Abstract:Aiming at the problems of complex operation and strong subjectivity in the viscosity measurement method of water-based ink, using Vis/NIR spectroscopy combined with chemometric methods, the prediction model of water-based ink viscosity was established to realize the rapid non-destructive detection of the viscosity of water-based inks. Firstly, the reflectance spectral data of water-based ink samples were obtained by the micro-spectrometer. The PLSR and PCR models were constructed based on the original full spectra and the preprocessing spectral data respectively, by analyzing and comparing the preprocessing effects of different preprocessing methods. Finally, the feature wavelengths were extracted by SPA and CARS, and the PLSR model was established based on characteristic spectra to predict the viscosity of water-based ink. The results show that 4 characteristic wavelengths were extracted by the SPA algorithm from full spectra. The SNV-SPA-PLS model was simplified significantly, which improved the prediction efficiency and had the best prediction performance (Rp2=0.9992, RMSEP=0.0732). The results show that the application of spectral analysis techniques for the viscosity of water-based ink detection is effective and feasible, which provides a new method for further viscosity testing of water-based inks by spectral analysis techniques and a technical basis for improving the quality and stability of printed materials.