The use of hydrogen in port fuel injection (PFI) engines faces challenges related to abnormal combustions that must be addressed, especially in transient operation. The in-cylinder air-to-fuel ratio and the amount of trapped exhaust gas have a significant impact on the probability of abnormal combustion as well as NOx emissions, and should be real-time monitored in hydrogen engines. Thus, the real-time estimation of the composition and thermodynamic state of the trapped gas mixture is crucial during transient operations, although highly challenging.
This study proposes an on-line real-time physics-based MIMO (Multi-Input-Multi-Output) model to accurately estimate the amount of trapped air and exhaust gas in the cylinder at the intake valve closing (IVC) event, based on the instantaneous in-cylinder pressure measurement. With proper estimation accuracy, the injector can be controlled to correctly provide the amount of fuel necessary to achieve the target air-to-fuel ratio (AFR) and reduce the probability of abnormal combustion events. Moreover, the proposed model includes an online controller that corrects the estimation delay by means of a single-cycle prediction, adjusting the injection process and preventing from over- and under- estimation of air and fuel trapped masses. This technique can be applied to a wide variety of engines, reducing calibration efforts at the test bench.
The proposed model, developed in the MATLAB/Simulink framework, is modular and physics-based, meaning that it requires a few calibrations in the form of heat exchange coefficients and can be easily varied to account for different factors. The physics-based model for the estimation of the amount of air and EGR was validated against 1-D numerical results and experimental data for a hydrogen-fueled PFI engine prototype in both steady-state and transient conditions. Results show average errors below 4.24% when estimating IVC trapped air, EGR and temperature in steady operation. Moreover, the model effectiveness was validated in open (without injection control) and closed loop (with injection control) in two transient profiles, showing limited AFR over/under shoots and proper load-following when controlling the fuel injector.