Influence of pre-chamber nozzle and main chamber geometry on ammonia combustion: A combined experimental and predictive study

• Advances the field of combustion science through enhanced methodologies for LBV measurement techniques. • The analysis will contribute to understand the NOx formation mechanisms. • Finally, current findings will encourage in designing new technologies and policies for cutting emissions. The drive to decarbonize transportation has spurred interest in carbon–neutral fuels. Ammonia (NH 3 ) is a compelling hydrogen carrier and direct fuel, but its low burning velocity and narrow flammability range presents significant combustion challenges. The current work presents a hybrid experimental-machine learning (ML) methodology to address these limitations. The primary novelty of this research lies in experimental design (Stage-I), which includes Pre-combustion chamber (PCC) nozzle geometry (diameters and shapes) and Main-combustion chamber (MCC) without slits, is subdivided into two cases. Case-I determined that a diameter of 10.03 mm achieved the highest combustion pressure (1.31 MPa at ER 0.8). Subsequently, Case-II investigated this optimal diameter with the addition of slits resulted in a further 14 % performance improvement over the without slits benchmark. Stage-II employs Random Forest (RF), Artificial Neural Network (ANN), and Deep Neural Network (DNN) models to predict the heat release rate (HRR). Since RF illustrates low error among experimental values, and it acts as a benchmark for further predicting error evaluated by ANN and DNN at hidden equivalence ratios. Root Mean Square Error (RMSE) and R 2 (coefficient of determination) for ANN and DNN have been illustrated with values of 23.10, 0.87,45.24 and0.51 respectively. Since ML values provides better fitment with experimental data, future studies may prioritize ML models, cutting down on expensive experiments and time consumption.