Optimization of process variables during torrefaction of coal/biomass/waste tyre blends: Application of artificial neural network & response surface methodology

Torrefaction models to predict solid yield (SY), enhancement factor (EF) and higher heating value (HHV) of blends of coal and biomass, and waste are limited in the literature. In this study, Artificial Neural Network (ANN) and Response Surface Methodology (RSM) were used to determine the optimum torrefaction process conditions (temperature, time, and blend ratio) and optimum fuel blend based on the torrefaction variables (HHV, EF, and SY). Pine sawdust (PSD), sugarcane bagasse (SCB), and corn cob (CC) were the biomass materials used. Three torrefaction temperatures (200, 250, & 300 ^OC), torrefaction times (30, 45, and 60 min), and blend ratios (50:50, 71:29, & 80:20) for coal/biomass/waste-tyre blends were studied. The HHV of the fuels that was determined experimentally was used to estimate the EF and SY. A Levenberg-Marquardt back-propagation algorithm ANN and RSM, Box-Behnken Design (BBD) techniques were applied to study the torrefaction process. Coal-to-Torrefied fuel ratio of 50:50 at 300 ^OC and 45 min was the optimum process condition, while Coal + Torrefied PSD was the optimum fuel with the HHV and EF of 28.27 MJ/kg and 1.41, respectively. The HHV of the optimum fuel (torrefied) is around 10% higher than the raw fuel. Fuel quality enhancement of Coal + Torrefied WT via torrefaction was not promising. A second-order polynomial and ANN models were developed for each fuel, but the ANN model was more reliable than RSM model considering the R², AAD and PRE results.