Sustainable Circularity and Digitalization Approach to Bio-waste Management: a Comparative Analysis of ANN and Modified Gompertz Model

Abstract This study presents a novel integration of biochar-enhanced anaerobic digestion (AD) with predictive artificial intelligence (AI) modelling, using JMP Pro 13 to compare multi-node Artificial Neural Network (ANN) configurations (1, 3, 5, and 7 nodes) against the Modified Gompertz model for forecasting biomethane yield. Biomass substrates including cow dung, food waste, goat manure, and brewery waste were assessed in both mono- and co-digestion with biochar. ANN models were trained on 60% of the dataset and validated on 40%, with a hold-back proportion of 0.333 to reduce overfitting. Model performance was evaluated using R², RMSE, and MAPE metrics. Higher-node ANN models (5 and 7) outperformed others for complex substrates, achieving R² values above 0.99 and prediction errors below 20%, while the Modified Gompertz model was more effective for simpler or co-digested inputs. Biochar characteristics were analyzed through FTIR, FE-SEM, and EDX, with cow dung biochar showing the highest carbon content (72.17%) and favorable porosity. The study highlights ANN’s potential as a robust tool for biogas prediction and underscores the role of biochar in improving AD efficiency and carbon sequestration. This dual modelling and materials approach offers a scalable solution for optimizing bioenergy recovery from organic waste within circular economy frameworks.