Intelligent Grading of Surrounding Rock Based on Multisource Heterogeneous Data Fusion
Jiayao Chen, Hongwei Huang, Mingliang Zhou · 2025
Accurate assessment of surrounding rock quality is essential for ensuring stability in tunneling projects. Traditional rock mass classification systems integrate engineering observations, measurements, and expert judgment to evaluate rock mass quality. However, their applicability is often challenged by the complexity and uncertainty inherent in tunnel excavation. This study constructs a multisource heterogeneous database to comprehensively characterize rock mass features and proposes a data-driven machine learning approach to benchmark the rock mass rating (RMR) system. A 13-dimensional dataset is developed, incorporating key rock mass attributes, with RMR values serving as the target output. Four machine learning models—classification and regression tree (CART), multilayer perceptron (MLP), gradient boosting regression tree (GBRT), and random forest (RF)—are employed to model the relationship between input features and rock mass quality. To enhance model performance, hyperparameter optimization is conducted using the tree-structured Parzen estimator (TPE) algorithm. Comparative analysis of model performance enables the identification of the most effective predictive framework while ranking feature importance. The proposed methodology offers a robust, data-driven approach for rock mass classification, facilitating more accurate and efficient assessments in tunneling engineering.