Uncertainty analysis of mechanical response of surrounding rock in deeply buried tunnels using machine learning and global optimisation algorithms

Uncertainty is an inherent characteristic of geotechnical engineering. As deep rock mechanics continues to develop, the safety of the surrounding rock in underground engineering becomes increasingly crucial. This study puts forward a novel framework that integrates the AdaBoost algorithm, reliability method (FORM), and simplical homology global optimisation (SHGO) algorithm. Applied in the D2 laboratory of the second phase project of the Jinping Underground Laboratory (CJPL – II) in China, the framework proved effective. The AdaBoost algorithm forecast the rock failure degree (RFD) and displacement (DSP) of 224 monitoring points in the D2 laboratory, with a relative error of less than 10% compared to numerical simulation. Five variables for uncertainty analysis were determined via sensitivity analysis. The developed AdaBoost surrogate model calculated the reliability index of surrounding rock response. Compared with the traditional response surface method (RSM−FORM), the relative error was less than 5%. Further correlation analysis on these five variables showed that positive correlations enhance the reliability index of surrounding rock response, while negative correlations do the opposite. This study confirms the effectiveness of the proposed framework in uncertainty analysis of geotechnical engineering.