Modeling the biomass pelleting process using artificial neural network

Artificial neural networks are powerful tools for modeling of extrusion processes of biomass materials. In order to study the pelleting process, composted municipal solid waste (MSW) pellets were produced under controlled conditions. The aims of the study were to investigate how the extrusion parameters affect density of the formed pellets and to explain the application of artificial neural networks for density prediction. The effects of independent variables, including the raw material moisture content (35 to 45% (wet basis)), particles size (0.3 to 1.5 mm), speed of piston (2 to 10 mm/s), and die length (8 to 12 mm) on pellet density, were determined. The results showed that four layers perceptron network with training algorithm of back propagation, hyperbolic tangential activation function, Delta training rule with ten neurons in first hidden layer and four neurons in second hidden layer had the best performance for prediction of pellet density. The minimum root mean square error (RMSE) and coefficient of determination (R2) for the multilayer perceptron (MLP) network were 0.01732 and 0.972, respectively.

کلیدواژه ها:

Extrusion parameters ، Biomass pellet ، Density ، Artificial neural network