You are on page 1of 18
8 — Testing of Masonry Structures for Seismic Assessment G. Michele Calvi, M.EERI, Gregory R. Kingsley, MERI, and Guido Magenes ABSTRACT The experimental evaluation of strength, deformability, and energy dissipation capacity of unreinforced masonry buildings subjected to seismic loadings presents unique and complex problems, bot for laboratory and eld evaluations, The paper addresses these problems, focusing on the relative merits and roles of several experimental techniques, including quasistatic, dynamic, and pseudodynamic loadings at full and reduced scale. INTRODUCTION The hazard presented by unreinforced masonry (URM) buildings in seismic zones has generated significant interest in the development of assessment, analysis, and retrofit methods which are appropriate for these structures. The formidable inventory and inherent diversity of these buildings demands a rational yet simple approach to assessment, well supported and validated by experimental research. However, experimental evaluation of unreinforced masonry, both in the field and in the laboratory, presents some unique problems, and must be approached somewhat differently than other types of structures. This paper outlines some of the eitical issues, and proposes some possible solutions, drawing on the experience ofan ongoing coordinated research effort on seismic response of URM structures, including experimental work at several institutions, (among them the University of Pavia, Italy, the University of Ilinois, Urbana, the Joint Research Center of the European Communities, Ispra, and the National Laboratory of Civil Engineering, Lisbon). Problems associated with both small and large tests on materials, components and complete systems will be addressed, with reference to quasistatic, dynamic, and pseudodynamic tests. The fundamental aspects of URM building response can be divided into sub-categories roughly according to scale: Structural Form, Unreinforced masonry structures are first distinguished from other building types by their structural form and the relative stiffness of the elements. In the most general terms, typical unreinforced masonry buildings are likely to be composed of several loadbearing masonry walls arranged in orthogonal planes, with relatively flexible floor diaphragms. When subjected to a seismic ground motion, the foundation transmits the seismic motion from the ground to the stiffest elements, the in-plane structural walls. The structural walls excite the diaphragms with a motion which has now been filtered by the wall response, and the diaphragms in turn excite the out-of-plane walls. When the diaphragms are flexible, the out- (GMC, GM) Dept. of Structural Mechanics, University of Pavia, via Abbiategrasso 211, 27100 Pavia, Italy (GRK) TSDC of Colorado, 805 14th Street, Golden, Colorado 80401, USA us (©Farthquake Spectra, Volume 12, No. 1, February 1996 146 G.M. Calvi, G. R. Kingsley, and G. Magenes of-plane walls are likely to see displacements amplified significantly from those of the in-plane walls, The mechanism of resistance follows the reverse order, with the inertia loads generated by out-of-plane wall motion being reacted by the diaphragms, and so on down to the foundation. The lateral load resistance is provided entirely by the in-plane walls. Observed seismic damage in URM structures offen includes out-of plane failures of walls, driven by excessive deflections of diaphragms and insufficient connections between them, but the in- plane walls provide the stability necessary to avoid collapse. The first problem in sting is then determine the components and subsystems which most closely represent realistic structural forms, and to determine the appropriate loads, filtered as they are by adjacent structural components, which should be applied to them, Mass and Stiffness. Unlike framed structures, a significant proportion of the mass in URM buildings is distributed throughout the walls rather than concentrated in the floors, particularly in older structures with very thick walls. The classical idealization of a building as a series of lumped masses connected by flexible elements is therefore suspect. Even if a reasonable lumped mass approximation can be made, the rapid and often highly localized stiffness degradation typical of masonry results in changing dynamic mode shapes and hence the mass lumping, ‘Sindarly, the stiffness of the diaphragms is typically not sufficient for them to be idealized as “rigid relative to the walls, and the standard model for floors of framed systems must also be abandoned. These characteristics have particular relevance to both quasistatic and pseudodynamic testing of structural systems, since dynamic mode shapes and even mass matrices cannot be assumed to remain constant as damage progresses. Masonry Components. As noted above, lateral load resistance is provided by in-plane walls. Usually, these walls are pierced by windows or doors, leaving a series of smaller piers to provide both the gravity and lateral load resisting systems. The behavior of such URM components differs dramatically from frame-type structures or reinforced concrete structural wall buildings. Mechanisms of lateral load resistance depend primarily on the pier geometry and the magnitude of vertical loads, and then on the characteristics of the brick/mortar interface. Flexural response tends to be dominated by rocking of piers rather than “beam” type behavior, and failure is generally characterized by shear, either in a diagonal tension. mode or by sliding on the predefined planes of the ae joints. Furthermore, masonry is highly sensitive to a conditions, and the effect of axial loads. This sensitivity makes masonry component behavior difficult to predict, and also difficult to replicate inthe laboratory. Mi Unreinforced masonry distinguishes itself from other materials most obviously by the tremendous variety of its constituent materials, and the importance of their interaction in the characterization of the structural response. The properties of individual components (brick and mortar) and, more importantly, their interaction must be understood before the behavior of components or structural systems can be predicted. For example, a relatively slight change in the characteristics of mortar bond can completely change the failure mode, load capacity, and displacement capacity of an entire masonry building. This raises the question: what material properties should be measured in Taboratory” ‘tests, and what methods should be used both in the laboratory and in the field? Nondestructive Evaluation. Finally, the analysis of URM structures is motivated almost exclusively by the need to assess the condition of existing buildings. Test methods for masonry must include nondestructive and / or mildly destructive tests which can be accurately applied in the field to quantitatively assess the critical parameters of masonry behavior. Mani tests are available, some of them standardized and others not, varying from localized maté tests to full scale dynamic vibration tests on existing buildings. The question remains as which of these lead most directly and accurately to a quantitative understanding of seismic resistance of existing buildings. ‘Testing of Masonry Structures for Seismic Assessment 47 Assessment. The seismic assessment of an unreinforced masonry structure entails consideration of all levels of scale outlined above: materials, components, and the complete system. Typically, the assessment of diaphragm flexibility, wall connections, and out-of-plane wall stability will be a concern, and measures will be taken to reduce out-of-plane wall failures by tying walls to diaphragms and stiffening diaphragms. Following this, the evaluation of the in-plane lateral load capacity of systems of masonry walls or piers will be of primary ortance, since they will define the final failure mode of the structure. To help put the subject of masonry testing into context, let us examine this particular problem in some detail. For the purposes of discussion, the failure mechanisms of masonry piers ~ rocking, sliding, and diagonal shear -- can be described very simply, sacrificing some accuracy to the purpose of emphasizing the relative importance of various parameters on the response of piers [16]. The maximum horizontal shear which can be resisted by a rocking pier may be approximated ay DtP(,__P ) Yaw e 0852 2 where D isthe per width, #7 isthe pier height the pier thickness, p=P/(D) isthe mea eae vertical stress on the pier due tothe axial load P fas the compresive strength of tad ys a parameter which describes the boundary conditions, taking a value of 1 when the” pie i fied on one end and fee to rotate onthe other, and a value of 2 when the pier is fixed at s. The capacity of a pier undergoing sliding along a horizontal joint may be approximated by V,= uP @ where 4 represents the sliding coefficient of friction of the masonry joint. The shear failure of URM piers associated with diagonal cracking is difficult to describe with a simple expression. For the sake of discussion, it is however reasonable to assume that the maximum shear can be described by an expression in the form LaPt fp, Po. gi V,=2 fi+e ; = 4 Su D where f,, represents a conventional tensile strength of masonry (not to be confused with the tensile strength of bed joints) to be determined by shear tests on wall specimens, and b is an empirically based parameter which is equal to the pier aspect ratio within a range of 1.0 to 1.5, and is constant outside these bounds [24]. The above expressions give an immediate indication of the parameters which must be evaluated in order to make a quantitative assessment of the seismic lateral load capacity of URM piers. The pier dimensions, for example, play a strong role in determining its likely failure mechanism and load capacity; unfortunately, the remaining parameters may not be so easy to measure as the aspect ratio, » lOsbsi5 @) ‘THE ROLE OF LABORATORY MATERIAL TESTS, For seismic assessment, the utility of results from individual material tests relies entirely on the availability of analytical models which can reliably relate the properties of materials to the properties of the masonry assemblage. At present, although much progress has been made in the recent years in the field of constitutive models, it seems that there is not yet enough confidence in the general reliability of these methods. When historical masonry is considered, 48 G. M. Calvi, G. R. Kingsley, and G. Magenes the variety of configurations and materials which can be encountered is such that a model which is satisfactory for one type of masonry can be inappropriate for another. In the case of clay brick masonry, recent and satisfactory examples of refined cyclic constitutive models implemented in nonlinear finite element codes can be found in [11]. There are also more Simplified models avaliable to correlate the pr properties of bricks, mortar and joints to those of masonry. For example, some models for in-plane failure domains of brick miry which have been shown to correlate well with experiments can be found in [12,17]. ‘These models are applicable for safety checks both for design and assessment TEST PROCEDURES AND INTERPRETATION OF EXPERIMENTAL RESULTS Since masonry is a composite material, tests on materials must aim not only to characterize the units and mortar as separate materials, but also their interaction, namely the brick-mortar interface properties. These tests are, in most cases, not particularly expensive or difficult, since they require only limited loads and standardized equipment, but the layout and the instrumentation of the test can become rather complex if detailed information on deformations is sought, as might be required to calibrate sophisticated numerical models. Masonry material tests are often characterized by a wide scatter of results, so a sufficient number of repetitions must be conducted to obtain a satisfactory statistical characterization of the mechanical properties. This scatter may, however, be attributable in some part to insufficiently controlled laboratory environmental conditions, manufacturing and testing procedures. In particular, masonry can be very susceptible to variability due to workmanship It is important to understand that this factor should be isolated and treated independently along with other test variables inasmuch as possible, and that laboratory tests should seek to reflect uniform conditions of workmanship. Although real buildings are not built with such. strict controls, it is desirable to limit the possible causes of scatter when laboratory tests are used to validate or calibrate numerical models: the good or bad performance of a numerical mode should not be affected by factors which do not have anything to do with the capabilities of the model. Indivic Tests on Units and Moi Some tests on units and mortar ae rather common, and thir procedures codified: typical tests are those to measure compression of units, cor sion of mortar, actos measure onesie seg of nis, gongs Sng of moras procedures is still ongoing, since many aspects of the procedure can have a strong effect on the results. For instance, the measured strength and stifthess depend critically on the dimensions of the specimen and the machine-specimen interface. Moreover, measures of deformations have been shown to have more than 100% variation, depending on the Sy device (e.g. linear displacement transducers or electrical resistance strain s) and of the support of the ‘ranaducers [6 Therefore in view ofthe variability ofthe results, and in expectation of a complete und. ‘the experimental procedures, analytical models which are very sensitive to Geformabaity parameters of the single constituent materials will necessarily require calibration with full scale structural component tests. /Tensile St Int Fric Tests on mortar alone cannot reveal the most important characteristics of the masonry joint, since these can be strongly influenced by the properties of the masonry units, most importantly the absorption properties. In joints made er gears absorptive units, the shear and tensile bond strength can be much less than those of the mortar alone. ‘Thus, the units can be the critical factor in determining the properties of the mortar joint, which in turn have strong influence on the tenant the masonry assemblage in shear and tension. Tensile tests normal to the joints relate primarily to the flexural behavior of the wall (both in- plane and out-of-plane). The most commonly used and codified tests to measure the tensile Testing of Masonry Structures for Seismic Assessment 149 strength of ints are the bond wrench test and the cross-couplet test (e.g. ASTM C lo72-86 and ASTM 952-91). Typically this guar is subjected to igh ‘scatter, and its value is so low (compared to compression strength) that a no-tension behavior normal to joint can be considered a safe and sufficiently accurate hypothesis. Furthermore, under cyclic seismic actions, flexural cracks in bed joints are likely to develop early, and the ultimate strength and deformation will have negligible relationship to the tensile strength of the bed joints (as reflected by Equation (1)), For the study of the shear behavior of joints, both complex and simple methods can be adopted [5, 6,]. Accepting the linear Mohr-Coulomb envelope as a good model to describe the interaction of normal and shear stress (which has been shown to be acceptable on a limited range of low normal stresses), the strength of the joint can be described by means of the cohesion c and the friction coefficient su. The friction coefficient of the strength envelope is, in most cases, almost identical to the friction measured after failure when slip occurs along the joint surfaces. The degree to which normal stresses along a joint approach a uniform distribution is the most important element which distinguishes one setup from another. Cohesion is by far the most sensitive parameter to setup and stress distribution, while friction is less sensitive. Whatever setup is used, the it a of the test in terms of mean stresses and displacements or slips is acceptable. The shear-box test described in [5] is a setup which allows a good uniformity of stresses along the bed joints and full control of mean normal stress (which is kept constant) and relative shear displacement (the mean shear stress is measured) also for cyclic tests. The more simple triplet test [6] is an economical but sufficiently accurate alternative for monotonic tests, and ‘with some variation can also be adapted for cyclic tests. Work is still in progress to establish reliable correlation among the different test setups. Whenever failure envelopes for joints are determined, it must be kept in mind that they cannot be directly used as shear failure criteria for masonry, since they describe only a local phenometon: appropriate modifications should be yatroduced [17], The coefficient of ffiction ican however be used with some confidence to define the shear sliding condition (Equation (2)) ona pre-formed horizontal crack (e.g. due to flexure) along a bed joint. For in-plane shear failure of masonry as described by Equation (3), tests on structural ‘components (wall panels) still remain the only direct experimental way to quantify the conventional tensile strength fa. Compression of prisms or wallettes Prisms or wallettes can actually be considered as structural elements rather than materials, but most models make use of strength parameters and stress-strain relationships which assume that masonry acts as an equivalent homogeneous material, for which the prisms or wallettes become the smallest representative portion. Compressive strength and the initial tangent or secant modulus are the primary parameters used in simplified models. It must be clear that compression failures are seldom encountered in unreinforced masonry buildings subjected to earthquakes. More often, local crushing is associated with in-plane or out-of-plane flexural failures, which, in the usual case where gravity stresses are less then 20% of the compressive strength, are not very sensitive to variations of the compressive strength (see Equation (1)). Compression strength is mostly used as a general measure of the quality of the material, to ich other parameters (e.g. shear strength) are sometimes correlated by means of empirical ctiteria. Such correlations are generally weak, and offer only very approximate guidelines. Whenever possible this approach should be replaced, or at least checked, by direct testing of the shear strength of masonry. 150 G.M. Calvi, G. R. Kingsley, and G. Magenes Young’s modulus is necessary for an estimate of the stiffness of masonry. When evaluating the stiffness it is, however, often overlooked that in wall structures the shear modulus G is as influential if not more than Young’s modulus, and, due to the orthotropic nature of masonry, the correlation between G and E is very poor, particularly for old masonry. Gis also unrelated to Poisson’s ratio v. Empirical formulas which are commonly used to evaluate the shear modulus G from Young’s modulus E (e.g. G= 0.4E) can be very unreliable unless specifically calibrated for the type of masonry considered. Again, shear tests on wall panels are a necessary and much more reliable complement to compression tests, when an accurate stiffness evaluation of shear stressed walls is needed. As in tests on individual materials, deformation measurements on small masonry subassemblies can be affected by huge scatter, even when the gauge length includes several courses of bricks. This is due to local effects associated with the natural heterogeneity of the material. As a consequence, it is always advisable to test a sufficient number of specimens when accurate evaluation of the stress-strain relations is needed, and to use redundant instrumentation for checks on the measured deformations. In conclusion, it is apparent from Equations (1) to (3) that the properties of unreinforced masonry in compression play a secondary role in the behavior of components relative to the properties of the joints in shear or in combined shear and tension. Tests on joints can therefore be considered to take precedence over compressive strength tests in the strength characterization of unreinforced masonry subject to seismic forces, rather than the other way THE ROLE OF STRUCTURAL COMPONENT TESTS It was mentioned in the introduction how the in-plane resistance of masonry walls is the primary mechanism which allows buildings to withstand seismic action. For this reason a great effort has been made in the past by several investigators to characterize the in-plane behavior of masonry walls, both statically and dynamically. The large majority of these tests have been focused on masonry piers subjected to in-plane shear, i.e. walls which are loaded through their upper and lower base cross-sections, and which represent the portion of a building between two openings. Very little experimental information is available on coupling spandrels or deep beam behavior in unreinforced masonry. This can be in part justified by the fact that both post-earthquake assessments and shaking table tests on building models have shown that, once out-of-plane failures are prevented, the final collapse of URM buildings is associated to shear failure of the piers of a critical story (usually the ground story). It may also be presumed that the behavior of spandrel beams is similar to piers, but there is little justification for this considering that the geometry, axial load levels, and boundary conditions will be completely different, Failure or cracking in the spandrels or coupling beams may have an important consequence on the degree of coupling of the piers, but final collapse is due to pier failure. TEST PROCEDURES In-plane tests of masonry walls or piers are typically performed under a given constant axial load, and with the application of a monotonic or cyclic shear force or displacement, but other procedures have been used in the past (e.g. monotonic or cyclic diagonal compression) where both shear and axial load vary during the test. The boundary conditions at the top and bottom of the wall are usually either fixed-fixed (no rotation allowed, double-bending condition) or fixed-free (rotation allowed at the top, free cantilever single bending). Neither choice can be considered the closest to reality, but they are chosen because they reproduce unambiguous static or kinematic boundary conditions which facilitate the interpretation. The most general possible setup is shown in Figure 1 Testing of Masonry Structures for Seismic Assessment 151 Force or displacement controlled Displacement controlled fcators ie actuator <> Xe ll secinjen H Figure 1. Test set-up for combined actions of shear and compression on URM piers. Both quasi-static and dynamic tests are of interest. Possible dynamic test procedures are the use of a setup identical to static tests, increasing the rate of application of forces and/or displacements [25], or the use of a shaking table [15,18] It is recognized that a real seismic excitation is better simulated in dynamic tests, but quasi- static tests have several advantages with respect to dynamic shaking table tests: e.g., application of large forces to the specimen is easier, the test to collapse of large specimens or structures requires less expensive equipment, phenomena like cracking and spreading of damage can be observed more closely, and forces and displacements can be measured much more accurately. On the other hand, masonry exhibits rate-dependent behavior: propagation of cracking at constant load or at constant imposed displacement is often observed, so that quasi-static tests tend to show more extensive damage and lower strengths than dynamic tests. ‘As a consequence, differences between static and dynamic tests involve, in higher or lower measure: strength, stiffness, damage propagation, and energy dissipation. At present, the comparison of existing experimental data regarding the in-plane behavior of masonry walls subjected to shear do not yet allow the establishment of clear quantitative correlations of the results of dynamic and quasi-static tests, mainly because only few dynamic tests have been performed. Some quantitative correlation procedures for reinforced masonry have been suggested in [25]. In general, static cyclic testing is a conservative approach. Within static tests, the chosen displacement or load history (i.e. whether monotonic or cyclic, and with what cycle patterns) has some relevance on the results, mostly in the post-ci cling regime, where stiffness and strength degradation is affected by repetition of cycles. Given the maximum displacement, cyclic histories like those shown in Figures 2 (A) and (B) are more severe and thus conservative when compared to monotonic or random earth like histories. Pseudodynamic testing may also be considered, but it is questionable whether pseudodynamic tests are necessary for single-degree-of-freedom components [7], unless used as a complement to cyclic testing or for validation of numerical models. Recent pseudodynamic tests on URM walls have been performed to check the applicability of the procedure to stiff brittle structures [4] in anticipation of applications to multi-degree-of-freedom systems. 152 G.M.Calvi, G. R. Kingsley, and G. Magenes Figure 2. Typical load vs. displacement response and idealized crack pattern of URM piers exhibiting (A) diagonal shear behavior, and (B) rocking behavior as a result of a change in aspect ratio (from [4]). INTERPRETATION OF STRUCTURAL COMPONENT TESTS Within the framework of seismic assessment, the main merit of tests on structural components is to allow the understanding of some fundamental issues i.e. 4) the types of failure mechanisms of prevailing interest for seismic behavior, b) what are the main parameters (material properties, geometry and boundary conditions) which influence the type of failure mechanism. ©) what is the influence of the type of mechanism on the cyclic behavior of walls; The first issue has been already discussed in the introduction, distinguishing in-plane and out- of plane failure modes, flexural, diagonal shear, and shear sliding mechanisms. it may be worthwhile to add that what is commonly defined as diagonal shear cracking can be associated with different crack patterns, ranging from single wide-open isolated cracks to diffused cracking, as a combination of bed- and head-joint failure or brick shear-tension splitting [14,16,17]. Experiments have shown that, given the material properties, the main parameters which determine the failure mechanisms of masonry walls are the axial load, the aspect ratio and the boundary conditions at the ends, as also reflected by Equations (1) - (3). The relationshiy between the various modes is illustrated conceptually in Figures 3 (A) and 3(B) which show, respectively, the effect of changing aspect ratio and axial load on pier shear strength expressed in nondimensional form, according to Equations (1) - (3). Once the boundary conditions are determined, an increase in aspect ratio tends to lead to a rocking failure, while a decrease leads to shear-type failures (either diagonal cracking or shear sliding), as shown in Figure 2 and 3(A). Keeping the geometry and the boundary conditions constant, an increase in axial load leads to diagonal shear failure, while a low axial load tends to be associated with either sliding or rocking, Figure 3(B). Both quasi-static and dynamic tests [4, 14, 15] on masonry piers confirm the influence of these parameters. ‘Testing of Masonry Structures for Seismic Assessment 153 0.20 az Rocking 4 esses Shear oe a Sliding 0.15 a yey wom WBE Shear Wier ot) 00s 4 Sliding 0.03 4 aad T i $ odo Ode O10 81s "OH0 Pier Aspect Ratio (H/D) Pier Axial Load Ratio (P/(f'm Ot) Figure 3. Influence of (A) pier aspect ratio and (B) axial load on the theoretical shear capacity of URM piers responding in rocking, shear, and sliding modes. Bound: conditions are fixed-fixed (w=2), coefficient of friction 1=0.4, and fu, = 0.05f'm. In addition, the variation of the properties of the materials in terms of the relative strength of mortar joints and bricks can favor some mechanism with respect to another. It was shown in [15] how a higher mortar bond strength tends to inhibit shear failure in favor of a rocking/sliding failure. The possibility to predict the failure mechanism in a wall is quite important, since static and dynamic experiments have shown how rocking and sliding tend to be associated to good seismic behavior, and diagonal cracking t0 bette behavior. Rocking provides a limit fo the maximum shear and dissipates energy through impact, keeping the integrity of witl the exception of localized damage in the comers. Sliding on horizontal bed-joints is a strongly dissipative mechanism, and, similarly to r maintains the overall integrity of the wall, although it can also result in larger permanent displacements. Also, rocking and sliding often tend to occur simultaneously, since when rocking occurs the local degradation due to stress concentration at the corners tends to decrease the friction coefficient along pre-formed flexural cracks. Diagonal shear cracking, on the other hand, is associated with brittle failure which, under dynamic loading, can be catastrophic and explosive in nature. Under some conditions, the post cracking behavior in diagonally cracked specimens has been found to be almost “ductile” in the sense that stable cycles of displacement with a significant residual strength were possible. This behavior appears to be associated with two conditions: shear cracks due to joint failure rather than unit failure, and low aspect ratios (H/D<1) [3, 18]. The latter condition appears to be the most significant: low aspect ratios tend to show stair- stepped cracks which have a small inclination and allow a rather effective residual shear transfer across cracks. In these cases the post-cracking behavior was not reported to be catastrophic, and the vertical load carrying capacity was maintained, The limited number of dynamic tests on structural onents calls for further research. Shaking table tests [15,18] have confirmed that dynamic tests show analogies with static tests in terms of failure mechanism and in terms of interaction of such fundamental parameters as axial load or aspect ratio. However, besides the mentioned rate-effect which causes apparently higher strengths in dynamic tests, the interest in shaking-table tests is that they exhibit phenomena that static tests cannot reproduce, and that are not related necessarily to strain- Tate effects. As an example, dynamic impact phenomena and the interaction of horizontal and vertical acceleration components when rocking occurs can lead to shears which are higher 154 G. M. Calvi, G. R. Kingsley, and G. Magenes than static equilibrium for overturning would imply. Tis phenomenon, which was measured in the experiments described in [15], calls for a better understanding of the dynamics involved, which would translate possibly to suitable correction factors for the equilibrium equations, when appropriate, at the assessment/design level. As a conclusion to this iter it can be stated that, with respect to strength and stiffness of a ‘wall subjected to in plane shear, tests on panels are the most reliable information for the calibration of any analytical model to be used in seismic assessment. Since flexural strength is based on rather reliable and simple formulations, tests should aim to give information on shear failure. When testing capabilities are limited, the information which can be derived by a few monotonic diagonal compression tests on rectangular panels with at least two different aspect ratios (¢.g. H/D= 1 and H/D =2.0) is at present by far more valuable and reliable than any other information based on single material testing. Such tests can be performed in principle on any universal testing machine for structural building materials. THE ROLE OF STRUCTURAL SYSTEM TESTS Isolated structural component tests are the most direct and cost effective means to understand structural behavior; however, it is necessary on occasion to test a complete structural system in order to observe the response of the components under realistic combinations of actions (applied displacements, rotations, shear, bending moment, and axial load), and to verify analytical models applied to complex systems, To test a complete URM structural under realistic seismic force distributions requires consideration of several complex and sometimes conflicting aspects of response. In addition to the effects of loading rate, material properties, axial loads, and wall geometry di: above, the behavior of the complete system is also influenced by the nonlinear interaction of diverse components such as timber diap! and massive stone walls. Consideration of the nature of the interactions can shed some on both the expected response of URM buildings and on the most appropriate way to test them. TEST PROCEDURES Shaking Table Tests, Shaking table tests on reduced scale models constitute an invaluable source of information for the fing of the dynamic behavior of masonry buildings in the nonlinear range. In terms of loading, this kind of experiment is the closest to reality, and has several merits with respect to quasi-static testing. The forces and relative displacements in the building are generated by the dynamic response of the structure. The problem which is present in static tests, i.e. what displacement or force history shall be Peed age iver as completely bypassed. The distributed inertia forces involved can be modeled correctly, while in static tests concentrated loads must be applied and the structure must always be idealized as an ‘equivalent lumped mass system. Failure modes associated with out-of-plane response of walls can be reproduced correctly, while in static tests this is very difficult if not impossible to reproduce. On the other hand, the nearly arbitrary choice of the ground motion (frequency content, duration, peak acceleration, natural or artificially generated) is comparable to the arbitrary choices involved in the selection of a displacement or load pattern, Rate effects are solved only as long as the reduced scale materials have a similar rate sensitivity to the original. In led model masonry a tendency to an increase in bond strength is reported, with modifications of the crack pattern and failure mechanisms. The definition of the forces involved in dynamic tests can be made only through the measured accelerations: due to the complex dynamic response of the structural system the result can be very difficult if not impossible to understand in the time domain. In addition to this, in distributed mass systems ‘Testing of Masonry Structures for Seismic Assessment 155 the evaluation of inertia forces calculated from measured accelerations in discrete points introduces some degree of approximation. The measurement of relative displacements involves also several difficulties, which are increased when stiff wall structures are tested. As a result, an understanding of the behavior of the building in the familiar terms of forces and relative displacements, and a correlation of these quantities with the progress of cracking and damage is much more difficult in shaking table tests than in quasi-static tests. It can be thus stated that both shaking table tests and static tests are necessary for a full understanding of the behavior of complex systems. As a natural consequence of the authors’ experience in static testing ing on full scale structural systems, attention will be paid in the following to the problems related to this kind of testing. Gees In a quasistatic test on a full-scale structural system consisting of URM walls and flexible floor diaphragms, the critical problem is to select appropriate patterns of forces and/or displacements to apply to the structure. ~povted mode shapes and mans datrbutions Wak the poleatal for rapily Jewalng le s and mass tions. Wit potential for rapic stiffness in masonry walls and the associated change in participating mass, the relationship between the walls and floors may easily change over the duration of a few cycles. Thus it is difficult to define a fixed load pattern which may be considered to represent a dominant mode of response in many URM structures. - When two parallel walls in the plane of seismic force are lightly coupled by a very flexible development of the shear capacity in one of the walls immediately limits the force which may be introduced into the other by the diaphragm, and the ultimate capacity of the stronger wall may never be developed [1,9]. The behavior of even lightly coupled walls is ths interdependent, and the response of one cannot be considered independently of the Ver. - Ina multi-story structure, the very high initial stiffness of the walls results in a high degree of coupling between the displacements of each floor. This suggests that the even very small errors in the application of a fixed pattern of displacements to a test structure may result in highly unrealistic interstory shear force distributions. = In taller structures, the high stiffness of upper stories magnifies the po: mode contributions to the total base shear, even after the development Peron stories a [23]. This suggests that the application of a fixed force distribution to the structure may neglect critical aspects of the structural response to seismic ground motions. - URM buildings have a strong potential for brittle failures, globally or locally, and are re highly likely in any case to exhibit strength degradation once their maximum capacity has reached. The use of actuators in force control is therefore highly discouraged. It appears that the application of fixed displacement patterns is unacceptable for URM test structures, and at the same time that the use of force controlled actuators is also unacceptable. A compromise may be found, however, bearing in mind that the relative importance of each of the above points depends very much on the characteristics of the structure to be tested, and that the approach to every test may be different. Consider, for example, the structure in Figure 4, representing a full-scale URM building prototype with very flexible diaphragms, which was the subject of a quasistatic test at the University of Pavia. In order to define a realistic sequence and pattern of loading for the structure, a reduced scale model was first tested on the shaking table at the University of Illinois by Abrams (. The dynamic test revealed that the seismic loading followed a rather simple pattern for this structure. Because approximately 60% of the mass was concentrated in the flexible floors, and the mass in each of the two story levels was equal, the seismic forces 156 G. M. Calvi, G. R. Kingsley, and G. Magenes eSt Stwucruee Figure 4. Full-scale URM building prototype, with flexible diapragms, tested at the University of Pavia, and the 3/8 scale model tested on a shaking table at the University of Hlinois. at each level were nearly equal and usually in phase with each other. The structure was only two stories, so higher modes of vibration played a negligible part in the load distribution. On the other hand, displacements in the two parallel walls were often out of phase, and nearly independent of one another. As noted above, the forces in the stronger wall were limited by the maximum force achieved in the weaker wall, and thus failure only occurred on the weaker side, To replicate these conditions in the full-scale test, a hybrid approach of applied displacements and forces was used. Equal displacements were applied at the top of each of the two in-plane walls, thus eliminating coupling between the two walls through the diaphragm at that level. At the first story levels, a force equal to the force at the top of the wall was maintained, thus simulating the observed uniform loading pattern in each wall in the dynamic test. In the full-scale test, it was possible to reach the maximum load capacity of both in-plane walls, since loads were not limited by the diaphragm. For stability of the test, all actuators were in fact cperated in displacement control, necessitating the use of an iterative actuator control algorithm to apply the desired loading pattern [7,22]. Pseudodynamic tests The difficulty of choosing a realistic pattern of forces or displacements to apply in a quasistatic test suggests that pseudodynamic testing procedures would be appropriate for URM structures. This approach has been largely untried, however, and presents some unique problems of its own. Calvi and Nakashima [8] have addressed some of these problems, and proposed an approach to the pseudodynamic testing of masonry structures. Of particular importance is the relationship between the assumed distribution of lumped masses required for the analytical portion of the procedure, and the changing stiffness of the structure. As damage progresses, the dynamic mode shapes of the structure will also change, and thus the assumed participation of the distributed mass of the walls. The pseudodynamic testing algorithm must therefore make provision for such changes. The well known difficulties of error propagation in pseudodynamic testing of stiff, multi- degree-of-freedom structures must also be addressed in testing masonry buildings. Some Testing of Masonry Structures for Seismic Assessment 157 broblems may be circumvented through the use of high quality high esclution displacement measurement instrumentation, and high quality servo-controlled hydraulic actuators. A cedure developed at UCSD in which the controling actuator displacements are amplified Bom the structure displacements through the use of sof coupling devices [22] would also be applicable to URM structures. There remains the problem that the choice of input seismic ground motion for pseudodynamic tests is extremely broad, and thus the resulting displacement history is nearly arbitrary [7]. Note that in this case the pattern of displacements applied to the individual degrees-of- freedom will be realistic, but that both the pattern and sequence of displacements may vary dramatically depending on the choice of motion. This problem is well beyond the scope of this paper, but it may be noted it has been satisfactorily solved for the case of a reinforced leat ‘masonry structure through the use cofthe Generated Sequential Displacement me INTERPRETATION OF STRUCTURAL SYSTEM TESTS The same factors which can affect the behavior of materials and components can also influence the outcome of complete structural system tests, with additional factors arising due to the interaction of components. As an example of the former, a change in the mortar joint pro ies alone can completely change the mechanism of failure of a system, as has been in recent tests [1,9]. Consequently, structural system tests cannot be conducted wthout sufficient supporting material and component tests. ‘An important example of the effect of component interaction is the role played by deep spandrels, as discussed previously. Stiffness and strength of deep spandrels affect the boundary conditions and thus the moment/shear ratio of the piers, and affect the amount of vertical load in the piers generated by overturning. Results from the structural test of the two- story building discussed above indicated that the behavior of the spandrel beams can have a strong influence on the structural response, but at present there is insufficient experimental data to adequately quantify the role of spandrels in the nonlinear regime. Both component and fall scale tests are necessary to fill this gap. Instrumentation on structural system tests must be designed to (1) characterize the overall system response, (2) validate analytical models of components and systems, and (3) verify the response of individual components tested previously, including the boundary conditions between components wherever possible. Instruments which can give some indication of the distributions of loads among different elements should also be sought, but this task will be difficult, if not impossible, in most cases. THE ROLE OF NONDESTRUCTIVE EVALUATION It was stated in the introduction that the motivation for understanding the behavior of unreinforced masonry is the large stock of existing URM buildings in seismic zones. Laboratory testing of URM structures is conducted in support of the development of analytical models, both simple and complex, which can then be used to predict the response of existing structures, but no matter what the type of analytical model is used, some characterization of the material and component properties will be necessary, and to do this in the field requires nondestructive or partially destructive test methods which can be applied in- situ. Considerable effort has been applied in recent years to develop such methods (19], and different techniques are gaining acceptance around the world for a variety of applications. The following are among the most useful for the specific case of seismic assessment. 158 G.M. Calvi, G. R. Kingsley, and G. Magenes Incplace shear a= 0.35 MPa & a= 0.15 MPa Shear Stress (MPa) 2 8 g 8 ° Displacement (mm) 7 Figure 5. Comparison of sliding shear vs. displacement results from in-place shear tests and triplet tests conducted on identical materials. (o¢= stress applied by flatjacks, o = applied normal stress on the triplet test joint.) MATERIAL TESTS ingle Flatiack Tests for Mi t of In-sit ._As discussed above, all of the major es of lateral load resistance of masonry piers depend critically on the presence of vertical loads. Assessing the existing state of vertical stress in masonry walls is therefore a crucial part in the overall condition evaluation. In the simplest cases, this may be achieved adequately simply by estimating the size and weight of the structure above the pier of interest. If however, there is some ambiguity about the vertical load path, or any indication that settlement or previous load histories have redistributed gravity loads around the structure, a quantitative measurement can be achieved through use of the flatjack test (see ASTM C 1196- 91), Results of the test should be interpreted with care, however, remembering that the method only measures stress in the outer wythe of material, and that stress distributions through the wall thickness are not likely to be uniform. ‘Two Flatjack Tests for Measurement of Deformability. Two parallel flatjacks are often used in tandem to measure the deformability and estimate the strength of the masonry between them (see ASTM C 1196-91). Because the deformability serves primarily in the estimation of the initial structure period of vibration, and is not related to conditions of ultimate strength or displacement, the utility of this test may be considered to be less than that of the single flatjack test. ' Tests. Characterization of the joint shear strength is essential in defining the resistance of piers in a sliding shear mechanism. ‘The standard method used in the U.S. is the in-place shear test (see Uniform Building Code Standard 21-6) in which a single brick in a I is displaced horizontally. This test is conceptually sound, but susceptible to possible errors in its application, not the least of which is the inaccurate estimation of the vertical load on the test unit, which is directly proportional to the shear strength. An improved method is currently under consideration by ASTM in which the test brick is isolated above and below by flatjacks located several courses away from the test unit. The purpose is to control the vertical stress on the joint, to obtain a realistic measure of the coefficient of friction. Even in this case, the vertical load on the unit may not be uniform due to the voids cut adjacent to the shear plane, and due to the effects of dilatency as the unit is displaced. A comparison between ‘Testing of Masonry Structures for Seismic Assessment 159 sliding shear test results from both triplet tests and in-place shear tests conducted with flatjacks is shown in Figure 5. Materials used in both cases were identical. It is evident that the in-place test can significantly overestimate the shear strength as represented by the triplet, test, probably due to an underestimation of the actual vertical stress on the joint. in particular, while the flatjack test gives a reliable estimate of the coefficient of friction, it largely overestimates the cohesion, which should be corrected by means of a calibration factor. FREE AND FORCED VIBRATION TESTS Itis known that it is possible, in principle, to obtain stiffness, damping and boundary conditions of the members of a structural system by measuring its dynamic response and feeding an inverse linear analysis with the mass distribution and parameters of the dynamic response, such as natural frequencies and modes of vibration obtained from transfer functions between different points of the structure. The problem is only well conditioned and the results reliable when the number of unknowns is less than or equal to the number of equations to be satisfied. For example, if only the first four frequencies of vibration of a structure are ‘measured, not more than four structural parameter may be identified [10]. Three main practical applications of vibration testing of masonry structure can be envisaged: ~the experimental measure of the frst periods of vibration will immediately allow the perception ofthe soundness of a numerical mod - the comparison of the eee i decoration of the Goundary content, provided that a similar local damage has also ~ a measure of the apparent viscous damping can offer some insight into the dissipation phenomena characterizing the structural response. For example a larger damping may suggest that some joint friction and local impact are present. Often sources ci nonlinearity such as imperfect connections and cracking also contribute to the increase of the incoherence between signals detected by transducers mounted on the same member. The first application is most valuable when the complex geometry and mass distribution of the structural system (for example cathedrals, or complex building groups in old urban nucle!) makes its seismic behavior sensitive to several modes of vibration. The second application has been extensively and successfully used to identify the degree of fixity ofthe columns ofthe Leaning Tower of Pisa [21]. The third application is particularly useful when some strengthening intervention has been made, and it is desired to check that the structural response has improved. A lower damping normally corresponds to a structural response farther from damage and collapse. When new systems have been added and are supposed to combine their response with the structure, the first and third applications can both be conveniently applied. me refined dynamic identification procedures might be successfully applied in the case of very special masonry structures (e.g. simple structures such as masonry towers) or in the case ofa vary special environment (i.e. a laboratory) where a large number of experimental data can be reliably obtained. The latter case is illustrated by the full scale masonry prototype tested in Pavia [9,10], which showed frequencies of vibration in the range of 4 Hz to 22 Hz in the undamaged state, with apparent viscous damping approximately equal to 1.3 %. After having been sac ‘damaged, the frequencies of vibration decreased by aj prom 3 50 % the corresponding ee ‘changed in relative amplitude of vibration the apparent damping increased by 2.7 %. An accurate comparison between behavior and corresponding numerical results allowed the evaluation Gtsome taportant important aspects of the dynamic response such as the low coupling offered by the flexible floors and local damage (in the form of a local reduction in stiffness). 160 G. M. Calvi, G. R. Kingsley, and G. Magenes In the case of masonry towers, which can be modeled as a simple cantilever, more complex parameters have to be analyzed in a refined identification procedure, because of the low sensitivity of both eigenvalues and eigenvectors to a variation of the mechanical properties. Tt must also be recognized that a dynamic identification procedure can offer information on the linear response around a specific point of the global force-displacement curve, and only indirect indications of strength can be inferred from elastic modulus evaluations. Fast and inexpensive preliminary measures of the dynamic response can often be obtained without any artificial excitation (with ambient vibrations), particularly when the structure is relatively flexible (as in the case of a masonry tower). In other cases some measurable structural vibration can be obtained by simple means such as braking trucks, before going to a proper exciter. CONCLUSIONS Seismic assessment of URM structures must be based on a realistic model of structural behavior, including an understanding of the sequence of events that will be produced by the flow of forces and displacements throughout the structure. This implies the comprehension of local and global mechanisms of damage and collapse that can only be developed with an appropriate combination of experimental testing and numerical analysis. Research oriented testing, mainly by means of laboratory tests, has the role of defining material and member constitutive relations and of exploring the global response produced by the mutual influence of different structural members. Assessment practice must also be supported by testing, oriented towards the quantitative evaluation of selected fundamental fers of the materials and the structure. This is possible mainly through nondestructive in-situ testing, since only in rare cases can undisturbed samples of masonry be | brought to the laboratory for destructive testing, and it is in practice impossible to manufacture exact replicas of the original material. The role of both laboratory techniques and in-situ testing were addressed in this paper. ‘Some important issues can be summarized here: - The diverse nature of masonry materials demands experimental testing in support of seismic assessment, but that testing need not be excessive if it is well planned with specific consideration of URM mechanisms of resistance. - Among the failure mechanisms associated with seismic actions, shear failure modes are the most dangerous and most difficult to assess. Priority should be given to the collection of experimental information on shear strength and deformability. Laboratory tests (in particular shear-compression tests on wall panels) performed and collected in the literature constitute a fundamental source of information. In-situ joint shear testing (shove test) is very useful for safety checks, but proper correction and safety factors should be applied to compensate its tendency to overestimate the strength (cohesion) of the joint, associated to local dilation effects. - The sliding failure on pre-formed joint cracks depends essentially on the friction coefficient of mortar joints. In-situ shove tests can provide good estimates. Laboratory tests on triplets combined with shear tests on walls are a means to verify correlation of local and global _ behavior. In the case of pre-formed horizontal cracks, the coefficient of friction of the joint can be directly used to estimate the sliding condition of the wall. - Flexural or rocking failure modes of piers are not very sensitive to masonry st which in most cases of practice can therefore be assumed. An accurate experimental evaluation of the behavior in compression of masonry is secondary to shear behavior, unless vertical stresses are particularly high or the materials are severely deteriorated. Testing of Masonry Structures for Seismic Assessment 161 - Vertical stresses play a key role in all of the failure modes cited above. In most cases, load evaluation and simplified analysis are sufficient to ensure an accurate assessment . Single flatjack tests can be used in special cases of complex structures and heterogeneous materials. - For research oriented testing, it should be recognized that recent extensive studies have increased the comprehension of several local failure modes, and of the relationship between the element response and the global response. This general knowledge needs to be extended to additional masonry materials to define the limits separating different modes of failure. Extensive, well-planned parametric experimental studies should therefore be encouraged to allow the refinement of numerical analysis tools. Tests on full scale buildings are needed only in very special cases, to verify the ability to predict global response. ACKNOWLEDGMENTS The contributions of Dr.Alberto Pavese to discussions on vibration testing are strongly appreciated. The authors gratefully acknowledge the support of Gruppo Nazionale per la Difesa dai Terremoti_ in Italy and of the National Science Foundation which facilitated this international collaboration. REFERENCES 1. Abrams , D.P. and A.C. Costley , "Dynamic Response Measurements for URM Building Systems*, US-Italian Workshop on Guidelines for Seismic Evaluation and Rehabilitation of Unreinforced Masonry Buildings, Techn, Rep. NCEER-94-0021, NCEER-SUNY Buffalo, 1994, 2. Abrams, D. Canadian Masonry 1992, pp.235-246. 3. Abrams, D-P., “Strength and behavior of unreinforced masonry elements”, 10th World Conference on Earthquake Engineering, Madrid, 1992, pp. 3475-3480. 4. Anthoine, A., G. Magenes and G. Magonette, “Shear-compression testing and analysis of brick masonry walls”, 10th European Conference on Earthquake Engineering, Vienna, 1994. “Safe Limits for Lateral Capacity of Cracked URM Walls”, Proc. of 6th Symposium, Vol.1, University of Saskatchewan, Saskatoon, Canada, 5. Atkinson, RH., Amadei, B. P., Saeb, S. and Sture, S., “Response of Masonry Bed Joints in Direct Shear," ASCE Jnl. of Structural Div., Vol. 115, No. 9, Sept. 1989. 6. Binda, L., G. Mirabella Roberti, C. Tiraboschi, and S. Abbaneo, “Measuring Masonry Material Properties,” US-Italian Workshop on Guidelines for Seismic Evaluation and Rehabilitation of Unreinforced Masonry Buildings, Techn. Rep. NCEER-94-0021, NCEER-SUNY Buffalo, 1994. 7. Calvi, GM., and GR. Kingsley, “Problems and Certainties in the Experimental Simulation of the Seismic Response of MDOF Structures,” Engineering Structures, (accepted for publ. 1995), 8. Calvi, G-M., and M. Nakashima, “Sulla Sperimentazione Pseudodinamica di Strutture in Mera,” Ingegneria Sismica, Anno VIII - N. 3 - Settembre Dicembre 1991, (in Italian). 9. Calvi, G.M., and G. Magenes, "Experimental Research on Response of URM Buildis Systems," US-Italian Workshop on Guidelines for Seismic Evaluation and Rehabilitation of Unreinforced Masonry Buildings, Technical Report NCEER-94-0021, NCEER- SUNY Buffalo, 1994. 162 G. M.Calvi, G. R. Kingsley, and G. Magenes 10. Calvi, G.M., and A. Pavese, “Application of Dynamic Identification Techniques to a Brick Masonry Building Prototype,” 10th European Conference on Earthquake Engineering, Vienna, 1994. 11. Gambarotta, L., and S. Lagomarsino, “Modeling Unreinforced Brick Masonry Walls,” US-Italian Workshop on Guidelines for Seismic Evaluation and Rehabilitation of Umeintoreed Masonry Buildings, Technical Report NCEER-94-0021, NCEER-SUNY lo, 12, Ganz, H.R, and B. Thirlimann, “Failure Criteria for Masonry”, Sth Canadian Masonry Symposium, Vancouver, Canada, 1989, pp. 65-77. 13. Kingsley, G.R., “Evaluation and Retrofit of Unreinforced Masonry Buildings,” of the Third National Concrete and Masonry Engineering Conference, San Francisco, California, 1995. 14. Magenes, G., Calvi, G.M., “Cyclic behavior of brick masonry walls", 10th World Conference on Earthquake Engineering, Madrid, 1992, pp.3517-3522. 15. Magenes, G, and G-M. Calvi, “Shaking table tests on brick masonry walls”, 10th European Conference on Earthquake Engineering, Vienna, 1994. 16. Magenes, G.,Comportamento Simic di Murature di Mattoni: Resistenza Meccanismi di Rottura di Maschi Murari,” Doctoral Thesis, Dept. of Struct. Mech., Univ. of Pavia, Italy, 1992, (in Italian). 17. Mann, W., and H. Maller, “Failure of shear-stressed masonry - An enlarged theory, tests and application to shear als, Proc. ofthe British Ceramic Society, No. 30, September 1982. 18, Mann, W., G.Konig and A.Otes, “Tests of masonry walls subjected to seismic forces”, 8th Int. Brick /Block Masonry Conference, Dublin, 1988, Vol.2, Elsevier, pp. 764-773, 19. Noland, J.L., Kingsley, G.R,, Atkinson, R.H., "Utilization of Nondestructive Techniques in the Evaluation of Masonry," 8th Int. Brick/Block Masonry Conference, Dublin, 1988. 20. Paulson, T.J., and D.P. Abrams. "Correlation between Static and Dynamic Response of ‘Model Masonry Structures." Earthquake Spectra, Journal of ERI, Vol. 6, No. 3, August 1990. 21. Pavese A. “Caretterizzazione dinamica e modellazione della torre di Pisa”, Seminari su “Restauro delle strutture monumentali”, Napoli, Scuola di specializzazione in Restauro, 25 maggio 1995. 22. Seible, F., G.A. Hegemier, A. Igarashi, G.R. Kingsley "Simulated Seismic-Load Tests on Full-Scale Five-Story Masonry Building,” ASCE Jnl. of Struct. Eng., Vol. 120, No. 3, March, 1994 23. Seible, F., MIJN. Priestley, G.R. Kingsley, A.G. Kurkchubasche, "Seismic Response of a Full-Scale Five-Story Reinforced Masonry Building," ASCE Jnl. of Struct. Eng., Vol. 120, No. 3, March, 1994. 24, Tomazevic, M., “Masonry Structures in Seismic Areas - A State-of-the-art Report”, 9th European Conference on Earthq. Eng., Moscow, 1990, Vol. A, pp.246-302. 25. Tomazevic, M., and M.Lutman “In-Plane Behavior of Reinforced Masonry Walls Subjected to Cyclic Lateral Loads - Part Two: Analysis of Test Results” Report ZRMK/PI-92/08, Institute for Testing and Research in Materials and Structures, Ljubljana, Slovenia, 1993.

You might also like