Developing a digital twin for dam safety management

Abstract

A dynamic, reliable, and even automatic evaluation of dam performance (e.g., deformation) is essential for safety management of earth and rockfill dams. The digital twin has emerged as a promising tool for smart dam safety management. Many countries have launched projects to develop digital twins of dams, which represent a living model that can continuously learn from monitoring data from the physical counterpart and produce a complete and timely description of the dam’s performance. A practical framework is proposed in this study to develop digital twin of an earth or rockfill dam for predicting its mechanical responses. The proposed framework utilizes both a physics-based dam model and monitoring data to enhance the model’s performance through Bayesian updating. It is illustrated by an operational digital twin project, i.e., the Danjiangkou Digital Twin Project for developing the digital twin of the core-wall rockfill dam at the right bank. The proposed method enables three novel features for dam safety management, namely, real-time simulation, future forecast, and scenario projection of dam performance. The results showed the updated dam model predicted the dam crest settlements accurately with an RMSE as small as 4.20 mm, verifying the effectiveness of the proposed framework.