On-line sensor calibration transfer among electronic nose instruments for monitoring volatile organic chemicals in indoor air quality

Since the homogeneous linearity between multi-sensors systems which are called electronic noses (E-noses) designed using commercially available heated tin oxide sensors, a high performance of on-line calibration transfer among multiple E-nose instruments based on global affine transformation (GAT) and Kennard-Stone sequential algorithm (KSS) is presented and evaluated in this paper. GAT is achieved in terms of one single sensor by a robust weighted least square (RWLS) algorithm and KSS is studied for representative transfer sample subset selection from a large sample space. This paper consists of two aspects: calibration step (for responses of sensors) and prediction step (for gas concentration). Prediction is developed to evaluate the performance of calibration transfer. In prediction step, three artificial neural networks for concentration prediction of three analytes were trained based on an error back-propagation algorithm. Both implementations of the two aspects were operated on Matlab, preliminarily evaluated using hazardous formaldehyde as referenced gas and subsequently directly applied to quantify benzene and toluene which are measured by six E-nose instruments at specific gas experimental platform. Simulated and experimental results were found to be adequate and good precision and accuracy were obtained.