For damage localization (the third phase of the proposed approach) a laboratory simulation of a wall model is presented here. The replicate of historical constructions was built as reference, undamaged, state. Afterwards, progressive damage was induced and sequential modal identification analysis was performed at each damage stage, aiming at finding adequate correspondence between dynamic behavior and internal crack growth.
The wall model was built with clay bricks with 210 x 105 x 55 mm3, handmade in the Northern area of Portugal. The adopted clay bricks have low compression strength and the adopted Mapei®mortar for the joints has low mechanical properties for the joints,
trying to be representative of the materials used in the historical constructions. The wall has a length equal to 1.08 m, a height equal to 1.10 m, and a thickness equal to 0.105 m, which matches the bricks thickness, see Figure 7. The thickness of the joints is about 1.0 cm.
Fourteen Damage Scenarios (DS) were induced with the application of two static forces: a constant vertical force to replicate existing vertical compressive stresses and a horizontal force to produce shear stresses. The vertical load was transmitted through a group of three steel beams with appropriate devices to distribute uniformly the load to the wall. The last beam was direct glued to the wall. The in-plane stresses were applied with the aim of producing bending and shear cracks, i.e. to reproduce the common crack pattern present in the masonry piers that suffer earthquake actions.
To achieve several controlled cracks, three series of static tests were carried out. Figure 8 presents foo the last series (Series C) the test apparatus, the static response of the wall and the final crack pattern. From the static response is possible to observe the stiffness decrease, where the horizontal displacement indicated corresponds to the maximum or top value (LVDT 1).
The final crack pattern includes three cracks. One bending crack is at left bottom part, and another bending crack is at right top part of the wall, namely crack cj and c2, respectively. The third crack (c3) is a shear crack in the diagonal direction of the wall.
Between each DS, modal identification analysis using output-only techniques was performed. The selected sensors for the dynamic tests were the accelerometers and strain gauges with quarter bridge configuration. The strain gauges were selected to measure modal strains in order to directly estimate modal curvatures, which are quantities more sensitive to damage than the modal displacements (Ramos, 2007).
A regular grid of five vertical lines and seven horizontal lines was chosen for the accelerometers. In the case of the strains gauges, it was decided to use an array of three vertical lines and five horizontal lines. Figure 9 shows the location of the measuring points where the reference points are given in a grey box.
The accelerometers were bolted to aluminum plates that were directly glued to the wall. The vertical strain gauges have 12 cm of length and the horizontal strain gauges have 6 cm of length. This way, the vertical strain gauge crosses, at least, three joints and the horizontal strain gauge only one joint.
Two different excitation types were used during the identification tests: (a) natural and ambient noise present in the laboratory; and (b) random impact excitation in space and in time with a hammer. The second excitation type was used because it was impossible to estimate accurately modal curvatures with ambient
excitation. The impact forces were about 2% of the specimen mass.
The long-term changes of the environmental effects inside the laboratory where also studied. Figure 10 presents the average values for ambient and surface temperature and relative air humidity measured for each DS. The average temperature values were about 17° C with a low increasing trend. The average humidity is about 69% with ±7% of maximum amplitude. Compared with the values obtained in the Mogadouro Clock Tower, it is expected for values of that order that the dynamic response of the wall does not change significantly and therefore it is assumed that the environmental effects inside the laboratory can be neglected in the subsequent damage identification analysis.
For damage detection, the analysis was carried out by observing frequency sifts in the natural resonant frequencies. In the case of the damage localization analysis, a group of damage methods was selected aiming at providing an adequate approach for analysis. The selected methods are (see Doebling et al., 1996, and Montalvao et al., 2006, for details):
• The Mode Shape Curvature Method (MSCM);
• The Sum of all Curvature Errors method (SCE);
• The FE Model Updating method (FEMU).
The selection of a group of methods can be discussible. Up to now, there is no single method which gives accurate results for damage localization and for all types of structural systems (Farrar and Doebling, 1998; and Choi et al., 2005). Therefore, the main issue is to obtain a wide perspective ofthe problem and conclusions on damage identification, taking into account that different methods provide different results. If significant damage is present in the structure, the results provided from different methods would converge in the identification, giving more confidence to the analyst.
All the selected methods have one common aspect: they all use spatial modal information of the structure, through mass-scaled or non-scaled mode shapes 0 and respectively (or/and through mass-scaled or non-scaled curvatures mode shapes cj>" and q", respectively). As damage is a local phenomenon, these quantities are useful to locate damage, especially the modal curvatures.
5.3.1 Global damage detection The SSI/Principal Component method implemented in the tool ARTeMIS (SVS, 2006) was used to estimate the modal parameters for all the fourteen DS. As an example of the frequency shifts along the DS, Table 3 presents for the Test Series C (the last series) the frequency results for ambient excitation. The frequency values are presented together with the value ±2om as a 95% confidence interval, and the frequency differences Am to the respective Reference Scenario (RS).
In general, the significant frequency decrease, i.e. higher that 2om (given in a grey box on Table 3), where around the DS where the crack was visually localized, with exception of mode 1 for Series C, as can be seen in Table 3.
To conclude about the global detection results, it seems that the modal properties of the masonry specimen are sensitive to the damage progress.
Was this article helpful?