Sensitivity analysis and soft-computaional prediction of colour characteristics of dried tomatoes

Abstract

Excessive heating with hot-air oven dryers produces considerable losses in the quality of dried tomatoes, particularly in the organoleptic and colour characteristics. Thus, process parameters need to be optimised to minimise detrimental colour quality changes that might not be easily achieved using sophisticated colour detection devices. While a sizable number of studies on the drying of tomatoes, soft-computational modelling and sensitivity analysis of tomatoes' colour characteristics during convective hot-air drying using Adaptive Neuro-fuzzy Inference System (ANFIS) and Artificial Neural Network (ANN) are still unreported. Therefore, this work presents a soft-computing prediction of tomatoes' colour characteristics during convective hot-air drying. The tomatoes were pretreated, sliced, and dried at 40, 50, and 60?C. The colour characteristics (L*, a*, b*, a*/b* change in colour, browning index, hue, and chroma) before and after were determined, and the data was used to train ANN and ANFIS models. The model's predictive performance was determined by calculating the coefficient of determination (R2), Root Means Squared Error (RMSE), and Mean Absolute Error (MAE) between predicted and experimental results. The results showed a range of 26.83 - 43.27, 22.79 - 42.10, 16.99 - 33.72, 1.11 - 1.34, 16.70 - 42.71, 16.94 - 62.37, 28.43 - 53.94, and 0.84 - 0.93, respectively, for the colour characteristics. The ANFIS model demonstrates a meaningful relationship between colour changes and drying conditions with a higher R2 (0.9999) and lower RMSE (0.0452) and MAE (0.0312) than ANN. Thus, the ANFIS model is reliable for prediction and can be further used for fuzzy-based controller process design.