PSI - Issue 78

Marco Martino Rosso et al. / Procedia Structural Integrity 78 (2026) 301–308

304

acceleration (PGA) expressed in g units respectively of 0.046 g, 0.083 g and 0.086 g, reaching a scaled level of 25% of the entire waveform. The subsequent applied Irpinia shaking were scaled with factors of 50%, 75%, 100% and 125%, finally inducing structural damages due to high earthquake input excitation levels. On the other hand, the imposed white noise excitation consists of Gaussian noise with standard deviation around 0.01 g and a quite constant power spectral density of about -60dB in the range of 0-50 Hz. The initial white noise 2 and 3 allowed for performing experimental dynamic characterization of the initial conditions of the buildings, whilst the subsequent white noise histories were expected to provide a change in modal properties due to progressive and accumulated damage in the RC frames under testing.

Table 1. Excitation input protocol: sequence of scaled Irpinia waveform and white noise shaking inputs.

Day of Test

Shaking table test sequences

PGA [g]

Dev. Std. [g]

Duration [s]

27/06/2022

Irpinia Ladder 25% (0.046; 0.083; 0.086)

-

24·3=72

White Noise 2 White Noise 3 Irpinia 50% WhiteNoise4 Irpinia 75% White Noise 5 Irpinia 100% White Noise 6 Irpinia 125% White Noise 7

0.070 0.064 0.202 0.067 0.376 0.076 0.440 0.068 0.616 0.079

0.011 0.011

160 320

-

24

0.009

160

28/06/2022

-

24

0.011

160

-

24

0.012

160

-

24

0.011

160

Data collected with MonStr devices were provided through an HDF5 database. The acquisition system was set based on unit-based asynchronous sampling, meaning that every sensor works based on its own local clock and stores the absolute time in order to resynchronize all the sensors to the same common reference. However, due to technical reasons, sensors were placed with different local orientations of the three axial accelerometers. Therefore, the preprocessing of data is required to express all considered acceleration signals according to the common global reference system to ensure consistency across channels. The nominal sampling rate of MonStr devices is 1 kHz; however, since the chosen asynchronous acquisition procedure, all the sensors may start reading with a small delay, thus providing sometimes an uneven number of samplings and not perfectly uniformly timely-spaced distributed across the channels. Therefore, another required pre-processing step was to interpolate all channels numerically to a uniform, synchronized 1kHz time grid. According to PalChaudhuri et al. (2004), in order to respect the causality property of a system in the ordering of the acquired event, it would be advisable to synchronize all sensors to the fastest clock in order to maintain partial ordering of the event despite interpolating data. Moreover, in order to reduce the noise introduced by the interpolation procedure, the data were resampled to a time grid double of the original sampling frequency in order to avoid losing peaks of the acquired signals, and then decimated back to the actual sampling rate of 1 kHz, jointly with the application of an FIR anti-aliasing filter. 3. Results and Discussion The two RC frames were subjected to Irpinia earthquakes of increasing intensity to analyze the progression of damage in the two structures using a comparative approach based on OMA conducted over the alternate White Noise excitations. In order to first characterize the baseline modal parameters of the RC building specimens, a finite element model (FEM) was implemented using the Scientific Toolkit for OpenSees (STKO) software. The model was implemented using forceBeamColumn elements for the RC pillars, 4-node thick shell ASDShellQ4 elements for both the infills and the horizontal RC solid slabs, and truss elements for the short side steel tube bracing system. The