DEVELOPMENT OF PREDICTIVE MODELS FOR QUALITY CONTROL OF CARROTS DURING DRYING

This thesis research project is aimed at setting up prediction models based on NIR spectroscopy, for quality control of organic carrot discs (Daucus carota L., var. Romance) during hot-air drying process (horizontal flow) up to 8 h. Hot-water blanching was tested at 95°C for 1.5 min, as pre-treatment to control the occurrence of enzymatic browning during drying. Hot-water blanching had a positive impact on the appearance of the carrot discs. PLS regression showed good performances for the prediction of aw (RMSE = 0.04; R2 = 0.96), moisture (RMSE = 0.04; R2 = 0.98), SSC (RMSE = 4.32-4.40 °Brix; R2 =0.88), carotenoids (RMSE = 21.75-23.10; R2 = 0.96) and changes in color (RMSE = 1.40-1.46; R2 = 0.85-0.86) during drying. Also PLSDA classification showed very good metrics (total accuracy 92.38%) in recognising 3-drying steps, both for control and hot-water blanched samples. Features selection by iPLS and iPLSDA algorithms showed results better/equal than models based on full spectrum. For these results, the implementation of low-cost NIR sensors on drier device, seems feasible