Predicting Method of Rock Spalling around Deep Underground Caverns Based on Case Back Analysis

Rock spalling is a common phenomenon of local rock mass failure in large underground cavern projects under high geostress, which seriously threatens the stability of engineering and construction safety. Relying on the underground powerhouse projects on both the left and right banks of the Baihetan Hydropower Station in Southwest China, a numerical evaluation method for the depth of rock spalling based on case back analysis is proposed through the use of numerical simulation, parameter inversion, field testing, and case investigation. Firstly, by collecting and organizing field data, the typical distribution pattern of spalling during the excavation of the roof arch and sidewalls of the Baihetan left and right bank powerhouses is statistically analyzed. Secondly, by taking rock displacement and loosening zone depth as target variables, key rock mechanics parameters in different rock spalling segmentations are obtained through the combination of a large number of field test results and the genetic-neural network algorithm (GA-ANN). Thirdly, the numerical simulation and the evaluation index of rock fracture damage (RFD) are used to analyze the range and depth of the brittle failure zone of the surrounding rock after excavation based on the hard rock degradation model (RDM) applicable to deep rock engineering, and the results are compared with the observed rock spalling damage on-site. The results show that more than 91% of the rock spalling depths correspond to RFD thresholds ranging from 1.35 to 1.50. Finally, the threshold was used to analyze the excavation of the layer Ⅲ of the underground powerhouses on the left and right banks of the Baihetan Hydropower Station. The results showed that the predicted accuracy of the rock spalling failure depth reached over 88%, proving the good applicability of the model in practical engineering, indicating that the RFD can be used to effectively predict the depth of rock spalling failure. This study can provide important support for predicting the depth of rock mass failure in deep underground engineering.