Sensor fault diagnostics and prognostics using component model and time scale orthogonal expansions

Abstract

A method of diagnosing sensor faults for a heating, ventilation and air conditioning system includes the steps of creating a component model for a specific component within the system. The component model is created through the use of commonly available manufacturing data. Data within the system is input into the component model and compared to calculated and predicted values that are also calculated using the identical component models. Differences between the calculated and actual values is determined and compared to a threshold difference value. If the difference exceeds the threshold value, then a fault is detected. The specific type of sensor fault is determined using probability distribution analysis. Each type of sensor fault produces a different type of statistical deviation from normal distribution. By recognizing these patterns of deviations from the normal distribution, the specific type of fault such as electrical, intermittent or freezing of the sensor can be determined to provide initial information as to the severity and type of remedial action required.