Spectral separation and concentration prediction of dye components based on kernel‐ MCR model

Abstract The spectral analysis of multi‐component dye systems faces challenges due to overlapping absorption bands and nonlinear effects, making it difficult to accurately characterise individual components and predict their concentrations. To address this, the study proposes a Kernel‐based Multivariate Curve Resolution (Kernel‐MCR) model combined with Multiple Linear Regression (MLR) for high‐precision spectral separation and concentration prediction. Mixed samples of a ternary dye solution were prepared using an orthogonal experimental design, and their spectra were recorded. After preprocessing, the Kernel‐MCR method achieved component spectral separation, and MLR modelled the concentrations using the resolved relative concentration matrix. Performance was evaluated using Spectral Angle Mapper (SAM), Pearson Correlation Coefficient (PCC), coefficient of determination ( R ² ) and Root Mean Square Error (RMSE). The results showed that Kernel‐MCR effectively resolved overlapping spectra, with SAM angles below 15° between separated component spectra and true spectra. The MLR modelling method combined with Kernel‐MCR significantly outperforms traditional modelling approaches, effectively addressing issues such as spectral overlap. The R ² for predicting the concentration of each component has improved to over 0.98, while the Root Mean Square Error (RMSE) has been reduced by more than 37%, validating the accuracy and robustness of this method. This method enhances multi‐component dye analysis, offering a reliable solution for intelligent detection in printing and dyeing processes.