Term 252

AI-Powered Multi-Gas Detection System for Early Corrosion Detection in Refinery Pipelines

Project Type: Self-Initiated

Project Description

Refinery piping systems are vulnerable to corrosion, fouling, and gas leakage, which can lead to safety hazards, environmental risks, and unplanned shutdowns. Early detection of degradation mechanisms remains a critical industrial challenge. This project presents the final design of a compact multi-gas detection and AI-assisted diagnostic system for real-time refinery piping condition monitoring. The system extracts a controlled gas sample from refinery pipelines and measures the concentrations of hydrogen sulfide (H2S), carbon dioxide (CO2), methane (CH4), and oxygen (O2). These measurements are processed through a structured data pipeline and analyzed using an XGBoost-based machine learning model to classify pipeline condition and estimate Remaining Useful Life (RUL).