Digital twin with data-mechanism-fused model for smart excavation management

Abstract

The accurate assessment and effective management of deep excavation risk have faced longstanding challenges due to the highly complicated and uncertain construction process. A digital twin, designed with the data-mechanism-fused (DMF) physical and virtual models, is developed to solve problems by integrating Building Information Modeling (BIM), data mining (DM), and physical mechanisms. In the DMF physical model, a mechanical model is embedded into the digital twin to implement real-time interaction and inversion between field-measured and simulated data, thus revealing the evolution law of mechanical properties and creating a multi-source DMF database. In the virtual model, the random forest (RF) regression is applied to fully learn the multisource database and accurately predict retaining wall behaviors on behalf of excavation risk. The proposed digital twin facilitates practical applications to imitate physical construction process, predict excavation-induced behavior, and realize closed-loop risk management with a high degree of automation, intelligence, and reliability.