• No results found

C YCLE - TO - CYCLE AND CYLINDER - TO - CYLINDER DEPENDENCE

The effect of cylinder-to-cylinder deviations and also the cycle-to-cycle coupling and variations are important issues because these effects limit the obtainable working region. The engine relies on trapped hot residuals to attain auto ignition temperature.

6 Results

Due to the hot residuals the combustion in one cycle can affect the next in a much more extensive way than an SI or HCCI engine run with little or no residual gas.

For HCCI combustion with trapped residuals at low temperature conditions, periodic behavior has been observed [52] i.e. a late burning cycle due to low charge temperature generates high temperature residuals which elevates the charge temperature for the next cycle resulting in early combustion timing generating low charge temperature and so on. One way of analysing this phenomenon is to use calculate a correlation coefficient, as shown in (6.1), between two consecutive cycles.

The correlation coefficient (R) is a linear correlation coefficient between two parameters. In this case it is between CA50 of cycle i and CA50 of cycle j. If R = 0 then no correlation is found. If R = -1 then a perfect inverted linear correlation is found which implies that a high value is followed by a low etc.

) , ( ) , (

) , ) (

,

( C ii C j j

j i j C

i

R = × (6.1)

Figure 39 shows the correlation for the whole sweep in speed along the boundary between SACI and pure HCCI as shown in Figure 27. Between 2000 and 3500rpm some negative correlation can be seen. Especially at 3000rpm the negative correlation for Cylinders 2, 4 and 5 is very high. These cylinders also have relatively high COVimep. Cylinder 3 has, on the other hand, also high COVimep, but a low R, so a high COV does not have to imply a high correlation, but vice versa!

A correlation investigation is also done with an increased cycle separation for a vector containing 100 consecutive measured cycles; this is shown in Figure 40. The figure shows cylinder 5 for the measuring point at 3000rpm with an R of approximately -0.8, which is a strong negative correlation. Here an auto correlation is done for an offset of up to 55 cycles. It is obvious that the largest correlation is for a separation of one cycle, which is the same as shown in Figure 39. For an increased offset in cycles the correlation effect is dampened out, still with an increased negative correlation for every second cycle. Although the combustion timing fluctuates back and forth from one cycle to the next some slower variations in combustion timing due to parameters such as wall temperatures will decrease the correlation. Still a higher correlation is recognized again for a higher cycle separation. This is an indication of a disruption in the cycle-to-cycle behaviour, i.e. the periodic behaviour ceases and starts again, in a repetitive way. It should be noted though, that the correlation for higher cycle separation is overall low.

One reason for the self-extinction of the periodic phenomenon is that when the changes in combustion timing gets so high that partial misfire occurs, the very late combustion timing will no longer increase the charge temperature for the next cycle.

Instead the next cycle will also burn late. Here the phenomenon can be extinct or start again in counter phase to the earlier behaviour. If the amplitude gets even higher it will not be self stabilizing, instead of a very late combustion, misfire will occur. If misfire occurs no residuals are present for the next cycle to initiate combustion again.

A cross-correlation between the different cylinders for an increased number of cycles separation is conducted, also looking at CA50. The correlation from one cylinder to another cylinder in the next cycle is negligible. No communications between the cylinders have been observed in terms of combustion timing variation.

6 Results

1000 1500 2000 2500 3000 3500 4000

−0.8

−0.6

−0.4

−0.2 0 0.2 0.4 0.6

Correlation [CA50]

Speed [rpm]

Cyl 1 Cyl 2 Cyl 3 Cyl 4 Cyl 5 Cyl 6

Figure 39. Correlation between cycle n and cycle n+1

Figure 40. Correlation between cycle n and cycle n+k, k= 1:55

By running SACI it is possible to lower the load compared to what is possible with pure HCCI, but the increased residual rate will eventually limit the usage of spark assistance. When the load is decreased the exhaust temperature and thus the residual temperature goes down, although the amount of residuals is increased. To reach even lower load the effect from the spark needs to increase. To a certain extent it can be compensated for by advancing the spark timing. However, too advanced spark timing will again increase the cycle to cycle variations. The conditions in the vicinity of the spark plug deteriorate due to lower turbulence, gas composition and low temperature.

Increased combustion stability and increased low load capability by spark assistance is supported in later work by Xie et al. [54].

At lowest possible load with SACI combustion the correlation R is again investigated.

The correlation is shown in Figure 41. The oscillations are increased compared to the higher load shown in Figure 39 for pure HCCI. The oscillation for single cylinders also increases with speed. This is expected as the influence from the spark should be higher at lower speed since the load is higher and thus the residual fraction is lower.

In both correlations the case at 1000 rpm is run in SACI mode. Here it goes from no oscillation tendencies at the higher load case to a clear negative correlation for the lowest possible load. In this case it is cylinder 4 and 6 that oscillate and set the low load limit.

1000 1500 2000 2500 3000

−1

−0.8

−0.6

−0.4

−0.2 0

Correlation [CA50]

Speed [rpm]

Cyl 1 Cyl 2 Cyl 3 Cyl 4 Cyl 5 Cyl 6

Figure 41. Correlation coefficient [R] for spark assisted HCCI at low load.

6 Results