Model validation, fault model, FDD, fault detection and diagnosis
A methodology for evaluating the performance of fault detection and diagnostics (FDD) tools for unitary air-conditioners has been developed (Yuill and Braun 2013). The methodology uses laboratory measurements of systems with and without faults to test FDD tools’ effectiveness. A gray box modeling method capable of modeling systems with faults was developed by Cheung and Braun (2013a and 2013b) to provide input data, as an alternative to using laboratory data that had been collected. The simulation method was validated by direct comparison with experimental data, but a comparison of FDD evaluation results provides a more direct and useful validation of the model for its intended purpose. Eight different systems have been modeled using Cheung and Braun’s method. Six FDD tools were evaluated using both experimental and modeled inputs under the same environmental and fault conditions. The fault conditions include non-standard charging, heat exchanger fouling, loss of compressor volumetric efficiency, liquid line restriction, and the presence of non-condensable gas in the refrigerant. The model’s performance is characterized by comparing its outputs from the evaluation – false alarm rates, misdiagnosis rates, missed detection rates, and rates of undiagnosed faults – with the results based upon experimental data. The model is found to be highly suitable for its purpose.