Diagnostic Bayesian networks for diagnosing air handling units faults - part I : faults in dampers, fans, filters and sensors

Abstract

Faults in air handling units (AHUs) affect the building energy efficiency and indoor environmental quality significantly. There is still a lack of effective methods for diagnosing AHU faults automatically. In this study, a diagnostic Bayesian networks (DBNs)-based method is proposed to diagnose 28 faults, which cover most of common faults in AHUs. The basic idea is to fully utilize all diagnostic information in an information fusion way. The DBNs are developed based on a comprehensive survey of AHU fault detection and diagnosis (FDD) methods and fault patterns reported in three AHU FDD projects including NIST 6964, ASHRAE projects RP-1020 and RP-1312. The study is published in two parts. In the Part I, the methodology is described firstly. Four DBNs are developed to diagnose faults in fans, dampers, ducts, filters and sensors. There are 10 typical faults concerned and 14 fault detectors introduced. Evaluations are made using the experimental data from the ASHRAE Project RP-1312. Results show that the DBN-based method is effective in diagnosing faults even when the diagnostic information is uncertain and incomplete.