Detection and diagnosis for sensor fault in HVAC systems

Principal component analysis (PCA) is presented to detect single sensor faults in heating, ventilation and air conditioning (HVAC) systems. Three PCA models, based on energy balance and air side and water side flow pressure balances, respectively, are set up to detect whether there is any abnormality occurring in the systems. With any fault discovered, a joint angle plot, which compares the new fault vector with known ones in the library, can be used to isolate the faulty sensor indeed. As pre-diagnosis, some expert rules are used to improve the diagnosing process. With the strategy of joint angle plot combined with expert rules, the single sensor fault can be timely isolated on line. (c) 2006 Flsevier Ltd. All rights reserved.