Faulty sensor detection, identification and reconstruction of indoor air quality measurements in a subway station

Indoor air quality (IAQ) is important in subway stations because it can influence the health and comfort of passengers significantly. To effectively monitor and control the IAQ in subway stations, several key air pollutants data were collected by the air sampler and tele-monitoring system. In this study, an air pollutant prediction model based an adaptive network-based fuzzy inference system (ANFIS) was used to detect sensor fault, and a structured residual approach with maximum sensitivity (SRAMS) method was used to identify and reconstruct sensor faults existing in subway system. When a sensor failure was detected, the faulty sensor was identified using the exponential weighted moving average filtered squared residual (FSR). Four identification indices, including the identification index based on FSR (I_FSR), the identification index based on generalized likelihood ratio (I_GLR), the identification index based on cumulative sum of residuals (I_Qsum), and the identification index based on cumulative variances index (I_Vsum) were used to assist in identifying sensor faults. The best reconstructed sensor value can be estimated based on a given sensor fault direction. The drifting sensor failure was tested and the effectiveness of the proposed sensor validation procedure was verified.