A novel photoacoustic gas sensor for dual-component identification and concentration analysis

In this study, a neural network-assisted photoacoustic gas sensor is proposed that enables dual-component identification and concentration analysis of methane and ethylene, effectively addressing the issue of cross-interference in photoacoustic spectroscopy (PAS) technology. This sensor identifies the unknown photoacoustic second harmonic signal using a self-built photoacoustic deep neural network-component identification model, and then determines the composition of the gas sample. The traditional concentration fitting equation method and the self-built Photoacoustic Deep Neural Network-Concentration Regression Model are integrated to analyze the gas samples composed of single- and dual-component. The sensor demonstrates exceptionally high linearity, accuracy and robustness. Additionally, the minimum detection limits (MDLs) for a single-component are determined to be 0.28 ppm for methane and 1.56 ppm for ethylene. For dual-component detection, the MDLs are 8.86 ppm for methane and 4.55 ppm for ethylene. The promising results of the present study demonstrate that deep learning algorithm provides a more effective, accurate, and stable solution for elimination of cross-interference in photoacoustic gas sensor.