Quantitative Measurement of Gas Component Using Multisensor Array and NPSO-based LS-SVR

To solve the nonlinear response of semiconductor gas sensor and cross-sensitivity to the non-target gases, this paper studies gas sensor array and least square support vector regression (LS-SVR) based gas concentration measurement method. Methane (CH4), hydrogen (H-2) and their mixtures are selected as the target gases. A multi-sensor array is composed of four metal oxide semiconductor (MOS) gas sensors with properties of different sensitivity. LS-SVR is used to establish the quantitative analysis model of each gas component. Given the difficulty in selecting parameters of LS-SVR and the high computational complex in using cross-validation when modeling on each gas component, this paper proposes a niche particle swarm optimization (NPSO) based parameter optimization algorithm which can find the global optimal parameters of the established LS-SVR model of each gas component. Compared with other methods such as artificial neural networks (ANNs), this proposed method improves precision of concentration measurement, and it is particularly adequate for the quantitative detection of gas concentrations within small training samples.