Machine learning-assisted FTIR spectra to predict freeze-drying curve of food

• DOI: 10.1016/j.lwt.2024.115894
• Publication Date: 2024-02-25T10:58:53Z
• Authors: Xihui Liu; Baolong Feng; Hongyao Liu; Yutang Wang; Bowen Luo; Yan Yang; Qi Zhang; Zhipeng Wang; Ziqi Xu; Bailiang Li; Fengzhong Wang

With the demand for improving freeze-drying (FD) process efficiency and protecting product characteristics, intelligent robust analysis of parameters drawing effective FD curves has become development direction modern process. In this study, a prompt applicable prediction model was designed by FTIR associated with chemometrics. By using spectral preprocessing principal component analysis, 34 feature wavenumbers were extracted as input variables modeling to quantify parameters. Among 18 parameter models, artificial neural network adopted optimum temperature time pre-freezing desorption stages (R2 = 0.91, 0.83, 0.92, 0.84, RMSE 0.12, 0.13, 0.08, 0.10), random forest confirmed best sublimation stage 0.88, 0.77, 0.16). According prediction, samples selected verification that experimental results close 96% agreement output.