Fault detection and diagnosis in HVAC systems using analytical models

Faults that develop in the heat exchanger subsystems in air-conditioning installations can lead to increased energy costs and jeopardise thermal comfort. The sensor and control signals associated with these systems contain potentially valuable information about the condition of the system, and energy management and control systems are able to monitor and store these signals. In practice, the only checks made are to verify set-points are being maintained and that certain critical variables remain within predetermined limits. This approach may allow the detection of certain abrupt or catastrophic faults, but degradation faults often remain undetected until their effects become quite severe. This thesis investigates the appropriateness of using mathematical models to track the development of degradation faults. An approach is developed, which is based on the use of analytical models in conjunction with a recursive parameter estimation algorithm. A subset of the parameters of the models, which are closely related to faults, is estimated recursively. Significant deviations in the values of the estimated parameters from nominal values, which represent `correct operation', are used as an indication that the system has developed a fault. The extent of the deviation from the nominal values is used as an estimate of the degree of fault. This thesis develops the theory and examines the robustness of the parameter estimator using simulation-based testing. Results are also presented from testing the fault detection and diagnosis scheme with data obtained from a simulated air-conditioning system and from a full size test installation.