Using damping ratios inferred from accelerations recorded during earthquakes in instrumented buildings in California this paper evaluates the Rayleigh damping model commonly used in seismic analysis of buildings. Damping ratios were obtained from 119 seismic responses, coming from 24 buildings, measured in 46 earthquakes over the past 25 years by the California Strong Motion Instrumentation Program. All records were analyzed using a least-squares parametric system identification technique in the time domain. Using only damping ratios judged to be reliable, the variation of damping with modal frequencies is studied for all buildings. It is found that in 96% of the cases, modal damping ratios increase approximately linearly with frequency; showing that damping in buildings is best represented by using a model consisting of a constant damping ratio plus a stiffness-proportional damping ratio. No evidence was found to suggest that a mass-proportional model nor a combination of mass- and stiffness proportional damping is appropriate. It is shown that the Rayleigh damping model, when used in combination with assumptions commonly used in engineering practice, typically leads to overestimations of mean peak floor accelerations, peak floor velocities, and interstory drift ratio demands with respect to those computed using the damping model inferred from the recorded data. (C) 2017 Elsevier Ltd. All rights reserved.
