Artificial Intelligence System Analyzes Risks in Real Time in TBM-Excavated Tunnels
Geprode, a company specialized in the anticipation of geological risks, has developed an innovative Brazilian software solution based on Artificial Intelligence (AI) and machine learning, designed for tunnel risk analysis. The system generates and continuously updates, in real time, a synchronized database integrating instrumentation data and TBM (Tunnel Boring Machine) parameters, also known as the “Tatuzão”.
The Georisco research group — focused on environmental dynamics, risks, and territorial planning — compared data from already excavated tunnel sections with information from pre-excavation zones, enabling the anticipation of ground behavior in sections yet to be excavated. This approach accelerates decision-making processes and significantly reduces geological risks.
Georisco integrates geotechnical instrumentation data into an advanced 24/7 monitoring system, optimizing excavation parameters and defining behavioral patterns based on results obtained from excavated sections, thereby complementing traditional geotechnical investigations. The methodology is grounded in advanced data analysis using supervised learning algorithms, specifically KNN (K-Nearest Neighbors).
Based on project-defined alert levels and data generated by instrumentation, the system is capable of predicting the exact moment when alert thresholds will be reached. Monitoring begins in the pre-excavation stage. Key variables considered include excavation speed, TBM advance rate, historical settlement patterns, and current settlement values in excavated sections. These data are processed through extrapolation and interpolation techniques to predict settlement for each excavation ring advance and to project its future evolution.
At each excavation ring cycle, instrumentation data are reviewed and compared with new records, continuously refining the predictive model to anticipate when alert values will be exceeded. This approach enables timely decision-making, enhances operational safety, and significantly reduces geological risks.
Advanced analysis identifies recurring patterns in ground settlements and excavation parameters, enabling real-time adjustments to construction plans that increase project efficiency. This approach transforms manual analysis into an automated process, massively integrating geotechnical data and TBM parameters in real time to predict ground behavior in the pre-excavation section.
In 2024, Georisco was the winner of the challenge “New Technologies for Geological Prediction Ahead of the Tunnel Face in TBM,” part of the ACCIONA INNOVATION #Startups Program, organized by the construction company. The group is currently being validated through a pilot project with ACCIONA itself, applied to the tunnels of São Paulo Metro Line 6 – Orange Line.
For this project, the defined excavation section is 21 meters (twice the TBM excavation diameter). Within this real operating environment, it is possible to further enhance and refine the model by training it with additional real-world data, thereby maximizing its predictive capability.
The group operates under human supervision, ensuring that recommendations are implemented correctly. This combination of human expertise and advanced technology provides a robust solution to the challenges of tunnel construction, particularly in urban environments, while helping to transform geotechnical instrumentation from a reactive approach into a proactive one—by quantifying what has occurred, comparing it with historical data, and predicting future behavior.
With the platform, costs are expected to be reduced by up to 20% in geological risk mitigation, 15% through reduced impacts within the area of influence, and 7% in geotechnical instrumentation. In the specific test case, there was an improvement in the efficiency of risk alerts, with a 17% reduction in the number of false alerts received. Additionally, of the total predictive alerts issued, 74% were confirmed as actual events.
View the magazine featuring the award-winning articles: CLICK TO VIEW!