AbstractThe recent increase in international competition has spurred the application of on-board process diagnostics to consumer products for improved system performance and reduced maintenance costs. Failure detection and isolation strategies monitor a system's operation for degradations, and if detected, classify the failure source. In this paper, a real-time refrigerator model and two state estimators are designed to predict the system behavior for a model- based failure detection processor. The nonlinear fourth-order refrigerator compressor and condenser model provides an effective description of the system operation for diagnosis, while also minimizing problem complexity and computational time. Several recent nonlinear observer design strategies arc reviewed, and then applied to this model to reconstruct the system states. The evaluation of overall observer performance will provide insight into the application of these estimation strategies to nonlinear thermofluid systems. A number of design issues encountered while deriving the dynamic system model and applying the observer formulations will be discussed.