QUANTITATIVE MONITORING OF FAT BLOOM DEVELOPMENT IN MILK CHOCOLATE USING PORTABLE NIR SPECTROSCOPY AND CHEMOMETRICS

Fat bloom is a major quality defect in chocolate, characterized by the development of a whitish appearance on the surface as a consequence of changes in the fat crystalline structure during storage. This study investigated the potential of portable near-infrared (NIR) spectroscopy combined with chemometrics to quantitatively monitor fat bloom development in milk chocolate subjected to accelerated thermal cycling. Twenty milk chocolate bars were subjected to alternating temperatures of 20 °C and 30 °C every 6 h, resulting in four temperature steps per day. Measurements were performed at days 0, 3, 6, 9, 12, 15, 18, 21, 25, and 27. Surface color was measured in the CIELAB system, and the whiteness index (WI) was calculated as an instrumental indicator of the progressive surface whitening associated with fat bloom. During thermal cycling, progressive changes in surface color and an increase in WI were observed, reflecting the gradual development of the characteristic whitish appearance associated with fat bloom. NIR spectra were acquired using a portable spectrometer in the 900– 1700 nm range, with two measurements per chocolate bar averaged before chemometric analysis. Partial least squares regression (PLSR) was applied to predict WI from the NIR spectra. Spectra were preprocessed using a Savitzky–Golay first derivative (second-order polynomial, 15-point window) followed by mean centering. The chocolate bars were randomly divided by experimental unit into calibration (14 bars, 70%) and independent external validation (6 bars, 30%) sets, with all measurements from each bar retained within the same dataset. The PLSR model using five latent variables showed strong agreement between measured and NIR-predicted WI values, with R² values of 0.959, 0.956, and 0.923 for calibration, cross-validation, and external prediction, respectively. The corresponding RMSEC, RMSECV, and RMSEP values were 1.88, 1.96, and 2.66, respectively, with a prediction bias of 0.19. These results demonstrate that, once calibrated, the portable NIR-PLSR model can estimate WI directly from the NIR spectrum of a milk chocolate sample, without requiring an additional instrumental WI measurement. Thus, the predicted WI can be used as a quantitative indicator of the degree of surface whitening associated with different stages of fat bloom development, supporting portable NIR spectroscopy as a rapid and nondestructive tool for monitoring fat bloom progression.

Acknowledgments: The authors are thankful to the São Paulo Research Foundation (FAPESP) (project number 2024/22967-5 and 2023/16840-0).