Decoupling features and virtual sensors for diagnosis of faults in vapor compression air conditioners

This paper describes the development and evaluation of features and virtual sensors that form the basis of a methodology for detecting and diagnosing multiple-simultaneous faults in vapor compression air conditioning equipment. The features were developed based upon a physical understanding of the system, cost considerations, and heuristics derived from experimental data and modeling results. Virtual sensors were developed in order to reduce the cost of implementation. The validity of the features and virtual sensors was evaluated using measurements from a variety of different air conditioners tested in a laboratory environment. More detailed evaluation results are presented in separate papers.