Tolerant control for multiple faults of sensors in VAV systems

Principal component analysis, joint angle plots and reconstruction schemes are presented in this paper to detect, isolate and evaluate multiple sensors faults occurring in variable air volume (VAV) systems. Multi-level principal component analysis models, including system level and local level, are built to detect multiple faults occurring in VAV systems. As the initial detection, a system level model is used to discover the abnormalities in view of the whole systems. Two local level models are used to further confirm the occurrence of the faults. Moreover, with the multiple faults separated into different locations by the two local level detection models, joint angle plots are used, respectively, to isolate the faults one by one. Finally, the reconstruction scheme is used to estimate the magnitude of the bias to recover from the faulty operation