On the experimental side Ford's responsibility was the investigation of the knock behaviour of a production engine under different operating conditions and with different fuels. 12 special test fuels covering a large range of RON/MON were tested. In addition Shell provided a primary reference fuel with fully known chemistry. For comparison and for dyno setup and check Ford also used Ford standard test fuel (94.8 RON/ 84.4 MON).
The evaluation of the data showed that all the fuels behaved differently with regard to their wide-open throttle performance. This can be attributed to the differences in heat value, evaporation phenomena, stoichiometric air-fuel ratio and combustion characteristics. All effects were analysed separately. The influence of stoich AFR and heat value was discussed in detail. The effects of spark variation are shown in terms of indicated mean effective pressure, volumetric efficiency and combustion data (10% burn duration, 50% burn duration).
The correlation of knock sensitivity with RON and MON number respectively shows well known tendencies. A linear relationship between spark advance and RON number corresponds to well known tendencies. This correlation could be demonstrated for all engine speeds, however the correlation factor deteriorates with increasing engine speed. For the highest engine speed, which was investigated, the MON number shows a better correlation with regard to spark advance than the RON number. The evaluation of the Octane Index as proposed by Prof Kalghatgi/Shell shows a potential improvement to characterize fuel with regard to knock sensitivity. This index is based on a combination of RON and MON.
The computational part of the project included reference case computations. In a first step meshing was performed for further analysis with STAR-CD, since this commercial code offers a highly sophisticated combustion model.
Three operating points were chosen for reference simulation of the combustion process: 1500 rpm part load, 1500 rpm full load and 5000 rpm full load.
An assessment of the performance of the Star-CD ECFM combustion model was performed. The analysis of the above reference test cases showed that the combustion model parameters have to be adjusted individually for each operating point. If only the recommended model parameters were used this resulted in unacceptable differences between measured cylinder pressure and mass fraction burned curves.
In a second step meshing of the ports and the combustion chamber was also completed for subsequent analysis with FIRE.
Similar operating points were chosen for reference simulation of the combustion process and the knock onset: 1500 rpm, 2000 rpm and 5500 rpm, all at full load.
In a large range of CFD simulations different approaches for wall and heat transfer, turbulence modelling and combustion simulation were tried out.
As a summary it can be stated that the results with k-x-f model look more physically and reliable during gas exchange phase compared to k-e.
For the simulation of the combustion period an Eddy-Break-Up model was compared with the AVL ECFM approach. Comparison with measurements was based on mean in cylinder data such as cylinder pressure, mass fraction burned and in cylinder temperature. The simulation showed correct trends for all operating points, however changing model parameters are required for accurate prediction of the combustion period.
In addition a comparison of calculated temperature data with measured radical concentrations was done. The measured data were obtained within the MinKnock project at the optical engine at CNR, Naples. The correlation was very good. This is indicating that the calculated flame front propagation correlates well with real engine operation.
Finally knock simulation was performed using the Shell Auto-Ignition model. A criterion was found providing a good correlation between calculated radical mass fraction and measured knock onset. This seems to be a promising approach to predict knock probability in the future.