Printer Friendly

Reduced Convective Combustion Chamber Wall Heat Transfer Losses of Hydrogen-Fueled Engines by Vortex-Stratified Combustion - Part 1: Background and Optical Engine Observations.


With the conventional spark-ignited reciprocating ICE being under constant development and improvement for over a century, it has become apparent that a technological barrier is asymptotically approaching from being able to extract further step-wise thermal efficiency improvements. Doing so will necessitate novel approaches to reduce thermodynamic losses, namely incomplete and real combustion, wall heat transfer, pumping losses and friction, as shown in Eq. 1:

[mathematical expression not reproducible] (1)

Of these, Verhelst and Wallner (2009) [1], Eichlseder et al (2003) [2], Wimmer et al (2005) [3], Shudo (2007) [4] and several other researchers [5, 6, 7, 8, 9] identified wall-heat transfer as the predominant factor of reduced efficiency in hydrogen-fueled engines, and a key factor to its improvement. Oppenheim (2002) proposed an air-blast atomizer to direct the combustible charge away from the walls to prevent contact of the latter with hot combustion products and reduce heat transfer losses [10]. Heitland et al (1998) [11] showed several concepts of confining the combustion in a fireball surrounded by a layer of residual gas using air blast valves similar to those proposed by Oppenheim. Shudo and Oba (2009) [8] also stressed the importance of reducing heat transfer from the burning gas to the combustion chamber wall for improving the thermal efficiency in hydrogen engines. They found that the cooling loss fraction with hydrogen was about double that of methane (both operating stoichiometrically) over a range of tested ignition timings, and employed a split-injection technique to direct high equivalence ratio mixtures at a desirable position such as the spark plug, and slow the diffusion toward the walls in order to control the spatial distribution of the mixture for achieving cooling loss reduction and thermal efficiency improvement in hydrogen combustion engines.

Therefore, the main overall objective of the present work is the systematic mitigation of wall heat transfer losses by some means of imposing a designed regime of charge and flame extent stratification in the combustion chamber for reduced thermal convection. Here it is proposed to achieve this by a novel method of confining the fuel-concentrated charge in the geometric center, bounded in-between with an insulating layer of essentially pure air at the periphery adjacent to the walls that is substantially maintained during combustion. A theoretical framework describing the flow dynamics of the proposed system is provided. Schlieren observations from an experimental optically accessible engine are presented. The visualizations demonstrate the effectiveness of the vortex-stratified concept in setting up the vigorous rotational flow field that forms a radial density gradient on the prepared charge mixture before ignition, which indicates a gas species stratification. The second part that proceeds this paper provides an understanding of the basis for heat transfer loss reduction by correlating the experimental observations presented here with CFD analysis that illustrates the vortex-stratification of gas species in greater detail than is possible from schlieren images alone, and quantifies the heat flux, which is found to be 50% less compared to a combustion chamber design that actualizes a homogeneous mixture without specific charge motion directionality, with injection timing and all other model parameters remaining equal.


Wall Heat Transfer Losses

In the absence of appreciable bulk motion, turbulence, mixture composition non-uniformity and local temperature variations in the unburned charge, a premixed, spark-initiated flame front expands roughly spherically from the point source of ignition with a thin reaction sheet at the laminar flame speed [S.sub.L]. Increased turbulence levels cause a wrinkling and convolution of the flame front and greatly enhances the burning rate [12]. The flame front propagates throughout the combustion chamber until extinguished by one or more factors such as oxygen depletion, fuel availability within combustible limits, or by quenching. Norton and Vlachos (2003) state that flames are quenched because of two primary mechanisms: thermal and radical [13]. Thermal quenching dominates here and occurs when sufficient heat is removed through the walls that combustion cannot be self-sustained.

The prevailing burned gas temperature can be approximated as substantially constant throughout the combustion chamber but with a steep gradient in the thermal boundary layer adjacent to the wall until it matches with that of the gas-side of the wall surface, as shown with the long-dashed curve in Fig. 1. If the extent of the flame can be limited to some distance from the wall (red dashed line) - greater than the thermal boundary layer thickness and inherent flame quenching distance but without leaving any unnecessary unburned fuel to affect the combustion efficiency deleteriously - the temperature profile would take the shape as sketched qualitatively with the solid curve with a resultant reduced heat flux to the wall.

Vortex-Stratified Combustion

Figure 2 shows a 2D schematic of a cylindrical combustion chamber of radius R, although in general it could be of arbitrary geometry. Here, it is realized as separate from the main chamber containing the piston and cylinder, and connected with channels, of which there have been many different design embodiments over the years, variously called divided-chamber, auxiliary-chamber, turbulent-chamber, pre-chamber, swirl-chamber, etc. In the proposed process, a rotational flow would be imparted from incoming jets of air from the main cylinder entering into the combustion chamber through two side channels offset vertically from the center. The directly injected fuel would be concentrated within combustible limits - which also limits the extent of the flame - to a radius [r.sub.x], and the difference [delta] makes up the width of the air layer. The flow is said to be highly ordered because the channels and geometric symmetry impart momentum in 2-dimensions, while the component along the centerline axis is negligibly small. The resulting motion is one of bulk rotation and minimal cross flow.

Toulson et al (2010) gives an extensive review of pre-chamber initiated jet-ignition combustion systems derived from SI engines [14]. While the present work can be considered a form of turbulent jet or "torch" ignition system, important distinctions also become apparent:

1. The review focuses upon small pre-chamber systems whose volume comprises <3% of the total clearance volume. Here, is it in the order of 50% and furthermore, it is preferable that this proportion be maximized. In fact, in the ideal, limiting case, there is no divided chamber and there is no criteria for one at all; the principle described in this work can completely dispense with the separate volume and ostensibly be realized as an open combustion chamber of arbitrary shape formed by the geometry of the cylinder, head and piston crown that avoids additional surface area, as well as the throttling and thermal losses associated with the narrow connecting passages, thereby further improving the thermal efficiency in the same manner seen in direct-injected Diesel engines over the indirect-injected ones;

2. Torch-ignition systems are premised upon a stoichiometric to slightly rich fuel-air mixture inside the pre-chamber located within or in close vicinity to the spark plug electrode gap in order to establish a robust flame kernel, which in turn issues into and burns up a much leaner mixture in the main cylinder volume at faster rates than can be achieved with uniformly lean mixtures; the overall or global mixture trapped and burned in each cycle remains lean or stoichiometric. Here, all the fuel is injected into the divided chamber and the main cylinder is desirably composed of only pure air, like an indirect-injected Diesel engine. Therefore, the heat release takes place and completes substantially within this chamber, so that only already burned gas issues through the connecting channels, which transfers heat to the main cylinder via mass transfer, mixing, conduction and convection rather than by a secondary combustion process;

3. Pre-chamber concepts have one or more connecting orifices designed to produce a highly turbulent flow but no ordered motion within the chamber [12].

At least since the mid-1970s, there has been work on a different classification of divided-chamber SI engines resembling the Ricardo Comet swirl chambers commonly seen in indirect-injection Diesel engines [15, 16, 17]. Like the Diesel counterpart, charge air enters from the cylinder into the swirl chamber tangentially, imparting a rotational flow. Unlike the Diesel process, however, the main cylinder charge still contains a lean fuel-air mixture rather than air alone. Additional fuel is introduced directly into the swirl chamber by means of an injector or valve. Peterson and Alkidas (1983) [18] show that the velocity near TDC in a divided Ricardo Comet swirl chamber using a rapid compression machine is in the order of 15-20 [+ or -]5 m/s at the center and outer extents of the chamber, respectively. Since the velocity does not scale with the radius, the air motion cannot be characterized as a solid-body rotation. In contrast, the charge motion in this work will be shown to behave closely to solid-body rotation through a wide range of radii at up to 5-10 times higher tangential velocities, and by consequence much higher angular accelerations.

Arcoumanis et al. showed the advantages of extending the lean combustion limit and enhanced burn rate offered by local charge stratification near the spark plug within an overall ultra-lean rotating mixture. The constant volume chamber had a tangential port for admission of a lean charge and the rotating bulk flow had mean velocity of up to 8 m/s [19].

Although superficial similarities might be noted between the surveyed and the present works, the authors know of no literature or working concept to date that specifically exploit the combined mechanisms of the gravitational effects of a rapidly rotating vortex flow on different-density gases and the Coanda effect to achieve the proposed simultaneous charge confinement and stratification in premixed-charge engines.


Flow Dynamics

Diffusion J of any species combination is explained by Fick's first law as given in Eq. 2:

J = -D[nabla]c (2)

The diffusive coefficient is denoted D and the negative sign indicates that the mass flux follows along a decreasing spatial concentration gradient [nabla]c. If one considers two initially separated chambers each containing pure, quiescent gas of one type or another, when the partition is removed, diffusion of the two streams will take place at the prevailing diffusive coefficient - for hydrogen-air, this is given as 0.61E-4 [m.sup.2]/s at 300 K and 1 atm [1] - until the concentrations of both gases reach equilibrium spanning the combined volume, thus [nabla]c [right arrow] 0.

Turbulence and charge motion greatly enhance diffusion, which allow homogenization of the introduced fuel-air mixture over the space of the combustion chamber in the very short timescales encountered in engines. Heywood states that turbulence levels scale linearly with engine speed (specifically mean piston speed) [12]. Increased turbulent flame speeds and burning rates therefore follow, such that the combustion duration in SI engines have been observed not to increase dramatically in crank angle degrees at higher engine speeds, but take place proportionally quicker in the time domain.

Three predominant forces act upon the motion of a fluid: Buoyancy acts in the direction of Earth's gravity; molecular diffusion in all directions; and centripetal acceleration force radially from the axis of rotation. Where these forces are in balance, per Newton's first law of motion, there is nothing to compel a change in the state of motion of a particle or body. Importantly, diffusion can be hindered in the radial direction of a polar coordinate system if centripetal forces dominate. Also, diffusion can be slowed or reversed in a fixed frame if the velocity of one entraining species exceeds that at which diffusion proceeds; in effect, the rate at which species 1 (here assumed to be hydrogen filling the volume initially for illustration) diffuses into the entraining species 2 (e.g. air) is exceeded by the replenishment of air behind the interface of the two species.

Centripetal Flame Confinement

Lapsa and Dahm (2009) investigated hyperacceleration effects on turbulent combustion in premixed step-stabilized flames [20]. In that work, premixed propane and air is introduced into a U-shaped channel with a semicircular combustion zone at varying flow velocities, so that the centripetal acceleration is related by Eq. 3 and exceeds the range of [+ or -]10000 g.

[a.sub.c] [equivalent to] [v.sup.2]/r (3)

They report that high centripetal accelerations induce large body forces on the high-density cold reactants and hot, low-density products. The direction of the backward-facing step - radially outward or inward in the channel - have different effects, either to force high density reactants into the recirculation zone and lower density products radially inwards into the reactant stream in the first case where |[a.sub.c]| [right arrow] +[infinity], or vice versa. This first case gives rise to a virtuous centrifugal pumping mechanism - a region of vigorous mixing and flame convection across the channel at dramatically enhanced rates. Conversely, for |[a.sub.c]| < 0 using a radially inward step, as the acceleration level is increased the resulting body forces become sufficiently large that flame propagation is suppressed by the forced segregation of the reactants and products. The result is that the flame becomes nearly flat as large-scale distortions of the interface between high-density reactants and low-density products become essentially impossible; the flame is forced increasingly close to the radially inward wall as the body forces impede it from propagating across the channel. Although this latter case is considered inferior in terms of combustion efficiency for the U-shaped channel in the work of Lapsa and Dahm, here it is the desired mechanism to be exploited.

Coanda Effect

Dumitrache et al (2012) give a detailed analytical solution that approximates a two-dimensional Coanda flow for both laminar and turbulent regimes in curvilinear coordinates [21]. The phenomenon that is exploited is the tendency for a flow to attach to and follow over a convex surface: The incoming flow from the main cylinder, taken to be pure air, thus remains substantially attached to the peripheral wall of the combustion chamber. A critical point is reached where the flow detaches or lifts-off from the convex surface, which is related to the free-stream velocity, viscosity of the fluid and the radius of curvature. Here, the Coanda effect introduces an additional degree-of-freedom for the optimization of wall heat transfer. Convection is greatly affected by the turbulence level and therefore velocity fluctuations in the thermal boundary layer. Absent any stratification, with burning of the fuel-air mixture until the flame is quenched by the wall, convection is further enhanced by high flow velocities. In an optimized design, the Coanda effect could be exploited using fillet radii in the transition from the channels to the combustion chamber to set up an initial attachment of the pure air flow to the surrounding chamber walls to promote charge stratification and favor rotation in the desired direction. As the compression stroke proceeds and velocity increases, the flow would detach from the surface and the incoming fluid momentum is imparted to the inner radii of the rotating flow, forming another stratification, this time of tangential velocity with respect to the radius. Thus, the fuel-concentrated core would have a high velocity for rapid mixing and combustion, but the flow of pure air at the periphery would rotate at a relatively lower angular speed to inhibit convection.


The basic geometry of the conceived divided combustion chamber is a cylinder with connecting channels to the main cylinder volume that are vertically offset from the centerline to impart a rotating vortex flow in a counterclockwise direction from the frame of observation. This is illustrated in Fig. 3 (left-half). One side of the interface

between each channel and the chamber is radiused, while the other side is left sharp to promote attachment of the flow to the radial periphery of the cylindrical wall along the direction of rotation due to the Coanda effect. Ports in the two-stroke engine are not shown. For comparative purposes to the vortex-stratified approach described above, a combustion chamber design that represents current engine practice with homogeneous charge mixture is provided, as shown in Fig. 3 (right-half). The directly opposed layout of the side channels cause the incoming air jets to impinge upon one another during the compression stroke, resulting in high turbulence, but no particularly ordered motion as defined in the introduction - although noting that high-tumble designs are presently commonplace for SI engines.

The choices of significantly raised, transverse, cylindrically shaped divided chambers and straight, diametrically positioned channels are compromises between the need to realize optical access, and achieve thorough mixing and high turbulence levels of the prepared charge before ignition. The resulting long, narrow channels and right angles connecting the divided chamber to the main cylinder are also compromises to be able to realize optical access while achieving a high compression ratio, and also to introduce a well-developed flow into the combustion chamber with a minimum of undesired charge motion in other directions - cross-swirl and cross-tumble - apart from the plane of observation. The high flow velocities through the channels and 90[degrees] turns would result in high thermal and pressure losses. However, this was accepted because the objective was not for a geometry optimized for efficiency or performance per se, but rather to be able to study the flow phenomena in the visible cylindrical cross section of the combustion chamber. Both vortex and non-vortex designs are otherwise very similar, employing the same diameter (20 mm) and length (20 mm) of the cylindrical portion, and nearly identical total clearance volumes; the only significant differences being the vertical positioning of the connecting channels and, due to packaging constraints, orientations of the bespoke Synerject Strata gas injector and NGK CPR9EA extended nose insulator spark plug. Both designs yield calculated nominal geometric and effective (trapped) compression ratios of 10.5 [+ or -]0.25 and 7.3 [+ or -]0.17, respectively, and k-factor - defined here as the ratio of the volume of the cylindrical chamber and the total enclosed fluid clearance volume at piston TDC - of 0.5.

Figures 4 and 5 respectively show the exploded CAD rendering and assembled cylinder head on the engine test cell. The experimental visualization of both non-vortex and vortex combustion processes are described in the following.


The base engine for the test-cell work is a series-production 2009 model year two-stroke gasoline marine outboard with 9.9 HP (7.4 kW) rating. The basic specifications of the original engine are listed in Table 1. It was used in earlier investigations by the authors [22], and has been largely carried over into this work. There, the major changes included modification to operate on hydrogen directly injected into the cylinder; changing the premixed lubrication to crankcase oil injection; and replacing the production magneto ignition with a computer-controlled system. The gasoline carburetor is retained but not used. One of the two cylinders is inactive and sealed off; the remaining working cylinder is outfitted with optical access by means of 2-piece combustion chamber inserts for each of either the vortex or non-vortex designs bolted together to a common cylinder head plate and assembled to the engine crankcase as shown in the previous figures.

The engine controller is modeled in Matlab-Simulink and actuated using Opal-RT for the hardware-in-the-loop (HiL). The system controls ignition- and injection timing events, as well as the lube oil injection rate, all adjustable via a PC-based graphical user interface control panel.

An AVL 4CA1 crank angle encoder and Digilent NEXYS 4 Artix-7 FPGA board together act as system timekeeper to synchronize acquisition sampling with 0.5[degrees] CA resolution. The start and end switching of engine firing events, as well as shutter triggering for the camera image frames, are also managed by the combination of the controller programming, HiL, encoder and FPGA.

Schlieren Imaging

The schlieren method, which displays the density gradient in a fluid as varying luminosity, is employed for flow and combustion visualization. The works surveyed have been performed on engines with per cylinder displacements of 500 [cm.sup.3] and greater and, typically at engine speeds not exceeding 1500 RPM [23,24, 25, 26, 27], with one investigation performed at 2400 RPM [28]. As previously stated, turbulence levels scale with RPM - more precisely mean piston speed. However, at the time of Heywood, direct evidence had been limited to low- to mid-engine speeds; whether the structure becomes significantly different at high speed was not known [12] and direct observation of combustion at high engine speeds have been elusive.

High-Speed Camera

The camera is a Photron Fastcam SA-X2 digital high-speed camera with a CMOS image sensor [29]. The limiting specifications are shown in Table 2. The lens used is a Nikon AF-S VR Micro-Nikkor 105mm f/2.8G IF-ED macro-zoom.

Observation at engine speeds of up to 5000 RPM at 0.5[degrees] CA intervals requires a frame rate of 60000 FPS. The employed image resolution is 384x264. The fastest supported shutter time of 293 ns is chosen in order to capture the highly dynamic flow phenomena without the blurring common from conventional photography when taking fast-moving subjects. The very short shutter time therefore requires a very high luminous flux entering into the image sensor. This is even more important given that some fraction of the light from the observed subject is cut-off from the schlieren knife-edge. The following describes the light source.

Light Source

The light source employs a Luminus Devices PT-121-G-C11-MPF high-power green colored (wavelength centered at 525 nm) LED [30]. The emitting area is 4 x 3 mm and typical peak luminous flux is 4000 lumens with a driving voltage and current of 4.9 V (5.9 V max.) and 30 A (36 A max.), respectively, at 50% square wave duty cycle. The green light was found by Kaiser, Salazar and Hoops (2013) to reduce chromatic aberrations in the captured images [31]. The LED also has the advantage of generating concentrated, directed light with a repeatable, narrow spectral range compared to other forms like arc or gas discharge lighting. These properties also have benefits for reducing chromatic aberrations. With forced air cooling of the chip, the LED is able to survive operation with sustained high driving direct currents exceeding 30 amperes, rather than a pulsed square-wave power supply synchronized with camera shutter events, which would have greatly increased complications of the driver circuit design. The employed power supply is a Sorensen XPF 60-20D set in the tests nominally to 5 [+ or -]0.1 VDC and 27 [+ or -]3 A.

Figure 6 shows the installation of the single axis optical rail in the test cell. The light source is at the top of the image and the combustion chamber under study is located in the zone of parallel (collimated) light beams between two lenses. The knife-edge is placed at the focal point ahead of the camera lens. The length of the rail is 2 meters. The engine is connected to a Unico AC induction motor, which spins and holds the engine to the test operating speed, then firing events and acquisitions are toggled through the HiL controller from the test bench.

Non-Vortex Results

The unprocessed schlieren images for one cycle in the non-vortex combustion chamber are shown in Figures 7 and 8. Injection of hydrogen can be detected at the top of each frame in the first two rows of Fig. 7. The timed start of injection takes place at -90[degrees] CA for an energized duration of 10[degrees] CA at an average rail pressure of 10 bar absolute. There is visual evidence of the hydrogen jet entraining the combustion chamber outside of this duration as well. The reason is imperfect sealing of the poppet valve with the seat, permitting hydrogen under pressure, given its small molecular size, to leak through. As cylinder pressure rises during the compression stroke, the adverse pressure gradient eventually decreases, disappears and reverses. Increasing force of the cylinder pressure acting upon the exposed bottom surface area of the injector valve improves the sealing effectiveness until no more leakage becomes apparent.

In all tests, start of ignition takes place at TDC (0[degrees] CA). Intermittently varying intensities of the visible spark and cyclic variations in the spark duration are due to residual ignition energy, and in-cylinder conditions affecting ionization in the spark gap area. Visual investigations of the mixture preparation and combustion yield the following observations germane for consideration and qualitative validation of the subsequent CFD modeling and analysis:

1. The incoming flow from the main cylinder during the compression stroke via the diametrically opposed connecting channels deflects up or down in a consistent manner. The right-hand jet deflects toward the sparkplug ground electrode and the other downward. This may be caused by perturbations either by the asymmetry of the chamber geometry (e.g. presence of the spark plug and sensors) or pre-existing anisotropy in the flow. This behavior is consistent in all observed cycles and will also be shown to exist in the CFD models even though there is no geometric asymmetry to promote this;

2. Flame propagation takes place with similar structure and characteristics as has been widely observed in the literature, with the exception here of the very fine wrinkling and rapid rate; the flame is seen to engulf most of the visible cross-section only 5[degrees] CA after ignition start with local flame extinction apparent behind the front at +5.5[degrees] CA. This is indicative of the high degree of turbulence and confirms its scaling with engine speed, even at the high RPMs not observed until now;

3. Residual ignition energy causes a visible spark over some 10[degrees] CA, corresponding to over 300 [micro]s. This spark duration is carried over in subsequent numerical models.

Figures 9 and 10 show the unprocessed schlieren images for one cycle in the vortex combustion chamber. Fuel injection is highly obscured in part due to the 3-dimensional spray plume being projected onto 2-dimensional images focused on the sparkplug electrode with a shallow depth of field by the lens employed. The images are also obscured by artefacts that are not caused by the flow but rather by contamination on the glass window surfaces; these can be due to dirt, scratches, wiping streaks, condensed lube oil, carbon particles or other foreign matter. Beginning from about -64.5[degrees] CA, a dark streak can be seen at about the 8 o'clock position corresponding with the entry of air from the left side channel. The incoming air from the right-side channel is obscured by the sparkplug. By the end of the compression stroke in the last frames of Fig. 9, a circular bulge at the center of a diameter reaching to the tip of the ground electrode can be clearly distinguished. This indicates a radial density gradient and, per the ideal gas law, can be caused by local gradients of either pressure, temperature or gas constant. Pressure and temperature gradients are relatively small during the compression stroke, so the main cause of these gradients are due to the gas constant and, by consequence, the species fractions.

Upon start of ignition at the top left frame of Fig.10, a kernel can be seen forming in the spark gap. The density gradient cloud - due to the temperature rise of the burning gas of the forming flame front - grows rapidly, but unlike the non-vortex case that expands with a roughly monotonically increasing radius from the spark plug, propagates initially along the flow rotation and can be seen to curl inwards toward the center in a similar manner as observed by Lapsa and Dahm [27], preferentially consuming the charge in it and delaying contact of hot burned gas with the peripheral wall.

Cylinder Pressure and Heat Release Analysis

Cylinder pressures are measured using AVL GH12D transducers and an AVL Indismart acquisition system. One transducer is located directly in the divided chamber. Another transducer in the transfer port and a Bosch absolute pressure sensor placed in the exhaust port act as the pegging reference, the latter attached at the end of a 1 meter long copper tube to damp out fluctuations. Thermodynamic statistics for 100-cycle ensemble averages are summarized in Table 3, as are corresponding indicator diagrams in Fig. 11 and 12.

Both vortex and non-vortex cases show stepped pressure traces during combustion. This is directly visible in the schlieren imaging by pulsations at the interface of the combustion chamber and side channels after the main observable combustion event. These pressure pulsations arise because the divided chamber and main cylinder volume connected by long channels form a Helmholtz resonator system. The peak pressures are comparable in light of the small difference in compression ratio. On average, the vortex chamber has a slightly faster 10-50% MFB duration and reaches [p.sub.max] at an earlier crank angle, but the non-vortex case completes the second half of the heat release significantly more quickly. This is due to the very high turbulence of the impinging air jets that the vortex design lacks.

Looking at Figures 11, 12, 13, the vortex chamber shows a higher pressure and continually rising cumulative apparent heat release in the expansion stroke compared to the non-vortex counterpart. The log P-log V diagrams in Fig. 12 clearly show different slopes of the pressure traces during expansion, indicating differences in the expansion polytropic coefficient n. During compression, heat and mass losses make n lower than the isentropic exponent coefficient, analogously the specific heat ratio [gamma] = [C.sub.p]/[C.sub.v]. During expansion, heat and mass losses make n higher than [gamma] by a similar amount, but [gamma] is lower during expansion due to the higher temperature and changed gas composition from a mixture of fuel and air to burned products.

Neither n nor [gamma] remain constant throughout the compression and expansion strokes. Nevertheless, calculating the polytropic coefficients on the ensemble averaged cylinder pressure data give the mean values for the vortex and non-vortex cases as 1.305 and 1.401, respectively. A constant value of n = 1.35 is used for the determination of the apparent heat release curves for both cases.

As Figures 13 and 14 illustrate, the vortex and non-vortex combustion chambers indicate different behaviors after reaching their respective heat release peaks. While the vortex design shows the heat release curve to be gradually rising well into the expansion stroke, the non-vortex counterpart appears to be losing heat by the declining percentage of fuel mass fraction burned (which is an artefact of the chosen expansion polytropic coefficient, since MFB% cannot physically decline by definition due to irreversible combustion). A rising curve after the peak in the vortex case is due to the chosen n being higher than is the actual case, that is, overestimating the heat (and mass) losses. Conversely, a falling curve is due to n being too low, underestimating the losses. With the correct polytropic coefficients, the respective curves would hold horizontally constant. The observed behavior in the heat release curves is consistent with the relationship of the imposed polytropic coefficient value (1.35) with the counterparts calculated from the pressure indication data (1.305 and 1.401 for the vortex and non-vortex cases, respectively).

That the plots fall after the peak in the non-vortex case but rise with the vortex chamber indicate greater wall heat transfer losses in the former due to the deviation of n from [gamma].

Engines with divided combustion chambers tend to have greater thermal losses than non-divided counterparts on account of the higher surface area-to-volume ratio and high flow velocities through the connecting passages. However, as both concepts compared here are divided chambers with very similar geometries, the magnitude of the difference in polytropic coefficients evaluated at identical engine speeds and operating parameters between the two combustion chamber designs cannot be attributed solely to differences in heat transfer losses alone. While [gamma] during expansion is not known a priori, it certainly cannot be greater than n, which would imply heat transfer from the relatively cool walls to the hot burned gas. The significantly lower value of n for the vortex chamber - when [gamma] should not differ significantly to the non-vortex case on the reasonable approximation of comparable global equivalence ratios and mean bulk gas temperatures - could be explained by the presence of residual unburned hydrogen confined in the center of the combustion chamber. This occurs if the equivalence ratio in this core is richer than stoichiometric. Heat release first peaks and further burning stops when oxygen is depleted in this core. As the expansion stroke proceeds, the mixture comprising burned gas products and unburned fuel in the divided chamber get transported through the side channels into the main combustion chamber, where there is oxygen present to react with the still hot residual hydrogen. Evidence supporting this hypothesized phenomenon is elucidated upon in the numerical studies in the second part following this paper. On the other hand, the non-vortex design has a homogeneous, overall lean mixture; thus, heat release peaks and ceases once all the limiting fuel is burned.


A novel vortex-stratified combustion process has been visualized experimentally in a fired, optically accessible, 2-stroke [H.sub.2]DI engine, with the objective of reducing convective heat transfer losses to the surrounding walls during combustion. This reduction can be realized by confining the fuel-air mixture, and thus limiting the extent of the flame, to some distance away from the walls. This paper has explained the theoretical background of this proposed mechanism: This is achieved by air entering into a divided combustion chamber tangentially and preferentially along the circumference due to the Coanda effect. This sets up a vigorous, highly ordered rotational field upon the charge, but mixing with the directly injected fuel is outpaced by the replenishment of air at the periphery, which causes the fuel to be concentrated at the center. Combustion takes place with a flame propagation that initially follows the rotation of the bulk flow, but then curls radially inward toward the center as a result of centripetal body forces acting upon the different density burning and unburned gases.

Schlieren images have been recorded at up to 5000 RPM to compare visually the mixture formation and flame propagation characteristics of two combustion chamber designs - one that demonstrates the vortex approach introduced in this work, and another that represents a homogeneous mixture process without specific charge motion directionality. These visualizations show qualitatively that the vortex process is effective in preventing or delaying the flame from reaching the combustion chamber walls; this is supported by quantitative observations of 0-D cylinder pressure and apparent heat release measurements, which indicate lower thermal losses by way of a lower calculated polytropic coefficient value compared to the non-vortex case (1.305 vs 1.401). The schlieren observations and heat-loss findings shall be compared with CFD analyses in the second part of this work.


[1.] Verhelst, S., Wallner, T., "Hydrogen-Fueled Internal Combustion Engines," Prog. Energy Combust. Sci., 35(6):490-527, December 2009, doi:10.1016/j.pecs.2009.08.001.

[2.] Eichlseder, H., Wallner, T., Freymann, R., and Ringler, J., "The Potential of Hydrogen Internal Combustion Engines in a Future Mobility Scenario," SAE Technical Paper 2003-01-2267, 2003,

[3.] Wimmer, A., Wallner, T., Ringler, J., and Gerbig, F., "H2-Direct Injection - A Highly Promising Combustion Concept," SAE Technical Paper 2005-01-0108, 2005,

[4.] Shudo, T., "Improving thermal efficiency by reducing cooling losses in hydrogen combustion engines," Int J Hydrogen Energy 32 (17):4285-4293, 2007, doi:10.1016/j.ijhydene.2007.06.002.

[5.] Eichlseder, H., Grabner, P., Gerbig, F., Heller, K., "Advanced Combustion Concepts and Development Methods for Hydrogen IC Engines, FISITA 2008 World Automotive Congress, F2008-06-103, 2008, Munich, Germany.

[6.] Michl, J., Schenk, M., Rottengruber, H., Huhn, W., "Thermal Boundary Conditions in a Stoichiometric Operating Hydrogen Engine," FISITA 2008 World Automotive Congress, F2008-06-110, 2008, Munich, Germany.

[7.] Shudo, T., Nabetani, S., Nakajima, Y., "Analysis of the degree of constant volume and cooling loss in a spark ignition engine fuelled with hydrogen," International Journal of Engine Research, 2(1):81-92, February 2001, doi:10.1243/1468087011545361.

[8.] Shudo T., Oba S., "Mixture distribution measurements using laser induced breakdown spectroscopy in hydrogen direct injection stratified charge," Int J Hydrogen Energy 34(5):2488-2493, March 2009, doi:10.1016/j.ijhydene.2009.01.012.

[9.] Grabner, P., Eichlseder, H., Gerbig, F., Gerke, U., "Optimisation of a Hydrogen Internal Combustion Engine With Inner Mixture Formation," 1st International Symposium on Hydrogen Internal Combustion Engines, September 28-29, 2006, Graz, Austria.

[10.] Oppenheim, A., "Prospects for Combustion in Piston Engines," SAE Technical Paper 2002-01-0999, 2002,

[11.] Heitland, H., Rinne, G., and Wislocki, K., "Can the Best Fuel Economy of Today's Engines Still Be Improved?," SAE Technical Paper 981912, 1998,

[12.] Heywood, J. B., "Internal Combustion Engine Fundamentals," 1988, New York, McGraw-Hill.

[13.] Norton, D. G., Vlachos, D. G., "Combustion characteristics and flame stability at the microscale: a CFD study of premixed methane/air mixtures," Chemical Engineering Science 58(21):4871-4882, November 2003, doi:10.1016/j.ces.2002.12.005.

[14.] Toulson, E., Schock, H., and Attard, W., "A Review of Pre-Chamber Initiated Jet Ignition Combustion Systems," SAE Technical Paper 201001-2263, 2010,

[15.] Brandstetter, W., Decker, G., and Reichel, K., "The Water-Cooled Volkswagen PCI-Stratified Charge Engine," SAE Technical Paper 750869, 1975,

[16.] Brandstetter, W., "Uber neuere Arbeiten am VW-PCI- und VW-PCV-Schichtladeverfahren, "MTZ 38(7/8):327-330, 1977.

[17.] Brandstetter, W., "The Volkswagen Lean Burn PC-Engine Concept," SAE Technical Paper 800456, 1980,

[18.] Peterson, R. C., Alkidas, A. C., "A Visual Study of Divided-Chamber Diesel Combustion Using a Rapid Compression Machine," Combustion and Flame 53(1-3):65-81, November 1983, doi:10.1016/0010-2180(83)90007-X.

[19.] Arcoumanis, C., Hull, D., and Whitelaw, J., "An Approach to Charge Stratification in Lean-Burn, Spark- Ignition Engines," SAE Technical Paper 941878, 1994,

[20.] Lapsa, A. P., Dahm, W. J. A., "Hyperacceleration effects on turbulent combustion," Proc. of the Combustion Institute 32(2):1731-1738, 2009, doi:10.1016/j.proci.2008.05.038.

[21.] Dumitrache, A., Frunzulica, F., Ionescu, T. C., "Mathematical Modelling and Numerical Investigations on the Coanda Effect," In Nonlinearity, Bifurcation and Chaos - Theory and Applications, edited by Awrejcewicz, J., Hagedorn P., 2012, accessed 2015/02/01,, doi: 10.5772/50403.

[22.] Oh D., Plante J., "A Highly Efficient Small-Displacement Marine Two-Stroke H2DI Engine With Low Emissions," ASME. J. Eng. Gas Turbines Power 135(8):082001-082001-10, 2013, doi:10.1115/1.4023752.

[23.] Namazian, M., Hansen, S., Lyford-Pike, E., Sanchez-Barsse, J. et al., "Schlieren Visualization of the Flow and Density Fields in the Cylinder of a Spark-Ignition Engine," SAE Technical Paper 800044, 1980,

[24.] Salazar, V. and Kaiser, S., "Influence of the Flow Field on Flame Propagation in a Hydrogen-Fueled Internal Combustion Engine," SAE Int. J. Engines 4(2):2376-2394, 2011,

[25.] Weinrotter, M., Wintner, E., Iskra, K., Neger, T. et al., "Optical Diagnostics of Laser-Induced and Spark Plug-Assisted HCCI Combustion," SAE Technical Paper 2005-01-0129, 2005,

[26.] Baritaud, T., "High Speed Schlieren Visualization of Flame Initiation in a Lean Operating S.I. Engine," SAE Technical Paper 872152, 1987,

[27.] Meier, F., Kohler, J., Stolz, W., Bloss, W. et al., "Cycle-Resolved Hydrogen Flame Speed Measurements with High Speed Schlieren Technique in a Hydrogen Direct Injection SI Engine," SAE Technical Paper 942036, 1994,

[28.] Spicher, U., Kroger, H., and Ganser, J., "Detection of Knocking Combustion Using Simultaneously High-Speed Schlieren Cinematography and Multi Optical Fiber Technique," SAE Technical Paper 912312, 1991,

[29.], accessed 2016/02/24.

[30.] Luminus PT121 Product Datasheet,, accessed 2016/02/24.

[31.] Kaiser, S. A., Salazar, V. M., Hoops, A. A., "Schlieren measurements in the round cylinder of an optically accessible internal combustion engine," Applied Optics, 52(14): 3433-3443, 2013, doi:10.1364/AO.52.003433.


[a.sub.c] - Centripetal acceleration [m/[s.sup.2]]

c - Concentration [kmol/[m.sup.3]]

CA - Crank angle

CAD - Computer-aided design

CFD - Computational fluid dynamics

D - Diffusive coefficient [[m.sup.2]/s]

FPGA - Field-programmable gate array

FPS - Frames per second

ICE - Internal combustion engine

IO - Input-output

J - Diffusion flux [kmol/([m.sup.2]s)]

LED - Light emitting diode

MFB - Mass fraction burned

n - Polytropic coefficient [-]

p - Pressure [bar]

r - Radius [m]

[r.sub.x] - Radius of flammability or flame extent limit [mm]

R - Radius of cylindrical combustion chamber [mm]

RPM - Revolutions per minute

SI - Spark ignition

T - Temperature [K]

TDC - Top-dead center

v - Velocity [m/s]

V - Volume [[cm.sup.3]]

[gamma] - Specific heat ratio (also isentropic coefficient) [-]

[delta] - Distance between the wall and the limiting combustible charge or flame extent [mm]

[eta] - Efficiency [%]

David Oh, Martin Brouillette, and Jean-Sebastien Plante

Universite de Sherbrooke

Table 1. Base engine specifications

No. of cylinders and layout          2 inline, 180[degrees] crankshaft
Bore x stroke (displacement)         56 x 50 mm (246 [cm.sup.3])
Connecting rod length                90 mm
Compression ratio                    6.8:1 (trapped)
Scavenging type                      Crankcase
Porting arrangement                  Loop, Schnuerle
Lubrication type                     Fuel-oil premix, 100:1 v/v
Induction                            Carburetor, reed valves
Port height (transfer/exhaust) (*)   41/33 mm

(*) Measured from piston TDC position

Table 2. Camera specifications

Max. image resolution   1024 x 1024
Max. frame rate         1 000 000 FPS
Min. shutter time       1 / 3 410 526 sec (293 ns)
Recording bit depth     12-bit (4096 levels) grayscale

Table 3. Summary of thermodynamic statistics at 4500 RPM, 100 cycles.

           Vortex                         Non-Vortex

IMEP       Mean: 3.6 bar                  Mean: 3.04 bar
           Std. dev.: 0.13 bar            Std. dev.: 0.188 bar
MFB10%     Mean: 2.72[degrees] CA         Mean: 2.55[degrees] CA
           Std. dev.: 0.478[degrees] CA   Std. dev.: 0.912[degrees] CA
MFB50%     Mean: 5.55[degrees] CA         Mean: 6.01[degrees] CA
           Std. dev.: 0.468[degrees] CA   Std. dev.: 1.12[degrees] CA
MFB90%     Mean: 13.77[degrees] CA        Mean: 11.33[degrees] CA
           Std. dev.: 5.408 [degrees] CA  Std. dev.: 1.215 [degrees] CA
Comb       Mean: 11.05[degrees] CA        Mean: 8.78[degrees] CA
Pmax       Mean: 35.45 bar                Mean: 35.02 bar
           Std. dev.: 1.333 bar           Std. dev.: 2.019 bar
COPYRIGHT 2017 SAE International
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2017 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Oh, David; Brouillette, Martin; Plante, Jean-Sebastien
Publication:SAE International Journal of Engines
Article Type:Report
Date:Dec 1, 2017
Previous Article:A Numerical Investigation of Ignition of Ultra-Lean Premixed [H.sub.2]/Air Mixtures by Pre-Chamber Supersonic Hot Jet.
Next Article:Reduced Convective Combustion Chamber Wall Heat Transfer Losses of Hydrogen-Fueled Engines by Vortex-Stratified Combustion - Part 2: Numerical...

Terms of use | Privacy policy | Copyright © 2018 Farlex, Inc. | Feedback | For webmasters