A new method is proposed for predicting the in situ stress of local surrounding rock of large underground cavern group in high stress

various stress-induced fracturing

such as spalling of caverns and borehole breakouts are considered as the inversion information. This method quantitatively describes the location

depth or width of stress-type failure such as sheet wall and hole spalling of underground caverns

represent by the range of constant deviatoric stress greater than crack initiation stress of rockmass according to the elastic model calculation. The value or direction of local in-situ stress could be constrained based on the law of the measured geo-stress data

and the other in situ stress component are predicted by simulation modeling intelligent inversion method. This method is applied to predict the natural stress of rockmass near chainage 0+76 of the right bank underground powerhouse of Baihetan hydropower station. The predicted maximum principal stress is about 34 MPa

and the prediction rationality of in situ stress is verified by fracturing characteristics in other position through comparison of numerical simulation and field observation.