A real-time intelligent classification model using machine learning for tunnel surrounding rock and its application

ABSTRACT

Real-time and accurate prediction of surrounding rock grade is crucial for tunnel dynamic construction and design. However, the internationally accepted semi-quantitative methods (e.g. rock mass rating (RMR), Q, and basic quality (BQ)) cannot provide fast and accurate classification in construction. This study proposed an intelligent surrounding rock classification method and a tunnel information management system, which can predict the surrounding rock grade in real-time and accurately. A database is collected with 286 cases in China, including seven geological parameters and surrounding rock grades. Based on different training parameters, 12 classification models are established using VGGNet, ResNet, and support vector machine (SVM) algorithms. The accuracy of the SVM classifier is 93.02%, which performs better than the VGGNet and ResNet classifiers. Moreover, precision, recall, F-measure, receiver operating characteristic (ROC), and 20-case verification show that the SVM classification model has greater robustness in learning and generalising for small and imbalanced samples. Additionally, a tunnel information management system is developed with cloud technology, which can accurately predict the surrounding rock grade within 10 s. Overall, the achievements of this study can provide valuable references for real-time rock mass classification in traffic tunnels and underground powerhouses.

KEYWORDS:

• Drilling and blasting tunnel
• surrounding rock classification
• surrounding rock geological parameters
• machine learning
• tunnel information management system

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by the Science & Technology Department of Sichuan Province, China [grant number 2021YFS0317]; and the National Natural Science Foundation of China (NSFC) [grant number U19A20111]; and the Science & Technology Department of Sichuan Province, China [grant number 2021YJ0041].