General Geometric Reconstruction Method of Failure Criteria for Lattice Spring Model

Failure criteria are crucial in numerical methods for effectively resolving material failures, especially in geomaterials such as rock and concrete. A wide array of criteria, designed for different rock types and complex stress conditions, has been developed. Although Artificial Neural Network (ANN)-based failure criteria present a unified solution for incorporating these diverse criteria into numerical models, their opaque nature and heavy data demands frequently curtail their practical utility and computational efficiency. To address these issues, this paper introduces a geometric-based method for reconstructing existing failure criteria. This novel approach not only simulates the functionality of ANNs but also offers greater interpretability, simpler visualization, and reduced data requirements. The effectiveness of this reconstruction method in representing established failure criteria is validated, and its utility in a four-dimensional lattice spring model (4D-LSM) for modeling true triaxial conditions is showcased through multiple numerical examples.