A Dimensionless Geomechanical Core to Inform ML-AI Prediction of Seismic-Induced Landslides

The Geomechanical Core (GMC) is a novel, scalable, and modular computational framework designed as the central engine component of hybrid AI systems for predicting and mapping seismic-induced landslides. GMC is founded on three pillars: a generalized failure model, a dimensionless constitutive model, and a serial modular architecture. The generalized failure model is a physically consistent adaptive function that distributes the initial shear strength over a normalized logarithmic shear-surface template, accounting for potential strength reductions caused by pre-seismic straining. At the heart of the engine, the constitutive model reproduces nonlinear elastoplastic stress–strain behavior through the relationship between normalized seismic load and straining (ductility ratio). This dimensionless formulation integrates the Hardening Soil (HS) approach for slope deformation and the Limit Equilibrium (LE) method for triggering and sliding, enabling the computation of both reversible and irreversible excess pore-water pressures under seismic excitation. The GMC chain of modules can be configured to represent different sectors of a landslide mass, where interacting boxes exchange static stress for initialization and dynamically update deformation states during shaking. GMC demonstrates high computational efficiency and aims to bridge the gap between large-scale predictive modeling and advanced constitutive-based numerical simulations.