(562g) New Sensor Selection Methods for Chemiresistor Arrays

The appropriate selection of materials for sensors is critical to the successful development of sensor arrays that have high selectivity, high sensitivity and minimum error. In this work, two sensor selection methods are developed for chemiresistor arrays. One method is used to select the best sensor array to distinguish multiple samples based on maximizing the variance of responses of each sensor for the samples. Another method is used to select the best sensor array to predict the concentrations of the analytes in the mixture in one sample. There are two steps in the latter method: firstly, the sensitivity matrix is estimated for the given sample on the basis of the chemiresistor model; secondly, the mean square error (MSE) is used as the criterion to evaluate all combinations of the available sensors. The sensor set which has the minimum MSE is the best one for the given sample. The advantages of these two methods are discussed by comparing these methods with the principal component analysis (PCA) and the partial least square (PLS) method.