Intelligent dynamic control of shield parameters using a hybrid algorithm and digital twin platform

Abstract

This paper presents a digital twin (DT) platform integrated with an online optimization algorithm that combines Bayesian Optimization (BO), Categorical Boosting (CatBoost), and the Nondominated Sorting Genetic Algorithm (NSGA)-III. The platform enables multi-objective dynamic optimization of shield parameters under varying geological conditions. Using the Wuhan Metro as a case study, the effectiveness of the method is validated. The results demonstrate that: (1) the DT model accurately estimates shield machine performance, with an R² of no less than 0.957 on the test set across three geological conditions; (2) the online optimization significantly enhances shield machine performance, with a comprehensive optimization improvement of over 25 % across all conditions; (3) comparison of the constructed algorithm's accuracy, along with Shapley additive explanations, confirms the accuracy, interpretability, and universality of the proposed algorithm.