Ignition and flame inception are well recognised as affecting performance and stable operation of spark ignition engines. The very early stage of combustion is indeed the main source of cycle-to-cycle variability, in particular in gasoline direct injection (GDI) engines, where mixture formation may lead to non-homogenous air-to-fuel distributions, especially under some speed and load conditions.
From a numerical perspective, 3D modelling of combustion within Reynolds Averaged Navier Stokes (RANS) approaches is not sufficient to provide reliable information about cyclic variability, unless proper changes in the initial conditions of the flow transport equations are considered. Combustion models based on the flamelet concept prove being particularly suitable for the simulation of the energy conversion process in internal combustion engines, due to their low computational cost. These models include a transport equation for the flame surface density, which needs proper initialization. A flame collocation is indeed to be properly made when starting the calculations, often just based on the user's skill and without resorting to any quantitative data derived from experiments. However, the way to define initial conditions for cyclic variability prediction is often based on just statistical considerations.
This work aims at exploiting information derived from images collected in a single cylinder 4-stroke GDI engine to properly collocate the flame at the start of the combustion calculation. The considered engine is optically accessible through a wide fused-silica window fixed on the piston crown having a Bowditch design. Image processing methodologies are applied to evaluate local and integral luminous intensity, and flame morphology parameters. The collected data allows improving the numerical simulation and gaining hints about the main parameters defining the engine cyclic variability.