Recurrent adaptive neuro-fuzzy inference system for steam temperature estimation in distillation of essential oil extraction process

In this paper, recurrent adaptive neuro-fuzzy inference system (RANFIS) structure has been proposed to solve approximation problem in identifying a global model of steam temperature of packed distillation column in steam distillation essential oil extraction process. The input-output data is acquired from field experimentation via MATLAB Real-time Workshop (RTW) integrated to the plant. The derived RANFIS model is optimized in order to get the optimum ANFIS structure that includes the optimal number of membership function, fuzzy rules, data selection, epoch which gives low computation time and root means squared error (RMSE). Several experiments were carried out using both pseudo random binary sequence (PRBS) and noise as perturbation signals. Performance comparison of RANFIS with ARX model shows that RANFIS identification gives an excellent global modeling method with RMSE of 0.1778 and consumed less computation or training time.