Hydrogen PFI engines face abnormal combustion issues, especially during transient
operation. The air-to-fuel ratio and trapped exhaust gas significantly affect
combustion stability and NOx emissions, requiring continuous monitoring.
Real-time estimation of the trapped gas composition and thermodynamic state is
therefore crucial but challenging.
This work introduces a real-time, physics-based Multi-Input-Multi-Output (MIMO)
model for accurately estimating trapped air and exhaust gas mass at the intake
valve closing (IVC) event. In detail, the estimation model makes use of dynamic
in-cylinder and exhaust pressure measurements to accurately model mass flows and
heat exchange equations with 0.5 CAD resolution. This allows extremely high
fidelity when modelling the physical properties of the various chemical species
along the engine cycle. Moreover, the model calibration appears only in the form
of two coefficients implemented on a lookup table for twelve different operating
points, highlighting the small calibration effort.
The physics-based model for the estimation of the amount of air and EGR was
validated against 1-D numerical results for a hydrogen-fueled PFI engine
prototype developed in GT-Power environment. The validation process analyzes the
model accuracy in multiple steady-state and transient profiles, in terms of
in-cylinder trapped air and residuals. 165 steady cases and two transient
profiles of 1800 engine cycles each are studied. Results show the robustness and
accuracy of the model, allowing proper AFR control especially when integrating a
fuel-injection correcting controller. Indeed, value of normalized mean absolute
percentage error around 2% and 5% are reported for air and EGR estimation. The
model proves to be highly accurate even in fast-transient operation: however,
further improvements will be carried out to reduce maximum errors observed.