GENERATION OF SIMULATED EARTHQUAKE GROUND MOTIONS COMPATIBLE WITH MULTIPLE TARGET SPECTRA USING BAYESIAN OPTIMIZATION

This paper proposes an algorithm that utilizes Bayesian optimization to generate spectrum-compatible time history waveforms. This algorithm aims explicitly to prevent interference among mutual corrective waveforms. It dynamically adjusts the total number of corrective waves based on the expected improvement in the objective function. This capability effectively minimizes the objective function, even when generating simulated ground motions using observed phase characteristics. The algorithm is also designed to fit multi-damping target spectra accurately. The effectiveness of this approach is demonstrated through the generation of simulated ground motions compatible with four multi-damping target spectra using observed phase characteristics.