one central plate with an elongated hole and two lateral plates with round holes. In this connection, the central plate slides while the two lateral plates are restrained, determining symmetric loading conditions. In asymmetric friction connections, one plate is restrained while the other slides. This configuration determines asymmetric loading condition. A third plate with round holes may be used to close the connection as a cap plate. Additionally, shim layers made from various materials may be used between the three plates to increase the stability of the friction behavior. An example of the tested specimen, which is classified as AFC, is shown in Fig. 2.

Loo et al. (2012) initially used a numerical study to investigate the possibility of using SFC instead of hold-down for restraining timber shear walls against uplift, cap the force transmitted to the wall, and reduce inelastic damage. Consequently, Loo et al. (2014a, b) presented two experimental campaigns: the first on single SFCs and the second on full-scale timber walls. In Loo et al. (2014a) the focus was on the single connectors, with different materials tested as shim layers, such as brake lining and different alloys. An innovative connection without shim layers was also tested, where the central slotted element was made directly with the high-hardness alloy. They found that high-hardness alloys offered the best sliding performances, like Bisalloy 400, while steel versus steel sliding presented a more erratic behavior. In Loo et al. (2014b) the concept of SFC was applied to a full-scale timber wall, where the friction connections replaced the hold-downs. The results showed that these devices determined a global elastoplastic response of the entire wall under cycling loading. Loo et al. (2016) further extended the concept by modeling CLT walls and multistory buildings and by proposing a displacement-based design method. Loo et al. (2017) tested SFC by adopting different methods of surface preparation of the sliding interface. Surfaces cleaned from loose rust and mill scale yielded the best hysteresis curves in terms of stability. Fitzgerald et al. (2021) studied an SFC with self-tapping screws for CLT, which was then also used for a full-scale campaign in Fitzgerald et al. (2020). Chanchi Golondrino et al. (2020) investigated the performance of SFCs in steel buildings. Golondrino et al. (2012a, b) presented two experimental campaigns on AFCs. The first (Golondrino et al. 2012a) aimed at understanding the behavior of the AFC when using different materials as shim layers. Materials with high hardness led to extremely stable loops, low-hardness materials determined moderately stable loops, while an erratic behavior was observed with medium-hardness materials (hardness values similar to steel plates). The second (Golondrino et al. 2012b) established a dependence between stability and shape of the loop and the clamping force level. Golondrino et al. (2016, 2020) presented an experimental campaign on AFC focused on the bolt length effects. They found that increasing the bolt length determined a decrease in AFC strength and a more pronounced preload loss; conversely, the effective friction coefficient and its variability tended to decrease. Chanchí Golondrino et al. (2018) tested brake pads as shim material for AFC. Rodgers et al. (2018) studied the dynamic effects on AFC for applications in steel structures, which further led to the design proposals in Chanchi Golondrino et al. (2019).

Friction connections were selected as appropriate for the e-CLT seismic retrofitting intervention due to their highly efficient energy-dissipation mechanism, which contributes to reduce interstory drift damage to the existing building. The damage-free energy dissipation mechanism also ensures the durability of its structural efficiency even after seismic events. This feature avoids the need for the removal and replacement of the FC, which in turn would involve the removal of the attached CLT panel. After a seismic event, it is sufficient to check and eventually replace the bolts and the shim layers. Another important feature, different from viscoelastic devices, is that these connections present stable hysteresis loops that are not dependent on the speed (Jaisee et al. 2021).

This section illustrates the results of the testing campaign. Figs. 10–12 show the results in graphical form for all the specimens. For each specimen the first graph represents the force displacement loops, which provides insight on the overall behavior of the friction connection. The second graph visualizes the measured force plotted against time for an enhanced readability of the stability of the force throughout the experiments. The third graph presents the dissipated energy during testing, which is a useful parameter for the definition of the slip force because it is an increasing function. Table 3 summarizes the results by reporting significant values defined in the following paragraphs.

The definition of the slip force $F_{\mathrm{slip}}$ from the experimental data is neither straightforward nor unique. The author decided to use the same approach used by Loo et al. (2014a), which adopts a definition related to the dissipated energy $E$where $E_i$ = energy at the $i$th time step; and $F_i$ and ${\delta }_i$ = force and displacement at the same time step, respectively. The dissipated energy presents a strictly increasing trend, and is therefore a useful unambiguous parameter for the calculations, as opposed to the force value. The cumulative distance of travel $D$ is the sum of the displacement time steps

The slip force is defined as the work per unit of length

The standard deviation is calculated with reference to the $F_{\mathrm{slip}}$ value

The coefficient of variation (COV) and the stability parameter $\lambda$ are calculated as follows:

These two values express the loop stability and the sliding behavior nature: a low COV, and thus a high $\lambda$ value, corresponds to stable performance, resembling the ideal rectangular shape. The experimental friction coefficient $\mu$ is calculated as follows:where $F_{\mathrm{slip}}$ = slip force calculated in Eq. (3); $n_s$ = number of shear surfaces; $n_b$ = number of preloaded bolts; and $F_P$ = preload force from Table 2.

Fig. 13 attempts to capture an essential aspect of the tests on friction connections: the definition of $F_{\mathrm{slip}}$ is substantially dependent on the loading path, while its value is not constant during the test. The authors present the final values in Table 3 related to each test. They also plot their time dependency in Fig. 13. The $x$-axis represents the displacement step (see Table 2 for the list of the displacement steps of the various load protocols), while, on the $y$-axis, the circles represent the slip force calculated from the beginning of the test up to a specific displacement step. Fig. 13 manifests two sorts of scatter: the horizontal variation, relative to changes in slip force between successive displacement steps, and the vertical variation. The latter derives from the repetition of three cycles at the same displacement step: an appreciable vertical distance between the three circles representing $F_{\mathrm{slip}}$ indicates a considerable variation of force during the repetitions of the same displacement step. In Fig. 13, the authors also report the friction coefficients, which follow the same trend of the slip force.

The first test, the monotonic STD.1, evidenced the weak point of the STD design from the very beginning. The outward bend of the bottom profile suffered from big deformations, which determined the test to stop. The FEM analysis in Tardo et al. (2020) also confirms the presence of strong stress concentrations in that area; additional information is available in Hatletveit (2020) and Marthinsen (2020). The load–displacement curve shows that the slip force was approximately 60 kN [Fig. 10(a)]. However, when the deformation of the bottom profile increased, the bolts started pressing against the side of the holes, thus increasing the resisting force. The increased resisting force did not depend on friction solely, but it was a combination of friction and bolts-holes plasticization. The deformations originated from the eccentricity between the applied force and the bend of the bottom profile: the upper part deformed outward while the lower part deformed inward, as seen in Fig. 14(a). Another significant deformation was a relative twisting between the two sliding surfaces, which made the previous effect even worse and was caused by the eccentricity between the two preloaded bolts, also marked in Fig. 14(b). Testing of the STD design specimen, despite some expected issues, was important to have a base for comparison with further improvements. The goal was to understand if, with some improvements and small modifications of the STD design, it was possible to achieve satisfactory mechanical behavior with a design the offers the frontal mounting possibility.

In the STD.2, STD.3, and STD.4 tests, the preload force was lowered to 30% of the previous values and the speed was also reduced to $0.5\mathrm{mm}/\mathrm{s}$ in order to obtain a lower slip force and minor deformations of the specimen. The curves of these tests, depicted in Figs. 10(d, g, and j) and compared in Fig. 15, exhibit an extremely erratic shape, with unpredictable behavior: Tests STD.2 and STD.4 present similar shapes and level of forces, while STD.3 reached higher forces rapidly, thus causing the stopping of the test before the completion of the whole protocol. Both Tests STD.2 and STD.4 exhibited preload losses in the bolts at the end of the test, but STD.3 did not. This occurrence may be the motivation for such prominent differences in behavior. In both STD.2 and STD.4, the hysteresis loops are remarkably similar, presenting a sudden drop in force after the initial peaks, possibly indicating the instant of preload loss. Twisting deformations were more pronounced throughout the whole duration of Test STD.3, which was the one that retained the preload until the end of the test, if compared to STD.2 and STD.4. The STD.2 and STD.4 tests also presented twisting deformations, mostly during the initial part of the test. In all these experiments, significant deformations were always present in the outer bend of the bottom profile. The authors also observed a notable scraping of steel between the washers and the sides of the holes.

Test STD.5 had two modifications: an 8-mm steel cap plate to limit steel scraping and two shim layers made with 2-mm-thick aluminum to obtain more stable loops. The authors maintained this AFC configuration throughout the rest of the campaign. Nevertheless, the performances of STD.5 were still poor [Fig. 10(m)]. As observed in the previous cases, STD.5 exhibited considerable deformations and consequent force spikes, which did not allow the completion of the loading protocol. The force continued increasing, with the absence of preload losses at the end of the test. Again, the resisting force did not depend on the sole friction, but it was a combination of friction and bolts pushing against the holes. The values of the slip forces and friction coefficients of the STD tests, reported in Fig. 13, are then rather approximate, because the tests did not exhibit a pure frictional behavior.

Although the STD-R.1 specimen had additional reinforcements, the behavior was comparable to the STD.5 test [Fig. 11(a)]. The reinforcement in the boxlike part of the specimen contributed to isolate the weak spot evidenced during the FEM analyses even more: the outer bend of the bottom profile. Twisting was also present, and the large deformations and rapid force increase led the experiment to an early stop, as in the previous case.

STD-1H.1 showed an almost rectangular loop behavior, shown in Fig. 11(d). Twisting deformation was significantly reduced, indicating that a proper bolt arrangement in a single slotted centered hole effectively mitigated the detrimental phenomena identified for Specimens STD and STD-R. The lower force level also determined a lower bending deformation than in previous cases. The preload was lost at the end of the test. This aspect also emerges from observing the value of the slip force, which presents some initial peaks around 16–17 kN, later declining to $F_{\mathrm{slip}}=8.1\mathrm{kN}$, as shown in Table 3 and Fig. 13(g). In the same figure, it is possible to see that the friction coefficient value is unexpectedly small. This value depends on the assumption of the initial preload force value used for calculation, which in reality decreased during the test. In Test STD-1H.2, the authors increased the preload to obtain a higher slip force. Still, higher deformations also occurred. Twisting was more evident than in the previous case, even if more limited than in the STD specimens. The bending deformations were quite noticeable during this test, possibly due to the absence of significant preload losses. This phenomenon suggests that higher stresses arose inside the bottom profile. The force rapidly increased to the 50-kN limit value of the testing machine during the early cycles; therefore, the test stopped before the completion of the entire Load Protocol C.

The ALT specimen showed promising results. The loops in Fig. 12(a) closely resemble a rectangular shape, and the force level agrees with analytical predictions. The friction coefficient is in accordance with scientific literature (Golondrino et al. 2012a), approximately equal to 0.2 for aluminum, as observed from Fig. 13(i). Both twisting and bending deformations were almost absent during this test, meaning that the reduced eccentricity and the single bend played a vital role in obtaining stable loops, confirming the initial FEM results of Tardo et al. (2020). The loop shape presents a corner chipping when the load is changing direction: this is common in asymmetrical friction connections (Loo et al. 2014a; Fitzgerald et al. 2021) and depends on dragging phenomena of the cap plate. The preload force was then increased in the remaining tests on the ALT design, with the goal of reaching a 30-kN slip force. The ALT.2 results, shown in Fig. 12(d), present a lightly less stable behavior. Still, the slip force and friction values agree with the expected ones, as confirmed by Fig. 13(j). One significant difference in shape concerning the previous test regards the first cycles. There were lower initial resisting forces in Test ALT.1, stabilizing at higher values close to the slip force. Conversely, there were more prominent peaks in the initial cycles of Test ALT.2, which then gradually declined and reached the stable force level. This aspect is also evident when comparing Figs. 13(i and j): while in the first case, it is possible to seize an increasing trend of the plotted $F_{\mathrm{slip}}$ development, in the second case, after the peak, the trend is decreasing. Test ALT.3 included a change of speed: from 0.5 to $2\mathrm{mm}/\mathrm{s}$. The results prove that the change in speed did not have any noticeable effect on the connection behavior, as highlighted by the comparison in Fig. 16, thus confirming one positive feature of friction connections: being independent from the speed of loading. The specimen was also the same as in the previous test, indicating that the connection response does not suffer cumulative damages. Friction coefficient values again agree well with Golondrino et al. (2012a), who found a dynamic friction coefficient value of 0.18 for aluminum and a static value of 0.34. The present test provides a friction coefficient value of 0.21 (dynamic condition) and a value of 0.33 when referred to $F_{\mathrm{peak}}$ (static initial condition). ALT.4 was the last test, and in this case the displacement protocol reached the full 100-mm sliding capability of the slotted holes. The load–displacement graph shows that up to 50 mm, the behavior is the same as the previous tests, while for higher levels of displacements, there is a moderate loop instability. In these last three tests, slight twisting deformation was observed mainly at the change of direction of the load and not continuously as in the STD and STD-R design.

This section presents a possible approach for the simulation of the observed experimental behavior for inclusion in further nonlinear dynamic analysis. The performance of the AFC is mainly related to Coulomb friction. The function defining the friction force must take into account the observed force decrease with cycleswhere $F_{\mathrm{slip}}(ϵ)$ = slip force; $F_P$ = preload force; $\dot{d}$ = time derivative of the displacement; and $ϵ$ = cumulated dissipated energy. The friction coefficient is defined aswhere ${\mu }_0$ and $\xi$ = parameters estimated from the experimental data. As an example of application, Fig. 17 shows the superposition of experimental data and simulated model for the Test ALT.3; the parameters ${\mu }_0=0.33$ and $\xi =0.00005$ were estimated from an ordinary least-squares optimization.

The proposed model accurately reproduces the experimental behavior, including the effect of decreasing slip force values for higher cycles. The energy dissipation is higher than the experimental value, and this is caused by the nonrectangular shape of the experimental loops due to the corner chipping effect of asymmetric friction connections. The simple proposed model, on the other hand, generates loops with perfect rectangular shape, thus leading to higher dissipated energy values. The authors neglect this phenomenon at this stage, given the limited impact on the dissipated energy.

This paper presents a testing campaign on a novel FC for seismic retrofitting of an existing RC frame building. The entire system, called e-CLT, is part of the European Horizon (2020) project e-SAFE. The seismic retrofitting originates from the installation of CLT panels attached to the outside of a building with innovative friction connections. The connections are asymmetrical friction connections, composed of a couple of cold-bent steel profiles. One steel profile is connected to the CLT panel, while the other is connected to the RC beam. The two profiles are held together by preloaded bolts, which can slide in slotted holes and dissipate energy via friction. This paper focuses on the experimental tests of the FCs for the estimation of their dissipative capacity. The authors carried out an experimental campaign on four different geometries: the standard (STD), two variations [one with reinforcements (STD-R) and one with a single elongated hole (STD-1H)], and an alternative with reduced eccentricity and lower number of bends (ALT). The main findings were that STD and STD-R suffer from major deformations due to bending and twisting, and the reason lies in two features: an intrinsic weakness of the outer bend of the bottom profile, because of the eccentricity, and the presence of two not aligned elongated holes. STD-1H and ALT showed promising results, indicating that the arrangement of a single elongated hole solved some of the deformation problems and that the reduced eccentricity and lower number of bends of ALT design was effective in isolating the friction behavior. The authors obtained acceptable results using aluminum for the shim layer, whose friction coefficient, nearly equal to 0.2, agrees with literature formulations. Aluminum as shim material provided stable hysteresis curves, while its absence caused an erratic response (e.g., steel-to-steel friction). ALT tests also confirmed the independence from loading speed of the friction connections. A possible modeling strategy was presented based on an exponential Coulomb friction model that reproduces the observed experimental behavior. In general, the main conclusions are that

The major limitation of the experimental campaign presented in this paper is the limited number of specimens tested and replications. However, these tests allowed studying the system at a small scale related to the friction connection, which was useful to investigate the potential of the system. The tests confirmed that the performance parameters of the system are related to the geometry: the eccentricity and the number of bends. As future work, the authors aim at expanding the experimental campaign by investigating the issues that arose during the current campaign: choice of different structural arrangements, use of different materials as shim layers with higher hardness than steel, and developing hysteresis models for the complete e-CLT system (Aloisio et al. 2022a, b).

This section provides the technical drawings of the tested specimen, which include the details on the measure and locations of the holes for the sliding movement. Fig. 18 shows a technical drawing of the STD specimen, Fig. 19 of the STD-R specimen, Fig. 20 of the STD-1H specimen, and Fig. 21 of the ALT specimen.

Some or all data, models, or code that support the findings of this study are available from the corresponding author upon reasonable request.

The authors would like to thank Alberto Moretti and the Italian company Adveco s.r.l. for their relevant and continuous support during the prototyping stage and for providing the specimen for the testing campaign; the master’s student Mathilde Marthinsen for her contribution during the testing campaign; and the head engineer of NMBU Øyvind Hansen for his continuous support regarding the design and production of the test setup. This paper was carried out in the framework of the Energy and Seismic Affordable Renovation Solutions (e-SAFE) project, which has received funding from the European Union’s Horizon 2020 research and innovation program under Grant Agreement No. 893135. Neither the Executive Agency for Small and Medium-sized Enterprises (EASME) nor the European Commission is in any way responsible for any use that may be made of the information it contains.