Steady vs. Dynamic Contributions of Different Doped Conducting Polymers in the Principal Components of an Electronic Nose's Response

Abstract

Multivariate data analysis and machine-learning classification become popular tools to extract features without physical models for complex environments recognition. For electronic noses, time sampling over multiple sensors must be a fair compromise between a period sufficiently long to output a meaningful information pattern, and sufficiently short to minimize training time for practical applications. Particularly when reactivity's kinetics differ from thermodynamics' in sensitive materials, finding the best compromise to get the most from data is not obvious. Here, we investigate on the influence of data acquisition to improve or alter data clustering for molecular recognition on a conducting polymer electronic nose. We found out that waiting for the sensors to reach their steady state is not required for classification, and that reducing data acquisition down to the first dynamical information suffice to recognize molecular gases by principal component analysis with the same materials. Particularly for online inference, this study shows that a good sensing array is no array of good sensors, and that new figure-of-merits shall be defined for sensing hardware aiming machine-learning pattern-recognition rather than metrology.