A strategy based on the data mining (DM) method is developed to detect and diagnose sensor faults based on the past running performance data in heating, ventilating and air conditioning (HVAC) systems, combining a rough set approach and an artificial neural network (ANN). The reduced information is used to develop classification rules and train the neural network to infer appropriate parameters. The differences between measured thermodynamic states and predicted states obtained from models for normal performance (residuals) are used as performance indices for sensor fault detection and diagnosis. Real test results from a real HVAC system show that only the temperature and humidity measurements of many air handling units (AHU) can work very well as the measurements to distinguish simultaneous temperature sensor faults of the supply chilled water (SCW) and return chilled water (RCW).