Spectral Pattern of Chocolate Production: Early Detection of Quality Problems

Quality defects in chocolate can be classified as physicochemical quality problems and structural instability (blooming). Both problems can be detected after production and generally take time to become visible. Therefore, early detection tools are necessary to prevent economic losses in the food industry. In this study, FTIR spectra collected at different stages of manufacturing were combined into spectral patterns to predict future quality defects, such as physicochemical defects (improper water activity, color, whiteness index, and hardness) and structural instability (fat and sugar bloom), using chemometrics. Accordingly, for uniform and defective chocolate production, five different conching processes, three different tempering processes, and two different cooling processes were used, and the samples were stored at 15°C, 20°C, and 28°C. To evaluate the defective samples, differential scanning calorimetry (DSC), polarized light microscopy, visual appearance tests for structural instability, and physicochemical analyses were conducted. Results showed that storage at 28°C caused fat bloom in all samples, regardless of processing. Therefore, to evaluate the effects of production parameters on fat bloom, 20°C samples were used in partial least squares-discriminant analysis (PLS-DA) models using spectral pattern. On the other hand, sugar bloom just appeared in the samples stored at 15°C, which were used for sugar bloom prediction PLS-DA models. Although fat and sugar bloom required time (4-8 months), physicochemical defects (unsuitable water activity, hardness, and color) appeared even in new production. PLS-DA models can predict these defects in new production, as well as formation of fat and sugar bloom during storage with 100% sensitivity and specificity in prediction. PRACTICAL APPLICATIONS: The chocolate industry can use this FTIR-based early prediction approach to reduce discarded products and economic losses and to improve the final product quality in the production line. This practical solution can be extended to different food matrices for monitoring of food production.