Translation processes for wind pressures on low-rise buildings

Abstract

Wind pressure time series on low-rise buildings are characterized by translation processes calibrated to the wind tunnel data from facilities at University of Florida. Translation processes are nonlinear transformations of Gaussian processes which match exactly and approximately the target marginal distributions and covariance functions, respectively. Since the wind pressure data under consideration cannot be well fitted by standard distributions, we use mixtures of distributions with different properties in the central and tail regions. It is shown that the translation processes with the mixtures of distributions are robust and efficient models for wind pressure time series. Moreover, the models can be used as simulation tools to generate synthetic wind pressure data at a large number of pressure taps, whose sample properties are consistent with those of the experimental records.