Cognitive Twin for Tunnel Rock Masses with DFN-Based Flow Prediction and Decision Optimization

• Grant number: 25/24641-2
• Support Opportunities: Scholarships in Brazil - Post-Doctoral
• Start date: May 01, 2026
• End date: April 30, 2029
• Field of knowledge: Engineering - Civil Engineering - Geotechnical Engineering
• Principal Investigator: Marcos Massao Futai
• Grantee: Lucas Bellini Machado
• Host Institution: Escola Politécnica (EP). Universidade de São Paulo (USP). São Paulo , SP, Brazil

Abstract

Underground infrastructures are strategic assets for economic development and social well-being. Operational interruptions or failures in these structures can lead to significant economic, human, and environmental losses. Among them, tunnels exhibit singular behavior, as their performance depends directly on the surrounding rock mass, which simultaneously acts as a structural element and an aggressive medium, constantly subjected to water percolation. This interaction governs the stability, durability, and life cycle of the asset. In the specific case of tunnels excavated in fractured rock masses, the occurrence of water inflow is directly related to preferential flow paths defined by the fracture network and its interaction with the excavation. The use of emerging technological tools that enable understanding and monitoring of these systems is essential to mitigate risks and maintain structural performance. However, the current management of underground assets still relies on outdated practices that fail to reflect the complexity and importance of such structures. This research proposes the integration between discrete modeling and machine learning for the creation of Cognitive Digital Twins, which, beyond connecting the virtual and real assets, are capable of processing and optimizing data intelligently. In the first stage, advanced Discrete Fracture Network (DFN) modeling will be employed based on real data obtained from Terrestrial Laser Scanning (TLS) and geological mapping, combined with FISH codes in the 3DEC software to generate stochastic configurations optimized for connectivity and water flow, based on discharge calculations. In the following stages, the simulation results will be integrated into parametric and procedural modeling environments (Rhino/Grasshopper), enabling the application of machine learning (ML) algorithms to provide adaptive intelligence to the system. This approach will allow the creation of a Cognitive Digital Twin of the rock mass capable of correlating observed hydraulic behavior with stochastic fracture generations that best represent the real tunnel response, adjusting model parameters and optimizing support performance. (AU)