Abstract
Fault detection and diagnosis methods for air conditioning systems typically apply static models after filtering out transient data using a steady state filter. However, air conditioning systems operating in the field often do not achieve a meaningful steady state and therefore the models cannot be applied because only transient data is available. This paper proposes a solution to this problem by predicting the equilibrium point using an exponential regression. The transient response of many system parameters such as cooling capacity, airflow, and refrigerant mass flow may be approximated as a first order dynamic response because the thermal mass in the system dominates other higher order dynamics. The best-fit for a decaying exponential will therefore yield a prediction for the equilibrium point, and static models may then be applied, thus enabling the use of static models with transient data. The method’s performance is quantified using both randomly generated data (Monte Carlo simulations) and the measured response of a field-operating system during both fault-free and faulty operation.