Evaluation of Energy Consumption Pattern in Rice Processing Using Taguchi and Artificial Neural Network Approaches

Abstract

This study was designed to evaluate and model the impacts of processing parameters (steaming time, soaking time, paddy moisture content and soaking temperature) on the energy consumption of five rice varieties (NERICA 8, FARO 52, FARO 61, FARO 60 and FARO 44). Energy consumption in the cleaning, soaking, steaming, drying, dehusking, polishing and grading operations were estimated by fitting data on labour, fuel and electricity consumption, time and machine efficiency into standard equations to determine total energy consumption. The energy consumptions were separately modelled using Taguchi and Artificial Neural Network (ANN) techniques for each rice variety. The accuracy of models was determined using the coefficient of determination (R²) and Mean Square Error (MSE). Total energy consumption among the rice varieties varied significantly, ranging from 2.3 to 2.3 MJ for white rice, and 45.3 to 76.9 MJ for parboiled rice. Paddy moisture content was observed to be the most important process parameter that influenced energy consumption. Taguchi models were more accurate for energy consumption [R² (0.95-0.97); MSE (1.24-1.96)], than ANN [R² (0.93-0.94); MSE (3.21-3.52)]. The study established appropriate processing conditions that can guarantee minimum energy consumption for NERICA 8, FARO 52, FARO 61, FARO 60 and FARO 44.