Sustainable Monitoring of Indoor Air Pollutants in an Underground Subway Environment Using Self-Validating Soft Sensors

Abstract

The purpose of this study is to develop a self-validating soft sensor to improve the prediction performance of indoor air quality soft sensors in an underground subway station. The reconstruction-based self-validation method was proposed and implemented in order to: (1) determine the optimal number of principal components when building a principal component analysis training model, (2) enhance the diagnosis accuracy when identifying faulty sensors and (3) reconstruct faulty measurements in a straightforward manner. Two soft sensors based on partial least squares and recursive partial least squares models were developed and their prediction performance was compared in the cases of using faulty sensor measurements and using reconstructed sensor values. Two types of sensor faults including a bias fault and a drifting fault were evaluated using the proposed method. The monitoring results show that the developed sensor self-validation strategy has a powerful ability to correctly detect, identify and reconstruct the sensor faults in the subway system. In addition, the proposed self-validation soft sensing technique could achieve sustainable monitoring of indoor air pollutants in the underground subway environment, because the reconstructed values can be used to replace the measured data when sensor faults have been detected by the detection indices.