This study aims to establish an integrated safety risk control method for tunnel boring machine (TBM) construction that covers the entire construction scope of data acquisition, prediction, and optimization decision-making, which changes the previous decision-making process that relies on human experience and promotes the intelligence of TBMs.
This study introduces a digital twin cockpit that combines a prediction model and an optimization algorithm within a digital twin framework to tackle the problem faced by TBM operators in analyzing real-time monitoring information and reduce security risks. 3D modeling methods are used to generate the virtual model to further complete the interactive interface of the digital twin cockpit. The digital cockpit is equipped with Conditional Generative Adversarial Network (CGAN) and Non-dominated Sorting Genetic Algorithm-III (NSGA-III) to simulate the relationship between excavation parameters and safety indicators, and provide optimization suggestions for excavation.
The findings of this study indicate that (1) the proposed CGAN model can provide a reliable estimation of TBM excavation parameters with a high coefficient of determination (R-square) values reaching 0.94; (2) TBM optimization objectives can be effectively improved through the adopted NSGA-III algorithm, with an overall improvement of 27.07% and (3) the intelligent decision-making module of the digital twin cockpit can improve TBM process volatility and reduce extreme operating risks.
The main value of this study is a new intelligent control method that follows the integration of digital twins, machine learning and NSGA-III algorithms for the optimization of TBM excavation operations.
