Predicting Performance of Briquette Made from Millet Bran: A Neural Network Approach

Abstract

Millet bran possesses good fuel quality and can be successfully used as a professional feedstock for producing solid biofuel.  In this paper, a framework for developing an Artificial Neural Network (ANN) to estimate the performance of millet bran briquettes is presented by using experimental data to train, test, and validate the ANN. With the capacity of the developed multi-layer ANN, the effects of moisture content, temperature, and applied pressure on the density, durability, and impact resistance are predicted. Different cases considering three parameters as inputs to the ANN, namely, moisture content, temperature, and applied pressure were analyzed. The outputs of the ANN are the density, durability, and impact resistance for each of the input parameters separately. By comparing with the experimental values, it is shown that the ANN-based method can predict the data well with a Mean Square Error (MSE) value ~ 0.2%. Further, Multiple Linear Regression (MLR) model is used to check the efficiency of ANN prediction from which it is shown that the proposed ANN-based method provides useful guidance for the prediction of the physical parameters efficiently, with the least deviation and high accuracy.