Thermo-mechanical coupling analysis of APSE using submodels and neural networks

The ؤspِ Pillar Stability Experiment (APSE) is an in situ experiment for investigating the spalling mechanism under mechanical and thermal loading conditions in a crystalline rock. In this study, the thermo-mechanical behaviors in the APSE were investigated with three models: (1) a Full model with rough meshes for calculating the influence of tunnel excavation; (2) a Submodel with fine meshes for predicting the thermo-mechanical behavior in the pillar during the borehole drilling, heating, and cooling phases; and (3) a Thin model for modeling the effect of slot cutting for de-stressing around the pillar. In order to import the stresses calculated from the Full model to the Submodel and to define the complex thermal boundary conditions, artificial neural networks (NNs) were utilized. From this study, it was possible to conclude that the stepwise approach with the application of NNs was useful for predicting the complex response of the pillar under severe thermo-mechanical loading conditions.