Damping identification under non-ideal conditions: Performance assessment of the random decrement technique ranked by peak amplitude

Damping identified from response measurements of high-rise buildings often exhibits significant scatter, posing a challenge for reliable interpretation and practical use. The Random Decrement Technique (RDT) stands out for its robustness and its ability to identify amplitude-dependent damping by adjusting the triggering condition. This study investigates the performance of the RDT ranked by peak amplitude (Peak RDT) under conditions deviating from the method’s mathematical assumptions. The analysis considers constant and amplitude-dependent damping models as well as stationary and non-stationary excitation conditions. The identified damping values are assessed against a known input. Numerical studies employing a Monte Carlo approach are conducted to generate distributions of the damping ratio, from which accuracy and precision are evaluated. Under stationary loading conditions, the mean damping is accurate for the linear system across all relative amplitudes. For the non-linear system, the accuracy of mean damping strongly depends on the length of the segments considered. Sampling long segments results in an overly linearized damping function. For both systems, the precision of damping estimates is significantly reduced at low and high relative amplitudes due to poor approximations of the system’s free decay at these amplitudes. Three types of non-stationarity excitations are considered: time-varying mean, time-varying variance, and their combination. Due to the underlying assumptions of the Peak RDT, damping values derived from response data exhibiting a time-varying mean result in significant accuracy errors in both linear and non-linear systems. In contrast, the method demonstrates robustness when applied to response data with time-varying variance.