AI Could Predict Infrastructure Failures Weeks in Advance

• Proeksa Meas
• 11 September 2026
• Construction Focus
• 796 Views

AI-powered analytics are helping infrastructure engineers move from reacting to failures toward predicting problems before they become emergencies, according to Vahid Abdollahi, an Applied AI Scientist at Bentley Systems.

Traditional monitoring often alerts operators only after a measurement crosses a fixed threshold. AI, however, can analyze data from multiple sensors to identify unusual patterns and potentially provide four to six weeks of advance warning of dangerous conditions.

The process begins with reliable sensor data. AI systems must first identify faulty, drifting, or abnormal readings before analyzing genuine structural changes. Techniques such as Seasonal-Trend Decomposition can then separate normal seasonal behavior from unexplained changes that may indicate structural problems.

By analyzing relationships between sensors, AI can also distinguish equipment faults from real events. For example, changes in rainfall, ground pressure, and movement can reveal potential problems such as internal erosion before significant movement occurs.

Abdollahi emphasizes that AI is designed to support engineers, not replace them. Feature-importance analysis can show why a model has identified a risk, while generative AI can turn complex sensor data into clear, prioritized recommendations for engineers.

Bentley Systems’ iTwin IoT brings these capabilities together to help engineering teams reduce alarm fatigue, identify emerging problems earlier, and improve the safety, resilience, and longevity of critical infrastructure.

Ready to transform how you manage and monitor your infrastructure? To learn more about iTwin IoT, visit the iTwin IoT page

• Bentley
• iTwin
• AI
• IoT
• Bentley Systems