Monitoring and prediction of chocolate blooming using Vis-NIR and FT-IR hyperspectral imaging and machine learning techniques: A study on tempering and storage effects

ERİKLİOĞLU H., Castillo R. d. P., ÖZTOP H. M.

LWT, cilt.228, 2025 (SCI-Expanded, Scopus) identifier

  • Yayın Türü: Makale / Tam Makale
  • Cilt numarası: 228
  • Basım Tarihi: 2025
  • Doi Numarası: 10.1016/j.lwt.2025.118135
  • Dergi Adı: LWT
  • Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Academic Search Premier, PASCAL, Aquatic Science & Fisheries Abstracts (ASFA), BIOSIS, Biotechnology Research Abstracts, CAB Abstracts, Chemical Abstracts Core, Compendex, Food Science & Technology Abstracts, Veterinary Science Database, Directory of Open Access Journals
  • Anahtar Kelimeler: Blooming, Chemometrics, Chocolate, Fat bloom, FT-IR microscopy, Hyperspectral imaging, Machine learning, Quality control
  • Orta Doğu Teknik Üniversitesi Adresli: Evet

Özet

Chocolate blooming is one of the main issues in the chocolate industry. Fat blooming is the most common type of chocolate blooming. When blooming occurs, customer satisfaction significantly decreases because of the unpleasant look and undesired texture. The reason for blooming is generally lack of tempering or poor storage conditions. Since blooming occurs with time, it is not easy to detect, especially in the early stages. Therefore, it is necessary to develop tools to monitor chocolate blooming in initial stages and predict the blooming stage. Hyperspectral imaging is a non-destructive imaging technique that can reveal differences related to physical and chemical structure. In this research, commercial chocolate samples were collected and melted to produce untempered chocolate. All samples were remolded into coin size tablets and hyperspectral images were taken in 30 days’ time. Results showed that by using line scan, VIS-NIR hyperspectral camera (400–1000 nm), spectral signature differences were observable between tempered, untempered chocolate and different storage times. Prediction accuracy was assessed by the use of chemometric approaches such as k-Nearest Neighbor (KNN), Support Vector Machines (SVM) and Artificial Neural Networks (ANN). All three methods showed high performance, but Neural Networks predicted 99 % of the samples correctly with Savitzky-Golay (SG) second derivative preprocessing. In addition, Fourier-transform infrared (FT-IR) imaging confirmed compositional changes between bloomed and non-bloomed regions. Multivariate Curve Resolution–Alternating Least Squares (MCR-ALS) analysis of acquired images revealed distinct spectral contributions from bloomed and non-bloomed regions, highlighting compositional changes related to fat migration. The integration of hyperspectral imaging and chemometrics allowed early detection and monitoring of blooming stages, offering a potential solution for real-time chocolate quality control.