An intelligent framework for predicting excavation deformation risk in tunnels incorporating rockmass classification and cloud-model-enhanced modeling

Intelligent drilling-and-blasting tunneling is a key direction in tunnel engineering. Traditional decision-making modes rely on preliminary geological investigation and empirical judgment, and therefore they are unable to quantitatively assess the deformation risks of surrounding rock. Moreover, these modes are deficient in technical means for rapid acquisition of geological information, as well as for efficient analysis and dynamic decision making with the assistance of finite element methods and artificial intelligence algorithms. This paper proposes a computationally efficient intelligent framework that integrates rockmass classification with cloud-model-enhanced inference. On this basis, through the introduction of uncertainty analysis methods, an automated computational path is successfully established. The framework comprises three modules: (i) a machine-learning classifier using Measurement While Drilling data for rock zoning, (ii) a cloud-model–Gaussian inference engine accelerated by NUMBA for rapid parameter estimation, and (iii) a deformation-based prediction model for risk assessment. Case studies on tunneling projects validate the framework. An accuracy of 87.98% in rockmass classification is achieved by the proposed decomposition method of drilling parameters. A fine interpolation from sparse boreholes to a range of twice the tunnel diameter is accomplished by the uncertainty analysis process based on a cloud-model, and the generation speed of multi-parameter fields is increased by more than ten times through the combination with Numba computation. Furthermore, the XGBoost model based on SE-CNN achieved an accuracy rate of 94.5% in deformation prediction. By linking rockmass information to deformation prediction, this technical framework enhances the reliability of deformation assessment and risk prediction in tunnel construction.