Insights on the experimental dynamic behaviour and energy losses of rocking blocks under free vibration

ORIGINAL ARTICLE Received: 7 February 2025 / Accepted: 6 July 2025 © The Author(s) 2025 Carla Colombo [email protected] 1 Department of Civil Engineering, University of Minho, ISISE, ARISE, Guimarães, Portugal 2 Department of Civil and Environmental Engineering, The University of Auckland, Auckland, New Zealand 3 Clermont Auvergne INP, CNRS, Institut Pascal, Université Clermont Auvergne, Clermont-Ferrand F-63000, France Insights on the experimental dynamic behaviour and energy losses of rocking blocks under free vibration CarlaColombo1· GeorgiosVlachakis1· Anastasios I.Giouvanidis2· NathanaëlSavalle3· NunoMendes1· Paulo B.Lourenço1 Bulletin of Earthquake Engineering https://doi.org/10.1007/s10518-025-02224-8 Abstract This study focuses on the experimental response of free-standing hard limestone blocks under free vibration. The campaign includes 120 tests, varying the block’s height-to-width ratio and considering multiple specimens to account for the aleatoric variability of the phenomenon. The study offers full reconstruction of the three-dimensional free-rocking motion, giving insights into the influence of unintended geometrical asymmetries, as well as material and interface irregularities on the response. The paper revisits the experimental estimation of the angular Coefficient of Restitution (CoR) through different methodologies based on the angular velocities, potential energy, and potential and frictional energies. The results are compared with Housner’s theoretical model. Due to the sensitive nature of rocking motion, a statistical approach is employed. The findings indicate that Housner’s model provides statistically accurate predictions of energy losses for blocks of mediumto-high slenderness (i.e. with aspect ratio 5≤H/B≤10 while it becomes statistically inaccurate for very slender H/B>10 and very stocky H/B<5 blocks. Importantly, the study demonstrates that all three experimentally estimated CoRs statistically depend on the aspect ratio. Finally, the CoRs extracted from the potential energy and potential and frictional energies are found statistically dependent on the rocking amplitude, while the CoR extracted from the angular velocities show statistical independence. Overall, this study offers a pathway for more systematic analyses and improved predictions of energy losses during impacts. Keywords Free-rocking · Coefficient of restitution · Unreinforced masonry · Out-ofplane · Experimental campaign · Digital image correlation 1 3 Bulletin of Earthquake Engineering 1 Introduction Rocking is a phenomenon frequently observed in slender and rigid structures during earthquakes. Under strong shaking, such structures may uplift and pivot about their base, as seen in bridge piers (Cheng 2007; Dimitrakopoulos and Giouvanidis 2015; Giouvanidis and Dong 2020; Makris and Vassiliou 2013; Reggiani Manzo and Vassiliou 2019), water tanks (Housner 1963), obelisks, columns, and unanchored equipment (Konstantinidis and Makris 2009; Makris and Vassiliou 2013; Psycharis 2018). Rocking also occurs in unreinforced masonry structures, affecting façades (Cappelli et al. 2020; Lagomarsino 2015; Mauro et al. 2015; Vlachakis et al. 2020) and/or non-structural elements, such as parapets and statues (Cocuzza Avellino et al. 2021; Giaretton et al. 2016). Specifically, if the restraints provided by diaphragms and lateral walls are insufficient, localised macroblocks can detach from the structure and experience a dynamic behaviour resembling that of a rocking block (Casapulla et al. 2017). The work of Housner (1963) marked the first systematic analysis of the rocking motion of rigid structures, deriving the equations of motion and introducing an angular Coefficient of Restitution (CoR) to quantify energy losses. However, later experimental studies (Aslam et al. 1980; Priestley et al. 1978) revealed sensitivity of Housner’s model to input parameters, such as the geometric features of the block (Aslam et al. 1980; Lipscombe and Pellegrino 1993; Spanos and Koh 1985; Tso and Wong 1989; Yim et al. 1980). The idealised assumptions of perfect geometry, purely two-dimensional motion, continuous contact with the ground, and fully plastic impacts often fail to hold in practice, as real structures exhibit sliding, bouncing, and response variability due to interface imperfections and irregularities (Lipscombe and Pellegrino 1993; Peña et al. 2008). This has led to more advanced models incorporating three-dimensional motion (Bao and Konstantinidis 2020; Chatzis and Smyth 2012; D’Altri et al. 2024; Di Egidio et al. 2014; Várkonyi et al. 2022) and asymmetric geometries (Mathey et al. 2016; Purvance et al. 2008; Wittich and Hutchinson 2015). Another challenging aspect of rocking motion is associated with the energy losses due to impacts. Housner’s model captures the energy dissipation at impact through the angular CoR, derived under the assumption of angular momentum conservation. However, later studies showed that the CoR is sensitive and relies on complex dynamic phenomena, thus questioning the assumptions underlying Housner’s model (Lipscombe and Pellegrino 1993). In response, several refined models have been developed. Some introduce modifications to the location of the impact point (Chatzis et al. 2017; Kalliontzis et al. 2016; Ther and Kollár 2017), while others relax the assumption of purely plastic impacts by incorporating the kinematic normal Coefficient of Restitution (CoRN) alongside Coulomb’s friction law (Brogliato et al. 2012; Giouvanidis and Dimitrakopoulos 2017; Shenton and Jones 1991). Additionally, compliant (Chatzis and Smyth 2012; Koh et al. 1986; Psycharis and Jennings 1983; Spanos et al. 2017; Vlachakis et al. 2021) and hybrid impact models (Bao 2024; Chatterjee et al. 2022; Yilmaz et al. 2009; Zhang et al. 2013; Zhao et al. 2019) are proposed in the literature, simulating the process in greater detail as a continuous phenomenon but requiring input parameters that are difficult to identify experimentally, such as interface stiffness and interface damping (Vlachakis et al. 2023). Experimental free-rocking studies have played a key role in advancing the understanding of rocking motion and its associated energy dissipation mechanism. Early free-rocking tests (Aslam et al. 1980; Lipscombe and Pellegrino 1993; Priestley et al. 1978) compared 1 3 Bulletin of Earthquake Engineering observed CoR values with Housner’s predictions. Later work explored a range of materials and geometries, including precast columns (Cheng 2007), granite blocks (Peña et al. 2008), steel blocks (Mathey et al. 2018, 2016), concrete walls (Kalliontzis and Sritharan 2018), masonry façades (Costa et al. 2013; Sorrentino et al. 2011) and aluminium blocks (Čeh et al., Čeh et al. 2018). Among them, Costa et al. (2013), ElGawady et al. (2011), Peña et al. (2008), and Sorrentino et al. (2008) extracted the CoR based on the preservation between potential and kinetic energies. A few other studies have examined the effect on the response of varying aspect ratios (Čeh et al. 2018; ElGawady et al. 2011), interface materials (Spanos et al. 2017), and imperfections (Al Shawa et al. 2012; Bachmann et al. 2018; Kafle et al. 2011; Mathey et al. 2018, 2016). Asymmetrical configurations have also been considered (Cocuzza Avellino et al. 2021; Huang et al. 2021; Purvance et al. 2008). Overall, previous experimental studies have highlighted the high sensitivity of rocking motion to geometric irregularities and energy dissipation. As a matter of fact, significant variance is commonly observed in the CoR across different experimental campaigns. Moreover, the limited number of tests, small sample sizes, and narrow range of aspect ratios often hinder the ability to generalise findings. As a result, impact and damping models are typically evaluated partially, offering incomplete interpretations of the experimental data. This paper presents an extensive experimental campaign on the free-rocking motion of limestone blocks, involving 120 tests across twelve aspect ratios. Ten repetitions per aspect ratio support uncertainty quantification and generalisation of the results. The study investigates the three-dimensional motion of the blocks and the influence of asymmetries and interface irregularities. A particular focus is placed on the experimental estimation of the angular CoR, providing alternative extraction strategies. Moreover, the work introduces a statistical analysis to compare the experimentally estimated CoR with the theoretical prediction of Housner’s model and discusses dependencies on both the aspect ratio and the amplitude of rocking motion. The paper is structured as follows. Section 2 outlines the key aspects of Housner’s model. Section 3 details the experimental campaign and data acquisition system. Section 4 describes the post-processing methodology used to estimate key parameters related to the free-rocking motion. Section 5 presents the results of a representative free-rocking test, along with an overview of the experimental campaign. Section 6 offers a statistical analysis of the experimental CoR data and, finally, Section 7 provides the conclusions and key remarks. 2 Theoretical background Consider the rigid block shown in Fig. 1a, freely standing on a rigid base and characterised by its height H , width B , mass m , moment of inertia IO about its corner pivot points O and O’, radial distance R between the centre of mass and the pivot points, and slenderness angle α= tan −1( B/H ) . When tilted with an initial rotation smaller than α and let free, the block starts free-rocking around the pivot points on its base with a rotation angle θx (Fig. 1b). The equation of motion of the block is expressed as (Housner 1963): ¨ θx=−p2[sin(αsgn (θx)−θx)] (1) 1 3 Bulletin of Earthquake Engineering where the sgn() represents the signum function, ¨ θx is the angular acceleration of the block, and p is the frequency parameter of the block defined as p=√ mgR / I 0 , where g is the gravitational acceleration. For slender blocks with a low value of the slenderness angle α , Housner suggested the linearisation of Eq. (1) and provided a closed-form solution of the differential equation. Consequently, the period of the free vibration during rocking is computed as: T = 4 pcosh−1 [1 (1 − θ x,peak /α )] (2) where θx,peak corresponds to the peak rotation θx during the half-cycle of the free vibration. Fig. 1c shows that the period of the free vibration (Eq. (2)) is strongly and non-linearly dependent on the rotation amplitude θx,peak , increasing from zero to infinite as the θx,peak/α ratio goes from zero to one. During rocking, a change in the sign of θx marks the transition from smooth pivoting to abrupt impact, accompanied by sudden energy dissipation. Housner modelled this impact as an instantaneous event and introduced the angular CoR, defined as the ratio of angular velocities after ( ˙ θ+ x ) and before ( ˙ θ− x ) impact. Moreover, Housner proposed an impact model to estimate the value of the CoR based on non-smooth impulsive dynamics. Under angular momentum conservation, the resulting expression shows that the CoR depends solely on the block’s slenderness (i.e. height-to-width ratio), independent of its size, material, or interface properties: CoR Housner =1 −3 2 sin2 α (3) Fig. 1 Scheme of a planar rocking block: (a) resting, and (b) tilted configurations; and (c) period of vibration based on Eq. (2) 1 3 Bulletin of Earthquake Engineering 3 Experimental campaign and acquisition system The free-rocking experimental campaign is performed using 60 limestone blocks. A preliminary investigation was carried out to characterise the material properties of the limestone blocks and the limestone-to-limestone interfaces (Vlachakis et al. 2023). Table 1 summarises the average values of the characterisation campaign, including the density of the material ρ , the compressive strength fc , the modulus of elasticity E , the static friction coefficient µstatic , and the normal kn and tangential ks interface stiffness, which varied widely depending on the normal stress. The normal and tangential interface stiffness are mechanical parameters describing the dry contact behaviour, which are essential for researchers aiming to model the rocking behaviour of the tested blocks using compliant contact models. The results of this characterisation campaign suggest that it is reasonable to assume that the blocks behave as rigid bodies, with most of the flexibility concentrated at the interfaces. Further details on the characterisation campaign are available in Vlachakis et al. (2023). The limestone blocks have nominally identical plan dimensions but varying heights (Fig. 2a). Specifically, each block nominally measures 50 mm in width B and 150 mm in length L , while the height H ranges from 200 mm to 750 mm, varying in 50 mm increments. This results in 12 aspect ratios H/B ranging from 4 to 15. The broad variation of slenderness ( H/B ) of the specimens has been selected as it is representative of real-world rocking structures, as indicated in studies such as Casapulla et al. (2017), while the length-to-width ratio ( L/B ) equal to three is intended to minimise three-dimensional effects and promote a predominantly two-dimensional rocking response (Costa et al. 2013; Peña et al. 2008; Ther and Kollár 2017). Table 2 summarises the geometrical features of the blocks, including the measured dimensions with the Coefficient of Variation (CoV) shown in parentheses as a percentage, the geometrical ratios, and the number of blocks. Each aspect ratio group consists of five nominally identical blocks. Nevertheless, each block can be tested in two configurations by rotating it 180° around the x-axis, thereby allowing the two opposite horizontal surfaces ( B×L ), top and bottom, to be considered as distinct testing interfaces. This is important because the rocking response is highly sensitive to slight variations of the contact interface, which can significantly impact the response of the blocks. Therefore, each unique interface accounts for an independent test, resulting in a total of 120 tests conducted during the free-rocking campaign. Figure 2b provides an overview of the experimental setup. A free-standing block rests on a fixed bottom block made of the same limestone material, with the contact surface between them forming the interface (properties detailed in Table 1). To initiate a test, the top block is manually rotated and positioned at rest against an external retaining system. This retaining structure consists of a stiffened L-profile anchored to the substructure and it is designed to accommodate a threaded bar. As the bar is gradually screwed, it causes the block to rotate slowly until it reaches its slenderness angle, at which point the block begins to rock freely on its fixed base without any external excitation. This method is used in other experimental campaigns and has proven to be more effective than alternative approaches (Sorrentino et Limestone block Limestone-to-limestone interface ρ [kg/m3] 2240 µstatic [-] 0.70 fc [MPa] 47.6 kn [MPa/mm] 15 (at 0.1 MPa) − 100 (at 1 MPa) E [GPa] 32.7 ks [MPa/mm] 5 (at 0.1 MPa) − 30 (at 1 MPa) Table 1 Average mechanical properties of the limestone blocks and limestone-to-limestone interfaces 1 3 Bulletin of Earthquake Engineering al. 2011). The motion until rocking commences is disregarded, while the end of the motion is assumed when there is a higher than two orders of magnitude difference between the peak rotation amplitudes and the measurement noise. The free-rocking motion is captured using a four-camera Digital Image Correlation (DIC) acquisition system (Fig. 2c). The DIC is an optical, contactless measurement technique, which allows the estimation of the field of displacement of objects by comparing a series of subsequent images relative to a reference image (Sutton et al. 2009). The benefit of using this monitoring strategy relies on the capability of tracking the specimen at a sufficiently high speed without physically interfering with the experimental setup and reaching a measurement accuracy of around 0.01 mm. Accurate image tracking relies on stable lighting conditions and the quality of the speckle pattern applied to the specimen. In this study, a high-contrast grey-scale pattern is applied to two side surfaces of the block (Fig. 2b). The DIC system features stereoscopic cameras with 25 mm lenses, capable of recording 145 frames per second. The four cameras are positioned around the perimeter of the experimental setup, with each pair focused on one surface of the block (Fig. 2c). The DIC data processing is performed using the ISTRA 4D commercial software from Dantec Dynamics. Fig. 2 (a) Limestone blocks with varying heights, (b) side view of the experimental setup, (c) 3D scheme of the DIC acquisition system; and (d) displacement contour plot obtained from the DIC measurements 1 3 Bulletin of Earthquake Engineering From the relative 2D perspective of the four cameras, each point of an object surface can be reconstructed into a 3D perspective using the principles of stereo-triangulation (Fig. 2d). For the purpose of the current investigation, the information in terms of coordinates and displacement along the x-, y-, and z-axes of two gauge points per surface are extracted. Fig. 2c clarifies the axes, gauge points’ location and their denomination. Moreover, the displacement and coordinate values of each point are associated with an uncertainty parameter, which indicates the accuracy of the measurement. 4 Post-processing methodology 4.1 Three-dimensional motion Assuming the limestone blocks behave as rigid bodies, their three-dimensional motion can be described by six degrees of freedom, shown schematically in Fig. 3: three translations ux,GC , uy,GC , uz,GC , and three rotations θx , θy , θz , around the x-, y-, and z-axes. The reconstruction of the three-dimensional motion is achieved using the Kabsch algorithm (Lawrence et al. 2019), which processes the raw data from the DIC in the form of coordinates and displacements of the four gauge points shown in Fig. 2c. The three translations ( ux,GC , uy,GC and uz,GC ) are referred to the Geometric Centroid (GC) of the four gauge Table 2 Geometrical features of the free-standing blocks. The blocks’ dimensions are accompanied by the Coefficient Of Variation (CoV) reported in parentheses as a percentage. The listed aspect ratios refer to the nominal dimensions Block group Height H [mm] Width B [mm] Length L [mm] H/B [-] L/B [-] No. of blocks 1 202.4 (0.24%) 50.1 (0.26%) 153.2 (0.26%) 4 3 5 2 248.0 (0.36%) 49.8 (0.68%) 150.2 (0.27%) 5 3 5 3 299.8 (0.13%) 50.3 (0.62%) 150.0 (0.00%) 6 3 5 4 349.6 (0.14%) 50.7 (0.38%) 150.2 (0.27%) 7 3 5 5 398.6 (0.12%) 49.3 (0.37%) 151.0 (0.42%) 8 3 5 6 448.8 (0.17%) 49.3 (0.66%) 151.6 (0.32%) 9 3 5 7 498.4 (0.24%) 48.9 (0.28%) 151.2 (0.49%) 10 3 5 8 551.4 (0.25%) 49.4 (0.68%) 152.2 (0.49%) 11 3 5 9 598.0 (0.22%) 49.5 (0.42%) 152.0 (0.31%) 12 3 5 10 652.2 (0.44%) 49.2 (0.56%) 152.0 (0.59%) 13 3 5 11 701.8 (0.14%) 50.8 (2.84%) 153.4 (0.32%) 14 3 5 12 750.6 (0.32%) 51.6 (3.10%) 151.6 (1.07%) 15 3 5 1 3 Bulletin of Earthquake Engineering points and are calculated for each direction as the mean displacement of the four points. The computation of the rotations requires the estimation of the rotation matrix R , which is reconstructed using the singular value decomposition method (Lawrence et al. 2019). The rotations are derived from the rotation matrix R following the convention of the Tait-Bryan Euler angles as follows: θx= tan−1(−R23/R33) (4) θy= sin−1(R13) (5) θz= tan−1(−R12/R11) (6) where each element Rij corresponds to the components of the 3×3 rotation matrix R . Consequently, the Euler anglular velocities relative to each rotation ( ˙ θx , ˙ θy and ˙ θz ), are Fig. 3 Degrees of freedom of a rigid block: translations along the: (a) x-axis ( ux,GC ) (b) y-axis ( uy,GC ), and (c) z-axis ( uz,GC ); and rotations over the (d) x-axis ( θx ) (e) y-axis ( θy ) and (f) z-axis ( θz ) [Note: to resemble the actual rocking motion, the rotations θx and θy are represented as rotations around the pivot axes, i.e. coupled with translational components along the x and y axes, respectively] 1 3 Bulletin of Earthquake Engineering obtained by differentiating the rotations over time using the second-order accurate central differences. Similarly, the Euler angular accelerations ( ¨ θx , ¨ θy and ¨ θz ) are estimated by differentiating the Euler angular velocities. The actual geometry of each limestone block differs from the nominal geometry. This is due to several factors, such as variations in block dimensions resulting from the stonecutting process, shifts in the centre of gravity due to material heterogeneity, surface irregularities at the contact interfaces that can affect the position of the pivot points or create unintended inclinations of the free-standing block, interface accumulated damage during testing, and others (Costa et al. 2013; Kalliontzis and Sritharan 2018; Peña et al. 2008). To account for these imperfections, this study performs an experimental determination of the parameters p and α of each test, by using as reference the peak rotations and periods of each free-rocking half-cycle (Peña et al. 2008). Specifically, p and α are estimated by using the analytical expression proposed by Housner (Eq. (2)) and through a fitting process that minimises the root mean square error between measured and computed periods. During this fitting process, the half-cycles with positive and negative rotations are treated separately, thus resulting in p+ and α+ for θx>0 , and p− and α− for θx<0 . The post-processing methodology includes the computation of the potential U , kinetic K and total E energies of the rocking blocks as follows: U=mgR±[cos (α±−|θx|)−cos α±] (7) K=0 . 5I o,±˙ θ 2 x (8) E=U+K (9) where the subscript symbol “ ± ” refers to the positive and negative signs of rotation. R± is obtained from p± = √ mgR± Io, ± , where I o, ±=m 12 ( B 2+ H 2)+ mR ±2 , with B and H being the nominal values. Furthermore, the post-processing methodology involves the estimation of the Instantaneous Axis of Rotation (IAR), which represents the imaginary axis around which a rigid body is rotating at a specific time instant. The IAR is calculated for each time step using the three-dimensional motion and is inherently orthogonal to the velocity vector of any point. Furthermore, the IAR is used to experimentally estimate the coordinates of the pivot axes at the block’s base during rocking motion. Specifically, the y coordinate of the pivot axes, denoted as yIAR , is estimated as the median value of the y coordinate of the IAR, separately for positive and negative rotations θx . Finally, the estimation of sliding uy,pp (i.e. horizontal translation along the y-axis) and uplifting uz,pp (i.e. vertical translation along the z-axis) are calculated for the two points on the pivot axes at the centre of the block’s length ( xIAR =0 ). Sliding along the x-axis is considered negligible compared to the y-axis and thus it is ignored. 4.2 Energy losses Following the model of Housner, the energy dissipation of the free-rocking response is described using the angular CoR. As a first approach, the experimental quantification of the CoR may be defined as the ratio of the angular velocities right after and before any impact n: 1 3 Bulletin of Earthquake Engineering the maximum rotation, which affects the accurate estimation of the IAR. Fig. 8b presents a three-dimensional coordinate plot of the IAR over the entire time history, i.e. each line represents the IAR of each time instant. Similar to Fig. 8a, abrupt lines correspond to the noisy and inaccurate estimation of the IAR for the time instants with negligible angular velocity (i.e. at the peaks of the rocking rotation). As expected, Fig. 8b highlights the concentration of the IAR at the edges of the block, resembling the contact imprint of the block. Nevertheless, the experimental evidence also indicates that pivoting occurs across a distributed area of the contact interface, and not at the idealised edge (or pivot point in a 2D view). This aspect is further highlighted in Fig. 8c, which shows a histogram of the distribution of the yIAR for all the tests, together with the nominal width of the blocks. The frequency of appearances of y-coordinate of the IAR reaches a peak very close to the nominal width of the block but slightly shifted inwards. Specifically, the majority of the counts lie around 0.42 to 0.50 and demonstrate a peak at 0.46. This discrepancy may be attributed to imperfections and irregularities due to heterogeneities in the body and/or the contact interface, deviations between the nominal and actual geometry of the block (as shown in Fig. 6), and/or the fact that contact occurs across an area rather than along the idealised edge of the block due to the finite stiffness of the interface (Vlachakis et al. 2023). Finally, the long tail distribution of y-coordinates of the IAR beyond the edge threshold does not have a physical interpretation but is rather attributed to the noisy estimation of the IAR during rocking with negligible velocities. Fig. 8 Instantaneous Axis of Rotation (IAR): (a) y coordinate ( yIAR ) over time of the representative test, (b) 3D representation for the whole time history of the representative test, and (c) histogram plot of the y coordinate normalised by the width of the block ( yIAR/B ) for all the free-rocking tests 1 3 Bulletin of Earthquake Engineering 5.2 Energy losses Figure 9 presents a representative example of the three different methodologies for estimating the angular CoR, plotted against the normalised absolute rocking rotation at the peak after each corresponding impact |θx,peak|/α , where α is the nominal value. Specifically, while the CoR calculated from the angular velocities CoRvel yields a single value per impact, the estimation of the angular CoR from potential energy CoRU , and potential and frictional energies CoRU+F , varies depending on whether one impact ( CoRcons U , CoRcons U+F ) or two successive impacts ( CoRside U , CoRside U+F ) are considered (refer to Section 4.2). Additionally, Fig. 9 includes the angular CoRHousner of the Housner model estimated using Eq. (3), which is calculated for the nominal value of α for this specific test. CoRvel exhibits the most scattered behaviour. The same behaviour has been observed in previous research (Costa et al. 2013; Giresini et al. 2018), and it is attributed to the challenges associated with the accurate estimation of the angular velocities before and after each impact. Conversely, both CoRside U and CoRside U+F are nearly constant with negligible variance, showing almost identical values, which in this test, are generally higher than CoRHousner . In contrast, CoRcons U and CoRcons U+F show scattered results, which diverge significantly for low amplitudes. This is a consequence of the asymmetries in the rocking response. Therefore, the subsequent analysis will focus exclusively on the side angular CoR data ( CoRside U , CoRside U+F ). Moreover, the fact that CoRside U and CoRside U+F (similarly CoRcons U and CoRcons U+F ) are almost identical implies that the frictional energy has a marginal effect on the angular CoR value. Figure 10a presents an overview of the CoRside U+F extracted from all 120 tests. Specifically, the CoRside U+F is normalised by the CoRHousner (using the nominal α value in Eq. (3)), while the abscissa consists of the normalised absolute peak rocking rotation |θx,peak|/α after each impact. The different aspect ratios are visually distinguished using colour variaFig. 9 Experimental angular CoR for the representative test using the different methodologies, against the absolute normalised peak rocking rotation after each impact (  θx,peak /α ) 1 3 Bulletin of Earthquake Engineering tions. Fig. 10a shows that CoRside U+F/ CoRHousner exhibits rather constant values at high rotation amplitudes, and significant scatter at lower amplitudes, similar to the representative test in Fig. 9. This trend, which was also observed in previous studies (ElGawady et al. 2011), is due to the greater influence of the irregularities of the block and contact interface at the low response amplitudes and/or the acquisition accuracy of the measurements. Nonetheless, 90% of the CoR data lie within 0.95 and 1.02, indicating a small difference compared to Housner’s model. More specifically, it appears that blocks of medium-to-high aspect ratio closely follow CoRHousner (Eq. (3)), whereas, for stockier blocks, Housner theory tends to underestimate the experimental CoR and thus overestimate the energy losses. A histogram of the CoR distribution is provided in Fig. 10b, where most of the experimental results cluster around CoRHousner and have a marginal skewness towards lower values. Furthermore, the experimental data are predicted well by a Cauchy distribution with a median of 0.997 and a median absolute deviation of 0.009. To provide a more robust representation against outliers and clarify the central tendency of each test, Fig. 11 presents an overview of the angular CoR using only the median value of each test. Choosing one value per test balances out the undesired dependency of the CoR data within each test, i.e. a slender block experiences more response cycles, more impacts, and thus more CoR data compared to a stockier block. The three subplots in Fig. 11 differentiate the results of the angular CoR as extracted from (a) the angular velocity ( CoRvel ), (b) the potential energy ( CoRside U ), and (c) the potential and frictional energies ( CoRside U+F ), respectively. In each subplot, the black line indicates the CoRHousner of Housner’s model (Eq. (3)). Moreover, a boxplot is presented for each aspect ratio H/B , where the central solid dot corresponds to the median of the data, while the vertical lines indicate the first and third quartiles. Fig. 11a shows that the CoRvel is characterised by high variance per aspect ratio H/B , yet the median values deviate slightly from Housner’s model. CoRside U in Fig. 11b and CoRside U+F in Fig. 11c share common features between them, indicating that for the cases of this experimental campaign, the frictional energy dissipation has a minor influence on the estimation of the angular CoR. In particular, for H/B =4 , Fig. 10 (a) Experimental CoR of all the tests derived using the potential and frictional energies normalised by the Housner model’s predictions CoRside U+F /CoRHousner versus the absolute normalised peak rocking rotation after each impact (  θx,peak /α ), and (b) histogram plot of the normalised CoR distribution 1 3 Bulletin of Earthquake Engineering Housner’s model underpredicts the experimental CoR estimates of CoRside U and CoRside U+F , suggesting that the block dissipates less energy than the Housner model predicts (Kalliontzis et al. 2016). For medium aspect ratios 5≤H/B ≤11 , the median values of the experimental CoR closely align with the Housner model predictions, with the exception of H/B =7 , implying the accuracy of Housner’s model to capture the energy losses. However, for higher aspect ratios ( 12 ≤H/B ≤15 ), the experimental CoR shows consistently lower values than the Housner model, suggesting that the block dissipates more energy than the Housner model predicts (Čeh et al., 2018). From a physical perspective, the higher experimental CoR observed mainly in stockier blocks can be attributed to the inward shift of the impact point from the nominal edge, caused by compliance and convex irregularities at the contact interface, as indicated by the IAR results (Fig. 8) (Ther and Kollár 2017). On the other hand, the lower experimental CoR observed in more slender blocks may be linked to additional dissipative phenomena that are overlooked in the CoR estimation, such as plastic deformations, hysteresis at the interface (Vlachakis et al. 2023) or radiation damping. Finally, Fig. 12 presents the experimental results of the normal CoRN. Fig. 12a shows a three-dimensional view of the representative test, where the normal CoRN is plotted across the block’s interface and features two surfaces: the surface with positive y-coordinate corresponds to impacts with negative angular velocity ( ˙ θx<0 ), and vice versa. Moreover, each point of the two surfaces represents the median value of all the impacts of the test at the specific x and y coordinates. Fig. 12a shows that the normal CoRN approaches zero near the pivot edges of the block and tends to one at the centre. Interestingly, the two surfaces are not symmetric along the x-axis, corroborating the presence of a three-dimensional component during the impacts due to irregularities and asymmetries at the interface. Fig. 12b,c show two section views of the normal CoR for three x and y coordinates respectively, while an absolute reference coordinate system is considered, i.e. all the CoRN surfaces are plotted using the positive values of y and x. Moreover, each line represents the median value of all the impacts for all 120 tests. The prediction of Housner’s model is also shown with dashed lines, which assumes perfectly plastic corner impact and consequently the corresponding kinematic normal CoRN of any point at the contact interface is given Fig. 11 Experimental CoR distribution with respect to the aspect ratio H/B considering the median value per test, extracted based on (a) the angular velocity ( CoRvel ), (b) the potential energy ( CoRside U ), and (c) the potential and frictional energies ( CoRside U+F ) 1 3 Bulletin of Earthquake Engineering as: CoR N=CoRHousner ( 1 − 2|y| B )/( 1+2|y| B ) . Fig. 12b shows that the experimental normal CoRN surface follows a similar trend with respect to Housner’s model across the width of the block, but consistently providing a slightly lower prediction. Additionally, the experimental normal CoRN appears to be negative at the nominal edges of the blocks ( |y|/B=0.50 ), which is not physically admissible. This inconsistency indicates once more that the actual impact point is not exactly at the idealised edge of the blocks ( |y|/B=0.50 ), but rather lies slightly inwards, on average, between |y|/B=0.42 to |y|/B=0.45 . This observation corroborates the findings of Fig. 8c, where the distribution of the y/B coordinate of the IAR shows a peak between 0.42 and 0.5. Fig. 12c, in conjunction with Fig. 12b, shows that the normal CoRN presents a rather constant value along the length of the blocks |x|/L , thus highlighting that the asymmetry of the experimental normal CoRN surfaces observed in each unique test (such as Fig. 12a) vanishes when considering all 120 tests, hence, indicating an aleatoric uncertainty. Fig. 12 Normal CoRN: (a) 3D plot over the block’s interface using the experimental median surfaces for the representative test; and experimental median surfaces for all the tests: (b) across the blocks’ width B , at three x coordinates across the length L , and (c) across the blocks’ length L , at three y coordinates across the width B 1 3 Bulletin of Earthquake Engineering 6 Statistical analysis Section 5 estimated the angular CoR from the free-rocking experiments using three methods and compared the results with the model of Housner. Nevertheless, a statistical analysis is presented in this section to draw more robust conclusions. This includes a series of statistical tests that address the following questions: ●Is the experimental CoRexp statistically equal to Housner’s model prediction CoRHousner (Section 6.1)? ●Is CoRexp/CoRHousner statistically dependent on the aspect ratio H/B (Section 6.2)? ●Is CoRexp/CoRHousner statistically dependent on the rocking amplitude (Section 6.3)? The analyses are guided by a research question involving the definition of a null hypothesis H0 and an alternative hypothesis H1. The null hypothesis H0 is assumed to hold until there is statistical evidence against it, while the alternative hypothesis H1 represents the scenario in case H0 is rejected. The p-value is calculated to test the null hypothesis and compared with different confidence levels. It measures how strong the evidence is against the null hypothesis: the lower the p-value, the stronger the evidence against H0 (Wasserman 2013). Table 3 illustrates the confidence levels and the corresponding p-values adopted in this investigation. All analyses are conducted using the experimental CoRexp normalised by the Housner model CoRexp/CoRHousner as a variable. This implicitly considers the dependency of the CoR on the aspect ratio and permits mutual comparisons among blocks with different aspect ratios. Three experimental datasets are employed for the statistical analyses. The first one includes the angular CoRs at each impact of each test and counts 7430 values in total. It represents longitudinal data for each free-rocking test, as i) impacts within the same test are interdependent, and ii) the number of impacts per test varies. The second dataset uses the median value from each test, thus reducing the dataset to 120 values, and represents a crosssectional dataset where the measurements are independent. Finally, a third dataset is used for grouping the data for each aspect ratio using their median, resulting in 12 counts. Specifically, Sec. 6.1 employs the second dataset (120 counts), Sec. 6.2 uses both the second (120 counts) and third (12 counts) datasets, while Sec. 6.3 uses the first dataset (7430 counts). The investigation of the research questions requires the selection of appropriate statistical methods. In particular, a decision must be made between parametric or non-parametric methods, depending on whether the assumptions of normality, independence, and homogeneity of variance for the variables under study are met. More specifically, parametric methods can be applied if all these assumptions hold; otherwise, non-parametric methods are required. To assess the normality of the data, a Kolmogorov-Smirnov test (K-S) is conducted (Kolmogorov 1933; Smirnoff 1939). This test compares the sample distribution to an equivalent set of normally distributed data centred on the experimental median. The K-S test assumes as H0 that the data are normally distributed and evaluates whether there is sufficient Confidence level p-value Evidence 99% ≤0.01 Very strong against H0 95% (0.01–0.05] Strong against H0 90% (0.05–0.10] Medium-weak against H0 Less than 90% > 0.10 Small or none against H0 For visual clarity, the p-values presented in the following sections adopt the font styles reported in this table. Table 3 Evidence classification of the p-value scale for different confidence levels (Wasserman 2013) 1 3 Bulletin of Earthquake Engineering evidence to reject this assumption. Table 4 presents the results of the K-S tests for the three experimental CoRs, based solely on the cross-sectional dataset of 120 values (second dataset) and 12 values (third dataset), as the K-S test method assumes data independence. Given that the p-values for all the analyses are lower than the assumed confidence levels (Table 3), it can be concluded that the data significantly deviates from the normal distribution. Since one of the key assumptions of the parametric methods is not satisfied, the subsequent statistical tests are conducted with non-parametric methods. 6.1 Is CoRexp statistically equal to CoRHousner ? This section conducts a statistical analysis to investigate whether Housner’s CoRHousner can predict the experimental CoRexp . To statistically assess the potential equivalence between the predicted CoRHousner and the experimental CoRexp , the Wilcoxon signedrank test is employed (Wilcoxon 1945). In this context, the null hypothesis H0 states that CoRexp/CoRHousner =1 . The analysis considers the median CoRexp per test (i.e. the dataset with 120 counts), once for all H/B ratios ( 4≤H/B ≤15 ) and then for every H/B ratio individually. Table 5 presents the results of the Wilcoxon signed-rank test showing p-values for all three methods of extracting the CoRexp . When considering all H/B ratios, there is strong evidence against the assumed equivalence between Housner’s CoRHousner and the experimental CoRexp . However, interesting results emerge when the H/B ratios are considered individually, revealing negligible evidence against the null hypothesis that CoRHousner is statistically equal to CoRexp for 5≤H/B ≤10 , with the exception of H/B =7 , regardless of the method used to extract the experimental CoRexp . This statistically reinforces the efficiency of Housner’s CoRHousner to capture the energy losses of rocking structures of medium-to-high aspect ratios. On the contrary, the statistical analysis indicates strong or very strong evidence against the H0 for aspect ratios H/B ≥12 , implying that for very slender rocking structures, CoRHousner is not a statistically accurate predictor of the energy losses. Finally, for H/B =4 and H/B = 11 , the three methods provide inconclusive results, with the CoRvel showing a better correlation with the Housner model. This outcome is in agreement with Fig. 11, showing that Housner’s model tends to overpredict the energy losses for stocky blocks, while it underpredicts the energy losses for very slender blocks. 6.2 Is CoRexp/CoRHousner statistically dependent on the aspect ratio H/B? This section examines whether the normalised CoRexp/CoRHousner has any statistical dependency on the H/B ratio, in addition to the dependency introduced by the Housner model. Since the data are not normally distributed, the non-parametric Spearman’s rank correlation test is used (Spearman 1961). The Spearman’s test provides an associated p-value and a correlation coefficient r. The null hypothesis H0 indicates no correlation, thus stating Table 4 Computed p-values of the Kolmogorov-Smirnov test of normality CoR from median per test p-value CoR from median per H/B p-value CoRvel /CoRHousner < 0.01 CoRvel /CoRHousner < 0.01 CoR side U/ CoRHousner < 0.01 CoR side U/ CoRHousner < 0.01 CoR side U+F/ CoRHousner < 0.01 CoR side U+F/ CoRHousner < 0.01 1 3 Bulletin of Earthquake Engineering Table 5 Computed p-values of the wilcoxon signed-rank test, examining the ability of the housner model to predict the experimental CoR CoR from median per test H/B ratios 4–15 456 7 8 9 10 11 12 13 14 15 CoRvel /CoRHousner < 0.01 0.11 0.38 0.18 0.05 0.21 0.42 0.22 0.08 0.02 < 0.01 < 0.01 0.02 CoR side U/ CoRHousner < 0.01 < 0.01 0.85 0.06 0.05 0.07 0.42 0.13 0.02 0.01 < 0.01 < 0.01 < 0.01 CoR side U+F/ CoRHousner < 0.01 < 0.01 0.85 0.11 0.05 0.10 0.57 0.54 0.03 0.01 < 0.01 < 0.01 < 0.01 1 3 Bulletin of Earthquake Engineering the independence of the CoRexp/ CoR Housner to the aspect ratio H/B . The r-value measures the strength and direction of the monotonic relationship between two variables, ranging from −1 to +1, where 0 indicates no correlation. An r-value of −1 or + 1 implies a perfect negative or positive monotonic relationship, respectively. Table 6 presents the results of Spearman’s test for the median CoRexp data per test (i.e. the dataset with 120 counts) across all three extraction methods. In all cases, there is very strong evidence against H0 as observed by the p-values, indicating a statistically significant correlation with the aspect ratio H/B . Moreover, the value of r reveals a negative monotonic relationship, with values ranging from −0.25, for the CoRvel , to −0.46, for CoRside U and CoRside U+F . In accordance with Fig. 11, these findings suggest that the normalised CoRexp/CoRHousner is inversely dependent on the aspect ratio H/B , thus implying that Housner’s CoRHousner shows a stronger dependency on the aspect ratio than the experimental CoRexp . The dependency between the CoRexp/CoRHousner and the aspect ratio H/B is further examined by investigating whether a linear regression model can adequately capture the normalised experimental data. The regression is conducted on two datasets: i) the median CoRexp from each test (dataset with 120 counts), and ii) their median value for each aspect ratio H/B (dataset with 12 counts). Table 7 presents the results of a linear regression analysis, including the p-value and the coefficient of determination R2. The null hypothesis H0 posits that the slope of the relationship between CoRexp/CoRHousner and H/B ratio is zero, implying lack of linear correlation. The p-values in Table 7 reveal strong evidence against the lack of correlation for the dataset with the median of each test (dataset with 120 counts), with a coefficient of determination between 0.04 and 0.18. The coefficient of determination notably increases for the dataset that only considers the median value per aspect ratio H/B (dataset with 12 counts). This increase occurs for the CoRside U and CoRside U+F , up to 0.54 to 0.56 respectively, while the correlation is insignificant for the CoRvel , i.e. 0.15. A closer examination of the data reveals that the correlation is heavily influenced by the results of the aspect ratio H/B =7 . When this aspect ratio is excluded (i.e. results in parenthesis in Table 7), the p-value drops to < 0.01 for all CoRexp/ CoR Housner , and the linear regression model better fits the remaining data. This results in notably higher R2 values of 0.82, 0.67 and 0.68 for CoRvel , CoRside U and CoRside U+F , respectively. Fig. 13 illustrates the experimental data together with the results of the linear regression analysis from Table 7, highlighting that the fitted lines provide a similar estimate for both datasets (i.e. datasets considering the median value per test and the median value per H/B ). 6.3 Is CoRexp/CoRHousner statistically dependent on the rocking amplitude? This section investigates statistically whether the experimental angular CoRexp depends on the rocking amplitude, in contrast to Housner’s model, which assumes that CoRHousner (Eq. (3)) is constant and response-independent. Two rocking amplitudes are considered as candidate independent variables: i) the peak rotation of the succeeding half-cycle normalised by the slenderness angle |θx,peak|/α , and ii) the angular velocity at impact normalised by CoR from median per test p-value r-value CoRvel/CoRHousner < 0.01 −0.25 CoR side U/ CoRHousner < 0.01 −0.46 CoR side U+F/ CoRHousner < 0.01 −0.46 Table 6 Computed p-values and correlation coefficient r of the Spearman’s rank correlation test, examining the monotonicity of the normalised experimental CoR against the aspect ratio H/B 1 3 Bulletin of Earthquake Engineering Table 7 Computed p-value, coefficient of determination R2, intercept and slope, of the linear regression analysis between the CoRexp/CoRHousner and the aspect ratio H/B . [note: the values in parenthesis exclude H/B =7 ] CoR from median per test CoR from median per H/B p-value R2 value intercept slope p-value R2 value intercept slope CoRvel/CoRHousner 0.02 0.04 1.001 −0.0014 0.21 0.15 1.003 −0.0013 ( < 0.01) (0.82) (1.011) (−0.0019) CoR side U/ CoR Housner < 0.01 0.18 1.016 −0.0028 0.01 0.54 1.020 −0.0029 ( < 0.01) (0.67) (1.024) (−0.0032) CoR side U+F/ CoR Housner < 0.01 0.18 1.017 −0.0028 < 0.01 0.56 1.021 −0.0029 ( < 0.01) (0.68) (1.025) (−0.0032) 1 3 Bulletin of Earthquake Engineering K C KB = U D UA (A.2) Considering that K C =0 . 5· I0, ±·˙ θ 2 x,C and K B =0 . 5· I0, ∓·˙ θ 2 x,B , Eq. (A.2) yields: CoR cons U= ˙ θx,C ˙ θx,B = √ UD UA I0,∓ I0, ± = √ U + peak;n U− peak;n I0,∓ I0, ± (A.3) where the subscript symbol “ ± ” refers to the positive and negative signs of rotation, whereas the superscripts “+” and “–” denote the post-impact and pre-impact states, respectively. The aforementioned methodology can also be applied to estimate the angular CoR considering two successive impacts n and n +1, i.e. accounting for the peaks of the potential energy from the same side of motion. Similarly (see Fig. 15): K F KE = U G UD (A.4) Substituting UD from Eq. (A.2) into Eq. (A.4), considering K F =0 . 5· I 0 , ∓·˙ θ 2 x,F , K E =0 . 5· I 0 , ±·˙ θ 2 x,E and assuming that the angular CoR of the two impacts n and n +1 is the same, gives: CoR side U=  UG UA =     U+ peak;n+1 U− peak;n (A.5) Fig. 17 Schematic representation of the frictional energy losses estimation: (a) plan view of the block interface, (b) block displacement due to torsional rotation θz , (c) block displacement due to translation utrans along the y-axis, and (c) bilinear elasto-plastic sliding law 1 3 Bulletin of Earthquake Engineering A.3 Angular CoR from potential and frictional energies The third methodology for estimating the angular CoR builds on the approach presented in Sec. A.2 by accounting for both the energy dissipated during impact and the frictional energy losses during the rocking phase (see Fig. 17a). The estimation of the frictional energy dissipation from instant “A” to instant “B” WA→B (Fig. 15) relies on the calculation of the work due to sliding along the contact interface. Since sliding along the x-axis is considered negligible compared to sliding along the y-axis, the frictional work solely accounts for the latter and is computed as follows: W A→B=NFRIC L/ 2 ∫ − L/2 [max (|duθ+dutrans|−uy,elastic; 0)] dx (A.6) where NFRIC =mgµdyn is the sliding friction force at the contact interface, with µdyn representing the dynamic friction coefficient, herein assumed to be equal to the static friction coefficient µdyn = µstatic = 0.70 (Vlachakis et al. 2023). duθ= sin (θz,B −θz,A)x is the displacement of a point with coordinate x that experiences a finite rotation over the z-axis θz,B −θz,A (Fig. 17b). dutrans =uy,B −uy,A is the displacement of any point due to the finite translation uy,B −uy,A (Fig. 17c), and u y,elastic =N F RIC LBks is the elastic relative displacement at the contact interface before sliding occurs. The definition of uy,elastic assumes a bilinear elastoplastic sliding law (Fig. 17d), with ks defining the elastic tangential interface stiffness (Vlachakis et al. 2023) and τfric being the sliding shear stress. Note that Eq. (A.6) assumes that the block is constantly in contact with the base during sliding and has a uniform distribution of reaction forces. Considering the half-cycle before and after impact n, the balance of the potential, kinetic and frictional energies reads (see Fig. 15): K C KB = U D +W C→D UA−WA→B (A.7) Substituting the kinetic energy expressions, Eq. (A.7) yields: CoR cons U+F= ˙ θx,C ˙ θx,B = √ UD+WC→D UA − WA → B Io,∓ Io, ± = √ U + peak;n + W + imp→peak;n U− peak;n− W− peak→imp;n Io,∓ Io, ± (A.8) Similarly, for impact n +1, the balance of the potential, kinetic and frictional energies gives (Fig. 15): K F KE = U G +W F→G UD−WD→E (A.9) Substituting UD from Eq. (A.8) into Eq. (A.9) and assuming that the angular CoR of the two impacts n and n +1 is the same implies that: 1 3 Bulletin of Earthquake Engineering CoR side U+F=              WC→D + WD→E + +(WC→D+WD→E)2+4(UA−WA→B)(UG+WF→G)Io,± Io,∓2    2(UA−WA→B)Io,± Io,∓ = =                     W+ imp→peak;n+W− peak→imp;n+1+ +     W+ imp→peak;n+W− peak→imp;n+12+4U− peak;n−W− peak→imp;n U+ peak;n+1 +W+ imp→peak;n+1Io,± Io,∓2       2  U− peak;n−W− peak→imp;n  Io,± Io, ∓ (A.10) Acknowledgements This work is financed by national funds through FCT– Foundation for Science and Technology, under grant agreements PRT/BD/152830/2021, 2020.07325.BD attributed to the first and second authors, respectively. This study has been partly funded by the STAND4HERITAGE project (new STANDards FOR seismic assessment of built cultural HERITAGE) that has received funding from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation program (Grant No. 833123) as an Advanced Grant. This study was also partly funded by FCT/MECI through project 2022.05425.PTDC (RESISTANCE) and project national funds (PIDDAC) under the R&D Unit Institute for Sustainability and Innovation in Structural Engineering (ISISE), under reference UIDB/04029/2020 ( h t t p : / / d o i . o r g / 1 0 . 5 4 4 9 9 / U I D B / 0 4 0 2 9 / 2 0 2 0 ) , and under the Associate Laboratory Advanced Production and I n t e l l i g e n t Systems ARISE under reference LA/P/0112/2020. The opinions and conclusions presented in this paper are those of the authors and do not necessarily reflect the views of the sponsoring organisations. The authors also greatly acknowledge the technical support from the staff of the University of Minho laboratory during the experimental campaign. Author contributions All authors contributed to the conceptualisation and design of this study. Material preparation, data collection and analysis were performed by Carla Colombo and Georgios Vlachakis. The first draft of the manuscript was written by Carla Colombo, and all authors revised previous versions of the manuscript. All authors read and approved the final manuscript. Funding Open access funding provided by FCT|FCCN (b-on). This work is financed by national funds through FCT — Foundation for Science and Technology, under grant agreements PRT/BD/152830/2021, 2020.07325.BD attributed to the first and second authors, respectively. This study has been partly funded by the STAND4HERITAGE project (new STANDards FOR seismic assessment of built cultural HERITAGE) that has received funding from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation program (Grant No. 833123) as an Advanced Grant. This study was also partly funded by FCT / MECI through project 2022.05425.PTDC (RESISTANCE) and project national funds (PIDDAC) under the R&D Unit Institute for Sustainability and Innovation in Structural Engineering (ISISE), under reference UIDB / 04029/2020 (doi.org/10.54499/UIDB/04029/2020), and under the Associate Laboratory Advanced Production and Intelligent Systems ARISE under reference LA/P/0112/2020. The opinions and conclusions presented in this paper are those of the authors and do not necessarily reflect the views of the sponsoring organisations. Data availability The datasets generated and analysed in this study are available from the corresponding author upon reasonable request. Declarations Competing interests The authors have no relevant financial or non-financial interests to disclose. 1 3 Bulletin of Earthquake Engineering Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. References Al Shawa O, de Felice G, Mauro A, Sorrentino L (2012) Out-of-plane seismic behaviour of rocking masonry walls Omar. Earthq Eng Struct Dyn 41(5):949–968. https://doi.org/10.1002/eqe.1168. Aslam M, Fodden WG, Scalise DT (1980) Earthquake rocking response of rigid bodies. Journal Of The Structural Division, ASCE 106(ST2, Proc. Paper, 15182):377–392. https://doi.org/10.1061/jsdeag.0005363. Bachmann JA, Strand M, Vassiliou MF, Broccardo M, Stojadinović B (2018) Is rocking motion predictable? Earthq Eng Struct Dyn 47(2):535–552. https://doi.org/10.1002/eqe.2978. Bao Y (2024) A distributed Impact model for seismic analysis of planar rocking body. J Earthquake Eng 28(10):2781–2800. h t t p s : / / d o i . o r g / 1 0 . 1 0 8 0 / 1 3 6 3 2 4 6 9 . 2 0 2 4 . 2 3 0 6 5 9 2 . Bao Y, Konstantinidis D (2020) Dynamics of a sliding-rocking block considering impact with an adjacent wall. Earthq Eng Struct Dyn 49(5):498–523. https://doi.org/10.1002/eqe.3250. Brogliato B, Zhang H, Liu C (2012) Analysis of a generalized kinematic impact law for multibody-multicontact systems, with application to the planar rocking block and chains of balls. Multibody Syst Dyn 27(3):351–382. https://doi.org/10.1007/s11044-012-9301-3. Cappelli E, Di Egidio A, Vestroni F (2020) Analytical and experimental investigation of the behavior of a rocking masonry Tuff Wall. J. Eng. Mech. 146(6). h t t p s : / / d o i . o r g / 1 0 . 1 0 6 1 / ( a s c e ) e m . 1 9 4 3 - 7 8 8 9 . 0 0 0 1 7 7 5. Casapulla C, Giresini L, Lourenço PB (2017) Rocking and kinematic approaches for rigid block analysis of masonry walls: state of the art and recent developments. Buildings 7(3). h t t p s : / / d o i . o r g / 1 0 . 3 3 9 0 / b u i l d i n g s 7 0 3 0 0 6 9 . Čeh N, Jelenić G, Bićanić N (2018) Analysis of restitution in rocking of single rigid blocks. Acta Mech 229(11):4623–4642. https://doi.org/10.1007/s00707-018-2246-8. Charalampakis AE, Tsiatas GC, Tsopelas P (2022) New insights on rocking of rigid blocks: analytical solutions and exact energy-based overturning criteria. Earthq Eng Struct Dyn 51(9):1963–1993. h t t p s : / / d o i . o r g / 1 0 . 1 0 0 2 / e q e . 3 6 4 9 . Chatterjee A, Jain R, Bowling A (2022) Modeling and simulation of rocking Block dynamics subjected to Base motion using an energetic restitution Law. J Earthquake Eng 26(13):6610–6632. h t t p s : / / d o i . o r g / 1 0 . 1 0 8 0 / 1 3 6 3 2 4 6 9 . 2 0 2 1 . 1 9 2 7 9 0 3 . Chatzis MN, Espinosa MG, Smyth AW (2017) Examining the energy Loss in the inverted pendulum Model for rocking bodies. J. Eng. Mech. 143(5):1–12. h t t p s : / / d o i . o r g / 1 0 . 1 0 6 1 / ( a s c e ) e m . 1 9 4 3 - 7 8 8 9 . 0 0 0 1 2 0 5 . Chatzis MN, Smyth AW (2012) Robust modeling of the rocking problem. J. Eng. Mech. 138(3):247–262. h t t p s : / / d o i . o r g / 1 0 . 1 0 6 1 / ( a s c e ) e m . 1 9 4 3 - 7 8 8 9 . 0 0 0 0 3 2 9 . Cheng CT (2007) Energy dissipation in rocking bridge piers under free vibration tests. Earthq Eng Struct Dyn 36(4):503–518. https://doi.org/10.1002/eqe.640. Cocuzza Avellino G, Cannizzaro F, Di Martino A, Valenti R, Paternò E, Caliò I, Impollonia N (2021) Numerical and experimental response of free-standing art objects subjected to ground motion. Int J Archit Heritage 16(11):1666–1682. h t t p s : / / d o i . o r g / 1 0 . 1 0 8 0 / 1 5 5 8 3 0 5 8 . 2 0 2 1 . 1 9 0 2 0 1 9 . Corral E, Moreno RG, García MJG, Castejón C (2021) Nonlinear phenomena of contact in multibody systems dynamics: a review. Nonlinear Dyn 104(2):1269–1295. https://doi.org/10.1007/s11071-021-06344-z. Costa AA, Arêde A, Penna A, Costa A (2013) Free rocking response of a regular stone masonry wall with equivalent block approach: experimental and analytical evaluation. Earthq Eng Struct Dyn 42(15):2297– 2319. https://doi.org/10.1002/eqe.2327. D’Altri AM, Vlachakis G, de Miranda S, Lourenço PB (2024) Rocking block simulation based on numerical dissipation. Nonlinear Dyn 112(20):17843–17862. https://doi.org/10.1007/s11071-024-09974-1. Di Egidio A, Zulli D, Contento A (2014) Comparison between the seismic response of 2D and 3D models of rigid blocks. Earthquake Eng Eng Vibr 13(1):151–162. https://doi.org/10.1007/s11803-014-0219-z. Dimitrakopoulos EG, Giouvanidis AI (2015) Seismic response analysis of the planar rocking frame. J. Eng. Mech. 141(7). h t t p s : / / d o i . o r g / 1 0 . 1 0 6 1 / ( a s c e ) e m . 1 9 4 3 - 7 8 8 9 . 0 0 0 0 9 3 9 . 1 3 Bulletin of Earthquake Engineering ElGawady MA, Ma Q, Butterworth JW, Ingham JM (2011) Effects of interface material on the performance of free rocking blocks. Earthq Eng Struct Dyn 40(4):375–392. https://doi.org/10.1002/eqe.1025. Giaretton M, Dizhur D, Ingham JM (2016) Dynamic testing of as-built clay brick unreinforced masonry parapets. Eng Struct 127:676–685. h t t p s : / / d o i . o r g / 1 0 . 1 0 1 6 / j . e n g s t r u c t . 2 0 1 6 . 0 9 . 0 1 6 . Giouvanidis AI, Dimitrakopoulos EG (2017) Nonsmooth dynamic analysis of sticking impacts in rocking structures. Bull Earthquake Eng 15(5):2273–2304. https://doi.org/10.1007/s10518-016-0068-4. Giouvanidis AI, Dong Y (2020) Seismic loss and resilience assessment of single-column rocking bridges. Bull Earthquake Eng 18(9):4481–4513. https://doi.org/10.1007/s10518-020-00865-5. Giresini L, Sassu M, Sorrentino L (2018) In situ free-vibration tests on unrestrained and restrained rocking masonry walls. Earthq Eng Struct Dyn 47(15):3006–3025. https://doi.org/10.1002/eqe.3119. Goldsmith W (1960) Impact: the theory and physical behavior of colliding solids. Edward Arnold Ltd Good P (2013) Permutation tests: a practical guide to resampling methods for testing hypotheses. Springer Science & Business Media Housner GW (1963) The behavior of inverted pendulum structures during earthquakes. Bull Seismol Soc Am 53(2):403–417. https://doi.org/10.1785/BSSA0530020403. Huang B, Pan Q, Lu W, Shen F (2021) Free-rocking tests of a freestanding object with variation of center of gravity. Earthq Eng Struct Dyn 50(11):3015–3040. https://doi.org/10.1002/eqe.3498. Kafle B, Lam NTK, Gad EF, Wilson J (2011) Displacement controlled rocking behaviour of rigid objects. Earthq Eng Struct Dyn 40(15):1653–1669. https://doi.org/10.1002/eqe.1107. Kalliontzis D, Sritharan S (2018) Characterizing dynamic decay of motion of free-standing rocking members. Earthq Spectra 34(2):843–866. https://doi.org/10.1193/011217EQS013M. Kalliontzis D, Sritharan S, Schultz A (2016) Improved coefficient of restitution estimation for free rocking members. J Struct Eng 142(12). h t t p s : / / d o i . o r g / 1 0 . 1 0 6 1 / ( a s c e ) s t . 1 9 4 3 - 5 4 1 x . 0 0 0 1 5 9 8 . Koh A, Spanos D, Roesset M (1986) Harmonic rocking of rigid block on flexible foundation. Engineering Mechanics 112(11):1165–1180. h t t p s : / / d o i . o r g / 1 0 . 1 0 6 1 / ( A S C E ) 0 7 3 3 - 9 3 9 9 , ( 1 9 8 6 ) 1 1 2 : 1 1 , ( 1 1 6 5 ) . Kolmogorov A (1933) Sulla determinazione empirica di una legge distribuzione. Giornale Dell’Istituto Italiano Degli Attuari 4:89–91 Konstantinidis D, Makris N (2009) Experimental and analytical studies on the response of freestanding laboratory equipment to earthquake shaking. Earthq Eng Struct Dyn 38(6):827–848. h t t p s : / / d o i . o r g / 1 0 . 1 0 0 2 / e q e . 8 7 1 . Lagomarsino S (2015) Seismic assessment of rocking masonry structures. Bull Earthquake Eng 13(1):97– 128. https://doi.org/10.1007/s10518-014-9609-x. Lawrence J, Bernal J, Witzgall C (2019) A purely algebraic justification of the Kabsch-Umeyama algorithm. J Res Natl Inst Stand Technol 124:1–6. https://doi.org/10.48550/arXiv.1902.03138. Lipscombe PR, Pellegrino S (1993) Free rocking of prismatic block. J. Eng. Mech. 119(7):1387–1410. h t t p s : / / d o i . o r g / 1 0 . 1 0 6 1 / ( A S C E ) 0 7 3 3 - 9 3 9 9 , ( 1 9 9 3 ) 1 1 9 : 7 , ( 1 3 8 7 ) . Makris N, Vassiliou MF (2013) Planar rocking response and stability analysis of an array of free-standing columns capped with a freely supported rigid beam. Earthq Eng Struct Dyn 42(3):431–449. h t t p s : / / d o i . o r g / 1 0 . 1 0 0 2 / e q e . 2 2 2 2 . Mathey C, Feau C, Clair D, Baillet L, Fogli M (2018) Experimental and numerical analyses of variability in the responses of imperfect slender free rigid blocks under random dynamic excitations. Eng Struct 172:891–906. h t t p s : / / d o i . o r g / 1 0 . 1 0 1 6 / j . e n g s t r u c t . 2 0 1 8 . 0 6 . 0 6 4 . Mathey CF, Politopoulos I, Clair D, Baillet L, Fogli M (2016) Behavior of rigid blocks with geometrical defects under seismic motion: an experimental and numerical study. Earthq Eng Struct Dyn 45:2455– 2474. https://doi.org/10.1002/eqe.2773. Mauro A, de Felice G, DeJong MJ (2015) The relative dynamic resilience of masonry collapse mechanisms. Eng Struct 85:182–194. h t t p s : / / d o i . o r g / 1 0 . 1 0 1 6 / j . e n g s t r u c t . 2 0 1 4 . 1 1 . 0 2 1 . Peña F, Lourenço PB, Campos-Costa A (2008) Experimental dynamic behavior of free-standing multi-block structures under seismic loadings. J Earthquake Eng 12(6):953–979. h t t p s : / / d o i . o r g / 1 0 . 1 0 8 0 / 1 3 6 3 2 4 6 0 8 0 1 8 9 0 5 1 3 . Pfeiffer F, Glocker C (2000) Multibody dynamics with unilateral contacts. Springer Science & Business Media. https://doi.org/10.1002/9783527618385. Priestley MJN, Evison RJ, Carr AJ (1978) Seismic response of structures free to rock on their foundations. Bulletin Of The New Zealand Society For Earthquake Engineering 11(3):141–150. h t t p s : / / d o i . o r g / 1 0 . 5 4 5 9 / b n z s e e . 1 1 . 3 . 1 4 1 - 1 5 0 . Psycharis IN (2018) Seismic vulnerability of classical monuments. Geot Geol Earthquake 46:563–582. https://doi.org/10.1007/978-3-319-75741-4_24. Psycharis IN, Jennings PC (1983) Rocking of slender rigid bodies allowed to uplift. Earthq Eng Struct Dyn 11:57–76. https://doi.org/10.1002/eqe.4290110106. 1 3 Bulletin of Earthquake Engineering Purvance MD, Anooshehpoor A, Brune JN (2008) Freestanding block overturning fragilities: numerical simulation and experimental validation. Earthq Eng Struct Dyn 37(5):791–808. h t t p s : / / d o i . o r g / 1 0 . 1 0 0 2 / e q e . 7 8 9 . Reggiani Manzo N, Vassiliou MF (2019) Displacement-based analysis and design of rocking structures. Earthq Eng Struct Dyn 48(14):1613–1629. https://doi.org/10.1002/eqe.3217. Shenton HW, Jones NP (1991) Base excitation of rigid bodies. I: formulation. Engineering Mechanics 117(10):2286–2306. h t t p s : / / d o i . o r g / 1 0 . 1 0 6 1 / ( A S C E ) 0 7 3 3 - 9 3 9 9 , ( 1 9 9 1 ) 1 1 7 : 1 0 , ( 2 2 8 6 ) . Smirnoff N (1939) Sur les écarts de la courbe de distribution empirique. Matematicheskii Sbornik 48(1):3–26 Sorrentino L, AlShawa O, Decanini LD (2011) The relevance of energy damping in unreinforced masonry rocking mechanisms. Experimental and analytic investigations. Bull Earthquake Eng 9(5):1617–1642. https://doi.org/10.1007/s10518-011-9291-1. Sorrentino L, Masiani R, Benedetti S (2008) Experimental estimation of energy damping during free rocking of unreinforced masonry walls. First results. AIP Conf Proc 1020(PART 1):1888–1895. h t t p s : / / d o i . o r g / 1 0 . 1 0 6 3 / 1 . 2 9 6 3 8 2 5 . Spanos PD, Di Matteo A, Pirrotta A, Di Paola M (2017) Rocking of rigid block on nonlinear flexible foundation. Int J Non Linear Mech 94:362–374. h t t p s : / / d o i . o r g / 1 0 . 1 0 1 6 / j . i j n o n l i n m e c . 2 0 1 7 . 0 6 . 0 0 5 . Spanos PD, Koh A-S (1985) Rocking of rigid blocks due to harmonic shaking. Journal Engineering Mechanics 110(11):1627–1642. h t t p s : / / d o i . o r g / 1 0 . 1 0 6 1 / ( A S C E ) 0 7 3 3 - 9 3 9 9 , ( 1 9 8 4 ) 1 1 0 : 1 1 ( 1 6 2 7 ) . Spearman C (1961) The proof and measurement of association between two things. Am J Phychol 100(3/4):441–471 Sutton M, Orteu J-J, Hubert WS (2009) Image correlation for shape, motion and deformation measurements. In Springer Science & Business Media. https://doi.org/10.1007/978-0-387-78747-3. Ther T, Kollár LP (2017) Refinement of Housner’s model on rocking blocks. Bull Earthquake Eng 15(5):2305–2319. https://doi.org/10.1007/s10518-016-0048-8. Tso WK, Wong CM (1989) Steady state rocking response of rigid blocks part 1: analysis. Earthq Eng Struct Dyn 18(89):89–106. https://doi.org/10.1002/eqe.4290180109. Várkonyi PL, Kocsis M, Ther T (2022) Rigid impacts of three-dimensional rocking structures. Nonlinear Dyn 107(3):1839–1858. https://doi.org/10.1007/s11071-021-06934-x. Verbeke G, Molenberghs G (1997) Linear mixed models for longitudinal data. Springer New York Vlachakis G, Colombo C, Giouvanidis AI, Savalle N, Lourenço PB (2023) Experimental characterisation of dry-joint masonry structures: interface stiffness and interface damping. Const Buil Mater 392. h t t p s : / / d o i . o r g / 1 0 . 1 0 1 6 / j . c o n b u i l d m a t . 2 0 2 3 . 1 3 0 8 8 0 . Vlachakis G, Giouvanidis AI, Mehrotra A, Lourenço PB (2021) Numerical block-based simulation of rocking structures using a novel universal viscous damping Model. J. Eng. Mech. 147(11):04021089. h t t p s : / / d o i . o r g / 1 0 . 1 0 6 1 / ( a s c e ) e m . 1 9 4 3 - 7 8 8 9 . 0 0 0 1 9 8 5 . Vlachakis G, Vlachaki E, Lourenço PB (2020) Learning from failure: damage and failure of masonry structures, after the 2017 Lesvos earthquake (Greece). Eng Fail Anal 117:104803. h t t p s : / / d o i . o r g / 1 0 . 1 0 1 6 / j . e n g f a i l a n a l . 2 0 2 0 . 1 0 4 8 0 3. Wasserman L (2013) All of statistics: a concise course in statistical inference. Springer Science & Business Media Wilcoxon F (1945) Individual comparisons by ranking methods. Iometrics Bulletin 1(6):80–83. h t t p s : / / d o i . o r g / 1 0 . 2 3 0 7 / 3 0 0 1 9 6 8 . Wittich CE, Hutchinson TC (2015) Shake table tests of stiff, unattached, asymmetric structures. Earthq Eng Struct Dyn 44:2425–2443. https://doi.org/10.1002/eqe.2589. Shake. Yilmaz C, Gharib M, Hurmuzlu Y (2009) Solving frictionless rocking block problem with multiple impacts. Proceedings Of The Royal Society A: Mathematical, Physical And Engineering Sciences 465(2111):3323–3339. https://doi.org/10.1098/rspa.2009.0273. Yim C-S, Chopra AK, Penzien J (1980) Rocking response of rigid blocks to earthquakes. Earthq Eng Struct Dyn 8:565–587. https://doi.org/10.1002/eqe.4290080606. Zhang H, Brogliato B, Liu C (2013) Study of the planar rocking-block dynamics with coulomb friction: critical kinetic angles. J Comput Nonlinear Dyn 8(2):021002. https://doi.org/10.1115/1.4007056. Zhao Z, Liu C, Wang N (2019) Rocking dynamics of a planar rectangular block on a rigid surface. Multibody Syst Dyn 45(1):105–125. https://doi.org/10.1007/s11044-018-09643-3. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. 1 3