Multisensor surveys of tall historical buildings in high seismic hazard areas before and during a seismic sequence

G Model

ARTICLE IN PRESS

CULHER-2874; No. of Pages 12

Journal of Cultural Heritage xxx (2014) xxx–xxx

Available online at

ScienceDirect www.sciencedirect.com

Original article

Multisensor surveys of tall historical buildings in high seismic hazard areas before and during a seismic sequence Giordano Teza a,∗ , Arianna Pesci b , Sebastiano Trevisani c a b c

Department of Geosciences, University of Padua, Via Gradenigo, 6, 35131 Padova, Italy Istituto Nazionale di Geofisica e Vulcanologia, Sezione di Bologna, Via Donato Creti 12, 40128 Bologna, Italy Department of Architecture, Construction and Conservation, University IUAV of Venice, Dorsoduro 2206, Terese, 30135 Venezia, Italy

a r t i c l e

i n f o

Article history: Received 3 January 2014 Accepted 17 June 2014 Available online xxx Keywords: Terrestrial laser scanning Thermal imaging Operational modal analysis Stratigraphy Soil-structure interaction Masonry

a b s t r a c t A seismic sequence that included a moment magnitude MW = 5.9 earthquake struck three regions of Northern Italy (Emilia Romagna, Veneto and Lombardy) in May–June 2012. The sequence caused significant damage to several historical buildings and in some cases caused complete structural collapse. Cracks appeared in the belfry and cusp of the 69 m high, ∼3◦ leaning bell tower of Ficarolo (Rovigo). A project aimed at studying the geometry of the tower, possible local seismic amplification and soil-structure interaction began in early 2013 before the earthquake. The data were provided by terrestrial laser scanning, low-cost operational modal analysis and geophysical measurements. The repetition of the surveys during and after the seismic sequence, which was augmented by thermal imaging measurements, allowed an evaluation of the changes caused by the earthquake. In addition to an evaluation of the damage, the data allowed the development of a method based on fast and relatively low-cost measurements that provide useful information for cultural heritage management purposes. The results highlighted that the surveys can be carried out during a seismic emergency and that preventive measures can be carried out under reasonable time and budget constraints in high seismic hazard areas. © 2014 Published by Elsevier Masson SAS.

1. Research aim The 69-m high bell tower of Ficarolo (Rovigo province, Northern Italy) leans at a significant angle (∼3◦ in the shaft). To characterize the tower’s geometry in detail and evaluate possible soil-structure interaction (SSI), terrestrial laser scanning (TLS), simplified operational modal analysis (OMA) and geophysical surveys were performed at the beginning of 2012. After the Emilia Romagna seismic sequence (May–June 2012), new surveys were performed, and infrared thermography (IRT) measurements were also collected. A comparison between the data acquired before and after the seismic sequence led to two main results: • an estimate of earthquake-induced damage to the bell tower; • an evaluation of the amount of information that can be obtained by fast and relatively low-cost measurements that can be repeated during earthquake emergencies. The second point is particularly important from the viewpoint of cultural heritage preservation, particularly in Italy, where

∗ Corresponding author. Tel.: +39 34 08 51 83 35; fax: +39 04 98 27 91 34. E-mail address: [email protected] (G. Teza).

many historical buildings are located in areas of high seismic hazard.

2. Introduction Good cultural heritage management requires the periodic evaluation of the condition of historical buildings to highlight possible needs for remediation and restoration, particularly in areas of significant seismic risk. Such evaluations require data from several observational techniques, including non-destructive testing (NDT) methods and those that have minimal impacts on the structure. Accurate geometric modeling of a historical building provides information that is useful for cultural heritage preservation [1]. In particular, data provided by TLS allow the generation of a detailed 3D model of a complex building with an accuracy of ∼1.5 cm. Therefore, it is possible to evaluate deviations from the expected shape of the structure. If multitemporal data are available, variations in shape can be assessed [2]. For example, TLS data allow the detailed characterization of the shape of a slender leaning tower [3]. The IRT, or thermal imaging, is able to detect near surface defects or features of a masonry structure [4]. Examples of integrated use of IRT and TLS also exists. In particular, [5] shows a technique aimed at IRT data mapping on TLS-based digital model for building

http://dx.doi.org/10.1016/j.culher.2014.06.008 1296-2074/© 2014 Published by Elsevier Masson SAS.

Please cite this article in press as: G. Teza, et al., Multisensor surveys of tall historical buildings in high seismic hazard areas before and during a seismic sequence, Journal of Cultural Heritage (2014), http://dx.doi.org/10.1016/j.culher.2014.06.008

G Model CULHER-2874; No. of Pages 12

ARTICLE IN PRESS G. Teza et al. / Journal of Cultural Heritage xxx (2014) xxx–xxx

2

Fig. 1. Ficarolo bell tower: a: view from the church square; b: internal view of double shaft.

diagnostics purposes. The key factor of IRT is the recognition of differences in the thermal transfer efficiency between different parts of the object, which can be related to such factors as changes in material properties, voids, or water percolation behind the surface. Experimental modal analysis (EMA) is based on the evaluation of the response of a structure to vibrations and is aimed at evaluating the structure’s natural frequencies and the corresponding modal shapes and damping. These characteristics can change significantly as structural damage occurs [6]. The output-only EMA, or operational modal analysis (OMA), which uses ambient noise as the vibration source, also provides reasonable quality results [7]. The ground is also characterized by natural frequencies that depend on the stratigraphic setting. These frequencies can be used to evaluate seismic amplification both at the scales of a town and of a single building for seismic microzonation. Moreover, if a low order natural frequency of a building is similar to a natural frequency of the ground, soil-structure interaction (SSI) can cause other amplification effects [8]. Each of these techniques is able to quickly provide important and accurate but generally incomplete information about the conditions of a historical building and its interaction with the ground. To obtain a reliable and sufficiently complete assessment of these conditions, the data provided by these techniques must be used together.

3. The Ficarolo bell tower The architect Gaetano Barbieri designed the Ficarolo bell tower (BT) (Fig. 1) to complete the S. Antonino Martire Catholic church, which was built between 1763 and 1772. Construction of the BT began in 1777, but stability problems, including partial sinking of the foundation and leaning of the structure, interrupted construction. The tower was completed after changing the design and allowing the ground to settle naturally. To partially compensate for the inclination, the axis of the belfry and cusp was designed to be unaligned with the shaft, leading to a curved shape of the tower. The elegant brick masonry BT has a double shaft with a staircase between the inner and outer shafts (Fig. 1b), heavily protruding cornices that subdivide the shaft into three equal parts, each of which with two windows that open towards the church’s square, and corners of Doric-Tuscan pillars. The current mean leaning angles are

approximately 2.6–3◦ for the shaft and 1.9–2.2◦ for the belfry and cusp. The corresponding out-of-plumbs are approximately 2.4 m for the 46-m high shaft and 3.1 m for the cusp apex, which is located 66 m above the ground. The total height of the tower, including the cross, is 69 m. Because the combination of height and leaning angle is visually impressive, Ficarolo is also known as the “Pisa of Polesine” (Polesine is the Venetian bank of the Po River), referring to the well-known 55-m high, 4◦ leaning tower of Pisa. Because the leaning angle has increased with time, leading to worries about the long-term stability of the structure, stabilization work was carried out in 2003. In particular, micro-piles were placed to consolidate the foundations. A straight pendulum was also placed to continuously measure the leaning angle of the shaft. This instrument showed that the leaning angle did not change until the 2012 seismic sequence occurred. The inclinometer showed that the base was unchanged and that the upper part of the shaft had moved by 2.5 cm. More information about the 2012 seismic sequence can be found in the Online Supplementary Material (OSM), Section OSM1.

4. TLS surveys and geometric analysis 4.1. Data acquisition, georeferencing and modeling Three surveys were carried out on April 10 (before the earthquake), May 26 and June 8, 2012. An Optech ILRIS 3D ER instrument [9] was used in all three surveys, and similar viewpoints (VPs) were used so the data could be compared (Fig. 2). The ILRIS is a time-offlight (TOF) instrument that can acquire data at distances between 3 m and 1500 m with a 2.5 kHz pulse repetition frequency and 40 × 40◦ field of view (FOV). The accuracy of single point measurements is ∼7 mm at 100 m, but the corresponding modeling accuracy is ∼3–4 mm. The choice of an appropriate sampling step leads to a spatial resolution better than 20 mm at a distance of 100 m [10]. The climatic conditions were similar for each survey and, in particular, wind was always absent. To cover the entire BT at conditions that were not excessively far from normal incidence, the scans were taken from five VPs at distances of 45 m to 102 m and at spatial sampling steps of 0.9 cm to 1.6 cm (Fig. 2). Three scans, which focused on three partially overlapping vertical portions of the BT, were taken from VPs 1, 2 and

Please cite this article in press as: G. Teza, et al., Multisensor surveys of tall historical buildings in high seismic hazard areas before and during a seismic sequence, Journal of Cultural Heritage (2014), http://dx.doi.org/10.1016/j.culher.2014.06.008

G Model CULHER-2874; No. of Pages 12

ARTICLE IN PRESS G. Teza et al. / Journal of Cultural Heritage xxx (2014) xxx–xxx

3

ˆ on the xy-plane and ˆj be the unit Let nxy be the projection of n vector of the y-axis (north direction). The azimuth of the leaning direction, L , is provided by







L = cos−1 nxy · ˆj/ nxy 

(2)

or







L = 360◦ − cos−1 nxy · ˆj/ nxy 

Fig. 2. Map of Ficarolo with bell tower (including leaning direction), TLS viewpoints (1:5), GPS receiver positions (base and targets A:E), HVSR points (P1:P5 and C1:C3) and MASW/ReMi layouts (M1, M2). Insert: bell tower sides.

4, whereas only one scan was taken from VPs 3 and 5. To avoid registration-induced errors in the analysis of the shaft’s morphology, the VP and the region of interest of each scan were always chosen so an entire side of the shaft could be acquired with one scan. The partial point clouds for each survey were registered into a common reference frame with the surface-to-surface iterative closest point (ICP) algorithm implemented in the IMAlign module of the PolyWorks software package [11]. A triangulated model was also generated. The data from the first survey were georeferenced using differential static GPS measurements in the positions A:E shown in Fig. 2. Details about the process are shown in Section OSM2. The resulting mean accuracy of the single target location was 3 cm in the north and east directions and 5 cm in elevation, leading to a ∼0.1◦ error in the leaning angle. Because the straight pendulum showed that the 7.5-m high BT base was unchanged after the earthquakes, the point clouds of the other surveys were aligned to the first survey by ICPbased surface matching of the base, resulting in fully georeferenced data. 4.2. Geometric analysis The geometric analysis is intended to: • characterize the geometry of the BT, in particular the leaning angle and direction and their earthquake-induced variations; • evaluate the displacements; • evaluate morphological changes. 4.2.1. Overall bell tower geometry An inspection of the TLS data highlights that the BT can be subdivided into six parts: the base, three segments of the double shaft, the belfry and the cusp. The truncated pyramid-shaped base is composed of stone, and the other components are masonry. Each shaft segment is a 12.5-m high parallelepiped with a constant square cross-section and sides that are 6.2 m (I segment), 5.9 m (II) and 5.6 m (III) long. The 0.9◦ taper of the shaft is therefore obtained by simple reduction of the parallelepiped base. This suggests how the ˆ be the normal leaning angle and direction can be computed. Let n ˆ be the unit vector of the vector of a leaning planar surface and k z-axis. The leaning angle, L , is ˆ − 90◦ . ˆ · k) L = cos−1 (n

(1)

(3)

for a leaning direction towards E or W, respectively. The Ficarolo BT leans towards NNW, and therefore Eq. (2) is used. Table 1 summarizes the details of the parts of the BT, including estimates of the leaning angle (with respect to vertical) and the leaning direction (with respect to north) for each measurement session. The angular uncertainties come from both statistical analysis of the data and errors due to the alignment of the second and third model to the first model. As stated above, an additional uncertainty of the leaning angle of R = 0.1◦ is associated with the georeferencing error, but it is not considered here because the aim is the detection of possible changes of leaning angle. No significant changes of the leaning angle and direction occurred. 4.2.2. Displacements In general, a direct comparison between multitemporal point clouds and/or digital models does not allow an evaluation of small displacements because the noise (instrumental noise and alignment error) has the same magnitude of the variations that should be detected. Because the straight pendulum and topographical measurements indicated that the base position and shape were not changed by the earthquake, the point cloud obtained for each BT side after the earthquake was aligned to the corresponding point cloud from the first survey by overlapping the base. This approach minimized the alignment errors; [3] demonstrated that the worst alignment error with a small overlapping area (10%) is ∼5 mm. In this case, the overlap was at least 15%. Fig. 3 shows the differences between the point cloud of each BT side after the seismic sequence and the digital model generated with the data collected before the seismic sequence. A glance to the colored multitemporal point clouds allows the recognition of occurred changes (left panel of each BT side). Results of quantitative analyses are shown in the central and right panel for each side. A slight increase of the leaning angle of the upper part of side 2 (in the leaning direction, i.e., NNW) and a convexity (amplitude: 1.5–2 cm) of the corresponding first shaft segment indicate sharp deformation and bulging, respectively. A small displacement (∼1 cm) appears in the upper part of side 3 with a resulting slight change of inclination. This is a notable result because this side is parallel to the leaning direction. Moreover, a weak signal of belfry rotation can be seen in Table 1, although it is dominated by the error. This does not contradict the results of the leaning angle analysis. The detected differences are in the range 0–1 cm distributed along a height of 20 m and are therefore too small to identify significant changes in the leaning angle. A comparison between the first and third surveys indicates a change of the belfry leaning azimuth by ϕ = 0.17◦ ± 0.27◦ towards W, while a comparison between the first and second surveys indicates a change of ϕ = 0.13◦ ± 0.24◦ towards W. Despite the high level of uncertainty, the values indicate a trend. The results show that the leaning angle and direction can be evaluated based on averages of single architectural elements from the differences between multitemporal point clouds. 4.2.3. Morphological changes According to [2], the TLS-based deformation analysis was carried out by:

Please cite this article in press as: G. Teza, et al., Multisensor surveys of tall historical buildings in high seismic hazard areas before and during a seismic sequence, Journal of Cultural Heritage (2014), http://dx.doi.org/10.1016/j.culher.2014.06.008

G Model CULHER-2874; No. of Pages 12

ARTICLE IN PRESS G. Teza et al. / Journal of Cultural Heritage xxx (2014) xxx–xxx

4 Table 1 Some data about Ficarolo bell tower geometry. Bell tower part

Elevation range (m)

10 April 2012  (◦ )a

Base Shaft – I Shaft – II Shaft – III Belfry Over-belfry/cuspc a b c d

[0, 7.3] [7.3, 20.2] [20.2, 32.7] [32.7, 45.2] [45.2, 53.2] [53.2, 66.2]

2.60 2.70 2.94 2.78 2.17 1.90

26 May 2012 ϕ (◦ )b

± ± ± ± ± ±

0.10 0.05 0.04 0.04 0.10 0.10

321.35 321.23 321.28 321.11 321.18 320.9

 (◦ )a ± ± ± ± ± ±

0.13 0.18 0.12 0.17 0.16 0.4

2.60 2.76 2.99 2.80 2.25 1.97

8 June 2012 ϕ (◦ )b

± ± ± ± ± ±

0.10 0.10 0.08 0.10 0.11 0.25

321.35 321.31 321.34 321.18 321.31 321.1

 (◦ )a ± ± ± ± ± ±

0.13 0.21 0.17 0.21 0.17 0.3

2.60 2.73 2.93 2.83 2.22 2.11

ϕ (◦ )b ± ± ± ± ± ±

0.10 0.07 0.07 0.05 0.13 0.20d

321.35 321.28 321.33 321.19 321.35 321.1

± ± ± ± ± ±

0.13 0.21 0.14 0.19 0.22 0.3

Leaning angle (with respect to vertical direction). Leaning direction (with respect to North direction). A 3-m high cross is placed over the cusp (the total bell tower height is 69.2 m). The over-belfry shape is slightly changed because of temporary remediation works.

• creating primitives by interpolating the data for each side; • creating morphological maps by computing the differences between the points and the primitives; • iteratively recomputing the primitives by excluding the areas with large differences, leading to more reliable morphological maps. To allow for a reliable interpretation of the results, possible distortions due to non-optimal surveying should be taken into account.

The incidence angles were less than 40◦ for the shaft observations and less than 50◦ for the belfry and cusp observations. The ratio between the horizontal distance from the instrument to the target, d, and the target height, h, was d/h = 0.8 for the worst observation conditions. The data from Table 1 in [2], which hold if an ILRIS instrument is used, state that the incidence angle-induced distortions are less than 7 mm at the top of the shaft (45.2 m high) and 9 mm at the point above the belfry (60.2 m). The cusp was not studied because its geometry cannot be modeled by a simple surface.

Fig. 3. Earthquake-induced displacements (comparison between first and last scans). For each side, left panel: colored point clouds (i.e. a color for each point cloud); central panel: differences; right panel: particular of area characterized by higher variations.

Please cite this article in press as: G. Teza, et al., Multisensor surveys of tall historical buildings in high seismic hazard areas before and during a seismic sequence, Journal of Cultural Heritage (2014), http://dx.doi.org/10.1016/j.culher.2014.06.008

G Model CULHER-2874; No. of Pages 12

ARTICLE IN PRESS G. Teza et al. / Journal of Cultural Heritage xxx (2014) xxx–xxx

5

Fig. 4. Detailed morphological analyses of first segment of side 2 before and after earthquake, where some changes are highlighted. Panels a and b: pillars (corners); panels c and d: wall.

Each side of the shaft was subdivided into three segments with constant cross-sections. Each was studied independently by means of a point-plane comparison, where the plane (i.e., the reference primitive) was obtained by fitting the points of the wall, excluding the pillars (corners). The lowest segment of side 4 (towards NNW) was shaded by a house. The morphologies of all sides of the BT before and after the seismic sequence are shown in Fig. OSM3. The results show that changes occurred along the entire structure. The morphological analysis of the first segment of the shaft on side 2 is described in detail (Fig. 4, where fitting of planes to the wall and the corner points is carried out independently). Such an analysis shows that:

• the bulge existed prior to the earthquake and was most likely caused by compensation for the inclination (this side is orthogonal to the leaning direction); • the bulge was significantly different after the earthquake: the bulged area became ∼15 m2 from the initial ∼8 m2 with the same amplitude, i.e. 2 cm (range 13–15 cm with respect to the reference plane); • the change affects both the wall and the pillars; • changes in the area below the lowest window (from a concavity of ∼0.5 cm localized in ∼0.1 m2 to a concavity of ∼1 cm in ∼1 m2 area) can be detected by the morphological analysis.

It is important to note that the morphological analysis is able to detect changes that cannot be evaluated by the displacement evaluation.

5. Thermal imaging 5.1. Method An infrared thermography (IRT) camera converts the energy in the far infrared (FIR) band captured by a bolometer pixel during the integration time (e.g., 20 ms) to the temperature of the corresponding element of the observed surface. The IRT allows the recognition of damaged zones of a masonry wall because the corresponding local heat capacity and thermal transfer efficiency differ from those of the undamaged material, and these differences appear in the time history of the temperature pattern. Let Tij (tk ) be the temperature associated with pixel ij of the thermal image taken at time tk (k = 0, 1, · · ·, N) assuming that that N + 1 images are registered into the same reference frame and that a thermal transient (e.g., night cooling) is captured. The running, self-referenced thermal contrast (RTC) is defined by [12] as





RTCij (tk ) = Tij (tk ) − Tm,ij (tk ) ,

(4)

Please cite this article in press as: G. Teza, et al., Multisensor surveys of tall historical buildings in high seismic hazard areas before and during a seismic sequence, Journal of Cultural Heritage (2014), http://dx.doi.org/10.1016/j.culher.2014.06.008

G Model CULHER-2874; No. of Pages 12

ARTICLE IN PRESS G. Teza et al. / Journal of Cultural Heritage xxx (2014) xxx–xxx

6





where Tm,ij (tk ) is the mean temperature in the square neighborhood Tm,ij (tk ) with sides of 2m + 1 (in pixels) that surrounds this pixel and m is chosen on the basis of the expected size of damage features. The use of RTC instead of the temperature contrast has the advantage of reducing the effects of non-uniform heating. A pixel could be defective if |RTCij (tk )|≥m,ij (tk ),

(5)

where m,ij (tk ) is the standard deviation of the temperature in Tm,ij (tk ), and  depends on the signal to noise ratio of the thermal images. According to [13], when details about the RTC-based damage recognition are shown together with its MATLAB implementation, a pixel ij can be assumed to be defective if condition Eq. (5) holds in at least 75% of the thermal images taken during a thermal transient.

5.2. Measurements The measurements were taken during a period of night cooling with a FLIR T620 instrument [14], which is characterized by a 480 × 640 pixel bolometric array, a 25 × 19◦ field of view (FOV) and a 0.68 mrad angular resolution. To cover the entire BT, the IRT images were taken from four VPs at acquisition distances between 35 m (2.4 cm spatial resolution) and 80 m (5.4 cm). The data acquisition for one VP (shaft side 1) required the mosaicking of three partial images, and two VPs (sides 2 and 3) required the mosaicking of two images. For side 4, a single pose was sufficient for each acquisition time. For each pose, one image per hour was taken from June 14 at 20.00 to June 15 at 01.30. Calibration of thermal images is a preliminary task for each IRT analysis, as shown by [5], where a calibration method optimized for architectural survey is discussed. Therefore, the thermal images were post-calibrated based on the acquisition distance, the mean IR reflectance of the targets and the air temperature. The IRT analysis was carried out using the THIMRAN toolbox [13]. A value of m = 15 was chosen, leading to a kernel for the RTC computation of 31 pixels, i.e., 105 cm at a distance of 50 m (3.4 cm pixel size). This choice allowed for the detection of anomalous areas whose size (width in the case of a crack) is in the range from cm to approximately 1 m. After several trials,  = 0.8 was chosen. Finally, because the thermal behavior of the belfry/cusp was affected by the temporary wood and steel structures placed by firemen for safety, the IRT analysis was carried out only on the shaft. The analysis of side 1 of the shaft, which is oriented towards ESE and is therefore parallel to the leaning direction (Fig. 2), shows that some areas are affected by anomalous RTC. In particular, a 2 m long, 1.2-cm wide crack near the lower window was recognized (Fig. 5, area 1). Because the difference in thermal contrast is significant, and some bricks are broken, this crack affects at least one layer of bricks. An inspection of the TLS data shows that this crack existed before the earthquake. Another crack is present on side 1 (Fig. 5, area 2). Area 3, in the middle part of the shaft, is a false positive caused by the clock, and area 3 is related to bricks that were replaced during the 2003 restoration. Side 2, which is towards SSE and is therefore orthogonal to the leaning direction, contains two areas with anomalous RTC (Fig. 5, areas 5 and 6). These areas are also affected by moderate morphological changes (Fig. 4). No significant areas with anomalous RTC are recognized in the other two sides. In particular, the thermal behavior of the upper part of side 4, whose bricks partially lack mortar, appears to be completely normal. This agrees with the results of the TLS-based analysis and visual inspections performed by the firemen; the lack of mortar is only a surface defect and is not structurally significant.

Fig. 5. Results of thermal analysis. Pixels where RTC exceeds threshold in at least 75% of night cooling duration are colored in red. Some damage areas (labeled 1:6) are highlighted.

6. Operational modal analysis and geophysical measurements 6.1. Simplified operational modal analysis The vibration source for OMA measurements is ambient noise, i.e., weak ground motions due to both natural sources (e.g., atmospheric and oceanic perturbations) and anthropogenic sources (e.g., road traffic). The measurements require a network of vibration sensors that are placed on the structure, linked and synchronized. Because these measurements are expensive, OMA is systematically used only to test new or renovated structures and is rarely used to monitor the condition of cultural heritage structures. Assuming that stochastic noise is constant during a measurement session, a smaller number of non-synchronized moving digital tromographs can be used to perform a low-cost OMA. In this case, the frequencies and damping ratios can be evaluated, but the modal shapes are not sufficiently accurate to provide a damage index. This type of OMA is sufficient for an evaluation of soil-structure interaction (SSI). A digital, compact Micromed Tromino tromograph [15] was used to perform the non-synchronized vibration measurements. This instrument is equipped with three vibration sensors placed orthogonally to one another; each includes an electrodynamic transducer (velocimeter) and a capacitive transducer (accelerometer). The sampling frequency, fS , can be 128 Hz, 256 Hz, 512 Hz or 1024 Hz. The mass of the instrument is 1.1 kg. Although the mass of each sensor is very low (ten grams), the performance of the instrument, whose resonance frequency is 4.5 Hz, is acceptable in the typical frequency range of interest, i.e., from 0.1 Hz to several hundred Hz [16]. In each survey, three Tromino instruments were used to make asynchronous measurements at fS = 512 Hz for 24 minutes along two vertical profiles at the locations shown in Fig. 6. The first profile (called W-profile) was aligned vertically along the middle of side 2 (towards SSE, orthogonal to the leaning direction), with nine positions: W0 was located on the ground floor, W1 was on the

Please cite this article in press as: G. Teza, et al., Multisensor surveys of tall historical buildings in high seismic hazard areas before and during a seismic sequence, Journal of Cultural Heritage (2014), http://dx.doi.org/10.1016/j.culher.2014.06.008

G Model CULHER-2874; No. of Pages 12

ARTICLE IN PRESS G. Teza et al. / Journal of Cultural Heritage xxx (2014) xxx–xxx

7

Table 2 summarizes the results in terms of the natural frequencies. Because the temperature affects these frequencies, two winter (one before the seismic sequence) and two summer values are shown for each identified mode. Five modes are identified: the first bending mode along LD and OD, the second bending mode along LD and OD, and the first torsional mode. The results show that the earthquake reduced the natural frequencies by 5% for the first bending modes (ıfIB ≈ 5%) and 8% for the second bending and torsional modes (ıfIIB ≈ 8%, ıfIT ≈ 8%). Fig. 7 shows the spectra of the W-profile signals deconvolved with the ground motions for LD and OD, respectively, where the character of the modes can be easily understood. The first LD bending mode is easily detectable by the FDD analysis implemented in MACEC but cannot be observed if simple deconvolved spectra are observed, as in the standard approach implemented in the Grilla 6.4 package [15]. The use of advanced data analysis methods such as FDD is therefore recommended. Fig. 10 also shows the corresponding deformed shapes of a finite element model (FEM) of the bell tower obtained using COMSOL Multiphysics [20], which uses a tower geometry based on the TLS data and reasonable assumptions about the material. Unfortunately, no reliable estimates of damping ratios are available because the reference-based, covariance-driven stochastic subspace identification [6] implemented in MACEC is unable to identify stable modes with non-synchronized data. 6.2. Geophysical measurements

Fig. 6. Positions of Tromino instruments for operational modal analysis.

staircase, W2–W7 were on the window sills, and W8 was on the belfry floor. The second profile (C-profile) was aligned vertically along the western corner of side 4 (towards NNW), and included 11 positions: C0 was on the ground floor, C1–C9 were on the staircase landings, and C10 was on the belfry floor. The mean vertical steps were 5.6 m and 4.5 m for the W-profile and C-profile, respectively. The horizontal axes of the Tromino instrument were always parallel to the structure’s main axes, i.e., the leaning direction (LD) and the direction orthogonal to it (OD). Each complete survey (two profiles) was performed in less than five hours. To confirm that the ambient vibrations were stationary during the entire measurement session, the instrument at the ground floor position (W0 or C0) took repeated measurements as the other two instruments were moved along the profile. The data analysis was performed using frequency domain decomposition (FDD) [17] implemented in the MACEC software package [18]. Some details about FDD are shown in OSM4. To remove the effects of foundation motions due to SSI, the time series W1–W1 and C1–C10 were deconvolved with the time series W0 and C0, respectively [19]. To perform this in the MACEC software, the W0 and C0 time series were used as inputs for the Wprofile and C-profile data analysis, respectively. The analyses were performed for each component (LD and OD) separately.

Because the velocity VS of a shear wave in a material is proportional to the shear modulus, which in turn depends on the material stiffness, the vertical VS profile is similar to a vertical stiffness profile. Shear wave stratigraphic resonances can induce site amplification that occurs where a layer of soft sediments (e.g., unconsolidated fine grained sediments) overlies seismic bedrock (e.g., the effective bedrock or overconsolidated sediments). Moreover, if the natural frequency of a low order mode of a building, whose modal energy is relatively high, is similar to or slightly higher than a ground resonance frequency, the resonance coupling can cause SSI-induced amplification. Good cultural heritage management in an area that is characterized by significant seismic hazard requires an evaluation of possible site amplification phenomena and SSI [8]. The soil resonances are determined by the analysis of ambient noise using Nakamura’s horizontal to vertical spectral ratio (HVSR). The resonance frequency of the fundamental mode for a simple, homogenous layer with a shear wave velocity VS overlying bedrock at a depth h is [21] fR =

VS . 4h

(6)

The resonance can be recognized as the peak of the H/V curve, which maps the ratio between the amplitudes of the horizontal (H) and vertical (V) components of the ambient noise vs. frequency. If VS (respectively h) is available, Eq. (6) provides h (VS ). Eq. (6) does not hold in a multilayered system. If several peaks appear in the H/V curve due to the presence of several reflecting interfaces, the data inversion requires ad hoc models that simulate the propagation of surface waves in layered media [22]. The geological setting of Ficarolo area known from literature is described in Section OSM5. The measurements were taken with the Tromino instrument during 20 minute acquisition periods at a sampling frequency of 128 Hz. To detect possible lateral stratigraphic variations, five measurement points close to the BT were chosen (Fig. 2, points P1:P5). Moreover, the soil stratigraphy in areas of better seismic response, without the effect of pavement, was evaluated by taking measurements in a meadow located 150 m from the BT (C1:C3).

Please cite this article in press as: G. Teza, et al., Multisensor surveys of tall historical buildings in high seismic hazard areas before and during a seismic sequence, Journal of Cultural Heritage (2014), http://dx.doi.org/10.1016/j.culher.2014.06.008

G Model

ARTICLE IN PRESS

CULHER-2874; No. of Pages 12

G. Teza et al. / Journal of Cultural Heritage xxx (2014) xxx–xxx

8

Table 2 Natural frequencies related to the five lower modes of Ficarolo bell tower. Mode

Natural frequency (Hz) Winter season

First bending, LD First bending, OD Second bending, LD Second bending, OD First torsional

Summer season

February 22, 2012

February 26, 2013

June 8, 2012

July 11, 2013

0.57 0.60 2.3 2.3 3.4

0.54 0.57 2.1 2.1 3.1

0.55 0.57 2.1 2.0 3.0

0.55 0.58 2.1 2.0 3.0

LD: leaning direction; OD: orthogonal direction.

Fig. 8 shows the H/V curves that were obtained using the Grilla 6.4 package [15] for points P1, P2 and C1–C3. According to the SESAME criteria [23], strong peaks are detected at 0.9–1.0 Hz. For example, the P1 peak frequency is fP1 = 0.94 ± 0.06 Hz. A significantly lower peak at approximately 1.8 Hz is also present. A comparison between the H/V curves shows that at this site: • the velocity inversion effects due to pavement (porphyry and asphalt) do not affect the frequency band in the region below 10 Hz, which is the frequency region of interest for SSI analysis; • no significant lateral variations affect the stratigraphy of the shallow deposits; • the H/V curves for points P1 and P2 are not affected by the presence of buildings. The VS profile was evaluated by refraction microtremor analysis (ReMi) [24] and multichannel analysis of surface waves (MASW) [25]. The vibration sources for the ReMi and MASW were ambient noise and a dropped weight, respectively. The ReMi measurements were carried out along two linear arrays in the meadow (Fig. 2, M1, M2) using a SoilSpy Rosina digital system with 16 channels and 4.5 Hz vertical geophones [15]. The geophones were placed along a 48 m long line at 3 m spacing. The ambient vibrations were recorded for 15 minutes, and the dispersion curves where analyzed manually using 10 s time windows to collect the clear dispersion curves. The MASW measurements were performed on the same arrays with

the same instrument, which also has a trigger channel for active techniques. The obtained VS profiles are shown in Fig. OSM5. Fig. 9 shows the results of the iterative numerical modeling of H/V curve P2. The modeling was performed with Grilla 6.4 according to the method described by [22] to obtain a simplified VS profile that is compatible with the stratigraphic information and data provided by the MASW and ReMi surveys. A profile of unit tip bearing capacity provided by a 26-m depth penetrometric test conducted before the 2003 stabilization work is also shown. The low peak at 1.8 Hz is interpreted as an impedance contrast at ∼20 m depth at the transition from alternating silt and clay sediments to fine sand deposits that marks the upper part of the ∼30 m deep sandy Aquifer I (Fig. OSM4). The strong peak at approximately 0.9 Hz is ascribed to Aquifer II, whose upper part should be located at a depth of approximately 70 m. The wide dispersion of the 0.9 Hz peak could be related to the presence of deeper reflectors (e.g., Aquifers III and AIV) and also to possible velocity inversion due to hydrogeological confining units. The obtained VS for the deeper layers is 350 m/s, which is consistent with the value of 370 m/s determined by other studies in the area [26]. The mean VS in the upper 30 m is VS30 = 170 m/s. 7. Discussion This discussion focuses on the implications of the results of this study for both the specific case study and for cultural heritage

Fig. 7. Operational modal analysis. Amplitude spectral ratios (ASR) of W-chain signals with respect to W0 (ground level), i.e. W-chain signals deconvolved with the ground motions: a: LD, 22/02/2012; b: OD, 22/02/2012; c: LD, 26/02/2013; d: OD, 26/02/2013. Deformed shapes are also shown. Pecks recognizable by both spectral ratios and FDD are highlighted by red arrows. A frequency peak recognized by FDD only is highlighted by a blue arrow.

Please cite this article in press as: G. Teza, et al., Multisensor surveys of tall historical buildings in high seismic hazard areas before and during a seismic sequence, Journal of Cultural Heritage (2014), http://dx.doi.org/10.1016/j.culher.2014.06.008

G Model CULHER-2874; No. of Pages 12

ARTICLE IN PRESS G. Teza et al. / Journal of Cultural Heritage xxx (2014) xxx–xxx

9

Fig. 8. H/V curves: a and b: curves in range 0.1–128 Hz for five selected sites (P1, P2: Ficarolo square; C1: C3: meadow area); c and d: the same curves in range 0.1–10 Hz.

Fig. 9. Interpretation of HVSR results: a: real H/V curve and synthetic H/V curve related to obtained model (whole frequency band); b: particular of real and modeled H/V curve in the range 0.1–10 Hz; c: modeled stratigraphy and vertical VS profile; d: results of penetrometric analysis (reference).

Please cite this article in press as: G. Teza, et al., Multisensor surveys of tall historical buildings in high seismic hazard areas before and during a seismic sequence, Journal of Cultural Heritage (2014), http://dx.doi.org/10.1016/j.culher.2014.06.008

G Model CULHER-2874; No. of Pages 12 10

ARTICLE IN PRESS G. Teza et al. / Journal of Cultural Heritage xxx (2014) xxx–xxx

Fig. 10. Some selected areas in sides 1 and 2 of bell tower shaft. Areas 1 and 2: extended cracks; area 4: brick change; areas 5: brick change and bulging; area 6: bulging.

management in general. The main strengths and weaknesses of the approach used in this study are discussed below. The results of three types of geometric analysis were shown: (a) characterization of the overall geometry of the BT; (b) evaluation of the displacements; and (c) evaluation of the morphological changes. These analyses are not redundant; each provides a specific type of information. Analysis (a) describes the general shape of the structure and the large scale (tens of meters) leaning angle and direction. Analysis (b) provides a medium scale (meters) description of the earthquake-induced displacements, and analysis (c) provides small scale (tens of centimeters) information. In addition, analyses (a) and (c) provide information about the layout before the seismic sequence. For example, bulging of the first segment of side 2 of the shaft (orthogonal to the leaning direction) was detected by the displacement evaluation (Fig. 3), but it cannot be detected by a large scale analysis. Moreover, a comparison of the results of the morphological analyses indicates that the bulge existed before the earthquake and that significant changes occurred because of the seismic sequence (Fig. 4 and OSM3). In particular, Fig. 4 shows that the extension of the bulge increased significantly. Another interesting result is the recognition of a slight rotation of the belfry/cusp. The thermal imaging highlighted relatively deep cracks in sides 1 and 2 (Fig. 5, areas 1, 2, and 5) that likely affect an entire layer of the brick masonry. An inspection of the TLS data reveals that these cracks existed before the earthquake. Fig. 10 shows visual images of anomalous areas recognized by the IRT. It is interesting to note that the anomalous area 6 appears to be normal in the visual band. Nevertheless, the TLS data show that this area is affected by morphological changes, in particular bulging (Fig. 4 and OSM3). These results highlight both the importance of IRT in identifying damage and the fact that the technique provides a good interpretation of the data. The results provide information about phenomena that affect the surface and shallow layers of a wall, which generally require data to be obtained with other techniques. The simplified OMA surveys highlight decreases in the natural frequency of the low order modes that are related to the similar climatic conditions before and after the earthquake (February 22, 2012 and February 26, 2013). The percentage decreases for the first bending, second bending and first torsional modes are

ıfIB ≈ 5%, ıfIIB ≈ 8% and ıfIT ≈ 8%, respectively. The FEM calculations show that these changes are compatible with significant weakening of the belfry base, which is probable because the TLS data highlight a slight rotation of the belfry. A comparison between the results obtained during the summers (June 8, 2012 and July 11, 2013) indicates that the BT is currently stable and that no damage processes due to instability or the leaning angle are active. Significant damage has affected the BT belfry and cusp, whereas possible damage to BT shaft is of minor importance; however, further analyses of the shaft, and particularly the detected bulge, with other techniques are planned. The interpretation of the OMA results is also based on the TLS data. This confirms the importance of the joint use of data provided by several techniques in evaluating the condition of historical structures. The use of non-synchronized data provides a good estimation of the natural frequencies of the low order modes, which are the most important modes for a tall tower. The survey was planned before the earthquake to evaluate the SSI and provide data about the state of the BT but not for a complete structural characterization of the building. Nevertheless, the non-synchronized data do not allow an evaluation of the mode shapes, and this fact has led to some criticism. There are two responses to this criticism: • the method is designed to be extensively applied to historical buildings in areas of high seismic hazard areas, and limitations on financial and human resources often prevent the systematic use of networks of synchronized accelerometers; • during seismic emergencies, the use of synchronized accelerometers may be difficult or impossible (it should be noted that the June 2012 measurements was carried out with the assistance of firemen). Another problem is related to the unsuccessful evaluation of damping ratios that in principle should be accessible. This is most likely due to insufficient coupling between the instruments and the structure. The instrument was placed on the structure with sharp feet. Further research is necessary to design a better coupling system that maintains the simplicity of the setup, such as

Please cite this article in press as: G. Teza, et al., Multisensor surveys of tall historical buildings in high seismic hazard areas before and during a seismic sequence, Journal of Cultural Heritage (2014), http://dx.doi.org/10.1016/j.culher.2014.06.008

G Model CULHER-2874; No. of Pages 12

ARTICLE IN PRESS G. Teza et al. / Journal of Cultural Heritage xxx (2014) xxx–xxx

placing modeling clay between the instrument and the floor (Silvia Castellaro, personal communication). The amplitude of the H/V ratio, which is related to the acoustic impedance contrast between stratigraphic layers, is not an effective seismic amplification factor observed during an earthquake [23]. A narrow and high amplitude peak indicates a large impedance contrast between layers and therefore possible strong site amplification, particularly if a soft layer overlies the seismic bedrock. In this study, a soft layer overlies a stiffer layer at a depth of approximately 70 m (aquifer II, Fig. OSM4), but the corresponding peak is not high and is relatively broad (Figs. 8 and 9). This indicates that significant local amplification effects are not expected. The mean shear wave velocity in the upper 30 m is VS30 = 170 m/s, which corresponds to Eurocode8 ground type D [27]. It is important to note that the geophysical measurements must be performed near the building and, if possible, in unpaved areas to collect information about the homogeneity of the stratigraphy, exclude tromographic data that are affected by the presence of the building and identify paving-induced velocity inversion effects. The main resonance peaks of the soil are between 0.9 and 1 Hz. [8] demonstrated that a frequency difference between a building and the soil of less than 15% can lead to large amounts of damage due to SSI resonance, differences between 15 and 25% lead to medium damage and differences of more than 25% lead to negligible resonant coupling. The BT modes with relatively high modal energy have a frequency of 0.55 Hz, whereas the main resonance peak of the soil is between 0.9 and 1 Hz. Because the difference is 45%, SSI does not occur. Like many bell towers, the studied structure has a relatively simple geometry. More complex buildings can be analyzed with these techniques even though the corresponding surveys will be more complex and require more time and higher costs. In particular, the identification of a structure whose modes are non-orthogonal will not be simple. This must be taken into account. Another important factor that must be considered is the increase of SSI with building height depending on the local stratigraphy. Therefore, observations of tall and slender structures are very important. The first TLS survey took approximately six hours to complete because of GPS-based georeferencing, whereas the other surveys required approximately five hours. The IRT survey was carried out during a complete night cooling cycle (5.5 hours). Each OMA survey required five hours (in one case, the OMA survey was performed at the same time as a TLS session). The ReMi/MASW surveys were performed during an OMA session, and the Tromino-based geophysical measurements required a few hours. The TLS, IRT and geophysical surveys do not require any contact with the structure, whereas the presence of an operator inside a potentially damaged building is limited to placing the instrument and ascending/descending the stairs. In conclusion, the joint use of these techniques quickly and safely provides high quality information.

8. Conclusions Multitemporal laser scanning, thermal imaging, asynchronous operational modal analysis and geophysical measurements were performed on the Ficarolo leaning bell tower before and after the 2012 seismic sequence. The results show that despite the absence of significant site amplification and soil-structure interaction, the 2012 earthquake caused damage to the tower. Centimeter scale geometrical/morphological changes occurred in the shaft, and there is some evidence of torsion in the belfry. Moreover, the observed decreases in the natural frequency of the lower order modes (5–8%) are related to damage in the belfry and cusp. Remediation work is planned and will take these results into account.

11

The main value of this work is associated with the methodology: • the observational data can be combined to obtain an accurate assessment of both the state of a masonry building and its interaction with the ground; • the surveys are fast, safe and relatively inexpensive. The proposed method should be used systematically to prevent damage to historical buildings, particularly tall towers. Moreover, it is recommended that the surveys be repeated after large earthquakes to evaluate damage to the structures. Acknowledgements The authors wish to thank Monsignor Giancarlo Crepaldi, parish archpriest of S. Antonino Church in Ficarolo, Massimiliano Furini, architect of the Rovigo Episcopate, and Matteo Previato, technician of the Ficarolo Municipality, for the bell tower entry authorization and the information kindly provided. The authors are very grateful to firemen of Castelmassa Detachment (Rovigo) for their support in OMA measurements in seismic crisis conditions. The ReMi/MASW instrument was purchased within the project, cofinanced by Veneto Region, POR CRO FESR 2007-2013, action 1.1.1. Appendix A. Supplementary data Online Supplementary Material (OSM) associated with this article can be found, in the online version, at http://dx.doi.org/ 10.1016/j.culher.2014.06.008. References [1] L. De Luca, C. Busarayat, C. Stefani, P. Véron, M. Florenzano, A semantic-based platform for the digital analysis of architectural heritage, Comput. Graph. 35 (2) (2011) 227–241. [2] A. Pesci, G. Teza, E. Bonali, G. Casula, E. Boschi, A laser scanning-based method for fast estimation of seismic-induced building deformations, ISPRS J. Photogramm. Remote Sens. 79 (2013) 185–198. [3] A. Pesci, G. Casula, E. Boschi, Laser scanning the Garisenda and Asinelli towers in Bologna (Italy): detailed deformation patterns of two ancient leaning buildings, J. Cult. Herit. 12 (2) (2011) 117–127. [4] C. Meola, Infrared thermography of masonry structures, Infrared Phys. Techn. 49 (3) (2006) 228–233. [5] M.I. Alba, L. Barazzetti, M. Scaioni, E. Rosina, M. Previtali, Mapping infrared data on terrestrial laser scanning 3D models of buildings, Remote Sens. 2011 (39) (2011) 184–1870. [6] F.N. Catbas, D.L. Brown, A.E. Aktan, Use of modal flexibility for damage detection and condition assessment: case studies and demonstrations on large structures, J. Struct. Eng. ASCE 132 (11) (2006) 1699–1712. [7] B. Peeters, G. De Roeck, Reference-based stochastic subspace identification for output-only modal analysis, Mech. Syst. Signal Process. 13 (6) (1999) 855–878. [8] A. Gosar, Site effects and soil-structure resonance study in the Kobarid basin (NW Slovenia) using microtremors, Nat. Hazards Earth Syst. Sci. 10 (2010) 761–772. [9] Available at: http://www.optech.com/index.php/product/optech-ilris/, (Accessed: 13.06.2014). [10] A. Pesci, G. Teza, E. Bonali, Terrestrial laser scanner resolution: numerical simulations and experiments on spatial sampling optimization, Remote Sens. 3 (1) (2011) 167–184. [11] Available at: http://www.innovmetric.com/polyworks/3D-scanners/home.aspx, (Accessed: 13.06.2014). [12] M. Omar, M.I. Hassan, K. Saito, R. Alloo, IR self-referencing thermography for detection of in-depth defects, Infrared Phys. Technol. 46 (4) (2005) 283–289. [13] G. Teza, THIMRAN: A MATLAB toolbox for mosaicking, registration and analysis of thermal images for damage recognition in large bodies, J. Comput. Civil Eng. 28 (2014) 1–8 (04014017). [14] Available at: http://www.flir.com/cs/emea/en/view/?id=41967, (Accessed: 13.06.2014). [15] Available at: http://www.tromino.eu/, (Accessed: 13.06.2014). [16] J.-L. Chatelain, B. Guillier, Reliable fundamental frequencies of soils and buildings down to 0.1 Hz obtained from ambient vibration recordings with a 4.5-Hz sensor, Seism. Res. Lett. 84 (2) (2013) 199–209. [17] R. Brincker, L. Zhang, P. Andersen, Modal identification from ambient response using frequency domain decomposition, Smart Mater. Struct. 10 (3) (2001) 441–445. [18] Available at: http://bwk.kuleuven.be/bwm/macec, (Accessed: 13.06.2014).

Please cite this article in press as: G. Teza, et al., Multisensor surveys of tall historical buildings in high seismic hazard areas before and during a seismic sequence, Journal of Cultural Heritage (2014), http://dx.doi.org/10.1016/j.culher.2014.06.008

G Model CULHER-2874; No. of Pages 12 12

ARTICLE IN PRESS G. Teza et al. / Journal of Cultural Heritage xxx (2014) xxx–xxx

[19] S. Castellaro, F. Mulargia, How far from a building does the ground motion freefield start? The case of three famous towers and a modern building, B. Seism. Soc. Am. 100 (5A) (2010) 2080–2094. [20] Available at: http://www.comsol.com/, (Accessed: 13.06.2014). [21] Y. Nakamura, A method for dynamic characteristics estimation of subsurface using microtremors on the ground surface, Quart. Rep. RTRI Japan 33 (1989) 25–33. [22] S. Castellaro, F. Mulargia, L. Bianconi, Passive seismic stratigraphy: a new efficient, fast and economic technique, Geol. Tech. Ambient. 3 (2005) 76–97. [23] P.-Y. Bard, SESAME participants, SESAME European research project, Guidelines for the implementation of the H/V spectral ratio technique on ambient

[24] [25] [26] [27]

vibrations. Deliverable D23.12, Paris, 2004 (available at: ftp://ftp.geo.uib.no/ pub/seismo/SOFTWARE/SESAME/USER-GUIDELINES/SESAME-HV-UserGuidelines.pdf. [Accessed: 13.06.2014]). J.N. Louie, Faster, better: shear wave velocity to 100 meters depth from refraction microtremor arrays, B. Seism. Soc. Am. 91 (2) (2001) 347–364. C.B. Park, R.D. Miller, J. Xia, Multichannel analysis of surface waves (MASW), Geophysics 64 (3) (1999) 800–808. A. Amorosi, M. Pavesi, Aquifer stratigraphy from the middle-late Pleistocene succession of the Po Basin, Mem. Descr. Carta Geol. Italia XC (2010) 7–20. CEN, Eurocode 8: design of structures for earthquake resistance – Part 1: 2004, BSI, London, 2005.

Please cite this article in press as: G. Teza, et al., Multisensor surveys of tall historical buildings in high seismic hazard areas before and during a seismic sequence, Journal of Cultural Heritage (2014), http://dx.doi.org/10.1016/j.culher.2014.06.008