Sooting propensity of gasoline/diesel-ether blends: Experimental assessment and artificial neural network modeling

This study investigated the impact of ethers as oxygenated fuel additives, on reducing soot emissions from gasoline and diesel combustion. Soot formation, a significant environmental challenge, is heavily influenced by the molecular structure of fuels, necessitating a thorough assessment of the sooting tendencies of diesel/gasoline-ether blends to better understand and mitigate particulate matter (PM) emissions. The study employed measurements of the smoke point (SP), oxygen-extended sooting index (OESI), and threshold sooting index (TSI) to evaluate the sooting tendencies of these blends. Artificial intelligence (AI) models were developed using Artificial Neural Network (ANN) tools, based on SP measurements from forty blends with varying ether percentages in diesel/gasoline mixtures. Various features, such as functional groups, molecular weight, branching index, density, and molar ratios, were used as inputs, while the measured SPs, TSIs, and OESIs served as target outputs. Although SP and TSI are widely used to evaluate soot formation, they have limitations in capturing the role of oxygen in combustion chemistry. To address this gap, the OESI—an index that explicitly incorporates the effect of fuel-borne oxygen—was employed in this study to evaluate soot formation in ether-based blends with gasoline and diesel. Moreover, ANNs were applied to predict soot formation in untested blends with similar molecular structures, providing a robust predictive framework that complements experimental analysis. The results revealed a strong correlation between predicted and experimental indices, with correlation coefficients (R) of 0.96 for SP, 0.99 for TSI, and 0.98 for OESI, indicating high model accuracy. The respective mean absolute errors were 1.16, 1.00, and 4.92, confirming the reliability of the AI approach. Key molecular characteristics, such as aromaticity, branching, and molar ratios, were found to significantly influence sooting behavior. This study highlights the potential of AI-driven models in accurately predicting soot formation trends in fuel blends containing ethers, offering valuable insights for the design of cleaner and more sustainable fuels.