Wavelet‐Based Stochastic Model for the Generation of Fully Nonstationary Bidirectional Seismic Accelerograms

A new stochastic methodology is proposed for the generation of bidirectional horizontal components of artificial, fully nonstationary, site- and spectrum-compatible seismic accelerograms through a nonseparable process. The model operates in the time-frequency domain and combines spectral representation techniques with signal processing tools. The basis of the methodology involves the generation of spectrum-compatible stationary artificial accelerogram signals whose nonstationarity is then modeled with a time-frequency modulating function that is based on a seed ground motion record. At the core of the proposed methodology lies the use of the Continuous Wavelet Transform (CWT). Specifically, the CWT method is used to perform time-frequency analysis and to define the nonstationary component. The proposed methodology provides any required number of seismic accelerograms whose temporal and spectral modulation is consistent with the characteristics of the site of interest. Furthermore, the model is extended to the case of pairs of bidirectional horizontal components. This is accomplished by probabilistically generating two orthogonal spectra whose geometric mean spectrum is compatible with a target spectrum. This procedure also takes into account the correlation structure of spectral acceleration pairs in orthogonal directions at different periods based on empirical models. The bidirectional components are then generated with the proposed site and spectrum-based methodology. An online tool that implements the proposed methodology is freely provided.