Nonlinear regression algorithm for processing signals from semiconductor chemical sensors to provide selective detection of impurities in artificial air

A new method has been developed for processing the signal of changes in electrical conductivity $\Delta\sigma$ under temperature $(T)$ modulation of a chemical sensor for the selective determination of trace concentrations of ammonia, acetone, n-hexane, propane, toluene, and other impurities in air. The method consists in the fact that, in the range of precisely set concentrations $C$ of each of impurities $Y$, the signal $\Delta\sigma$ as a function of reciprocal temperature $z=10^3/T$ is interpolated using nonlinear regression by a set of parameterized functions $F_i(z,A_i,b_i,c_i,\dots)$, $i$ = 1–4, and the dependences for principal (concentration) parameters $A_{iY}(C)$ are plotted, which determine the so-called “selectivity portrait” of $Y$. Fitting into it, similar values for detected impurity $X$ confirm its identity with $Y$, and the common abscissa of all intersection points $A_{iX}$ level lines with $A_{iY}(C)$ defines the numerical value and unit of measurement for the $C_X$ concentration.