Uncertainty analysis of hydrogen spark-ignition engine parameters

Predictive engine simulations require rigorous assessment of how uncertainties in operating conditions and model parameters propagate through thermodynamic and combustion models. Despite the growing interest in uncertainty quantification (UQ) in engine research, existing efforts remain largely limited to measurement uncertainty or isolated sub-models, while comprehensive UQ applied to predictive system-level engine models is still lacking. This need is particularly acute for hydrogen/air spark-ignition combustion in spark-ignition engines (SIE), where strong thermo-diffusive effects and a high sensitivity to operating conditions challenge simulation robustness. This work proposes a complete UQ framework applied to a predictive 0D combustion model for SIE. A 0D version of the Coherent Flame Model (CFM), adapted to incorporate Markstein effects, is calibrated for propane and hydrogen. A surrogate-based UQ methodology is developed, combining global sensitivity analysis and uncertainty propagation to efficiently explore the influence of parameter uncertainty. This methodology is then applied to several uncertainty scenarios. First, a configuration representative of limited available data, characterized by large uncertainty ranges, is considered. Second, a configuration reflecting well-instrumented experimental conditions, with tighter uncertainty bounds, is analyzed. The results are strongly dependent on the selected scenario. In the data-limited case, the analysis identifies intake pressure, equivalence ratio, and compression ratio as the dominant contributors to output uncertainty, with hydrogen exhibiting increased sensitivity to mixture preparation and moderately higher overall uncertainty levels. In the second scenario, uncertainty levels are significantly reduced, and the dominant contributors are intake pressure and cylinder wall temperatures. Furthermore, the trapped mass is shown to depend jointly on intake pressure and wall temperature, demonstrating that Computational Fluid Dynamics (CFD) boundary-condition calibration cannot rely solely on intake-pressure adjustment. Overall, the study establishes a general methodology for uncertainty analysis of predictive 0D combustion models and provides actionable guidance for the specification of boundary conditions in CFD simulations.