In order to reach significant pressure rises and reduce experimental variability, gasoline vapour combustion investigation was performed in the combustion vessel. A fully pre-mixed gaseous mixture was prepared and in addition to pressure, heat release and emissions analysis, flame propagation speed was computed from direct high speed imaging. In current study the spherically expanding flame was analysed in terms of turbulent flame speed. According to Ferguson and Kirkpatrick [236] the relationship between the laminar and turbulent flame speed can be expressed as:
𝑆𝑡
𝑆𝑙 = 𝑎 (𝑈𝑡 𝑆𝑙)
𝑏
4.1
109
where St is the turbulent flame speed, Sl the laminar flame speed, Ut turbulent intensity and a & b are constant that depend on the geometry and specific conditions in the combustion chamber. The difference between the laminar and turbulent burning velocities can be 3-30 times [236]. Although laminar flame speed could be derived from the measurements, it was considered not to offer additional benefits as comparative fuel properties were under investigation and not each fuel separately.
Flame images were used to calculate the cross sectional area and the circumference of the propagating flame and on the assumption that the flame propagates spherically, radius increase per time step could be found. All raw images were background corrected and the threshold used for flame front detection was set based on the intensity of the specific combustion event. Electrode size was used for calibrating the pixels/mm value.
A sample of the image processing and time series of a combustion event can be seen in Figure 4.20. Figure 4.21 displays the radius change in time based on the two methods described. It can be seen that the circumference based measurement method produced calculated radius values twice as large as the area based method.
This was likely to have been caused by difficulties in flame edge detection. It was especially prominent feature at the beginning of the combustion event and at time instants after 35-40 ms, where high spark and flame intensities, respectively, caused misinterpretation of images due to limitations of the threshold values used.
Furthermore, a much steadier change in radius was found with area based calculations.
As such, it was decided that only the area based calculation of flame speed would be used, where the typical averaging time period was between 15-35 ms. Figure 4.22 displays typical pressure and heat release rate traces for gasoline vapour combustion. It can be observed that the peak pressure reached was two orders of magnitude larger than that for spray combustion. Additionally, repeatability in peak pressure and heat release rate over five combustions was found to be less than 2%
which was thought to be low enough for the current investigations. Furthermore, using a Lambda sensor, exhaust gas analysis showed that around 1.3% error in pre- combustion gas mixture composition was experienced over the same number of combustion events.
110
Figure 4.20: Time series of sample gasoline vapour combustion with initial pressure of 0.7 bar. Left to right, image processing procedure: raw image; background corrected image;
binarised image for area based radius calculation; binarised image for circumference based radius calculation
111
Figure 4.21: Radius change in time for area and circumference based calculation methods.
Time points t1-6 correspond to those shown in Figure 4.20
Initial base fuel characterisation was carried out under ambient initial pressure conditions. Although this was suitable for pressure/heat release rate and flame propagation speed analysis, it was found that emissions analysis suffered due to the output from mainly the oxygen sensor being flow dependent. It can be observed from Figure 4.22 that after a combustion event the pressure dropped very quickly to pre combustion levels. This meant that without opening air inlet to the combustion vessel,
Figure 4.22: Sample pressure and heat release rate traces for gasoline vapour combustion in combustion vessel at raised initial pressure and φ = 0.8
0
112
vacuum pressures were generated by the analyser which affected the readings.
Furthermore, allowing for extra air to be pulled into the vessel (to avoid vacuum pressures) diluted the exhaust gases quickly, and no stable readings could be recorded.
Therefore, initial pressure was raised to 0.7 bar which was found to be sufficient to allow for up to a 60 second period during which stable readings could be logged. Characteristic emissions recordings for raised initial pressure conditions can be seen in Figure 4.23.
Figure 4.23: Emissions readings from combustion vessel. At t1 data logging commences, at t2
the gas analyser sampling is activated and vacuum pressure is generated in gas supply line, at t3 the exhaust gases are released from the vessel and at t4 air inlet to the combustion vessel is opened to avoid vacuum pressures occurring. Typical area used for averaging is
displayed by the shaded area
Figure 4.24 displays the peak pressures for different equivalence ratios at atmospheric and raised initial pressure conditions. For both cases, the ignitability limits were found to be from around φ = 0.8 to φ = 1.8 and peak pressure occurred at around φ = 1.5 after which a sharp decrease in the pressure was observed. Due to soot production at higher than stoichiometric equivalence ratios, additive comparison tests were carried out under lean conditions. Any soot produced would give rise to errors in emissions analysis of subsequent combustion events, where burning of soot from the vessel walls could increase CO and CO2 readings. Moreover, this could affect
0
113
Figure 4.24: Peak pressures for atmospheric and raised initial pressure conditions at different equivalence ratios
pressure readings and constant depositing of soot on the vessel windows would mean significant timescales needed to clean the vessel. As such, tests were carried out at leanest possible condition which for raised initial pressure conditions meant 3 ml of gasoline fuel per vessel filling, giving an equivalence ratio of 0.79.