The European Commission legislation concerning the reduction of CO2 emission limits until 2020, has encouraged the car manufacturers to develop cleaner and more efficient combustion engines. Strategies for lean gasoline combustion using stratified operation in the lower part load range and ultra-lean homogeneous operation for higher engine loads have a huge potential to reduce fuel consumption and CO2 emissions. Improved de-throttling at low loads in combination with favourable gas properties and reduced wall heat losses are the main benefits. However, high-speed Particle Image Velocimetry (PIV) measurements have shown that the cycle-to-cycle variation (CCV) of the in- cylinder flow correlate significantly with the CCV of the indicated mean effective pressure (IMEP) in stratified engine operation.

The goal of this research work is to understand the 3D-flow phenomena in the chain of in-cylinder processes leading to the CCV of the IMEP. The current PhD thesis comprises a joint experimental and numerical investigation of the in-cylinder flow in a direct injection spark ignition (DISI) engine during stratified engine operation. High-speed PIV measurements are carried out quasi simultaneously in the central tumble and the intake valve plane in an optically accessible single-cylinder engine derived from the Mercedes-Benz’s M274 production engine. Multi-cycle Large-Eddy Simulation (LES) is performed on the same engine geometry with the research code AVBP co-developed by IFPEN and CERFACS. AVBP was chosen among other commercial Computational Fluid Dynamics (CFD) codes due to its low-dissipative explicit acoustic solver and high mesh quality standards.

As there is today no universal reliable LES validation strategy available, the question remains open on which level of detail a validation is required. A new PIV-guided LES validation strategy based on the local and global in-cylinder flow structures is proposed. The main targets are defined: First, to reproduce the experiments in terms of the mean flow structures and flow fluctuations; secondly, using conditional statistics to confirm the flow hypothesis derived from PIV that reveals the formation of a local ‘upward flow’ below the spark-plug impacting combustion; thirdly, to extract a more detailed understanding of the 3D in-cylinder flow structures leading to the formation of the ‘upward flow’; finally, to identify the key in-cylinder flow parameter influencing the CCV of the IMEP. In order to assess the full potential of LES, the flow prediction using the 2 nd and 3 rd -order centred in space convective schemes Lax-Wendroff and TTGC is also performed. In the experiments, the missing link between the CCV of the in-cylinder flow and the CCV of the IMEP is the air-fuel ratio (lambda) information within the cylinder. The Lagrangian spray simulation in LES of the Piezo- actuated Pintle-type injector (Bosch HDEV 4.1) is addressed in this research work. The primary droplet break-up process is derived from the transient internal-nozzle flow simulation computed with the Reynolds-Averaged Navier-Stokes equations (RANS). The initialisation of the spray is achieved through a coupling interface in the LES computational model. A methodology featuring a spray adaptive region is developed and validated against closed-volume chamber spray measurements. The integration of the validated spray methodology into the gas-exchange mesh is achieved using a new moving-mesh strategy specially developed for the explicit acoustic LES solver of AVBP. A multi- cycle LES with injection is performed in stratified engine operation with the aim to investigate the CCV of the key in-cylinder flow parameters before the injection and the CCV of the lambda distribution at ignition time.

A complete chain of cause-and-effect of the successive in-cylinder processes is derived, starting from the in-cylinder flow characteristics before the injection until the lambda distribution around the spark-plug at the start of combustion. Those findings increase significantly the level of knowledge and understanding of spray-guided combustion systems. Solutions are proposed to reduce the CCV of the IMEP in the current engine configuration. The PhD thesis underlines the necessary complementarity of PIV and LES to solve complex 3D-flow problem.