Concentration Estimation of Industrial Gases for Electronic Nose Applications

Sensors drift is one of the most critical challenges while designing an Electronic Nose System (ENS). The discrimination and quantification of gases in the presence of drift is challenging and requires either (i) system recalibration, (ii) domain transformations or (iii) data from target domain. This paper proposes a heuristic optimization technique integrated with a pattern recognition model to estimate the concentration of different industrial gases in the presence of small experimental drift. The proposed method is validated against an experimental data acquired with an array of 16 screen-protected gas sensors. Samples from 6 volatile compounds; ethylene, ethanol, ammonia, acetone, acetaldehyde and toluene are tested to validate the proposed solution. Besides giving accurate performance in terms of concentration estimation the proposed solution does not require system recalibration, domain transformations or target domain data and meanwhile it also reduces the computational complexity of the system.