Information density-based adaptive band selection for wide dynamic range gas quantification in infrared spectroscopy

Accurate gas quantification spanning from ppm-level to percentage-level concentrations is critically demanded in fields such as environmental monitoring and industrial safety. Although Fourier-transform infrared (FTIR) spectroscopy provides broadband multi-component detection capabilities, its application faces challenges from nonlinear spectral responses at elevated concentrations due to absorption saturation, instrument-limited resolution, and baseline reconstruction errors. To address this, we propose the information density-based adaptive band selection (ID-ABS) method. This model integrates spectral line intensity, absorption saturation characteristics, instrument line shape function, and baseline features to dynamically evaluate full-spectrum information density distribution, thereby determining optimal inversion parameters for each component. The workflow first calculates a synthetic calibration transmittance spectrum across all wavelengths based on initial parameters, then computes component-specific information density spectra to identify their respective optimal inversion bands. Parameters are iteratively updated via nonlinear multivariate regression until convergence. Experimental validation achieved methane quantification with a linear dynamic range of 3 × 10 7 (R 2 = 0.9998). The applicability of the ID-ABS model can be extended to other gases with infrared absorption features, significantly enhancing FTIR-based quantitative analysis capabilities for complex multi-component mixtures.