Calibration of the dynamic analysis of Tellus 2D

This paper adds dynamic analysis to Tellus 2D, an agent-driven plane-strain geotechnical finite element program,and verifies and validates it before it is used in design. A dynamic phase integrates the equations of motion withthe Newmark scheme from the stresses of any static phase, with consistent mass, Rayleigh damping per material andviscous, free-field, tied and compliant boundaries. The program reproduces exact solutions of wave propagation anddamping within 3%, the surface motion of DEEPSOIL within 1.50%, the spectra of Strata for ten records with amean log ratio of 0.008, the SHAKE91 sample problem within 2.03% and the period of a building in the PLAXIS2D tutorial within 5%, although its static deflection under HS small is 27% larger than in the tutorial. At the GarnerValley downhole array, a one-dimensional analysis of 12 records over-predicts the surface spectra by a mean logratio of 0.326 at 2% damping, which falls to 0.234 when the damping is matched at the site frequency. PM4Silt,implemented from the equations of its report, reaches the critical-state strength exactly and the cyclic strengths ofthe FLAC implementation within a factor of about 1.5 in the number of cycles. Applied to Austrian Dam in the1989 Loma Prieta earthquake, it gives excess pore pressures of the measured magnitude and crest accelerations of0.58–0.72 g, close to FLAC’s 0.7 g, against 1.0–1.8 g for models without cyclic pore pressure, but the crest settlementcould not be compared, because the loose unsaturated fill began to flow in strong shaking, which the implicit solutioncannot follow. A 54 km section of the Bangkok basin, run through the 2025 𝑀𝑤 7.7 Myanmar earthquake on oneGPU in 90 minutes, reproduces the recorded velocity of about 10 cm/s and the amplification near 1 s, and behaves asone-dimensional columns within 10%.