Machine learning-based prediction and optimization of performance and emission characteristics in CI engines fueled by pumpkin seed biodiesel with thermal barrier coating and cerium oxide nanoparticles

This study investigates the combined influence of a yttria-stabilized zirconia (YSZ)-based thermal barrier coating (TBC) and cerium oxide (CeO2) nanoparticles on the performance and emissions of a compression ignition (CI) engine operated using pumpkin seed biodiesel blends. A multidisciplinary approach integrating experimental evaluation, machine learning prediction, and statistical optimization is employed. The coating system comprises a YSZ–Al2O3–CeO2 composite layer applied via plasma spraying, while CeO2 nanoparticles (35–55 ppm) are dispersed in PSB–diesel blends using surfactant-assisted ultrasonication. Engine tests were conducted across biodiesel blends (B10–B30) and four load levels (50–100%). A feed-forward ANN with three input neurons, two hidden layers (optimized through grid search), and one output neuron was trained to predict SFC, BTE, CO, HC, NOx, and smoke with high accuracy (R2 > 0.99). A user-defined RSM (L27) design was used for multi-response optimization, constrained to maximize BTE while minimizing SFC and emissions. The optimal conditions, B30, 52% load, and 50 ppm CeO2, resulted in reduced SFC (0.316 kg/kWh), enhanced BTE (24.87%), and substantial emission reductions. Novelty arises from the integration of non-edible pumpkin seed biodiesel, CeO2-enhanced nanofuels, a composite TBC, and a hybrid ANN–RSM predictive–optimization framework, which collectively demonstrate a viable pathway toward cleaner and more efficient CI engine operation.