A robust fault detection and diagnosis strategy for centrifugal chillers

This paper presents a robust fault detection and diagnosis (FDD) strategy for centrifugal chillers. The strategy consists of a model-based chiller FDD scheme and a sensor fault detection, diagnosis, and estimation (FDD&E) scheme, which handle chiller faults and sensor faults, respectively. The sensor FDD&E scheme uses a PCA-based method (principal component analysis) to capture the correlations among the major measured variables in centrifugal chillers, as it performs well even in the presence of typical chiller faults. The chiller FDD scheme has been developed based on six physical performance indices, which are capable of describing the health condition of centrifugal chillers and, thus, indicating chiller faults. Only after all the sensors whose measurements are crucial to the chiller FDD are validated by the sensor FDD&E scheme is the chiller FDD scheme used to conduct the chiller system FDD. The strategy was validated using laboratory data from ASHRAE RP-1043 and field data from a centrifugal chiller in a real building.