This paper presents two case studies on the application of digital twins for predictive monitoring and anomaly detection in critical infrastructures: seismic monitoring for dams and sinkhole detection for railway tracks. The first case study focuses on the development of a surrogate model using neural networks to simulate seismic wave propagation and quantify uncertainties related to seismic hazards in dam infrastructure. By training a neural network on a dataset generated through the spectral element method, the model significantly reduces computational time and enables efficient global sensitivity analysis. This approach enhances the understanding of how key parameters influence seismic responses, supporting more reliable and responsive seismic monitoring systems for dams. The second case study addresses railway infrastructure, where sinkholes can lead to dangerous track deformations. Using high-resolution LiDAR data, we develop an automated system to detect ground deformations near railway tracks. By extracting ground points from LiDAR point clouds and generating Digital Elevation Models (DEMs), sinkholes are detected through a segmentation-based approach. Comparing different filtering algorithms, such as Voxel Based Ground Filtering (VBGF) and Cloth Simulation Filter (CSF), allows for optimizing the detection process and enhancing safety. Both case studies illustrate how digital twins can be employed for real-time monitoring, leveraging advanced data processing and predictive models to improve the safety and resilience of critical infrastructures.
Ababsa et al. (Fri,) studied this question.