AUTHOR=Ngesti Rahaju Sri Mumpuni , Handi  , Veza Ibham , Irianto  , Muhamad Said Mohd Farid , Roslan Muhammad Faizullizam 

TITLE=Acetone butanol ethanol (ABE)-diesel blends and ANN: performance, combustion, and emission prediction of ABE–diesel blends using ANN

JOURNAL=Frontiers in Energy Research

VOLUME=Volume 13 - 2025

YEAR=2025

URL=https://www.frontiersin.org/journals/energy-research/articles/10.3389/fenrg.2025.1707341

DOI=10.3389/fenrg.2025.1707341

ISSN=2296-598X

ABSTRACT=Acetone–Butanol–Ethanol (ABE) fuels and Artificial Neural Network (ANN) models have gained increasing attention. ABE–diesel blends are now being studied using ANN models. The application of ANN enables accurate prediction of diesel engine performance and emissions behavior fueled with ABE-diesel blends, showing the potential of ABE-based fuels for efficient and cleaner combustion with ANN. However, the dependence of traditional diesel fuels on fossil-derived sources continues to contribute significantly to particulate matter (PM) and nitrogen oxides (NOx) emissions. This challenge has motivated the use of bio-derived oxygenated fuels as practical drop-in solutions. Among these, ABE stands out as a promising candidate because of its favorable miscibility with diesel, established fermentation pathways, and inherent oxygen content that promotes cleaner and more efficient combustion. Yet, evidence regarding in-cylinder behavior and emission trade-offs at low blend ratios remains fragmented. This study experimentally investigates ABE–diesel blends at 5% (ABE5) and 10% (ABE10) by volume in a single-cylinder diesel engine operating at 1,200 rpm under five staged load conditions. In-cylinder pressure data and regulated emissions were measured to evaluate performance, combustion, and emission responses. These datasets were then used to train Cascade ANN and Elman ANN to predict brake specific fuel consumption (BSFC), brake thermal efficiency (BTE), and emissions (CO, CO2, NOx, HC) from several inputs. The results demonstrate that ABE5 consistently improved efficiency at light-to-medium loads, achieving up to 14.5% reduction in BSFC at 1 Nm and 17.5% increase in BTE compared to diesel. At higher loads (5 Nm), ABE5 converged with diesel, showing only a marginal penalty of −2.8% in BTE. In contrast, ABE10 showed mixed outcomes, with occasional gains at moderate loads but efficiency penalties at high loads (BSFC +6.5% vs. diesel at 5 Nm). Emission analysis revealed that ABE5 reduced CO by up to 48%, CO2 by 37% at low load and ∼16% at full load, NOx by up to 32% at full load, and HC by up to 22%, while ABE10 tended to increase HC at low-to-medium loads. ANN predictions achieved good accuracy, with correlation coefficients above 0.93 for all outputs. The Elman ANN captured nonlinear performance-emission dynamics effectively, while the Cascade ANN model provided slightly higher precision for efficiency metrics. Error metrics remained consistently low, with mean absolute percentage error below 1%. Overall, this study shows that low-level ABE blending (5%) offers a practical pathway to reduce fossil diesel consumption while improving efficiency and mitigating emissions.