Data‐driven 1D wave propagation for site response analysis

Prediction of ground motion triggered by earthquakes is a prime concern for both the seismology community and geotechnical earthquake engineering one. The subfield occupied with such problem termed site response analysis (SRA), its one-dimensional flavor (1D-SRA) being particularly popular given simplicity. Despite simple geometrical setting, paramount challenge remains when it comes to numerically consider intense shaking in soft upper soil strata crust: how mathematically model high-strain, dissipative, potentially rate-dependent behavior. Both heuristics models phenomenological constitutive laws have been developed meet challenge, neither them exempt either numerical limitations or physical inadequacies both. We propose herein bring novel data-driven paradigm bear, thus giving away need construct behavior altogether. Data-driven computational mechanics (DDCM) solid that gaining popularity; particular, multiscale version relies on studying microstructure (in case soil, representative volumes containing grains) populate dataset later used inform at macroscale. This article presents first application DDCM 1D-SRA: first, we demonstrate capacity handle wave propagation problems using discrete datasets, obtained via sampling grain ensembles element method (DEM), lieu law then apply specifically analyze harmonic waves deposit overlies rigid bedrock. this displays elastic yet non-linear depth-dependent due evolving overburden pressure. validate implementation comparison regular finite elements analyses (FEA), discuss benefits DD, traditional amplification functions are recovered DDCM. Finally, number future work opportunities lie ahead proof-of-concept, chiefly terms site-specific studies incorporating more complex realistic traits project was open-source software relevant code freely made available other researchers.