Research and application of a QCL Faraday rotation spectroscopy NO nonlinear inversion model based on machine learning

Aiming at the detection needs of the large range of NO concentration changes in mobile source exhaust, this study develops a machine-learning-based nonlinear concentration inversion model for QCL Faraday rotation spectroscopy. The model was trained using 44 spectra and evaluated through 40 runs, and the PSO strategy was compared with GWO, SA, and ACA, showing that the PSO achieved the best model performance with a correlation of 0.9983±0.0016, absolute error within ±20<i>p</i><i>p</i><i>m</i>, and relative error within 2%. Validation against a HORIBA OBS-ONE under RDE conditions showed high consistency, demonstrating that the proposed method enables accurate, sensitive, and wide-range NO measurements for mobile source emissions.