Introduction
Masonry structures display high vulnerability to seismic action—threatening human lives, built assets, and a major part of our cultural heritage (
Bruneau 1994;
Penna et al. 2014). Among the observed types of failure, out-of-plane (OOP) collapse is the most frequent—particularly in the case of façade walls with poor connections to transversal elements or diaphragms (
Ingham and Griffith 2011;
Vlachakis et al. 2020, among others). Analysis of these OOP collapse mechanisms is typically conducted using simplified limit analysis procedures (
D’Ayala and Speranza 2003;
Vaculik et al. 2014), whereby the capacity of the mechanisms is evaluated through calculation of an equivalent static lateral force required to trigger each mechanism and eventually cause collapse of the structure. However, forced-based approaches can be overconservative because they tend to neglect the dynamic reserve of stability, particularly in the case of large-scale structures, which undergo significant displacements before overturning (
Godio and Beyer 2019;
Sorrentino et al. 2017).
To that end, the employment of theoretical rocking dynamics has been proposed to evaluate the OOP dynamic stability of masonry walls (e.g.,
Casapulla et al. 2017 and references therein). Following Housner’s (
1963) seminal work on the single rigid rocking block, the equations that describe the motion of masonry walls (simulated as rigid blocks) have been developed using rocking (Lagrangian) dynamics. Furthermore, the influence on the response of both seismological parameters (
Giouvanidis and Dimitrakopoulos 2018) and certain structural characteristics, such as the presence of additional loads due to, e.g., masses from floor and/or roof elements, or thrust from vaults (
Giresini et al. 2015;
Mauro et al. 2015), tie bars (
AlShawa et al. 2019;
Mauro et al. 2015), and transverse walls (
Giresini and Sassu 2017;
Al Shawa et al. 2012;
Sorrentino et al. 2011), as well as the formulation of a two-block mechanism, which occurs in the case of walls restrained by floors or a roof (
Mauro et al. 2015;
Sorrentino et al. 2011) has been extensively investigated.
In such a methodological framework, energy dissipation is assumed to occur entirely at impact, and is accounted for by the coefficient of restitution (COR) which correlates the angular velocity of the structure before and after impact. The COR may be determined analytically through the assumption of conservation of angular momentum, after specifying the geometry of the block and the points of impact (
Hogan 1992;
Housner 1963). Experimental investigations have evaluated the accuracy of the COR calculated using the classical rocking theory, and despite some discrepancies observed, it appears that the overall energy loss of rocking blocks can be adequately captured (
Bachmann et al. 2018;
Cappelli et al. 2020;
Chatzis et al. 2017;
Costa et al. 2013;
Kalliontzis and Sritharan 2018;
Lipscombe and Pellegrino 1993). However, as the complexity of the structure increases, e.g., more degrees of freedom, different boundary conditions, or introduction of flexible interfaces, among others, the classical rocking theory becomes complicated. Thus, alternative analytical and numerical models have been proposed to capture the transient nonlinear dynamic response of various rocking configurations (
Giouvanidis and Dimitrakopoulos 2017a;
Mehrotra and Dejong 2020;
Spanos et al. 2001).
At the same time, recent developments in computational approaches used for modeling masonry structures are gaining momentum, particularly block-based models, which have been found capable of reproducing the dynamic response of masonry walls while also being able to model masonry texture and interaction with surrounding structural elements. These include the finite-element (FE) method (
D’Altri et al. 2019), the discrete-element (DE) method (
Lemos 2019), and multibody dynamics (
Portioli and Cascini 2018). However, despite their widespread use, applications of these models usually lack a reliable treatment of the energy loss due to the nonsmooth behavior of impacts during rocking motion (
de Felice et al. 2017;
Sarhosis et al. 2019;
Vassiliou et al. 2021). Thus, viscous damping models are usually adopted. Viscous models are mathematical artifices with a continuous nature, seemingly in complete contrast with the impulsive energy loss suggested by the classical rocking theory. This characteristic appears to be a source of uncertainty in numerical simulations of rocking bodies (
AlShawa et al. 2017;
Lemos and Campos Costa 2017;
Malomo et al. 2021), making damping one of the main parameters that needs to be adjusted to obtain a better fit to a reference response rather than a quite consistent method, such as the classical rocking theory (
Housner 1963).
The main objective of the present study is to bridge the gap between the well-established energy loss of the classical rocking theory (
Housner 1963) and the treatment of damping of block-based numerical models. This is conducted following a phenomenological calibration of a viscous damping model to mimic the classical rocking (COR) theory, and pertinent ready-to-use expressions are proposed. More specifically, two of the most common problems evident in masonry structures are investigated: two-sided and one-side rocking (Fig.
1).
Two-sided rocking behavior represents the motion of a parapet, gable or boundary wall rocking over their foundation when subjected to ground excitation, whereas one-sided rocking behavior describes the motion of masonry façades with insufficient connections to transversal walls. Through a series of over a thousand numerical simulations, this work proposes ready-to-use expressions that agree with the well-established classical rocking theory and efficiently capture the energy loss during rocking motion of any numerical block-based structure. Importantly, the results of this study serve as the basis for a more rational and holistic approach to model (1) multi-degree-of-freedom rocking structures, (2) the interaction of complex geometries and boundary conditions with the rocking response, and (3) the inclusion of material nonlinearities.
To this end, the main dynamic characteristics of both approaches, i.e., classical rocking theory and the numerical block-based modeling, are firstly outlined, paving the way for an appropriate selection of the viscous damping model. During the calibration process of the proposed viscous damping model, special attention is given to its universality via its application in fundamentally different numerical modeling software, as well as its capability of including modifications of the classical rocking theory suggested by previous experimental and theoretical studies (e.g.,
Kalliontzis et al. 2016;
Sorrentino et al. 2011;
Ther and Kollar 2017, among others). Finally, the performance of the proposed numerical viscous damping model is evaluated against experimental campaigns available in literature, highlighting both its merits and shortcomings.
Analytical Modeling of Rocking Structures
In principle, during a strong ground motion, rocking action activates the structure’s rotational inertia and offers a favorable seismic isolation effect, which relieves the structure from deformation and ultimately damage. Rocking becomes evident in a variety of structural configurations, e.g., from bridges (
Giouvanidis and Dimitrakopoulos 2017b;
Giouvanidis and Dong 2020;
Routledge et al. 2020) and buildings (
Bachmann et al. 2017) to cultural heritage structures (
Mehrotra and DeJong 2018) and classical monuments (
Di Egidio and Contento 2009;
Psycharis et al. 2013). However, the main concern in rocking dynamics is the existence of large displacements (or equivalently rotations) during rocking motion, which might cause instability and eventually overturning of the structure. Therefore, the associated challenges behind this peculiar seismic behavior should be properly addressed with advanced analytical and numerical simulations that can capture with high fidelity the transient nonlinear dynamic behavior of rocking structures.
Consider the rocking block of Fig.
1 standing free on a rigid ground. During an earthquake, rocking commences when the seismic demand due to the horizontal ground excitation becomes equal to the seismic resistance due to the gravitational and inertial forces acting on the body. This condition yields the minimum ground acceleration necessary for rocking to initiate
(
Housner 1963), where
is the acceleration of gravity and
is the slenderness of the block
(Fig.
1).
After rocking initiates, the equation that describes the motion of the rocking block of Fig.
1 can be expressed as follows (
Housner 1963):
where
= (horizontal) ground acceleration; positive and negative signs = clockwise and counterclockwise rotation
, respectively; and
= frequency parameter of the block, defined as
with
representing the mass of the block, and
representing the rotational moment of inertia with respect to the pivot points. In the case of a rectangular block,
simplifies to
, where
is the diagonal distance from the center of mass of the block to the pivot point (Fig.
1).
The importance of the parameter
, being inversely proportional to the square root of the size
, has been studied in the past (e.g.,
Makris 2014, among others). From a physical perspective, the frequency parameter
refers to the pendulum frequency of the block as if it is hanging from its pivot point (
DeJong and Dimitrakopoulos 2014), and not the classical natural frequency, which usually measures cycles of vibration per second. This discrepancy stems from the fact that the natural frequency and period of the rocking motion is amplitude dependent, and thus, unsuitable for characterizing a structure (
Housner 1963)
During rocking, the smooth motion of the block is interrupted by nonsmooth impacts at its contact/pivot points. The classical rocking theory (
Housner 1963) considers impact as an instantaneous event, where the system is characterized by infinite stiffness (Fig.
2). When impact occurs, energy is lost. Ignoring bouncing and assuming the block sustains pure rocking motion (i.e., no sliding at the contact interface), the energy loss at impact is captured by COR,
, acting as radiation damping. The COR connects the preimpact with the postimpact angular velocity
.
When the rectangular block of Fig.
1(a) undergoes two-sided rocking motion (denoted as
henceforth), conservation of angular momentum yields (
Housner 1963)
From Eq. (
3), COR depends entirely upon the geometry (i.e., slenderness) of the block. Thus, the material properties (e.g., mass or stiffness) are irrelevant to the damping phenomenon during impact.
However, there are cases where the block during its smooth rocking motion, apart from the impact with the ground, also comes into contact with an adjacent wall representing masonry façades inadequately connected with the transversal walls. In such cases, the block exhibits one-sided rocking motion (denoted as
henceforth) as shown in Fig.
1(b). Bao and Konstantinidis (
2020) treated the impact with the adjacent sidewall as an additional event characterized by a separate/independent COR. Similarly, Sorrentino et al. (
2011) assumed three consecutive impacts taking place in close but distinct time instants. Specifically, an impact at the base of the block [i.e., Point 2 in Fig.
1(b)], followed by an impact at the upper corner [i.e., Point 3 in Fig.
1(b)], and finally an additional impact at the base [i.e., Point 1 in Fig.
1(b)], resulting in a postimpact rocking rotation around the same pivot point as the preimpact rocking rotation but in the opposite direction (i.e.,
). Under these assumptions, COR, which captures the impact of the rectangular rocking block of Fig.
1(b) with the transverse wall utilizing conservation of angular momentum, becomes
Therefore the total energy loss when the block of Fig.
1(b) undergoes one-sided rocking motion can be expressed as a lumped COR considering all three impacts (
Sorrentino et al. 2011)
Adopted Viscous Damping Model
The main objective of the present work is to adopt a viscous damping model that is able to reproduce with high fidelity the impulsive dynamics’ energy loss characteristics and subsequently ensure dynamic equivalence between the classical rocking theory (
Housner 1963) and the numerical viscous damping model in a universal manner, i.e., applicable to different rocking structures of different materials simulated in different finite-element/discrete-element software. Recall that a unilateral dashpot definition provides the most convenient basis to meet this goal. This section provides a closer look at the performance of the adopted viscous damping model through a series of free-rocking simulations of the structure shown in Fig.
1 undergoing both two-sided and one-sided rocking motion.
The analytical model adopted herein is based on the classical rocking theory (
Housner 1963). Specifically, an event-based approach is adopted, according to which the motion of the rocking block of Fig.
1 can be decomposed into a smooth motion interrupted by nonsmooth contact events (i.e., impacts). The smooth rocking motion of the block is obtained through the solution of the differential equation of Eq. (
1) using mathematical programming in MATLAB version 2017a (
MathWorks 1992). Whenever impact occurs, the integration stops and the coefficient of restitution [Eq. (
3) for the two-sided rocking case or Eq. (
5) for the one-sided rocking case] is applied to determine the new initial conditions for the next iteration.
The same problem is also formulated in a finite element environment, i.e., ABAQUS CAE version 2019 (
Simulia 2012), a widely used FEM software well-suited for block-based simulations that include contact phenomena. A schematic representation of the numerical models is depicted in Fig.
5. Within ABAQUS CAE, the explicit solution scheme is adopted instead of an implicit scheme because it is preferable to proceed with the solution over small time increments. The main reason behind this choice is the reduced computational cost that the explicit scheme offers. Moreover, the contacting bodies of the problem are only forced to interact when they come into contact. The normal behavior of the contact area follows a linear stiffness pressure-overclosure relationship (i.e., soft contact approach). The tangential behavior adopts a penalty friction formulation with elastic stiffness and a friction coefficient that defines the slipping criterion. An artificially high value is given to the friction coefficient to avoid possible sliding at the contact interface, and dilatancy effects are neglected. For the case of the one-sided rocking problem of Fig.
5(b), distinct contact properties are set to govern the different body interactions, i.e., contact with the base (
) and with the sidewall (
). Finally, a viscous damping ratio
is directly assigned to the contact interfaces following the unilateral dashpot definition discussed previously, i.e.,
at the base and
at the sidewall.
Fig.
6 shows the response of the analytical and numerical models through a series of free-rocking simulations for both two-sided [Figs.
6(a and b)] and one-side [Figs.
6(c and d)] rocking motions. Fig.
6 compares the two modeling approaches in terms of the rocking response history [Figs.
6(a and c)] and the total energy content expressed as the sum of the potential and kinetic energies of the system [Figs.
6(b and d)]. In Fig.
6, the initial rotation of the block is set to
, and the block has height
and base width
. For the analytical model, the COR that characterizes impact with the base is taken as
[Eq. (
3)], and the additional impact with the transversal wall, for the case of one-sided rocking, is captured through the
[Eq. (
4)]. For the numerical model, the damping ratio at the base is considered
[based on Eq. (
11)], whereas for the one-sided rocking case, the damping ratio during contact with the sidewall is
[based on Eq. (
13)].
Fig.
6 reveals the remarkable agreement between the two modeling approaches. Observe that only for lower rocking amplitudes and only for the two-sided rocking case [Figs.
6(a and b)] do the two examined models differ, with the numerical model dissipating energy slightly faster than the analytical model. Most importantly, from Figs.
6(b and d), the numerical (viscous damping) model shows a similar behavior with the analytical model, i.e., negligible energy loss during the smooth rocking motion phase, whereas steplike energy loss appears at each impact due to the unilateral dashpot formulation adopted herein. At the same time, the finite amount of energy dissipation at each impact is controlled by the damping ratio
. Therefore, by adjusting the damping ratio of the viscous damping model, energy equivalence between the two modeling approaches can be achieved. To this end, a calibration methodology is presented in the next section, with the aim of providing generalized predictive
relationships applicable to a variety of block-based (numerically simulated) structures that undergo either one-sided or two-sided rocking motion.
Importantly, even though the proposed numerical viscous damping model is presented using ABAQUS CAE, its applicability is universal and extended in other commonly used software for the numerical analysis of block-based rocking models. Fig.
7 presents a comparative investigation on the response of the rocking block of Fig.
5(a) derived by two additional commonly used block-based software, DIANA FEA version 10.4 (
DIANA 2017), and 3DEC version 7.0 (
Itasca 2013) through free-rocking simulations.
The three examined packages of modeling software are fundamentally different; however, the responses predicted by the adopted software are in almost excellent agreement. In particular, DIANA FEA adopts an implicit scheme to solve dynamic problems (compared with the explicit scheme adopted in ABAQUS CAE). Even though implicit schemes allow the use of larger time steps, the relatively high frequency at impact [Eq. (
7)] imposes a strict upper limit to the adopted time step. At the same time, even if this restriction is disregarded, to control spurious high-frequency oscillations, artificial numerical dissipation is introduced, e.g., through the use of the Hilbert-Hughes-Taylor (HHT) algorithm (
Hilber et al. 1977). Thus, relatively small time steps are eventually used in order for the viscous dissipation to dominate the artificial one.
To treat contact, DIANA FEA uses a zero-thickness interface element formulation, characterized by an elastic normal and tangential stiffness. The constitutive model of the interface is based on plasticity, following a nonassociative flow rule (with zero dilatancy) and a classical Mohr-Coulomb failure criterion. The physical damping is assigned to the interface elements based on a Rayleigh definition input. Because a unilateral SDR viscous damping is of interest, the alpha factor is set to zero and the beta factor is assigned accordingly to provide the desired
value at the
of Eq. (
7).
The software 3DEC, on the other hand, adopts an explicit solution scheme, which makes use of a central-difference algorithm to solve the equations of motion (
Lemos 2019). This software is based on the discrete-element method whereby the structure is modeled as an assembly of discrete rigid or deformable blocks. Rigid blocks are used in this paper, as a result of which system deformability is assumed to be entirely concentrated at the interfaces between blocks. Contact between blocks is modeled using zero-thickness nonlinear interface springs (point contacts); thus, no joint or interface elements are defined. As in the case of ABAQUS CAE and DIANA FEA, the soft contact approach is also used here, with the extent of interpenetration between blocks being controlled by the interface stiffness, and the nonlinear response is controlled via the tensile strength in the normal direction (set to zero herein) and the Coulomb-slip joint model in the tangential direction, which depends in turn on the values specified for the cohesion (zero in this case) and the friction angle (set artificially high) (
Pulatsu et al. 2019). Finally, stiffness proportional damping is used, which is specified through the damping ratio
and the frequency (in this case,
) at which it acts.
Fig.
7 illustrates a good agreement among the three simulation methodologies despite being fundamentally different (as explained previously). This conclusion highlights the universal applicability of the proposed numerical viscous damping model. The models used are readily available within each software package, and no additional implementations are required to reproduce the current results.
Conclusions
This paper presented a novel viscous damping model that derives equivalence between the damping of numerical block-based models and the impulsive nature of energy loss of the classical rocking theory. Specifically, new predictive equations are proposed and calibrated through over 1,000 free-rocking numerical simulations, which correlate the damping ratio of the numerical model to the coefficient of restitution of the statistically accurate classical rocking theory, for both the two-sided and one-sided rocking cases. The former is representative of the behavior of structures such as parapet walls, which are free to rock in both positive and negative directions, with impact only against the base, whereas the latter is more commonly observed in the case of façades, which are generally free to rock in only one direction, with impact taking place against the base and transverse walls.
The performance of the new model is evaluated through comparisons with experimental tests from the literature. In general, the novel numerical model performs reasonably well for the two-sided free-rocking case, with only slightly faster energy dissipation than its experimental counterpart. The two response-histories are in very good agreement in terms of both phase and peak amplitudes for the first cycles of motion.
In the case of forced rocking, the numerical model is again able to reproduce the experimental response with good accuracy in terms of oscillation phase and peak response, as well as potential overturning. In fact, for rocking amplitudes in the range of 30% to 80% of the critical (overturning) rotation (), which is also the range of practical engineering interest, the response is adequately captured for both free- and forced-rocking motion. For higher amplitudes, the response is slightly underdamped, which, however, is on the engineering safe or more conservative side.
In the case of one-sided rocking, a fair comparison is again observed for the free-rocking case, with the numerical model once again damping out marginally faster than the experimental tests. Further, the sensitivity of the response to the gap at the interface between the structure and the sidewall is also highlighted. However, this work reveals how adequately well the proposed numerical model captures the experimental response despite the major influence of the gap (on the response). This is also emphasized through the forced-rocking simulations, where the assumed value of the gap is found to strongly influence the predictions of the numerical model. Nevertheless, the proposed numerical (viscous damping) model shows adequate adaptability and robustness to capture the peak amplitude (albeit with some differences in phase) as well as the overall decay of the rocking motion.
Importantly, although the numerical simulations presented in this paper have been conducted in ABAQUS CAE, the universality of the proposed viscous damping model enables it to be used in fundamentally different numerical simulation software used for the analysis of block-based rocking models (i.e., DIANA FEA and 3DEC)—thus highlighting its broad applicability.
In conclusion, this work offers reliable and ready-to-use expressions that capture the energy loss mechanism of any block-based numerical model that undergoes either one-sided or two-sided planar rocking motion—in agreement with the well-established classical rocking theory. Importantly, the results of the present study serve as the basis for a more rational and holistic approach to model multi-degree-of-freedom rocking structures, the interaction of complex geometries and boundary conditions with the rocking response, and/or the inclusion of material nonlinearities, among others. Even though the proposed numerical (viscous damping) model was evaluated against the two most commonly observed out-of-plane collapse mechanisms of unreinforced masonry structures, to further enhance its applicability, a more thorough investigation against additional collapse mechanisms (e.g., vertical spanning strip wall or the corner mechanism) is required. Such analyses, however, should be accompanied by extensive experimental campaigns, which are currently lacking in the literature and thus are topics of future research.