1 Introduction

The results presented in this manuscript are part of a wider research project aimed at assessing the vulnerability of buildings typical of the Groningen region (located in the Northeast Netherlands). This area, historically not prone to tectonic ground motions, during the last two decades has been subjected to seismic events induced by reservoir depletion due to gas extraction. The most severe event was an earthquake of local magnitude 3.6 that occurred on August 16th, 2012, near Huizinge, above the central part of the Groningen gas field (Bourne et al. 2015). Buildings not specifically designed for seismic actions are thus now exposed to this type of low intensity shaking. Unreinforced masonry (URM) buildings represent the large majority of the local existing building stock (almost 90%).

Currently, very limited data are available on the seismic response of construction typologies similar to those of the Dutch practice. An experimental campaign, starting in 2015, aimed at investigating the performance of structural components, assemblies and systems typical of building typologies present in the Groningen area. The testing campaign included in situ mechanical characterization tests (Tondelli et al. 2015) and laboratory tests comprising: characterisation tests performed on bricks, mortar and small masonry assemblies; in-plane cyclic shear-compression (Graziotti et al. 2016a) and dynamic out-of-plane tests on full-scale masonry piers (Graziotti et al. 2016b). Two full-scale shaking table tests have been conducted in 2015 and 2016 on two different URM building typologies on the testing facilities of the EUCENTRE laboratory. The first one is described in this manuscript, while information on the second one are available in Graziotti et al. (2016c, 2017).

The shaking table test presented in this article was designed to address several open questions related to the seismic behaviour of terraced houses that constitute the majority of the Dutch URM building stock, mainly with residential purposes. They are usually two-storey buildings with openings on only two of their sides, consisting of several structurally independent side-by-side units (4–6). The greatest part of this architectural typology is built with cavity walls, a construction system that became widespread after World War II.

A cavity wall building is a type of construction where an air gap is left between the two leaves of bricks. Sometimes insulating material is inserted in the cavity. The external leaf of a cavity wall is often a brick veneer wall without any load bearing function, while the internal leaf is the loadbearing one, carrying the vertical loads transmitted by the floors and the roof. It is common for the inner leaf to be constructed with different materials than the outer leaf. In several European countries, an example of this solution is to have the inner wall made of calcium silicate brick/blocks, while the outer wall uses clay bricks. Wythes on either side of a cavity wall are typically connected by regularly spaced metal cavity ties, which can vary in material, shape and spacing. Because of their relatively light weight, good thermal insulation properties and effective protection against driving rain, cavity walls are widely used in Central and Northern Europe countries, especially for residential construction. Information on the seismic behaviour of cavity walls is quite limited, and mostly related to earthquakes occurring in Australia (Newcastle, 1989) and New Zealand (Christchurch sequence, 2010–2011) (Ingham and Griffith 2011). Furthermore, a shaking table test on a cavity-wall building specimen with loadbearing concrete blocks was performed by Degée et al. (2008).

The shaking table test presented in this manuscript aims at studying the seismic response of this type of building. In particular, relevant results of the experimental study are: drift limits for different performance levels, damage evolution for increasing shaking levels; storey accelerations amplified along the building height, and displacement/drift profiles for increasing shaking intensity. In particular, specific attention was paid to the dynamic performance of the roof structure and on the possible activation of gable out-of-plane mechanisms. With the aim of reproducing the seismic behaviour of an existing URM terraced house, an incremental dynamic test was carried out up to the near collapse conditions of the specimen. This work presents the geometric and mechanical characteristics of the specimen (Sect. 2), the input motions applied to the shaking table, the testing protocol and the instrumentation (Sect. 3). Section 4 discusses the performance of the building under ground motion excitation, the test results in terms of damage evolution, hysteretic responses, performance of the roof structure and the identification of the local and global limit states (Sect. 4).

2 Specimen characteristics

The building specimen was intended to represent the end-unit of a URM cavity-wall terraced house of the late ’70s. This residential typology is characterised by wide openings on the front and back sides. The transverse walls, that separate units, are double-wythe cavity walls without any openings. Internal transverse walls are composed of a couple of loadbearing walls, carrying most of the vertical loads coming from floors and roof and, therefore, they are capable of resisting significant in-plane lateral forces. Houses built with this common configuration are expected to be characterised by two very different seismic behaviours in the two principal directions. These structures are generally more flexible and vulnerable in the longitudinal direction. For this reason, a unidirectional shaking table test was carried out by applying base excitations along this direction. Figure 1 shows the front view of a classic terraced house and its plan view.

Fig. 1
figure 1

A typical terraced house in Loppersum, Groningen, NL: a illustration of the front façade; b plan view

Full size image

Adjacent units are generally structurally detached, and the discontinuous slabs rest only on the loadbearing walls of the individual units. Each unit is therefore completely self-supported by transverse walls and structurally independent from the other units. The only common walls are the outer veneer walls. For this reason, it was possible to test on the shaking table a representative sub-volume (one end-unit) of an entire terraced house (as shown in the coloured part of Fig. 1a). The first floor is generally made of a reinforced concrete (RC) slab, while the second floor is either a RC or timber diaphragm.

The presence of a timber roof usually dominates over other typical roofing solutions for this building typology.

2.1 Geometry of the specimen

The test-house was a full-scale two-storey building, with a timber roof and RC slabs. The specimen was built directly on the shaking table of the EUCENTRE laboratory (Fig. 2a). It was 5.82 m long, 5.46 m wide (slightly reduced compared to the typical width found in the building stock, due to the shaking table dimensions) and 7.76 m high with a total mass M of 56.4 t. The walls, supported by a steel–concrete composite foundation, consisted of two unreinforced masonry leaves. The inner loadbearing leaf was made of calcium silicate (CS) bricks whereas the external leaf was a clay brick veneer without any loadbearing function. The two pre-cast RC floor slabs (with a mass of 10.3 t and 11 t for the first and second floor, respectively) were supported only by the two transverse (North and South) inner CS walls. The inner CS masonry was continuous along the entire perimeter of the house, while the outer clay brick leaf was not present in the South façade, simply because the specimen was meant to represent the end-unit of a system of row houses. Pictures of the specimen after the end of the construction are shown in Fig. 2, while Fig. 3 depicts the ground and first floor plan views of the specimen.

Fig. 2
figure 2

Views of the full-scale specimen: a North-West (from top); b South-West; c North-West (from bottom)

Full size image
Fig. 3
figure 3

a Plan view of the ground floor, and b the first floor of the specimen. Arrows indicate the assumed positive direction of the shaking table motion. Units of cm

Full size image

An air gap of 80 mm was left between the two leaves, as usually seen in common practice. L-shaped steel ties with a diameter of 3.1 mm and a length of 200 mm were inserted in the 10-mm-thick mortar bed-joints during the laying of the bricks, ensuring the connection between the two masonry leaves (the location of the steel ties is showed by blue dots in Fig. 4). The L hook side was embedded in the inner CS walls for a length of 70 mm, while the “zig-zag” extremity was embedded in the clay masonry for a length of 50 mm (Fig. 5a, d). Two gable walls in the transverse façades (North and South) supported a 43° pitched timber roof. In the ground storey, pre-cast reinforced concrete lintels were placed above the openings on both inner and outer walls. The dimensions of the lintels were 160 × 100 mm for CS walls and 110 × 100 mm for clay walls. Lintels were 1.33 and 2.22 m long for shorter and wider openings, respectively.

Fig. 4
figure 4

Elevation views of the specimen’s inner CS leaf. Units of cm

Full size image
Fig. 5
figure 5

Construction details of the specimen: a positioning of cavity steel ties; b building phase of inner CS leaf; c laying of the second floor slab; d geometry of steel tie; e construction of CS leaves at second floor level; f safety steel frame

Full size image

A rigid steel-frame was installed in the interior of the test-house. This structure served mainly as a safety system, providing support for the two slabs in case of partial or global collapse of the specimen, as well as a rigid reference system for a direct measure of the floors, walls and roof displacements (Fig. 5f). The frame was not in contact with the building, since its columns passed through 450-mm-square holes in the two slabs, large enough to accommodate significant lateral displacements of the specimen (Fig. 3c).

2.2 Building construction details

It is well known among the engineering community that construction details can significantly affect the seismic response of a structure, especially a URM building. Observation of damage caused by major earthquakes, as well as laboratory tests (Tomaževič et al. 1991; Magenes et al. 2014) have shown that the role of the connections between horizontal and vertical structural elements is of primary importance for ensuring a good structural performance. The construction details of the specimen were representative of the Dutch common practice of the ‘60s and ‘70s. Figure 5 presents pictures captured during the construction phase of the specimen.

The connection between the first floor slab and the inner CS longitudinal leaves (East and West walls) was ensured by means of 6-mm-diameter threaded bars, the position of which is indicated by red dots in Fig. 4. Details of this connection are also shown in Fig. 5c, d. This solution was meant to represent a common technique associated with a cast-in situ RC slab where the bars were embedded in the masonry before casting the slab. Since the construction in the laboratory environment required the slabs to be precast, the connectors were pre-inserted in the concrete and then anchored to the masonry in a second phase. As recurrent in the building stock, an air gap (hole sleeve in Fig. 6c) was left in order to guarantee differential thermal expansion of the components. There was no direct connection between the outer clay veneer wall and the first-floor slab.

Fig. 6
figure 6

Details of the connections between the precast RC slabs and the longitudinal CS walls: (a, b) second floor, and (c, d) first floor level

Full size image

The second-floor slab was not directly supported by the longitudinal walls (East and West); the gap between the slab and the inner CS longitudinal walls was filled with mortar after the removal of the temporary supports and the attainment of the deflection of the slab. Similarly, the timber wall plates were not in contact with the longitudinal clay walls (East and West), but they were attached to the edge of the second floor slab by means of 100-cm-spaced 10-mm-diameter threaded bars, while the resulting gap between the beams and the top of the veneer was afterwards filled with mortar (Fig. 6a, b). Such details were adopted in order to reproduce a loading configuration common in the building stock. This solution resulted in almost no vertical load being transmitted to the longitudinal walls under static conditions.

The timber roof was a simple structure consisting of one ridge beam, two wall plates on top of the longitudinal outer leaves and two girders per side between the ridge beam and the wall plates, at approximately every 1.2 m. Tongue and groove planks, with a width equal to 182 mm and a thickness of 18 mm, were nailed on top by means of two 60 × 2 mm nails at each intersection (Fig. 7a). The timber beams of the roof were supported by the transverse inner CS leaves (North and South gables), whereas this connection was further reinforced by the presence of L-shaped steel anchors, as shown in Fig. 7c. The roof was completed by the installation of clay tiles and the total mass of the finished roof was 2.8 t. The in-plane stiffness of the timber diaphragm was essentially provided by the nailed connections between beams and planks, as well as by the effectiveness of the tongue and groove joints.

Fig. 7
figure 7

Details of the roof structure: a geometry of the timber diaphragm; b connection between the timber beams and the South gable; c steel anchors

Full size image

2.3 Mechanical properties of materials

Part of the experimental campaign that was carried out at the laboratory of the University of Pavia, allowed the mechanical properties of the building materials employed for the construction of the specimen to be determined. It comprised strength tests on units and mortar samples, as well as tests on small masonry assemblages, such as compression tests, bond wrench tests and shear tests on triplets. CS and clay units were tested in compression according to EN 772-1 (2000). The dimensions of the CS units were 212 × 102 × 71 mm. The clay bricks were perforated with ten vertical holes, they had a void ratio of 17% and dimensions of 211 × 100 × 50 mm. The flexural and compressive strength of the mortar were determined according to the prescriptions of EN 1015-11 (1999). Six masonry wallettes made of CS and clay bricks were tested in compression in the direction perpendicular to the horizontal bed-joints, according to EN 1052-1 (1998). These tests allowed the determination of the compressive strength of masonry (f m ), as well as the secant elastic modulus of masonry at 33% of the compressive strength (E m ). Bond wrench tests on twenty CS and twenty clay masonry wallettes were performed in order to determine the bond strength of masonry, according to EN 1052-5 (1998). Specimens of both types of masonry were also subjected to the shear test for the determination of the initial shear strength (f v0) and the friction coefficient (μ), according to the guidelines given by EN 1052-3 (1998).

A parallel testing campaign was conducted at the Delft University of Technology (TU Delft) on specimens built using the same materials (Esposito et al. 2016). In particular, tests that allowed the determination of the tensile load capacity of the steel ties connecting the two masonry leaves were performed by Messali et al. (2016). They found that the pull-in and pull-out strengths of the “zigzag” tie extremity (Fig. 5d) embedded in clay masonry specimens, considering an overburden pressure of 0.3 MPa, was higher than the strengths associated with the hook extremity embedded in CS specimens and subjected to the same imposed pressure. The average pull-out and push-in strengths recorded for CS specimens were approximately 1.46 and 1.09 kN, respectively. Moreover, the tensile ultimate capacity of the steel anchors was approximately 4.3 kN. The concrete used to cast the two slabs had an average compressive strength of 29.5 MPa. The masses of the test unit are summarized in Table 1, while Table 2 lists the masonry mechanical properties resulting from material characterization tests.

Table 1 Summary of structural masses, units of t
Full size table
Table 2 Masonry mechanical properties
Full size table

3 Testing protocol and instrumentation

3.1 Input signals

The specimen was subjected to a sequence of incremental dynamic tests. A series of table motions of increasing intensity were applied with the objective of assessing the ultimate capacity and failure modes of the building. Since the shaking table is uniaxial, the seismic input was applied parallel to the longitudinal direction (North–South) of the tested building, exciting out-of-plane the loadbearing transverse walls (North and South façades). The selected input motions aimed at being representative of expected ground motion in the region of Groningen. A detailed study on the seismic hazard characteristics identified two main scenarios with return periods of 50 and 500 years (see Appendix B2 of Graziotti et al. 2015). Two records EQ1 and EQ2 with 5–75% significant duration of 0.375 s and 1.72 s and a PGA of 0.095 g and 0.159 g, respectively, were finally selected to be representative of the two scenarios. Their smooth response spectra were considered ideal for a higher control of both the shaking table and the response of the structure. Figure 8 shows the theoretical acceleration time-histories of the experimental inputs and their acceleration response spectra.

Fig. 8
figure 8

EQ1 and EQ2 input signals: a acceleration time histories, and b acceleration response spectra

Full size image

3.2 Testing protocol

The sequence of incremental dynamic tests was performed by gradually increasing the intensity of the two ground motions with EQ1 being applied first, followed by EQ2. Table 3 presents the applied testing sequence specifying the input typology, the intensity and the comparison between nominal and recorded PGAs and 5% elastic spectral accelerations, Sa(T 1,1) at the fundamental period T 1,1 = 0.17 s of the undamaged structure (calculated by means of a dynamic identification test, see further details in Sect. 4.4). Further intensity measures (IMs) listed in Table 3 are the peak ground velocity PGV, the 5%-damping elastic spectral displacement obtained from the recorded base acceleration, and the modified Housner intensity (mHI). The mHI has shown very good correlations with the nonlinear displacement demand induced in short period URM structures (Graziotti et al. 2016d). It is defined as the integral of the pseudo-velocity elastic response spectrum between a structural period of 0.1 s and 0.5 s (which correspond to the range of periods of interest of the tested specimen):

Table 3 Summary of the testing sequence
Full size table
$$mHI\left( {5\% } \right) = \int\limits_{0.1}^{0.5} {\text{PSV}} \left( {5\% ,T} \right){\text{d}}T$$
(1)

The shaded sections in Table 3 refer to the tests with increasing intensity. It is worth noting that such tests have been often proceeded by tests of the same typology, but with reduced intensity for the purpose of shake table calibration (e.g. tests #6, #11, #12, #13 and #18). The intensity level of these calibration tests (labelled with a C in the test name, e.g. EQ1-50%C) has been chosen in order to prevent further damage or deterioration in the specimen. In general, a good agreement between nominal and recorded quantities have been observed, with a slight overshoot of the recorded spectral acceleration corresponding to the initial fundamental period T 1,1. Each test with increasing intensity was alternated by random noise tests (RNDM), which, by means of a dynamic identification procedure, allowed the changes in the dynamic properties of the structure to be detected as the damage level increased. In particular, the following sections report the evolution of the fundamental period denoted as T 1,i (where i is the test identification number as reported in Table 3). The incremental testing sequence has been stopped after the attainment of a near collapse condition in order to prevent a global collapse of the specimen that could have caused damage to the laboratory facilities.

3.3 Instrumentation

In order to detect and monitor the structural response under different levels of input motion, several instruments were installed on the building. The location and typology of the instrumentation was determined based on the identification of the critical zones and on the physical quantity to be recorded. The instrumentation consisted of 33 accelerometers and 30 displacement transducers. Figure 9a shows the locations of the three types of employed accelerometers (uni-, bi- and tri-axial). The accelerometers were installed on both inner and outer leaves, as well as on the two floors and the ridge beam of the roof. Figure 9b shows, instead, the displacement transducers installed on the specimen: 10 wire and 20 linear variable displacement transducers (LVDTs). The displacements measured between the specimen and the rigid frame were considered equivalent to the relative displacements with respect to the shaking table surface. In particular, wire potentiometers were installed in order to record the out-of-plane response of the North and South façades at the mid-height of the first and second storeys and the gable. The LVDTs were, instead, utilized to monitor directly the longitudinal and transverse displacement of the first and second slabs. The displacements of some points of the external façades and internal walls were monitored by a 3D optical acquisition system (see Appendix 2). These data allow to compute the differential displacement between inner and outer leaves.

Fig. 9
figure 9

Locations of the instrumentation: a accelerometers, and b displacement transducers (letters indicate the component at which the transducers is attached to: S slab; F frame; IL inner leaf; OL outer leaf; FB foundation beam; T shaking table; R roof ridge beam; L laboratory floor)

Full size image

4 Test results

The following sections report the main results of the shaking table tests. In particular, some issues related to the global seismic response of the tested building are discussed, in terms of the observed crack patterns, the deformed shapes and the hysteretic behaviour. To summarise briefly: the building sustained shaking of PGA = 0.14 g (EQ1 150%) with no visible damage and was in a near-collapse state after testing at PGA = 0.31 g (EQ2 200%), when the test sequence was stopped. Videos of the applied testing sequence are available online (Eucentre 2015a).

4.1 Shaking table performance

The comparison between the theoretical response spectra and those obtained from the accelerations recorded on the specimen’s foundation, shows a general good match. A very slight overshooting of low period spectral ordinates was noticed in all the tests. A 15% undershooting of spectral acceleration in the high period range was observed only in the test EQ2-200%. In the same test a considerable amplification peak occurred at a period of T = 0.18 s. The sudden change of the specimen dynamic characteristics (the fundamental period was doubled), due to its heavy damage and its interaction with the table, did not allow a perfect match of the target spectrum. The comparison of the acceleration response spectra for the tests of EQ2-100% (PGA = 0.17 g) and EQ2-200% (PGA = 0.307 g) is shown in Fig. 10.

Fig. 10
figure 10

Comparison of the acceleration response spectrum of the recorded base acceleration against the target one for testing under a EQ2-100% (PGA = 0.17 g), and b EQ2-200% (PGA = 0.307 g)

Full size image

4.2 Damage evolution

At the end of each stage of the shaking table testing sequence, detailed surveys were carried out for the report of every possible evidence of damage having affected the structure (Figs. 11, 12). During the testing under the first scenario seismic excitations (EQ1 scaled from 25% PGA = 0.024 g to 150% PGA = 0.137 g), the building did not experience any noticeable damage. The specimen suffered only slight damage that became visible just after testing under EQ2-100% test (PGA = 0.17 g). The formation of a few cracks was observed at the base of the first storey inner-leaf corner piers, associated mainly with their flexural behaviour. The observed damage did not change significantly after testing at EQ2-125% (PGA = 0.194 g).

Fig. 11
figure 11

Crack pattern evolution of the inner CS walls

Full size image
Fig. 12
figure 12

Crack pattern evolution of the outer clay walls

Full size image

The first significant cracks observed in the CS masonry of the second storey were recorded after the test EQ2-150% (PGA = 0.243 g). They were mainly horizontal cracks observed just below the interface between masonry piers and the second floor level slab, as mapped in Fig. 11. A horizontal crack developed along the base of the squat pier of the second storey, on the West side, indicative of the pier’s bending-rocking response. This crack was further extended with a stair-stepped diagonal pattern to the centre of the adjacent spandrel. Until this intensity level no damage in the two transverse walls was detected.

The building experienced a substantial level of damage (compared to that observed under lower intensity shaking) after the test EQ2-200% (PGA = 0.307 g). At this shaking level a global response of the structure was triggered, as evidenced by the formation of new cracks or the elongation of pre-existing ones, identified on every one of the piers, as shown in Fig. 11. A detailed survey of the building was conducted and revealed extensive damage in the spandrels of the calcium silicate masonry. In particular, the formation of wide diagonal cracks (starting from the corners of the openings), with sliding of the mortar joints and de-cohesion of blocks were observed (Fig. 13e). In addition, the horizontal cracks located at the top of the second storey piers were extended, reaching a maximum residual sliding of 15 mm.

Fig. 13
figure 13

3D view of the observed crack pattern at the instant of attainment of the peak second floor displacement: a pounding of the ridge beam on the North CS gable; b flexural cracks on the top of the second storey; c flexural cracks on the bottom of a first storey pier; d sliding at the interface of the second storey pier and the slab; e de-cohesion of masonry blocks and diagonal shear cracks through the joints; f stair-stepped cracks in the transverse walls; g shear failure of the veneer’s short spandrel with the formation of X-shaped crack pattern, h flexural cracks in the veneer’s long spandrel; i sliding at the interface of the veneer and the timber wall plate

Full size image

As far as the damage reported in the transversal walls is concerned, the formation of 45° stair-stepped diagonal cracks (no greater than 1.2 mm) was clearly observed. This could be associated with the activation of an out-of-plane two-way bending mechanism. Focusing on the gables, horizontal cracks along their base were apparent (one or two layers above the second floor level), indicative of an out-of-plane overturning mechanism activated at the gable level. Other cracks were also identified at the locations where the timber beams of the roof were supported on the gable walls. Cracks around these beams were due to interaction of the beams with the supporting masonry gable walls (Figs. 11, 13a).

Regarding the damage noticed in the veneer walls, perceptible cracks developed only during the last test, EQ2-200% (PGA = 0.307 g). In particular, the long spandrel of the eastern façade developed a flexural mechanism with vertical cracks at both ends, originating from the concrete lintels (Figs. 12, 13h), whereas the shorter spandrel presented failure in shear, forming the characteristic X-shape crack pattern (Figs. 12, 13g). On the western side, large stair-stepped shear cracks were observed, such as those crossing the entire short spandrel with an angle of 45°. To a great extent, most of the deformations were absorbed by sliding of the concrete lintels with respect to the masonry supports, as well as sliding at the interface of the roof wall plates and the second storey masonry piers (Fig. 13i). In the northern veneer, the only cracks observed were located at the second floor level. As they extended along the entire length, they were associated with the tendency of the gable wall to develop an out-of-plane overturning mechanism (Fig. 12).

Figure 14 reports the evolution of the maximum residual crack width measured after the end of every test. The same quantity is also plotted versus the peak and residual correspondent inter-storey drift ratio (θ and θ res , respectively). A higher residual crack width was measured in the ground storey (i.e. 1st storey). This was due to two main factors: the higher drift demand (see Fig. 20) at lower levels and the concentration of the 2nd storey deformation in the interface between the CS wall and the top floor slab. Regarding the first floor, where the slab displacement was completely accommodated by the deformations of the piers, a good correlation between θ and crack width was observed. The relation between the crack width and the θ res was found to be almost linear for both storeys.

Fig. 14
figure 14

Evolution of the maximum residual width of the observed cracks as a function of the: a testing sequence, b peak IDR, and c residual IDR

Full size image

4.3 Deformed shapes

Deformed shapes in elevation have been generated by plotting the horizontal displacements recorded by the traditional potentiometers mounted on the floors and the wire potentiometers located at the level of the storeys’ mid-height and ridge of the roof. Figure 15 represents the out-of-plane deflected shape of a longitudinal cross section of the specimen at the instant of peak second floor displacement for EQ2-100% and EQ2-200%, respectively.

Fig. 15
figure 15

Deflected shapes of the specimen during the tests a EQ2-100% (PGA = 0.17 g) and b EQ2-200% (PGA = 0.307 g). Displacement units of mm

Full size image

The deformed shapes changed significantly according to the ground motion intensity level and the state of deterioration of the specimen. In both cases, the higher drifts were observed at the roof level. This sub-structure was significantly more flexible. The initial response (similar from EQ1-25% to EQ2-150%, herein represented by the Fig. 15a) was instead characterised by a higher drift demand in the first storey with the second floor remaining almost rigid and experiencing a very low drift demand. During the last test (EQ2-200%), the specimen exhibited similar inter-storey drifts in both storeys, resulting in an almost linear trend (similar to the one corresponding to the first mode of vibration), associated with rocking response of the slender piers over the height of the building, as illustrated in Fig. 15b.

Furthermore, during the EQ2-200% test, after the failure of the interface between the top of the clay wall and the timber wall plates (Fig. 13i), a clear relative displacement was observed between the CS wall and the clay veneer, showing that the presence of cavity ties was not sufficient to ensure their collaboration. Most of the ties were permanently bent at the end of the tests. The inner loadbearing CS structure displaced significantly, while the southern portion of the East and West veneer walls was not involved in such an oscillation. A video of the test shows clearly this phenomenon (Eucentre 2015b).

4.4 Hysteretic responses

The evolution of the specimen’s hysteretic response is shown in Fig. 16, in terms of base shear, V, versus global drift of the first two storeys, \(\tilde{\theta }\), through all the tests. The global drift is defined as the relative displacement of the second floor slab divided by its distance from the base, given by:

Fig. 16
figure 16

Evolution of the global hysteretic response in terms of base shear versus global drift ratio (left); Backbone curve in terms of base shear coefficient (right)

Full size image
$$\tilde{\theta } = \frac{{\Delta_{2} }}{{h_{1} + h_{2} }} = \frac{{\Delta_{2} }}{{5440\left[ {\text{mm}} \right]}}$$
(2)

The time histories of the base shear have been computed as the sum of the products of each acceleration recording times the tributary mass of the corresponding accelerometer. Masses are assumed to be lumped at the accelerometer locations. The mass of the masonry body from the foundation level to the mid-height of the ground storey (at 1.38 m from the base) was assigned to the ground floor (and hence multiplied by the base acceleration time history).

The base shear coefficient BSC is defined as:

$$BSC = \frac{V}{M \cdot g}$$
(3)

where \(M \cdot g\) is the total weight of the specimen.

In each plot of Fig. 16, the hysteretic response of preceding tests is reported in grey. The white dots represent the positive and negative peak force responses with the corresponding displacements. The proportion between the two axes of all the plots is the same. In this way, the progressive specimen stiffness degradation and the consequent fundamental period elongation are appreciable.

The EQ2 input induced a more pronounced asymmetry in the specimen response with respect to the EQ1 earthquake. The displacement demand in the negative direction (towards South), indeed, was rather higher than the one in the positive direction. The first significant nonlinearity in the hysteretic response is observed during testing under EQ2-150% (PGA = 0.243 g), associated with the occurrence of spread flexural cracks in the inner CS walls. During the test at EQ2-200% (PGA of 0.307 g), a large nonlinear behaviour was observed associated with extended damage to the specimen, highlighted by the dramatic enlargement of the hysteresis loops, and the consequent significant increase of the specimen’s fundamental period of vibration.

An ultimate global drift ratio \(\tilde{\theta }\) = 0.7% was reached, while a shear deformation of the roof diaphragm γ R  = 1.5% was observed for the significantly more flexible roof structure. The maximum base shear V max attained was approximately 139 kN, corresponding to a base shear coefficient BSC max  = 0.25. The dynamic force–displacement backbone curve can be obtained by connecting the peak points of the experimental curves. In other words, it is defined as the plot of the maximum resisted base shear, V max , and the corresponding global drift, \(\tilde{\theta }\), for each stage of testing. The last point of both the positive and negative branch was obtained as the pair of the maximum drift attained and the corresponding base shear. A force “plateau” in the specimen capacity was reached in both directions. The attainment of the higher base shear occurred for sway towards the negative direction (towards the single-leaf side, South). In particular, the base shear attained for southward motion (V − MAX  = 139.5 kN) was 37% higher than the force reported for motion towards the double-leaf side of the structure (V + MAX  = 101.6 kN). The asymmetry in the envelope response curve could be attributed to the northward “spike” of the applied accelerogram EQ2 and to the asymmetry of the structure.

4.5 Response of the roof structure

The gable-roof system response was of particular interest for further investigation. The behaviour of the roof was acknowledged as one of the main factors that has driven the response of the substructure during the evolution of the dynamic tests, while the testing procedure ended because of the very large deflections of the gables. The detailed response of the roof in the course of the shaking table testing is illustrated in Fig. 17, in terms of acceleration versus relative displacement curves. The first quantity regards the acceleration, a R , recorded by the accelerometers located at the ridge beam level, whereas the second refers to the relative displacement of the ridge, δ R , with respect to the second floor level. The slope of the dashed line is representative of the effective stiffness, K R , the gable-roof system, while its ever-decreasing trend indicates that the roof diaphragm undergoes a significant stiffness degradation. Trends for the progressive stiffness degradation, defined as the ratio between the current degraded stiffness, K Ri , and the initial stiffness, K R1, can be derived and plotted as a function of the maximum in-plane shear deformation, γ Rmax that the roof diaphragm undergoes during each test, as shown in Fig. 18. The relative roof displacement, δ R , of Figs. 17 and 18 is calculated from the relative ridge displacement (with respect to the second floor) by removing the residual displacements. Similarly, for generating the plots of Fig. 18, the roof shear deformations, \(\tilde{\gamma }_{R}\), was computed after subtracting the residual shear deformations, \(\gamma_{R,res}\), since they resulted in curves biased towards the right, and should not be confused with the roof shear deformations, \(\gamma_{R}\), reported in Fig. 20. The roof shear deformation is computed as the relative ridge displacement divided by the inclined length of the roof pitch, L R  = 3.61 m.

Fig. 17
figure 17

Evolution of the roof hysteretic response in terms of acceleration versus relative displacement at the ridge beam level (left); Backbone curve (right)

Full size image
Fig. 18
figure 18

a Roof stiffness degradation as a function of the maximum attained in-plane shear deformation; b Envelope of the force–displacement responses of the roof

Full size image

The inertia force of the entire roof system, F R , system could be estimated by attributing a representative portion of the total mass of the gable-roof system to the ridge beam level. The lumped mass assumed at the top of the roof was equal to one-third of the self-weight of the gable-walls plus half of the weight of the roof, estimated around 2.6 t. Figure 17 reports the force–displacement response of the roof structure, as well as the resulting backbone curve of the system, defined by the peak points of the experimental curves (plot of the maximum attained force, F R , against the corresponding relative ridge displacement, δ R , occurring at the same instant, for each stage of testing).

The envelope of the force–displacement responses displays no indication of strength degradation, which confirms diaphragm flexibility and the absence of observable structural failures in the roof. The plots on Fig. 18 show that the roof exhibited an almost linear elastic behaviour up to a displacement of approximately 4 mm, with a stiffness, K a , equal to approximately 3.2 kN/mm. Beyond this value the roof entered into a nonlinear phase characterized by a higher dissipation of energy and a reduced stiffness of K b  ≈ 0.12 kN/mm. The wide hysteresis loops demonstrate that the diaphragm is capable of dissipating considerable amounts of energy when subjected to lateral loading.

In order to determine performance parameters of the roof that could be further exploited to investigate its seismic response, it was necessary to appropriately characterize the force–displacement data using a consistent and rational methodology. In the absence of a universally accepted method, the performance of the system could be captured using a bilinear idealization of the backbone response curve. As reported by Peralta et al. (2004) and Wilson et al. (2014), the response can be approximated by a bilinear representation by applying the principle of hysteretic energy conservation, imposing at the same time the following constraints: the curve should pass through zero load and displacement; the ultimate displacement, δ u , could be taken as the maximum experimental displacement of the secondary linear branch; and the secondary stiffness should be computed as the global gradient of the approximate linear portion of the experimental envelope curve (here, observed for displacement amplitudes above 10 mm). Figure 18 illustrates the key performance parameters for the roof of the tested house, consisting in the initial stiffness, K a , the secondary stiffness, K b , the effective yield displacement, δ Ry , and the corresponding yield load, F Ry , for both positive and negative displacements.

Because of the composite nature of the roof structure it was difficult to fully single out the experimental response of the timber diaphragm. The experimental data acquired from the tests could only be used to infer conclusions for the roof system response when examined as an ensemble, composed of the gable walls and the timber diaphragm (constructed by boards fastened perpendicular to timber joists).

4.6 Identification of the specimen damage limit states

In this section, the identification of global quantitative thresholds that adequately describe the overall structural damage state of the building, is attempted. The roof sub-structure damage evolution is treated in a separate Sect. 4.2 and not included in this one. The seismic performance of existing buildings is usually evaluated through four damage limit states as proposed, for example, by Calvi (1999): DL1: no damage, DL2: minor structural damage and moderate non-structural damage (still usable building), DL3: significant structural damage and extensive non-structural damage, DL4: severe damage leading to demolition. Due to the high non-linearity characterising URM buildings leading to difficulties in distinguish between DL1 and DL2, Calvi (1999) suggested to condense them into a unified damage state. Recently, Lagomarsino and Cattari (2015) proposed a multiscale approach for the definition of the damage thresholds related to each of the four performance levels; at a global scale, the damage levels are identified on the pushover curve according to the fraction of resistant base shear attained, at a sub-system scale such thresholds are defined in terms of inter-storey drift; at the structural element scale, the seismic performance is evaluated according to the percentage of piers and spandrels exceeding a pre-defined deformation limit condition.

This section compares such damage limits with the actual damage observed through the testing stages of the present experimental test. Difficulties arise in the definition of clear damage states mainly due to two factors: the progressive accumulation of damage and the limited number of tests.

Figure 19 shows the global response of the building in terms of global drift \(\tilde{\theta }\) (Eq. 2) and base shear coefficient BSC (Eq. 3). The global drift \(\tilde{\theta }\) (as the displacement of the second floor) is not the best engineering demand parameter, EDP, but it could be useful to give a general idea of the specimen performance in terms of deformation achieved. The white dots represent the points of maximum resisted base shear, V max , and the corresponding global drift, \(\tilde{\theta }\), for each stage of testing (notice that this point is lower than the maximum global drift achieved in the correspondent test, \(\tilde{\theta }_{max}\)); the successive corner points of the black solid line are local peaks achieved in the last test EQ2-200%, while the black dot represents \(\tilde{\theta }_{max}\) recorded during the test EQ2-200%. The different limit states, defining the thresholds between damage states, are defined as follows. Figure 19 plots them associated with views of the West side inner CS wall crack patterns.

Fig. 19
figure 19

Definition of damage limits on the experimental backbone curve, illustration of the corresponding damage extent on the West side inner wall

Full size image

DL1 is defined as the maximum achieved level of displacement with no visible damage. The inspection after the execution of test EQ1-150% (PGA = 0.137 g) did not report any cracks. The structure could be considered as fully operational. The maximum recorded global drift was \(\tilde{\theta }_{max}\) = 0.047%, while the maximum inter-storey drift, recorded at first floor level, was \(\theta_{1}\) = 0.07%.

DL2 is defined as the maximum achieved level of displacement with minor/slight structural damage. The observed damage could be easily repaired (maximum crack residual not higher than 1 mm, Baggio et al. 2007) for a possible immediate occupancy. In particular, this damage limit was achieved during the test EQ2-100% (PGA = 0.17 g), when the cracks appeared at the bottom of the S-W pier of the first storey (\(\tilde{\theta }_{max}\) = 0.073%, \(\theta_{1}\) = 0.12%). The determination of DL2 on a global scale is very sensitive to engineering judgment. In this particular case, it was associated with EQ2-100% test because during the following run (i.e. EQ2-125%) the residual crack width reached 2 mm, even though this damage was still limited to the S-W corner of the building.

DL3 is defined as the maximum achieved level of displacement with moderate structural damage (but still repairable). This state was associated with damage observed in all the piers contributing to the longitudinal resistance of the specimen after test EQ2-150% (PGA = 0.243 g). A posteriori, it was interesting to notice that this run was the first one to demand the full exploitation of the specimen lateral strength. DL3 could be considered as a life safety limit state. The maximum residual width of the crack was 4 mm. The behaviour was characterized by a peak global drift \(\tilde{\theta }\) = 0.23%, and a first storey drift \(\theta_{1}\) = 0.34%. Beyond this limit, the house could not be repaired economically.

DL4 is defined as the maximum displacement reached by the specimen before the decision to stop the test due to a near collapse condition. The definition of a clear near collapse limit state is, hence, not trivial as for the case of the other limit states. The limit could be considered as a collapse-prevention threshold. Moreover, due to the significant reduction of stiffness, small variations in the input intensities could lead to significantly different peak displacements. The observed heavy structural damage (in piers and spandrels of inner and outer leaves) suggests that repairing a house that has reached this limit state may not be convenient. During the last test EQ2-200% (PGA = 0.307 g), a peak global drift \(\tilde{\theta }_{max}\) = 0.729% and first storey drift \(\theta_{1}\) = 0.88% were achieved. After this test, the maximum residual crack width was 5 mm.

Table 4 compares the experimental and analytical damage limits as proposed by Calvi (1999) and by Lagomarsino and Cattari (2015). In particular, a comparison in terms of sub-system scale variable (i.e. interstorey drift \(\theta_{1}\)) and global scale variable (i.e. V/V max ratio) is proposed for each damage state.

Table 4 Comparison of the experimental and theoretical damage limits proposed by Calvi (1999) and Lagomarsino and Cattari (2015)
Full size table

In general, there is a very good agreement between the damage thresholds defined based on the experimental observations and those proposed in the considered analytical approaches. Only the collapse-prevention limit, DL4, is underestimated by both criteria. This may be due to the fact that the analytical approaches take into account a possible shear failure, e.g. Calvi (1999) refers to Magenes and Calvi (1997), while the response of the building under examination is dominated by flexural/rocking behaviour, typically associated with a higher displacement capacity (e.g. Graziotti et al. 2016a). As the experimental limit states are associated with a dynamic building response governed by bending/rocking mechanisms the softening which can be observed in the force–displacement envelope (Fig. 19) is much less pronounced than the one assumed in the analytical approach by Lagomarsino and Cattari (2015).

The backbone curve has been further idealised by means of a bilinear approximation based on the equal energy criterion as prescribed by NTC08 (2008). The ultimate strength in terms of base shear coefficient was BSC b  = 0.241 and the ultimate global drift \(\tilde{\theta }_{u,b}\) = 0.73%, this deformation coincides with the peak global drift \(\tilde{\theta }_{max}\) achieved in the last test EQ2-200%. The bi-linearization procedure proposed by NTC08 (2008) has been developed in order to simply characterize the capacity curve of a building. In this case, the bi-linear idealisation has not been truncated in correspondence to a drop V/V max  = 0.8 (as prescribed by NTC08 in case of pushover analysis) but it was extended to the actual maximum displacement achieved (without collapse) during the shaking table test. The “yielding” point corresponds to global drift of \(\tilde{\theta }_{y,b}\) = 0.079%. It is worth noticing that the quantitative definition of DL2 almost coincides with the end of the linear elastic range of the bilinear curve whereas DL3 almost corresponds to the maximum lateral force.

4.7 Derivation of engineering demand parameters according to the specimen performance

EDPs, such as peak inter-storey drift ratio (IDR), residual inter-storey drift ratio (RIDR) or peak floor acceleration (PFA) are important synthetic measures of the seismic behaviour of a building under a given earthquake. The selection of proper EDPs is a crucial point in order to characterize the performance of a structure. The analysis of data derived from shaking table tests, as those herein presented, is a good chance to directly correlate the physical observed damage with EDPs. Figure 20 shows a series of parameters related to the building performance. It is worth remarking that the specimen performance has been influenced by the progressive accumulation of damage during the entire testing sequence, since the test was incremental. This should be taken into account when a correlation between EDPs and intensity measures (IM) is formulated.

Fig. 20
figure 20

Summary of the performance of the building specimen

Full size image

Figure 20 reports the building performance in terms of peak displacements (Δ 1 , Δ 2 and Δ R ), IDR (θ 1 , θ 2 and roof diaphragm shear deformation γ R ) usually strictly connected to the in-plane damage occurring in structural elements like piers and spandrels, and RIDR very often associated with a general damage and damage accumulation. The response in terms of PFA/PGA is also shown. This EDP could be correlated with the OOP performance of masonry (or more in general secondary) components or the damage occurring to acceleration sensitive non-structural components.

The evolution of the building fundamental period of vibration during all test phases is also shown. The fundamental period evolution is calculated by means of dynamic modal identifications performed before each strong motion by means of low amplitude RDNM excitations (see Table 3). The peaks in the power spectral density can generally be assumed to represent either peaks in the excitation spectrum or normal modes of the structure (Pick Picking method). The normal modes were determined from the identification of the peaks in the power spectral density, the analysis of the phase angles and the computation of the ordinary coherence function. The Peak Picking method used in this study was the one extended by Brincker et al. (2000, 2001) that introduced the so-called Frequency Domain Decomposition method. The basis of the method is the Singular Value Decomposition of the response spectral density matrix into a matrix of singular values and an orthogonal complex matrix containing the mode shape vectors of each spectral peak. Once the frequencies of vibration were defined, the mode shape components were computed from the amplitude of the cross-spectra normalized to the maximum component, with the direction of motion derived from the phase angles from the cross spectra between channels. The first mode of vibration of the undamaged building has been identified at a fundamental period T 1,1 = 0.17 s, for the inner walls system only. The first period of the external veneer walls is assumed to be presumably close to the same value. Figure 15, representing the deformed shapes under earthquake-type excitation, also well depicts the calculated deformed shape of a first mode type of behaviour, with the longitudinal walls responding in-plane and the gable walls overturning out-of-plane, parallel to the direction of the shaking table motion. More details regarding the modal identification outcomes are available on Graziotti et al. (2015).

The IDR associated with the first floor, θ 1, was systematically higher than the one of the second floor, θ 2, up to the test EQ2-200% (PGA = 0.307 g). Attaining the DL4 condition, they reached a similar maximum value of approximately 0.88%. The first damage limit state (DL2), where damage has been observed in the first storey piers, is associated with a first floor drift of θ 1 = 0.12%, while the severe damage limit state (DL3) with the exploitation of the specimen full capacity is associated with inter-storey drift ratios θ 1 = 0.34% and θ 2 = 0.18% for the first and second storey, respectively. From the same plot, looking at the evolution of γ R , it is also noticeable that the roof substructure seems to experience non-linearity starting from early stages of the test (see also Fig. 18a). Residual inter-storey drifts (RIDR) have been noticed after the end of the testing phase EQ2-100%, with the attainment of DL2.

The plot of the floor acceleration amplification, AMP i , shows a progressive decrease, starting from values around 1.5 in the first tests to values close to 1 in the last tests. In accordance to the very limited θ 2 an almost negligible amplification has been recorded between the first and the second floor. In the EQ2-200% test the observed two-way out-of-plane cracks in all the North and South walls developed after the specimen has been subjected to floor acceleration PFA >0.3 g. This EDP could be considered as a first crack damage state for the OOP walls (further research are ongoing on this topic). The roof structure amplified the ground acceleration by a factor of 5 in the first runs down to a factor of 2 in the last test.

The results of the present experimental tests allow also the EDPs and the observed damage to be related with a seismic intensity measure (i.e. the PGA) for the input motions selected according to the hazard study. This could represent a reference for a sanity check of structural analyses on similar buildings.

5 Conclusions

The presented work was part of an extensive experimental campaign aimed at assessing the seismic vulnerability of Dutch URM buildings. It presents results of a unidirectional shaking table test performed on a full-scale specimen representative of a Dutch two-storey URM building with cavity walls and timber roof. The building specimen was intended to represent the end-unit of a terraced house of the late ’70s, without any specific seismic detailing. The loadbearing walls were built with a 10-cm-thick calcium silicate URM, while three out of the four façades were completed by a clay veneer connected to the calcium silicate walls by means of steel ties. The materials were characterized by mechanical characteristics compatible with the ones found in the building stock. The specimen was subjected to incremental input motions representative of two different induced seismicity scenarios characterized by smooth response spectra and a short significant duration. The processing of the recorded signals, both in terms of accelerations and displacements sustained by the tested structure, allowed the evaluation of the seismic resistance and displacement demand at each stage of testing. All the recorded signals (accelerations, displacements, videos) can be requested online (http://www.eucentre.it/nam-project).

The loadbearing structure exhibited a box-type global response thanks to the presence of the rigid concrete slabs, which engaged the longitudinal walls and prevented the occurrence of local out-of-plane failure mechanisms in the transverse walls of the 1st and 2nd stories, no torsional effect was recorded. As a consequence, the full in-plane capacity of the longitudinal walls was exploited. Four damage states were identified and compared with some of the theoretical proposals available in literature, with good agreement. In summary, the building withstood the input motion with a PGA of 0.17 g with little damage (maximum first inter-storey drift θ 1 = 0.12%) and was in the near-collapse state at a PGA of 0.31 g (θ 1 = 0.88%). No significant shear damage occurred in the masonry piers, which were in general slender, and their response was mainly governed by rocking, whereas sliding occurred at the top of masonry walls parallel to the table motion. A substantial compatibility of displacements was observed between the inner and outer walls up to the near collapse state. During the last run (PGA = 0.31 g), the two substructures moved almost independently and, as the stiffness contribution of the external clay masonry was reduced, the displacement demand of the internal structure increased. The fundamental period of the structure after the tests was almost 3.5 larger than the initial undamaged one. Furthermore, some diagonal stepped cracks were observed in the transversal load bearing walls due to the out-of-plane excitation.

The structure was characterized by a very flexible roof. A study of its dynamic behaviour is proposed in the manuscript. The timber diaphragm was subjected to a maximum shear deformation of almost 1.5%. Values for the amplifications of accelerations are also given herein.

Despite the high flexibility and the consequent vulnerability of roof system, the shaking table tests were able to fully exploit all the strength of the loadbearing structure. The maximum base shear coefficient was almost 0.25. The hysteretic plots, the large amount of experimental data derived from the dynamic tests (available upon request), the series of tests on smaller structural assemblies and characterization tests on materials constitute a useful basis for the development and calibration of numerical models that can reproduce the response of structures with different configurations. These calibrated models, thanks to the identification of different damage limit states herein presented, will be a reference for the vulnerability studies of the Groningen building stock.