A simulation based fault diagnosis strategy using extended heat flow models (HFM)

Because of the complexity and diversification of today’s HVAC systems, Fault Detection and Diagnosis (FDD) systems have become necessary to reduce maintenance cost and to provide building energy efficiency, but with a minimum of engineering cost. Based on a generic method of generating Fault Detection (FD) systems from Building Information Models (BIM), we now propose an  extension of the underlying Heat Flow Model (HFM) to implement a diagnosis engine and thus create the complete software system called HFM-FDD. The diagnosis uses an associative network to map the dynamically reported failure rule vectors to a small set of probable faults. The associative network is automatically created at every time-step through fault simulation that takes the current conditions such as outdoor temperature, set-points, and occupancy into account. While keeping the engineering costs low, the shown diagnosis result is very promising.