On the Detection of Volatile Organic Compounds (VOCs) Using Machine Learning and FTIR Spectroscopy for Air Quality Monitoring

Air pollution refers to the release of pollutants into the air that is due to energy production and consumption that release VOCs such as the BTEX group. Real-time monitoring of these VOCs in factories, homes, and roads is important to avoid exceeding the harmful limits. These limits are in the sub-ppm range is not straightforward to detect using low-cost miniaturized FTIR spectrometers. In this work, we investigate the application of various machine learning techniques on the infrared absorption spectra of the VOCs. The algorithms are implemented with the target of creating a model to differentiate the VOCs and their corresponding concentrations, potentially in the open air. The effect of the signal-to-noise ratio (SNR) of the spectrometer on the accuracy of the model are studied and found that the accuracy can reach higher than 90 %, when the resolution is 60 cm −1 , the SNR is higher than 27 dB and the path length is 15 m. Having large training data set is found to be not needed for large SNR.