Model-free data-driven computational mechanics for deformation analysis of fractured rock masses

The deformation and stress distribution of fractured rock masses under external loads remain key scientific challenges, primarily due to their inherent heterogeneity, discontinuity, and anisotropy. To address this, this study proposes a model-free data-driven computational mechanics for deformation analysis of fractured rock masses, which can more accurately capture their mechanical response under external loading. This method bypasses traditional constitutive modeling by directly incorporating experimental data, including rock stress-strain data, joint normal stress-displacement data, and tangential stress-tangential displacement data. Solutions are obtained by minimizing the distance between these data points and those satisfying the conservation equations (i.e. equilibrium and geometric equations). Specifically, the fractured rock mass is regarded as a binary system consisting of rocks and joints, with the rocks represented by solid elements and the joints simulated through Goodman elements to capture discontinuous deformation behavior accurately. The validity and computational accuracy of the model-free data-driven computational mechanics for deformation analysis of fractured rock masses are verified by comparing the data-driven solution with a reference solution based on the constitutive model through typical numerical calculations. Furthermore, validation with real experimental data confirms that the method offers significant advantages in analyzing the nonlinear mechanical behavior of fractured rock masses, providing a unified and practical numerical approach for studying the deformation response of fractured rock masses.