Modeling the prediction of hydrogen production by co‐gasification of plastic and rubber wastes using machine learning algorithms

Abstract

This study aimed to investigate the application of radial basis function (RBF) and multilayer perceptron (MLP) artificial neural networks for modeling hydrogen production by co-gasification of rubber and plastic wastes. Both the RBF and MLP neural networks were configured by determining the best-hidden neurons that could offer optimized performance. Based on the best-hidden neurons, a model architecture of 4-16-1, 4-20-1, 4-17-1, and 4-3-1 was obtained for RBF (with standard activation function), RBF (with ordinary activation function), one-layer MLP, and two-layer MLP, respectively, indicating the number of input nodes, the hidden neurons, and the output nodes. The predicted hydrogen production from the co-gasification closely agrees with the observed values. The 1-layer MLP with R2 of .990 displayed the best performance with all the input parameters having a significant influence on 9 the model output. The neural network algorithm obtained in this study could be implemented in the eventuality of making a vital decision in the process operation of the co-gasification process for hydrogen production.