Time-dependent estimators for on-line monitoring of full-scale structures under ambient excitation

• New bias-compensated estimators were proposed for the identification of time-varying systems.
• An output-only procedure was tailored to the spectrogram representation to provide on-line estimates of instantaneous frequency, damping and amplitude.
• Correction curves for spectrogram-based damping estimators were derived from a comprehensive set of Monte Carlo simulations.
• An experimental application to a full-scale masonry structure was presented.

The study described in this paper revisits the concept of instantaneous identification of system parameters based on time–frequency representations. In order to overcome the distortion caused by the time–frequency analysis window, optimal bias-compensated estimators are introduced. In particular, bias-compensated estimators are conceived specifically to provide on-line estimates for time-varying linear parameters such as instantaneous frequency and damping. With respect to previous studies, which relied on the concept of optimal time–frequency representation, the novel procedure corrects on-line estimates provided by standard representations. Afterwards, a practical application to a full-scale structure is presented. The church of “Madonnina della Neve” in Savigliano (Cuneo province, Italy) is a masonry building that exhibits defects in the connections between structural parts and visible cracks in lateral masonry walls. As a consequence, the global behaviour observed on this structure demonstrates significant flexibility in both the longitudinal and transversal directions. Recently cracks have worsened due to vibrations induced by traffic and heavy vehicles. Proposed bias-compensated estimators were used to analyse the unusual time-varying behaviour of the masonry structure, as observed from the response measured during the transit of vehicles and trucks.