Partial least squares regression analysis of natural gas-like mixtures absorption spectra collected using a mid-infrared supercontinuum broadband source

We report on a broadband gas sensor based on direct absorption spectroscopy in the mid-infrared range for the simultaneous detection of methane, ethane, and propane in natural gas-like mixtures. The system employs a broadband supercontinuum light source, coupled with an absorption cell and an optical spectrum analyzer with a resolution of 0.5 cm-1. This configuration enables reconstruction of the full absorption bands of the target alkanes, which exhibit significant spectral overlap in the 2.8-3.2 μm spectral region. A comparative study between multiple linear regression (MLR) and partial least-squares regression (PLSR) was conducted to determine the concentration of each individual component. The results highlight the superior performance of PLSR in the presence of unbalanced concentration ratios (1:10) among the three alkanes, achieving mean prediction accuracy of 98%, 93% and 94% for methane, ethane, and propane, respectively.