Transient pattern analysis for fault detection and diagnosis of HVAC systems

Abstract Modern building HVAC systems are complex and consist of a large number of interconnected sub-systems and components. In the event of a fault, it becomes very difficult for the operator to locate and isolate the faulty component in such large systems using conventional fault detection methods. In this study, transient pattern analysis is explored as a tool for fault detection and diagnosis of an HVAC system. Several tests involving different fault replications were conducted in an environmental chamber test facility. The results show that the evolution of fault residuals forms clear and distinct patterns that can be used to isolate faults. It was found that the time needed to reach steady state for a typical building HVAC system is at least 50–60 min. This means incorrect diagnosis of faults can happen during online monitoring if the transient pattern responses are not considered in the fault detection and diagnosis analysis.