The study on detection method of binary mixed gas based on MOS sensor arrays

Abstract

The accuracy of the metal oxide (MOS) gas sensor for mixed gas detection is reduced due to cross-sensitivity. This paper presents a new method for detecting binary mixed gas. This method preprocesses multi-channel signal data from gas sensor array using kernel principal component analysis (KPCA). Then the target gas is recognized by the support vector machine (SVM). The backpropagation artificial neural network (BP-ANN) was used to quantitative analysis of mixed gas. Genetic algorithm (GA) has the advantage of a global search to optimize the weight and threshold of the BP-ANN. The experimental samples of ethanol and acetone as the target gases were designed and realized to verify the proposed binary mixture gas detection method. Compared with Principal component analysis (PCA)for feature extraction, the recognition accuracy of gas species improved by 4.54%, reaching 98.27%. Compared with the concentration estimation method of the BP-ANN, the estimation accuracy is improved. The concentration estimation errors of ethanol and acetone increased by 1.72% and 0.94% respectively.