Black Box Modeling of Steam Distillation Essential Oil Extraction System Using NNARX Structure

Abstract

This paper evaluates the Neural Network AutoRegressive with eXogenous (NNARX) structure in modeling the steam distillation essential oil extraction. The model order will be selected based on Rissanen’s Minimum Description Length (MDL) information criterion. In the training of NNARX model, both unregularized and regularized models will be assessed. There are three regularization levels of the weight decay that will be implemented in this work. The number of hidden neuron and iteration will be optimized before the training session. The testing of the trained model will be based on R2, adjusted-R2, NMSE, RMSE, residual histogram and correlation tests. All results will be compared and evaluated with respect to the testing data.