Partager

Document IIF

Prédiction de l’état de l’eau lors de la lyophilisation à l’aide d’algorithmes d’apprentissage automatique de type génétiques et d’optimisation de baleine (WAO).

Prediction of water status during freeze drying with genetic and whale algorithms of machine learning.

Auteurs : YOUNAS S., AHMAD M., ARQAM M., MANZOOR M. S., WANG X., LI W., LIU C., ALI Z., SABIR M. A., IMRAN M.

Type d'article : Article de la RIF

Résumé

Freeze-drying (FD) of foods significantly alters nutritional composition which are strongly related to water status. This study aims to compare the prediction, stability and robustness of more advanced machine learning models such as genetic algorithms (GA) and whale optimization (WO) to overcome, enhance the stability and error limitations of partial least square (PLS), back propagation neural network (BPNN), and support vector machine (SVM) with complex spectra. Multispectral imaging spectra of visible-near infrared (Vis-NIR) (405–970 nm) of water fractions of Lentinus edodes coupled GA and WO models. Low-field nuclear magnetic resonance (LF-NMR) showed that FD removed 90.55 % of total water (TW) within 36 h of drying. However, a rapid sublimation rate was observed during first 12 h (74.20 %), while only 16.35 % of water evaporated in next 24 h due to the presence of less amount of free water (FW). Advancement of models improved the stability and prediction accuracy of water status. Developed models showed more precise prediction and robustness in terms of ratio of prediction to deviation (RPD). For TW, GA and WO of BPNN found the most improved and stable model R2p = 0.9799 and 0.9556, respectively. GA-SVM has shown excellent results in IW with R2p = 0.9432 and lower RMSEP = 6.376. Additionally, GA-BPNN obtained excellent RPD of 5.0545, confirming robustness and stability with complex spectral data. In conclusion, research provides an excellent opportunity for integrating advanced optimization approached and identify the complex Vis-NIR spectroscopic date with potential expansion in future work in assessment of water status during processing.

Documents disponibles

Format PDF

Pages : 12 p.

Disponible

  • Prix public

    20 €

  • Prix membre*

    Gratuit

* meilleur tarif applicable selon le type d'adhésion (voir le détail des avantages des adhésions individuelles et collectives)

Détails

  • Titre original : Prediction of water status during freeze drying with genetic and whale algorithms of machine learning.
  • Identifiant de la fiche : 30034330
  • Langues : Anglais
  • Sujet : Technologie
  • Source : International Journal of Refrigeration - Revue Internationale du Froid - vol. 178
  • Date d'édition : 10/2025
  • DOI : http://dx.doi.org/https://doi.org/10.1016/j.ijrefrig.2025.06.011

Liens


Voir d'autres articles du même numéro (8)
Voir la source

Indexation