Artificial Neural Network Approach for Estimating Biochar Yield from Biomass Composition and Pyrolysis Temperature

Warit Abi Nurazaq (Politeknik Negeri Jember)
Suluh Pambudi (King Mongkut's Institute of Technology Ladkrabang)
Novangga Adi Mulyono (Politeknik Negeri Jember)

Jurnal Teknik Terapan (J-TETA)
Vol. 4 No. 1 (2025): April

Biochar yield from biomass pyrolysis is influenced by complex interactions among feedstock properties and pyrolysis conditions. This study proposes the generation of an artificial neural network (ANN) model to predict biochar yield using input variables including volatile matter, fixed carbon, ash content, elemental composition (C, H, O, N), and temperature on pyrolysis process. A multilayer perceptron (MLP) network was trained using experimental data collected from various biomass sources. The model achieved high performance, with correlation coefficients (R2) of 0.98812 for training, 0.96529 for validation, and 0.94148 for testing. Mean squared error (MSE) analysis showed optimal validation performance at epoch 31, while the error histogram and regression plots confirmed strong predictive accuracy across all datasets. These results demonstrate that ANN is a powerful tool for modeling biochar production, offering a reliable and efficient alternative to labor-intensive experimental methods.