External cavity quantum cascade laser absorption spectroscopy based on convolutional neural network algorithm for volatile organic compounds analysis.

Volatile organic compounds (VOCs) are organic compounds having significant impacts on environmental and human health. In this paper, a laser absorption spectroscopy gas sensor system based on an external cavity quantum cascade laser was developed for analyzing multiple VOCs simultaneously. Acetone (C3H6O) and ethanol (C2H6O), as two typical VOCs, were selected as the objects of analysis. Given the overlapping interference issue of C3H6O and C2H6O absorption spectra, a one-dimensional spectral separation convolutional neural network algorithm was developed for VOC spectral signal processing. The experimental results indicated that the spectral fitting residuals were within 0.0028, achieving an 8.77-fold improvement over the traditional partial least squares method. The correlation coefficients for concentration predictions of C3H6O and C2H6O can be improved from 0.9991 and 0.9989 to 0.9996 and 0.9996, respectively, which shows good reliability of the developed algorithm model for both blended absorption spectrum separation and concentration retrieval in the simultaneous detection of multiple gas components.