Drying Kinetic Models: Performance Evaluation under Auto-Correlated Observations

Abstract

The standard drying kinetic models like Lewis and Pages models assume that error terms of fitted models are uncorrelated to each other, which may not hold in reality as the observations are measured on successive time intervals. The best computational solution is to incorporate the correlated error structure into the model fitting process. The present study evaluated the performance of drying kinetic models with auto-correlated errors and compared with the standard drying kinetic models using different goodness of fit statistics obtained from the modified models. Validation study showed that Lewis model with auto-correlated errors was best fitted model for the real time data on moisture ratio of Malabar tongue sole fish than standard Lewis model. The estimated drying constant of the fitted model was 0.09 and auto-correlation coefficient was -0.29. The fitted model had higher R2 value (0.94) and lower standard error (0.01) for estimated parameters of the model when compared to the standard Lewis model.