Quantitative recognition of volatile organics by fuzzy inference system based on discrete wavelet transform of SAW sensor transients

This paper explores nonlinearity in polymer coated surface acoustic wave (SAW) sensor responses for achieving enhanced quantitative vapor recognition capability. The analyses are based on the SAW sensor transients under step concentration exposure of volatile organic compounds. The sensor nonlinearity results from viscoelastic nature of films, and it depends on the film thickness and viscoelastic parameters of the film coating. The method of analysis involves representation of transient data in terms of discrete wavelet approximation coefficients and then application of Sugeno type fuzzy inference system based on fuzzy c-means clustering in wavelet space for simultaneous classification and concentration estimation. The simulation experiments are based on polyisobutylene (PIB) coating and exposure to 7 target vapors (chloroform, chlorobenzene, o-dichlorobenzene, n-Heptane, toluene, n-hexane and n-octane). By analyzing data as a function of film thickness for lossless and lossy conditions of film viscoelasticity it is found that there exists an optimum region of film thickness over which both the vapor classification rates and the concentration estimates suffer minimum error.