Performance, emission, and neural network analysis of Simmondsia chinensis biodiesel as a sustainable fuel in a compression ignition engine: A multi-objective approach

The automotive sector faces increasing challenges from strict emission regulations and rising fossil fuel demand, driving the transition toward alternative energy sources for internal combustion (IC) engines. This study investigates the optimization of a Common Rail Direct Injection (CRDI) engine fueled with a ternary biodiesel blend and antioxidant additive, where methyl acetate (MA) was incorporated to improve fuel stability and reduce emissions. The optimized blend, DBMA20+10 % EGR (30 % SCB, 50 % diesel, 20 % MA), was evaluated under varying pilot injection timings (35°, 40°, 45° bTDC) and pilot fuel injection quantities (10–20 %). The engine tests revealed that the combination of 20 % PFIQ at 45° bTDC PIT delivered the best results, improving brake thermal efficiency and significantly reducing brake-specific fuel consumption (20.37 %), smoke (24.51 %), Hydrocarbon (40.19 %), Carbon monoxide (52.41 %), and Oxides of Nitrogen (13.34 %) compared with DBMA20+10 % EGR under stock conditions. To complement experimental analysis, artificial intelligence techniques were applied. The Artificial Neural Network model achieved reliable prediction accuracy (R² = 0.8524–0.9631), while the Long Short-Term Memory (LSTM) network demonstrated superior performance (R² = 0.9473–0.9965) with prediction errors below 6 %. These results confirm the potential of LSTM-based optimization for accurately forecasting engine performance and emissions, thereby reducing experimental efforts and facilitating effective tuning of CRDI engines for alternative fuel applications.