SubseaAI — Real-Time Pipeline Leak Detection

Pipeline leaks in underwater oil and gas systems are often detected after days or weeks due to slow and expensive inspection methods. These delays lead to massive financial losses and environmental damage. We wanted to build a system that can detect leaks instantly and continuously without manual intervention. What it does SubseaAI is an AI-powered pipeline monitoring system that detects leaks in seconds using real-time multi-sensor data. It combines temperature, pressure, and acoustic signals to identify anomalies and generate high-confidence alerts. The system also pinpoints the exact leak location, enabling faster response and prevention. How we built it We developed a full-stack system combining simulation, AI models, and a live dashboard. A physics-based simulator generates realistic sensor data (temperature, pressure, acoustic) AI models analyze anomalies using Isolation Forest Dempster-Shafer theory is used for multi-sensor fusion A Kalman filter is used to estimate and track leak location FastAPI backend handles APIs and real-time communication A browser-based dashboard visualizes all sensor data and alerts Challenges we ran into Simulating realistic physical behavior for multiple sensors Combining different sensor outputs into a single reliable decision Maintaining real-time performance with multiple data streams Designing a system simple enough to demonstrate clearly What we learned Multi-sensor fusion significantly improves detection accuracy Real-time systems require careful balance between speed and reliability Clear visualization is critical for understanding complex systems Future scope Integration with real hardware sensors Deployment in offshore environments Scaling to monitor long-distance pipeline networks

Built With

• python