Real‐time framework to detect leaks in industrial heat exchanger networks

Abstract The importance of detecting oil leaks in a refinery from a safety perspective is crucial. Nevertheless, undetected or unresolved oil leaks can lead to equipment failures, production disruptions, and costly repairs, which can significantly impact the refinery's overall safety, operations, and profitability. This study aims to leverage data‐driven multivariate statistical techniques to detect leaks in heat exchangers at a local refinery. The study developed a set of offline models using static Principal Component Analysis (PCA), dynamic PCA, static Projection to Latent Structures (PLS), and dynamic PLS. These models were then deployed online to monitor the heat exchanger's performance and promptly detect any future leaks in real‐time. To demonstrate the feasibility of the proposed approach, all the steps were implemented using a commercial software platform, specifically Aspen ProMV® from AspenTech. The results of the study demonstrated that static PCA, dynamic PCA, and dynamic PLS succeeded in detecting the leaks 50 days (with a 95% confidence limit) and 19 days (with a 99% confidence limit) before the facility operators found them. By implementing these data‐driven techniques, refineries can enhance their ability to automatically detect and address oil leaks promptly, improving overall safety, operational efficiency, and compliance with regulatory requirements