• No results found

Engine configuration and experimental procedures

5.2 Formaldehyde and hydroxyl in HCCI combustion

5.2.2 Engine configuration and experimental procedures

The engine used in the experiments is a 0.5 l single-cylinder optical HCCI engine. The engine was equipped with a port-fuel injection system, which generates a mainly homogeneous charge.

The fuel used was a blend of equal mass percentages of iso-octane and n-heptane, and the air-fuel ratio for all engine runs was roughly 3.0-3.1.

Figure 5.27 The optical engine used for the experiment.

75

A wide-band lambda sensor was used to measure the air/fuel ratio. In order to maintain 50%

heat released at 5-6 CAD ATDC, the inlet air temperature was governed. This temperature was 120°C at startup and was decreased to 80-90°C at the end of the measurement, as more and more heat was accumulated in engine inlet and cylinder walls.

Table 5.6 Engine specifications.

Type Four-valve one-cylinder optical HCCI

Engine speed 1200 rpm Displacement (1 cylinder) 0.5 l

Bore 81 mm

Stroke 93 mm

Compression ratio 12:1

Fuel 50% isooctane

50% n-heptane

To provide optical access from below, the piston was elongated, equipped with a quarts window, and a 45° mirror was mounted beneath the transparent piston. Horizontal access to the upper part of the cylinder liner was provided through a quartz cylinder liner. Thus, the laser sheet was sent horizontally through the quartz cylinder liner, and the measured signal was detected from below through the window in the piston. The path of the laser sheet through the cylinder is as shown in Figure 5.28.

Figure 5.28 The path of the laser beam through the cylinder.

Two laser sources, firing in sequence with a pulse separation of 1 micro-second, are used for the two LIF measurements. The laser used for OH-LIF is working at 283 nm wavelength with 30

1 cm

76

mJ pulse energy. The laser used for formaldehyde LIF is working at 355 nm with 75 mJ pulse energy. The two laser pulses are superimposed on top of each other and formed to horizontal laser sheets with a width of 40 mm covering half the cylinder bore. The LIF signals from the two probed species are separated before they are collected on two ICCD camera systems equipped with filters appropriate for formaldehyde- and OH LIF signal, respectively. The optical set-up of the LIF system and the engine are presented in detail in an earlier paper [67]. In that work and in references there-in additional information about the formaldehyde- and OH LIF technique can also be found.

Experimental procedures and results

LIF measurements were performed in a time range starting well before the onset of low-temperature reactions and continuing beyond the major, high low-temperature heat release (HTHR). For every measured crank angle position 20 image-pair of LIF signals from formalde-hyde and hydroxyl radicals were collected, and the corresponding pressure trace was recorded.

Figure 5.29 Superpositioned LIF-images from start of low-temperature combustion till the end of high temperature regime. Formaldehyde is shown in green and hydroxyl in red. The images were digitized in order to clearly show the in-stantaneous structure of each species.

In order to compensate for background signals, e.g. from lubricant oil and scattered light, back-ground images were obtained from motored cycles. This backback-ground was then subtracted from the measured signal before the data was evaluated. For clarity, the presented image-pairs of hy-droxyl and formaldehyde signals were color-coded and superimposed to one single image show-ing the simultaneous signals of both OH and CH2O. This allows gaining knowledge on the spatial structures of the combustion progress. In these combined images, LIF signals correspond-ing to formaldehyde are colored green (light-grey) while OH signals are colored red (dark-grey).

  – 17 CAD – 12 CAD – 7 CAD +2 CAD

+6 CAD +10 CAD +15 CAD +21CAD

77

Examples of superimposed LIF images are found in Figure 5.29 and of averaged cycle resolved signals in Figure 5.30.

Figure 5.30 Averaged experimental formaldehyde and hydroxyl signals, and heat release.

Formaldehyde and hydroxyl signals are normalized (a.u. = arbitrary units).

Conditional filtering of LIF-images

The maximum peaks in the measurement noise are well above 10 % of the maximum measured signal at any measurement, whereas the average noise levels are below that. A large portion of the noise found in these images origin from the CCD and amplifier in the camera used in the mea-surements. This noise is white noise, meaning that it is non-periodical; no specific frequency can be associated with it.

In order to generate joint probability density plots (PDP) (see “Calculation results – relative distributions of hydroxyl radicals and formaldehyde from SRM calculations” and “Calculations results – comparison of particle/pixel distributions” for more information about joint PDPs) from the LIF data, it was necessary to remove noise from the raw images. Otherwise the joint PDPs would contain information that is not useful, and the joint PDPs would display stochastic variations in the noise rather than demonstrating systematic changes in the combustion process.

Since substantial parts of the noise is white, no periodicity could be associated with it and thus signal processing through Fourier transformation is not an adequate method. Instead conditional filtering was used, a straight-forward method that could easily be implemented in Matlab. If correctly implemented there are with this method only minor and less critical distortions for these images of the remaining signal. Caution must be taken however that the threshold levels are carefully distributed and set, otherwise the filtering will give substantial impact on the PDPs.

The best way to determine whether the filtering was correctly implemented is to compare fil-tered PDPs with unfilfil-tered.

0 0,2 0,4 0,6 0,8 1

-30 -20 -10 CAD 0 10 20

OHandCH2OIntensity(a.u.)

0 5 10 15 20 25

RoHR(J/CAD)

CH2O OH RoHR

78

In this method, a number of preset threshold levels for OH-signal and CH2O-signal were set up.

The threshold levels were investigated in descending order. The pixels in the LIF-images were systematically investigated and compared to surrounding pixels 2 positions away. If a pixel ex-ceeding a certain threshold level was surrounded by two pixels in either direction also exex-ceeding the same threshold, it was considered a signal and left retaining in the image. Otherwise the pixel signal was cropped down to the nearest lower threshold level and then again compared to its surrounding pixels.

Figure 5.31 Left: Remaining signal from valves in CH2O-signal at -27 CAD TDC af-ter filaf-tering. Right: Signal (afaf-ter filaf-tering) in the OH frequency area during the growth phase of CH2O (-14 CAD) and thus long before the occurrence of the major heat release. Shown is a trace that was not possible to filter, remaining throughout most parts of the combustion. This signal is believed to origin from broadband dispersing ketones.

A lowest threshold level was also set, below which no signals were considered. This lowest thre-shold level supposedly filters the background noise, not really related to the act of measuring but rather to the signal independent noise in the measurement equipment; CCD thermal noise, CCD on-chip electronic noise, and amplifier noise. In this case the cut-off level was set to 4% of the maximum concentration of either species.

The conditional filtering method is quite efficient, and what are only partially filtered are reflec-tions from the contour of the valves as seen in Figure 5.31. Luckily this contribution does not affect to any noticeable degree the joint PDPs. Moreover there was also found some faint but clearly visible signal in the OH frequencies when the presence of OH was not expected (Figure 5.31). This signal occurs in the region crossed by the laser beam and starts almost where the first signs of low-temperature oxidation is found, and disappears during the main heat release. It is

79

believed that this signal comes from broad dispersing ketones. This signal could not be filtered away, and a visible imprint from this signal was found in the joint PDPs.