Leveraging AI for Refinery Leak Detection

A US oil and gas company operates a vast network of refineries nationwide that are responsible for processing and distributing fuel. Small leaks in equipment at this scale can quickly translate into significant product loss, operational disruption, environmental exposure, and regulatory scrutiny. Existing monitoring processes were constrained by hardware limitations and fragmented detection capabilities, resulting in field teams often being alerted only after a leak had begun and several gallons of product were already lost. The client needed a reliable, automated, and scalable solution capable of detecting leaks earlier without requiring costly hardware overhauls.

The Aimpoint Digital team developed and implemented a robust computer vision solution using Azure Machine Learning (AML) to enable proactive leak detection at enterprise scale. The system analyzes live video feeds at both the edge (on-site IoT devices) and in the cloud, leveraging a hybrid approach that combines state-of-the-art deep learning–based computer vision models with classical computer vision techniques to accurately detect the visible leak events.

To ensure long-term stability, we also implemented robust monitoring and evaluation frameworks to continuously measure model performance, track detection accuracy, and enable iterative improvement at enterprise scale. The solution was packaged for scalable IoT deployment, allowing seamless rollout across hundreds of facilities.