A data-driven method for the deformation analysis of layered rocks

Abstract

This paper proposes a data-driven method for the deformation analysis of layered rocks, which consists of generating a stress–strain database and using a data-driven computational solution. The method does not require defining the material's constitutive relationship to conduct analysis of layered rock deformation under loading of the same material. First, the data-driven identification (DDI) algorithm infers and builds a stress‒strain database of the material based on the strain field and loading force. Then, this database is used to calculate the response of the same material structure with arbitrary geometry and boundary conditions using data-driven computational mechanics (DDCM). The specific workflow of the method is demonstrated, and the computational accuracy and reliability are verified through an experimental application example. The method naturally combines the DDI algorithm and the DDCM solver, providing a new concept for analysing the deformation of layered rocks. Through this method, it is possible to conduct more accurate deformation analysis of layered rocks without defining their constitutive relationships. This has significant engineering application value in the design of excavations for layered rock slopes, foundations, and underground caverns.