A Constrained Minimization Approach for Concentration Estimation in Acid–Base Neutralization Reactions Using Near-Infrared Light

We aim to achieve concentration imaging within a Y-shaped microfluidic chip using NIR (near-infrared) light. Based on absorbance difference spectra acquired by FT-IR (Fourier-transform infrared spectroscopy), we constructed regression models suitable for concentration prediction. To enhance predictive performance, we applied PLS-CARS (Partial least square regression-competitive adaptive reweighted sampling) for optimal wavelength selection and determine the regression coefficients. Furthermore, we developed model transfer techniques to bridge the FT-IR and imaging systems, along with pixel-wise calibration transfer methods. we formulated and solved a constrained optimization problem incorporating physical constraints to address issues such as spatial discontinuities and negative concentration values in the estimated distributions.