Prediction of the changes in physicochemical properties of key lime juice during IR thermal processing by artificial neural networks

Thermal processing of the key lime juice leads to the inactivation of pectin methylesterase (PME) and the degradation of ascorbic acid (AA). These changes affect directly the cloud stability and color of the juice. In this study, an artificial neural network (ANN) model was applied for designing and developing an intelligent system for prediction of the thermal processing effects on the physicochemical properties of key lime juice during conventional and infrared (IR) heating. The inputs of this network were time and temperature and the outputs were changes in PME activity, AA content, browning index (BI) and also cloud stability of the juice. The feed-forward neural network with a logarithmic transfer function, Levenberg–Marquardt training algorithm and eight neurons in the hidden layer (topology 2-8-4) was chosen as the best ANN model (R2> 0.95, RMSE=0.47 and SE=0.28). The predicted values using the optimal ANN model vs. experimental values represented a correlation coefficient higher than 0.95 and 0.90 during IR and conventional thermal processing, respectively. This model can therefore be applied in prediction of the effects of thermal processing on the physicochemical properties of the lime juice in pilot plants, processing factories and online monitoring.