Predicting Laminar Burning Velocities of Oxygenated Fuels Using a Structure–Thermophysical Neural Network Framework

Accurate prediction of laminar burning velocity (LBV) is vital for understanding fuel reactivity and enabling the design of efficient, low-emission combustion systems. In this study, an artificial neural network (ANN) model was developed to predict the LBVs of oxygenated fuels–including alcohols, ethers, aldehydes, ketones, esters and hydrocarbons–using a comprehensive dataset of 3,320 experimental measurements. The input space combined structural descriptors (molecular functional groups, molecular weight, and branching index) with thermophysical parameters (equivalence ratio (ϕ), temperature, and pressure), allowing the model to capture both molecular and environmental influences on flame propagation. The optimized ANN achieved a best validation performance of mean squared error (MSE) = 7.6 × 10^{− 4}, and regression analysis demonstrated excellent agreement between predicted and experimental values, with correlation coefficient (R) = 0.9840 (training), 0.9792 (validation), 0.9808 (test), and 0.9827 (overall). The corresponding root mean squared error (RMSE) was approximately 0.02 m/s, confirming strong predictive accuracy across all data subsets. For benchmarking, a multiple linear regression (MLR) model using the same descriptor set was also developed which achieved significantly lower predictive performance (overall R² ≈0.60 and RMSE ≈0.08 m/s), demonstrating the necessity of nonlinear modeling to capture structure–thermophysical interactions governing LBV. Predicted values closely aligned with experimental measurements over the entire LBV range, with over 95% of residuals within ±0.03 m/s. These results establish the ANN as a robust and generalizable predictive framework, demonstrating the effectiveness of combining structural and thermophysical descriptors for LBV estimation. The model provides valuable insights for fuel screening, kinetic mechanism development, and clean combustion technology design.