A model-based online fault detection and diagnosis strategy for centrifugal chiller systems

The paper presents an online adaptive strategy for the fault detection and diagnosis of centrifugal chiller systems. The strategy is developed based on six physical performance indexes. These performance indexes have the capability to describe the health condition of centrifugal chillers and particularly to account for existing chiller faults. A set of rules for faults and their impacts on the six performance indexes are deduced from theoretical analysis, and then serve as the fault classifier. The benchmarks of the performance indexes are provided by simplified reference models, whose parameter identification is simple. In addition, an online adaptive scheme is developed, by analyzing uncertainty coming from both model-fitting errors and measurement errors, to estimate and update the thresholds for detecting abnormal performance indexes. The FDD strategy is validated by both field data collected from a real building chiller system and by laboratory data provided by an ASHRAE research project.