Selection of training data for modeling essential oil extraction system using NNARX structure

In this work, the suitable training data for modeling the steam distillation essential oil extraction system is presented. The data were collected from the self-refilling distillation column using RTD sensor with associated signal conditioning circuit. The control signal is on/off The heating system implementing 1.5 kW electrical immersion heater. The power switching is performed using zero-crossing solid-state-relay. The input signals are the PRBS with different probabilities. There are 3 situations of data to be investigated. Since the system is highly-nonlinear, it is expected that the training data that covers the full operating condition will be the optimum training data. These data are separated into training and testing data by interlacing technique, which make the total number of data 6. For each data, the model order selection is based on ARX structure and MDL information criterion. These data are cross-validated between each others and the validation results are presented and concluded. The performance indexes are the percentages of R² , adjusted-R² and NMSE.