Extracting selectivity from the transient responses of a single coated gas sensor to analyte mixtures using multivariate analysis-based signal processing

In this work, selectivity to aromatic hydrocarbons is extracted from the transient responses of a single coated gas sensor to mixtures using multivariate analysis-based signal processing. The sensor device is coated with a polymer-plasticizer blend thin film, selected for its good affinity and partial selectivity to the targeted hydrocarbons. Single analytes are first characterized with respect to their sensitivities and response time constants. These parameters are then used as inputs to an exponentially weighted-recursive least squares estimation (EW-RLSE) algorithm to identify and quantify the various analytes in gas mixture samples. It is found that, accurate extraction of the various concentrations of individual analytes in a mixture from the transient sensor response require a minimum sampling rate; the lower the sampling time (up to a critical value), the finer the data, and the higher the estimated concentration accuracy. The influence of sampling rate on the accuracy of analyte quantification is presented for time steps of 1, 2, and 3 seconds. Using this approach, identification and estimated concentration values of single analytes were obtained from mixtures with errors within ±3
^%
to
$\pm 4.5{\%}$
for BTEX (benzene, toluene, ethylbenzene, and xylenes) analytes. This work shows that for some specific applications, accurate detection and estimation of analyte concentrations can be achieved without the use of multiple coatings/sensor arrays or complex procedures involving synthesizing metal-organic frameworks (MOFs) or fabricating components of micro-gas chromatography (µGC) systems.