Development of Engine and Fuel Co-Optimization Platform in modeFRONTIER

State of the art spark ignited gasoline engines achieve thermal efficiencies above 46 % e.g. due to friction optimized crank trains, high in-cylinder tumble flow and direct fuel injection. Further improvements of thermal efficiency are expected from lean combustion, higher compression ratio and new fuel blends. The new fuel components are selected for high laminar flame speed and high knock resistance. Additionally, the energy and emission balance of the fuel production is becoming more important in the selection process of new fuel components.

Quasi-dimensional Stochastic Reactor Model simulations with detailed chemistry allow to consider the thermochemistry of any (new) fuel component(s). Based on the specifications of the fuel, surrogate models are developed to accurately predict the laminar flame speed, auto-ignition in the end gas and emission formation. An innovative tabulated chemistry approach is utilized to generate dual-fuel laminar flame speed and combustion chemistry look-up tables. This allows to change the fuel composition on the fly and reduce the simulation duration by a margin. The modeFRONTIER optimization tool serves as the platform to develop the engine and fuel co-optimization process. The process is based on genetic algorithms, metamodels and the Stochastic Reactor Model using tabulated chemistry. The engine simulation is considered as a multi-dimensional optimization problem with 5-10 input parameters and 1-3 objectives. The Non-dominated Sorting Genetic Algorithm II developed by Deb and co-workers is applied for various technical problems and showed a robust and efficient performance. The extension of the genetic algorithm with metamodels improves the convergence of the Pareto Front to the theoretical best solution. The Stochastic Reactor Model with tabulated chemistry considers the complex turbulence-chemistry interaction, which drives the combustion and emission formation in spark ignited engines. Further, optimization durations of 5 – 10h on 5 CPU cores with 5000 designs are possible.

The aim of this work is the multi-objective optimization of four different fuel components blended with RON95 E10 fuel regarding high thermal efficiency and low CO2 emissions. Experimental measurements of a single cylinder research engine operated at eight fired operating points with RON95 E10 fuel are used to train and validate the simulation model. The RON95 E10 fuel is described by a 4-component surrogate and blended with Methanol, Hydrogen, Toluene and Nitromethane. The different fuel blends are optimized for three operating points at 1500 rpm 15 bar IMEP, 2000 rpm 20 bar IMEP and 2500 rpm 15 bar IMEP with high compression ratios and lean combustion.