With today's competitive and complex environment which results from rapid industrial development, air quality monitoring is becoming a necessity. Devising devices that provide reliable, cost-effective, and fast monitoring of indoor/in-car harmful chemical compounds is of paramount importance for governments as well as individuals. Sensors array systems or commonly called electronic nose (E-nose) systems have been used in various fields of consumer applications. Owing to their versatility and ease of use, these systems can be an adequate alternative for indoor/in-car air quality monitoring. In this study a novel self-made and cost-effective electronic nose aiming at quantifying five indoor/in-car harmful gases (formaldehyde, benzene, CO, NO2, toluene), has been devised and implemented at the college of electronic and communication engineering of Chongqing University, China. A hybrid genetic algorithm support machine vector regression (GA-LSSVMR) model is used for pattern recognition and concentrations estimation. With absolute relative errors of prediction (MAREP) less than 10%, these models outperform those based on hybrid genetic algorithm backpropagation neural network regression (GA-BPNNR). Furthermore, the best regression models were embedded into the system for real-time concentration estimation; our system's predictions mostly agree with those of specific gas detectors. The product will therefore be a good alternative for indoor/in-car air quality monitoring.

No abstract

No abstract