Artificial Neural Network Modelling for Slow Pyrolysis Process of Biochar from Banana Peels and Its Effect on O/C Ratio

Abstract

Banana peels are one of the most underutilized agriculture wastes, having a great potential to be converted into various alternative products in many industries. In agriculture, biochar derived from banana peels can be used as soil amendment and soil conditioner. In this study, slow pyrolysis was chosen among the several thermochemical techniques to produce biochar, due to its ability to achieve low oxygen to carbon (O/C) ratio. Artificial neural network (ANN) modelling of the slow pyrolysis process was implemented based on the experimental data gathered. Feedforward artificial neural network (FANN), a type of ANN, was used to create the model based on a single hidden layer network using MATLAB™ software. A combination of up to 25 hidden neurons in the hidden layer as well as 10 different training algorithms were tested using iterative training method. Levenberg-Marquardt training algorithm with 24 hidden nodes showed the best results, with the highest determination coefficient (R2) of 0.99858, low output MSE of 0.04467 and lowest validation MSE of 0.00001.KeywordsBiocharSlow pyrolysisBanana peelsFeedforward artificial neural network