Polymer selection for SAW sensor array based electronic noses by fuzzy c-means clustering of partition coefficients: Model studies on detection of freshness and spoilage of milk and fish

The paper presents a method for selection of polymers for SAW vapor sensing arrays. The method applies fuzzy c-means (FCM) clustering algorithm on the solvation data for vapor-polymer sorptive interactions. The data set consists of equilibrium partition coefficients for target vapors in a large set of prospective polymers. The FCM algorithm sorts out polymers of similar characteristics into c fuzzy clusters, and subset of polymers representing the centers of these clusters are selected for making the sensor array. The partition coefficient data are calculated by using Abraham's linear-solvation-energy relationship (LSER), and the clustering algorithm considers the polymers as objects and the vapors as observables. The value of c≥2 is determined by repeated clustering with c incremented by 1 until a common subset of polymers emerges that appears in every successive FCM implementation. The method is validated by simulation of SAW sensor arrays for detection of freshness and spoilage marker volatiles in headspace of milk and fish. The identification and concentration estimation are done by applying the radial basis function neural network on simulation data. It is suggested that the present FCM method of polymer selection prior to sensor fabrication could be helpful in optimizing performance and reducing cost of SAW electronic noses for various applications.