PSI - Issue 58

Dipendra Gautam et al. / Procedia Structural Integrity 58 (2024) 102–108 D. Gautam et al. / Structural Integrity Procedia 00 (2019) 000–000

103

2

1. Introduction System identification is the process of mathematical model building for a dynamic system from the recorded time series data. Both input-output and output-only system identification techniques are widely used across various branches of engineering. The fundamental application of system identification is confined to the characterization of dynamic properties; however, there is a growing trend in the use of system identification based damage detection as well. Ambient vibration based system identification is extensively used worldwide for nonparametric as well as parametric system identification applications (D’Ambrisi et al., 2012; Guler et al., 2008; Lacanna et al., 2019; Michel et al., 2018; Varum et al., 2017). Furthermore, ambient vibration based tracking of the dynamic characteristics through periodic testing is also reported by some researchers (Pereira et al., 2022; Vidal et al., 2014). Finite element model updating applications are also exemplified by several researchers using ambient vibration data (Bassoli et al., 2018; Foti et al., 2012; Sivori et al., 2020). Given the uncertainties in material characteristics of historical as well as common masonry structures, numerical models may observe unprecedented inaccuracies, leading to skewed modal properties. In this regard, structural system identification is used to juxtapose and calibrate the results from operational modal analysis and numerical modeling. Ambient vibration time series records are also extensively used to characterize masonry structures (Diaferio et al., 2016; Foti et al., 2014; Gallipoli et al., 2009; Moisidi et al., 2018; Rupakhety et al., 2022). Although a vast literature spectrum exists in the use of system identification, tracking dynamic properties to assess the fluctuations in dynamic characteristics to identify damage aggravation is not reported in the existing literature, to the best of our knowledge. We present a damage aggravation scenario in a large neoclassical masonry building that was damaged by the 2015 Gorkha, Nepal earthquake. From two measurements at the same location four years apart, the modal frequency fluctuation is detected using a state space parametric system identification approach. Damage aggravation is also identified using the correlation between the mode shapes. 2. Materials and methods The case study structure is the Ananda Niketan neoclassical masonry building currently used by the Institute of Engineering, Tribhuvan University. The building is a three storied building with a central courtyard and an N-S projecting extension, which is parallel to the Y direction in Fig. 1. The building is around 12 m tall Greco-Roman brick masonry in mud mortar building. Further details regarding the construction system and material properties of neoclassical buildings in Kathmandu can be found in the papers by Adhikari et al. (2019) and Gautam et al. (2023). The building has outward dimensions of approximately 35×42 m with an internal central courtyard of 16.5×20 m. The elongated wing has dimensions of around 30×13 m. The building was first constructed in 1892 and was affected by the great Eastern Nepal earthquake of 1934 (M w ~8.1). During the 2015 Gorkha earthquake sequence, the building sustained damage grade D3 per the European Macroseismic Scale (EMS-98) damage grading system (Grunthal, 1998). Although the building sustained a D3 damage level, no interventions were installed until May 2023. We took the first vibration records in May 2019 and the second in May 2023 using ETNA-2 triaxial accelerometers manufactured by Kinemetrics Inc., California. Time synchronization was done using externally connected global positioning system (GPS) devices. Time series records were taken for an hour in the same location. As time series records were taken in operational conditions, some transients were observed due to movements. Otherwise, no major disturbances were noticed during vibration testing. The time series records were processed using numerical algorithms for subspace state space system identification (N4SID) approach proposed by Van Overschee and De Moor (1994). The algorithm is a stable and noniterative parametric identification technique that uses oblique projections of the subspaces from the block Hankel matrix for modal parameter estimation. The algorithm is executed in MATLAB ® 2022b (MathWorks Inc., 2022). We used a Tukey window for signal tapering and a fourth-order Butterworth filter for signal filtration. A stabilization plot was made, and an optimal model order was selected. Due to space limitations, the algorithm/pseudocode is not presented in this paper. Interested readers are directed to the original contribution for further details. We also extracted mode shapes from the signals for both measurements and tested the modal assurance criterion (MAC) matrix (Allemang and Brown, 1982) for damage aggravation detection. The MAC value can be obtained as follows: