A physics-based model for real-time prediction of ignition delays of multi-pulse fuel injections in direct-injection diesel engines

Abstract

Real-time prediction of in-cylinder combustion parameters is very important for robust combustion control in any internal combustion engine. Very little information is available in the literature for modeling the ignition delay period of multiple injections that occur in modern direct-injection diesel engines. Knowledge of the ignition delay period in diesel engines with multiple injections is of primary interest due to its impact on pressure rise during subsequent combustion, combustion noise and pollutant formation. In this work, a physics-based ignition delay prediction methodology has been proposed by suitably simplifying an approach available in the literature. The time taken by the fuel-spray tip to reach the liquid length is considered as the physical delay period of any particular injection pulse. An equation has been developed for predicting the saturation temperature at this location based on the temperature and pressure at the start of injection. Thus, iterative procedures are avoided, which makes the methodology suitable for real-time engine control. The chemical delay was modeled by assuming a global reaction mechanism while using the Arrhenius-type equation. Experiments were conducted on a fully instrumented state-of-the-art common-rail diesel engine test facility for providing inputs to develop the methodology. The thermodynamic condition before the main injection was obtained by modeling the pilot combustion phase using the Wiebe function. Thus, the ignition delays of both pilot and main injections could be predicted based on rail pressure, injection timing, injection duration, manifold pressure and temperature which are normally used as inputs to the engine control unit. When the methodology was applied to predict the ignition delays in three different common-rail diesel engines, the ignition delays of pilot and main combustion phases could be predicted within an error band of ±25, ±50 and ±80 µs, respectively, without further tuning. This method can hence be used in real-time engine controllers and hardware-in-the-loop systems.