Towards predictive maintenance: A performance evaluation framework for cooling towers in HVAC systems

Cooling towers are essential for heat rejection in heating, ventilation, and air conditioning (HVAC) systems, especially for large scale buildings, but have historically received less research attention compared with other HVAC equipment. However, with the increasing adoption of “multi-tower – multi-pump – multi-chiller” configurations and the widespread integration of variable frequency drives (VFDs) in cooling towers and condenser water pumps for the purpose of energy saving, the demand for the operational optimization and predictive maintenance of cooling towers has grown significantly. To address this need, this study proposes a performance evaluation framework toward predictive maintenance, integrating both physics-informed and data-driven approaches. The framework enables in situ thermal performance assessment and early detection of potential degradation using operational data, without requiring system shutdowns. Maintenance decisions are guided by two key components: (1) characteristic curves derived from Merkel theory, serving as a benchmark for performance evaluation, and (2) model predictive accuracy and prediction interval (PI) reliability metrics, which indicate performance degradation and potential maintenance benefits. The proposed framework was validated using real-world operational data from a data center cooling plant. Compared to existing evaluation methods, it eliminates the reliance on complex lookup tables, interpolation, or intrusive testing. By providing a scalable and practical solution, the framework supports energy-efficient and reliable cooling tower operation, and serves as a foundation for predictive maintenance deployment.