Fabio Berni

The statistical tendency of an optically accessible single-cylinder direct-injection spark-ignition engine to undergo borderline/medium knocking combustion is investigated using 3D-CFD. Focus is made on the role of fuel surrogate formulation for the characterization of anti-knock quality and flame speed of the actual fuel. An in-house methodology is used to design surrogates able to emulate laminar flame speed and autoignition delay times of the injected fuel. Two different surrogates, characterized by increasing level of complexity, are compared. The most complex one (six components) improves the representation of the real fuel, highlighting the crucial role of accurate fuel kinetics to predict flame propagation and unburnt mixture reactivity. A devoted chemical mechanism including the oxidation pathways for all the species in the surrogate is also purposely developed for the current analysis. Knock is investigated using a proprietary statistical knock model (GruMo-UNIMORE Statistical Knock Model, GK-PDF), which can infer the probability of knocking events within a RANS formalism. Predicted statistical distributions are compared to measured counterparts. The proposed numerical/experimental comparison demonstrates the possibility to efficiently integrate complex-chemistry driven information in 3D-CFD combustion simulations without online solving chemical reactions: a combination of laminar flame speed correlations, ignition delay look-up tables, and a statistics-based knock model is adopted to estimate the percentage of knocking cycles in a GDI engine while limiting the computational cost of the simulations.

In order to reduce development costs and time-to market, 1D and 3D CFD tools can support engine design providing reliable estimations of the tailpipe emissions. In particular, 3D-CFD in-cylinder simulations can evaluate formation of both soot and gaseous pollutants inside the combustion chamber. The main issue in such kind of simulations is the validation against experimental findings. In fact, the complexity of the emission measurements does not allow a straightforward one-to-one comparison between numerical and experimental results. Therefore the present paper aims at providing, on the one hand, a robust numerical framework for both gaseous and solid emissions, on the other hand a dedicated post-processing for a fair comparison between simulations and experiments. From a numerical standpoint, a simplified approach is dedicated to gaseous emissions, while a more detailed one is reserved to soot modeling. The latter is based on the Sectional Method, whose reaction rates are tabulated following 0D chemical kinetic simulations of a purposely designed surrogate in a constant pressure reactor. Simulations and experiments proposed in the present analysis are referred to a high-performance turbocharged direct-injection spark-ignition engine operated at part-load and low rpms. On equal performance, revving speed and mean mixture quality, different injection timings are investigated. The developed numerical approach and post-processing ensure a good agreement between simulations and experiments.

In this work, a turbocharged spark ignition engine equipped with a variable valve actuation device is investigated to numerically optimize the Brake specific fuel consumption (BSFC) at different loads and speeds by employing a genetic algorithm. The engine is preliminary analyzed at the test bench under both part and full load operations and different valve strategies. A system schematization is realized in a 1D code. The developed model is integrated with user-defined sub-models for the description of the in-cylinder processes, and then is validated over the measurements. A 3D CFD model of a single cylinder is developed in a commercial code and validated against experimental mean in-cylinder pressure and combustion indicators. The validated 1D engine model is coupled to an external optimizer, to identify the optimal calibration, performing multi-variable and multi-objective optimizations with the adoption of the MOGA genetic algorithm. The latter aims at minimizing the BSFC in a BMEP sweep, at fixed speed, while controlling the load through the Inlet Valve Closure (IVC) at fully opened throttle valve. The optimization results show that an advanced control of the intake valve strategy allows a maximum BSFC advantage of 26% at medium/high loads and medium speeds, if compared to the manufacturer-advised engine calibration. The outcomes of the optimization process are also confirmed by the 3D CFD tool. The latter not only contributes to the tuning of the 1D model, but it also provides an in-depth on detailed 3D aspects, such as turbulence and knock, that could not be assessed via a simplified 1D approach. The presented methodology represents a valuable tool to refine the virtual calibration of VVA engines and to support the design phase, thus remarkably reducing the experimental efforts. Moreover, it is a promising example of integration between 1D and 3D CFD tools.