Data-driven and association rule mining-based fault diagnosis and action mechanism analysis for building chillers

•We proposed a method for both fault diagnosis and action mechanism analysis in building chillers.•The method is developed based on the classification based on association (CBA) and association rule mining.•Seven common faults of chiller are well identified with an overall diagnosis accuracy of 90.15%.•The mined diagnostic rules match well with domain knowledge and the fault action mechanisms are concluded.•The discrepant rule analysis can provide proper reference for multiple faults decoupling. Developing advanced fault detection and diagnosis (FDD) techniques for building chillers is becoming increasingly essential for building energy saving. Previous FDD studies have mainly concentrated on the model performance, while fewer studies have examined the chiller fault action mechanism. This paper, therefore, proposes a method that can conduct both fault diagnosis and fault action mechanism explanation of building chillers. The method is data-driven-based and can be trained by system operational data based on the classification based on association (CBA) algorithm. Also, it is qualitatively based because system operational rules can be extracted from the diagnostic model by association rule mining. The fault diagnosis process and the fault action mechanism on the chiller system can be then understood by rule interpretation. The experimental chiller data of ASHRAE RP-1043 is used to validate the effectiveness of the proposed method, and the results show that the CBA-based fault diagnosis model can well identify seven common chiller faults with an overall diagnostic accuracy of 90.15%. In this work, the key rules of each fault are extracted and visualized. The mined rules can be well interpreted by domain knowledge, and the action mechanisms of seven faults are concluded. Moreover, the discrepant rule analysis can provide a proper reference for multiple fault decoupling. The knowledge discovered from the fault diagnosis process is valuable for the development of FDD researches and shortcuts for field application.