ยป
Predictive
maintenance model for cooling water heat exchangers
Fouling
This technical paper presents a prediction approach for cooling water heat exchanger healthiness or cleanliness that aims to facilitate proactive maintenance in industrial operations. Traditional maintenance approaches which rely on time-based measures, resulting in unwarranted repairs and cost.
Koay Tze How, Nor Naqiah M Nazari Anuar, Lim Beng Lin, Sy M Hafiz B Al-Idrus, PETRONAS
To address the outlined issue, a prediction model for cooling water heat exchangers is proposed, leveraging machine learning algorithms to analyse 1-year historical data and predict the health condition of the heat exchangers. The model utilizes diverse parameters, which are heat transfer coefficient, heat transfer area, maximum skin temperature and temperature inlet hot fluid. These data are collected from the heat exchanger in the specific business unit and used to develop a prediction model. The results showcase a promising method for forecasting the heat exchangers health status, offering valuable insights for proactive maintenance, and preventing unforeseen failures. This paper discusses on the methodology used for developing the prediction model, the utilization of the confirmation model based on inspection reports, and the subsequent implementation of the model. This study serves as a foundation for future research and development efforts towards the implementation of an advanced and robust heat exchanger healthiness prediction model for the industry.
Introduction In petrochemical plants, various equipment like heat exchangers operate simultaneously, continuous monitoring is crucial to prevent unplanned shutdowns that could cost millions in revenue losses. Digital twins for cooling water( CW) heat exchangers are being implemented, utilizing simulation and revalidation studies to monitor operating conditions. This helps the plant to manage maintenance and replacement effectively.
Digital twin heat exchanger healthiness monitoring Oil and gas industry lacks digital twin platform for heat exchangers without hard sensors. Current monitoring methods are limited and often inaccurate, relying on manual calculations. A new digital twin is needed. It is developed with integration of ICON Symmetry( ICON) for process simulation, Heat Transfer Research, Inc( HTRI) for thermal analysis, and Microsoft Azure Machine Learning( Azure) for fouling prediction. ICON simulates the cooling water network, HTRI uses ICON process parameters to calculate heat exchanger thermal performance, and Azure model predicts fouling based on HTRI output results such as skin temperature, fluid temperature, velocity, etc. All data feeds into a dashboard for real-time heat exchanger healthiness monitoring.
HTRI automation The Digital Twin leverages HTRI automation as its core engine. It efficiently processes extensive equipment operational data simulated from ICON. This automation enables real-time analysis and instantaneous results generation, facilitating continuous equipment health monitoring and fouling prediction capabilities. This automated approach represents a substantial improvement in operational efficiency, eliminating
Figure 1. Heat exchangers monitoring methodology. www. heat-exchanger-world. com Heat Exchanger World July 2025
45