Abstract
Operational modal analysis is a robust and practical approach to structural health monitoring which assumes white noise as input. Therefore, the accuracy of this method can be compromised when dealing with colored unknown excitations, in which, for example, harmonic loads induced by the operation of mechanical equipment, may affect the modal parameter estimation. This study aims to address the challenge of identifying both natural and forcing frequencies of a complex building by exploiting the potentiality of both the spectral kurtosis analysis and stochastic subspace identification technique. The first one is based on the evaluation of a statistical quantity characterized by low values when data are stationary and Gaussian and high values when specific frequencies and nonstationarity are present in the signals. It allows the detection of harmonics, transients, and repetitive impulses in the frequency domain. Its combined use with the stochastic subspace identification technique enables us to effectively identify and separate harmonic‐induced vibrations from structural response to ambient white noise. This approach can lead to a more accurate modal parameter estimation that has been investigated in this work through numerical and experimental analyses carried out on the Cardinal Massaia hospital building in Asti, Italy. An experimental daily dynamic campaign has been carried out to acquire accelerations in operational conditions including disturbances due to machinery like elevators and air conditioners. The combined use of kurtosis analysis and stochastic subspace identification techniques has been used to process a large dataset of noisy measurements acquired in operational conditions. Five different measurement setups have been implemented, each one composed of 14 sensors. Notwithstanding the complexity of the case study under investigation for both the structural configuration and difficulties in the experimental data acquisition, this approach allowed to distinguish natural from forcing frequencies, highlighting its accuracy and robustness.