Fault Detection and Isolation of automotive Air Conditioning systems using first principle models

• Adopt a control-oriented model of the air conditioning system for fault diagnosis.
• Adapt the actuator/sensor/parametric fault diagnosis to a well-established H∞H∞ filter framework.
• Investigate gain-scheduled possibility by varying working point.
• Test filter robustness against model uncertainty.
• Validate algorithm on air conditioning model with satisfactory performance.

Although model-based Fault Detection and Isolation (FDI) has become a common design tool in automotive fields, its application to automotive Air Conditioning (A/C) systems based upon vapor compression cycles is limited due to the lack of control-oriented models characterizing the refrigerant phase change. The emergence of Moving Boundary Method (MBM) illuminates a promising way of assisting FDI scheme development, because common faults in automotive A/C systems, such as compressor fault, pressure transducer fault and fouling fault, can be easily incorporated by the control-oriented model developed. Out of various observed-based FDI methods, the H∞H∞ filter technique, due to its robustness to model uncertainties and external disturbances, is chosen for designing FDI scheme over actuator/sensor/parameter faults. The model and the filter are connected closed-loop by an H∞H∞ controller gain-scheduled to meet different cooling loads. From the closed-loop analysis results, the H∞H∞ filter is capable of detecting and isolating actuator/sensor faults, as well as estimating parameter faults, even if external disturbances imposed on the air side of the evaporator exist.