Elsevier

Combustion and Flame

Volume 136, Issue 3, February 2004, Pages 283-302
Combustion and Flame

The nature of early flame development in a lean-burn stratified-charge spark-ignition engine

https://doi.org/10.1016/j.combustflame.2003.08.011Get rights and content

Abstract

The operating range of lean-burn SI engines is limited by the level of cycle-by-cycle variability in the early flame development, which typically corresponds to the 0–5% mass fraction burned. An experimental investigation was undertaken to study this flame variability in an optical, stratified-charge, SI engine close to the lean limit of stable operation (A/F=22). Double-exposed flame images acquired through either a pentroof window (“tumble plane” of view) or the piston crown (“swirl plane” of view) were processed to calculate the intra-cycle flame growth and convection rates under 1500 RPM low-load conditions. Projected flame-boundary analysis was also performed to investigate the effect of flame shape/wrinkling on the subsequent timing of 5% mass fraction burned on a cycle-by-cycle basis. The images showed that the flame always preserved its shape while growing in size (even when it had been initiated with a highly convoluted shape); image processing demonstrated the manner with which the flame-growth speed varied as the flame propagated and approached the pentroof and piston-crown walls for slow, “typical” or fast burning cycles. It was found that it was beneficial to have a high convection velocity in the swirl plane of flow during the first 10° CA after ignition timing (corresponding to less than 0.1% mass fraction burned), but after this stage it was beneficial to have a moderate convection velocity for the flame. However, on the tumble plane of flow, a high convection velocity was preferable up to 30° CA after ignition timing (corresponding, typically, to 1% mass fraction burned). Slow development of a flame was associated with higher stretch rates for the same flame radius than fast-developing flames during the period of growth from 3 to 6 mm in radius (about 0.1–1% mass fraction burned). Extended analysis of the projected flame front's shape and its wrinkling showed that the fastest lean-condition flames had contour characteristics similar to those of the flames recorded for stoichiometric conditions. This suggested that the fastest lean flames on a cycle-by-cycle basis might have been richer than the average in the vicinity of the spark plug at ignition.

Introduction

Lean-burn spark-ignition (SI) engines are an attractive means for powering vehicles from the environmental point of view because of their higher efficiency and lower specific emissions compared to stoichiometrically charged piston engines [1]. Broadly, it is advantageous to make such engines operate as lean as possible, but as the air to fuel (A/F) ratio of the mixture is increased above stoichiometric, cycle-by-cycle variations in flame development and, consequently, in indicated mean effective pressure (IMEP) become more prominent and eventually limit the range of lean-burn operation, because these cause very poor driveability [2] for a car. Therefore, the ultimate aim of engine research in this area is to provide the understanding necessary to extend the current limits of A/F ratio at which engine designs operate.

It has been known for some time that lean-burn engines must have charge stratification, in the sense that the ensemble-averaged mixture near the spark plug must be richer than the spatial average in the cylinder [3]. However, it was only recently that research demonstrated the flow structures which lead to successful lean-burn engine designs. In this context, the studies of Berckmüller et al. [4], Hardalupas et al. [5], and Carabateas et al. [6] can be quoted; they all used engine geometries very similar to the one of the present study. The former used laser-induced fluorescence (LIF) to measure the A/F ratio in planes near the spark plug throughout the intake and compression strokes of an optical single cylinder research engine. The latter two publications used phase Doppler anemometry (PDA) and laser Doppler velocimetry (LDV), respectively, to measure the in-cylinder droplet size and velocity, as well as gas flow velocity. These measurements demonstrated the existence of two major “tumble” and “swirl” circulatory flow patterns about axes, respectively, parallel to and perpendicular to the axis of the cylinder, while recently, Aleiferis et al. [7] presented visualization of the resulting combined three-dimensional flow pattern.

The implications of these three-dimensional flow structures for extension of the lean-burn limit of engine operation need to be examined in conjunction with the cyclic behavior of the combustion process. Of the four main stages of combustion in SI engines (i.e., spark and flame initiation, initial flame-kernel development, fully developed turbulent flame propagation, and flame termination), it is only the first two that are important in terms of cyclic variations in IMEP [1]. Specifically, there is a strong correlation between large IMEP and small values of the crank angle (CA) at which 5% of the mass fraction is burned (MFB) (θXb5%) under low-load operations [8]. The figure of 5% is arbitrary in the sense that it has been determined by experimental convenience: it is hard to measure lower values of MFB from the indicator diagram because of uncertainties in the measured in-cylinder pressure. Therefore, different techniques must be employed to investigate the speed of initial flame-kernel growth at CA degrees closer to ignition timing than θXb5%. Imaging of the initial flame-kernel growth has proved particularly useful in this context, because the minimum flame size that can be adequately resolved by imaging corresponds to tenths of one percent of MFB, but most of the reported work has been conducted mainly under skip-firing conditions either in square-piston engines or with 2-valve engine heads and disk-shaped combustion chambers, based mainly on the Schlieren technique (e.g., [9], [10]). Combustion in pentroof-head geometries has been investigated by direct imaging of the light emitted by the flame using image-intensified CCD cameras [11], [12], [13], [14]. These studies presented some limited relationships between calculated flame areas and peak in-cylinder pressure (Pmax), combustion duration, and θXb5% or θXb10%. The recently published work of Aleiferis et al. [15] was concerned with extensive stereoscopic CCD imaging of the flame in a pentroof-type stratified-charge SI engine, which permitted estimation of the flames volume during the critical stage of 0–5% MFB. Note that the timing of 5% MFB corresponded to about 40° CA after ignition timing (AIT) for A/F=22, low load (30% volumetric efficiency), and 1500 RPM in the engine investigated. It was found that the “quality” of a cycle was determined within the short duration of the spark event (≈20° CA), at the end of which the flame had grown, on average, to a radius of 4 mm, corresponding to less than 0.2% MFB. Specifically, it was found that the strength of the correlation between, for example, kernel volume and θXb5% decreased from almost unity at 40° CA AIT to 0.5 at 10° CA AIT, being as high as 0.75–0.8 at 20° CA AIT.

On the basis of experiments with homogeneous mixtures of air + fuel, controlled levels of turbulence and no “mean” velocity in combustion bombs (e.g., [16]) or in simple compression machines (e.g., [17]), it is known that the effect of turbulence on flame shape/wrinkling is critical in producing fast flame-kernel growth and, historically, such effects have been associated with cyclic variations in IMEP [2]. However, Lecordier [17] observed in a rapid compression arrangement that, although spatially averaged measurements indicated that higher levels of imposed turbulence (mapped by particle image velocimetry, PIV) resulted in higher burning rates (quantified by laser-sheet flame tomography), individual cycles showed results occasionally contradicting this trend and he concluded that it is important to look for factors associated with such averaged and individual-cycle effects. Additionally, the LDV studies in SI engines have been contradictory in terms of whether it is the local “mean” (i.e., ensemble averaged) or RMS velocities around the spark plug at ignition timing that influences the subsequent burning rate on a cycle-by-cycle basis [18], [19], [20]. For the engine used in the current study, neither mean nor RMS velocity at locations close to the spark plug was found to affect θXb5% or IMEP [8] significantly on a cycle-by-cycle basis, despite the findings of previous studies on engines of similar configuration [20]. This remains a surprising conclusion in terms both of intuitive expectation of the effects of turbulence on flame growth due to wrinkling of the flame and of other velocity-related effects, such as either heat loss to the spark plug electrodes or effect of stretch rates. However, it could be argued that measurements of individual velocity components at isolated points and ensemble averaged at particular instants of crank-angle phase may be deficient in revealing the effects of spatially and temporally distributed turbulence on the critical period (up to 40° CA) and region (up to 10 mm in radius) over which initial flame-kernel growth takes place (and which may include effects of flame interaction with the pent-roof walls). Thus, it is possible to hypothesize that individual flame kernels can, during their initial propagation, change their growth rates over the first 30°–40° CA due to turbulence-induced change in flame shape/wrinkling, due to variations in the turbulent flow field encountered by the growing flame, due to changing amounts of heat lost to the electrode, and due to magnitudes of flame stretch.

The aim of this work was to establish the relative importance of the variables, derived from flame images, on the kernel's rate of growth over the first 40° CA AIT. Drawing from earlier studies and extending the single flame-image-per-ignition-stroke technique of Aleiferis et al. [15], the current study is concerned with measurements of temporal and spatial growth of individual flames on a consecutive, cycle-by-cycle basis, using variable-lapse, double CCD imaging per cycle. The objectives were to quantify the flame-kernel growth and convection rates, the latter acting as a convenient (even if imperfect) measure of heat loss to the electrodes, of small-scale flame wrinkling and of large-scale flame distortion on the timing of 5% MFB in a close-to-production, continuously firing, lean-burn, four-stroke SI engine. The advantage of making measurements in a reciprocating engine, as opposed to a “rapid compression” machine or a combustion bomb, is that this has levels of turbulence and mean convection which are representative of the configuration of technical importance. We analyzed the behavior of flames in the two principal planes of in-cylinder flow motion (the so-called “tumble” and “swirl” planes) and compared this behavior extensively with published flame-growth data from other engines. In this way, we sought to establish what, if any, were the effects of the “whole” flow field, including its temporal development, on kernel growth in a stratified-charge lean-burn SI engine.

Section snippets

Engine and techniques

A single-cylinder research engine with optical access to the combustion chamber was used for the present work. This engine employed the Honda single over-head camshaft (SOHC) variable valve timing and lift electronic control (VTEC) mechanism for the inlet valves, which was introduced by Horie et al. [21] and is known as the VTEC-E mechanism. The research engine retained most of the characteristics of the Honda VTEC-E 1.5-liter, 4-valve per cylinder, pentroof-geometry mass-production engine,

Results and discussion

Figure 3 presents typical MFB traces and illustrates the definition of the CA of 5% MFB (θXb5%). Further, Fig. 4 shows the relationship between IMEP and θXb5% for stoichiometric (A/F=15) and lean (A/F=22) conditions. It can be seen that θXb5% occurred at about 20°–25° CA AIT for A/F=15 (340°–345° CA), but it took almost double that time for θXb5% to occur when A/F was set to 22 (i.e., 40° CA AIT or 360° CA). Both data sets were acquired with the same spark advance of 40° CA, so that the flames

Summary and conclusions

The operating range of lean-burn SI engines is limited by the level of cyclic variability in the early development of the flame, typically corresponding to 0–5% of the mass having burned. An experimental investigation was undertaken to study the flame behavior in an optical stratified-charge SI engine close to the lean limit of stable operation (A/F=22, Φ=0.68, port-injected isooctane). Double-exposed flame images, acquired through either a pentroof window (“tumble plane” of view) or the piston

References (58)

  • S.C. Bates

    Combust. Flame

    (1991)
  • J.B. Cole et al.

    Proc. Combust. Inst.

    (1981)
  • D. Bradley

    Proc. Combust. Inst.

    (1992)
  • G.P. Beretta et al.

    Combust. Flame

    (1983)
  • P. Clavin

    Prog. Energy Combust. Sci.

    (1985)
  • G.J. Germane, C.G. Wood, C.C. Hess, SAE paper 831694...
  • N. Ozdor, M. Dulger, E. Sher, SAE paper 940987...
  • A.A. Quader, SAE paper 820131...
  • M. Berckmüller, N.P. Tait, D.A. Greenhalgh, SAE paper 970826...
  • Y. Hardalupas, A.M.K.P. Taylor, J.H. Whitelaw, K. Ishii, H. Miyano, Y. Urata, SAE paper 950507...
  • N.E. Carabateas, A.M.K.P. Taylor, J.H. Whitelaw, K. Ishii, K. Yoshida, M. Matsuki, SAE paper 961923...
  • P.G. Aleiferis et al.
  • J.R. Ault et al.
  • T.A. Baritaud, SAE paper 872152...
  • S. Pischinger, J.B. Heywood, SAE paper 900021...
  • S.C. Bates, SAE paper 890154...
  • S.C. Bates, SAE paper 892086...
  • C. Arcoumanis, C.-S. Bae, J.H. Whitelaw, H.M. Xu, SAE paper 942052...
  • P.G. Aleiferis, A.M.K.P. Taylor, J.H. Whitelaw, K. Ishii, Y. Urata, SAE paper 2000-01-1207...
  • R.G. Abdel-Gayed et al.

    Proc. R. Soc. London Ser. A

    (1987)
  • B. Lecordier, Doctorat Nouveau Régime, Université de Rouen, 1997 (in...
  • B. Johansson, SAE paper 940990...
  • J.F. Le Coz, SAE paper 920517...
  • K. Horie, K. Nishizawa, T. Ogawa, S. Akazaki, K. Miura, SAE paper 920455...
  • K.N.C. Bray et al.
  • L. Gillespie, Ph.D. thesis, University of Leeds,...
  • G.F.W. Ziegler, A.P. Zettlitz, P. Meinhardt, R. Herweg, R. Maly, W. Pfister, SAE paper 881634...
  • A.O. Zur Loye, F.V. Bracco, SAE paper 870454...
  • R.A. Hicks, M. Lawes, C.G.W. Sheppard, B.J. Whitaker, SAE paper 941992...
  • Cited by (161)

    View all citing articles on Scopus
    1

    Current address: Department of Mechanical Engineering, University College London, Torrington Place, London WC1E 7JE, UK.

    View full text