POD-based filters for the representation of random loads on structures

Random loads on structures are often represented as stationary Gaussian multi-variate random processes, whose components are correlated with each other according to characteristic patterns. Numerous techniques usually applied in structural dynamics require having the input constituted by a vector of independent random processes or even white noises. The application of such techniques in many practical contexts requires the introduction of suitable pre-filters to reproduce the exact correlation and harmonic content of the real external excitation. The present paper describes the development and the implementation in the time and in the frequency domain of digital filters based on the proper orthogonal decomposition (POD). Two different approximated approaches are discussed and verified through numerical examples.