A fault-detection method is proposed in this article to detect system-level incipient faults in HVAC&R systems. This method adopts performance indexes to indicate the health statuses of subsystems of HVAC systems. The support vector regression algorithm is used to develop the reference performance index models. The exponentially weighted moving average control chart is introduced to detect the residuals between the measured performance indexes and benchmark performance indexes. Validations are made on a simulation platform for a super-rise commercial building in Hong Kong. Three typical subsystems are considered in this study, i.e., the cooling tower system, chillers, and heat exchanger system. Results show that significant improvements are achieved on detecting incipient faults. The support vector regression based performance index models have higher accuracies than that of multiple linear regression based models. The exponentially weighted moving average based fault-detection method shows superior capacity for detecting incipient faults compared with t-statistic-based methods.
ASJC Scopus subject areas
- Building and Construction