The Impact of Low Octane Primary Reference Fuel on HCCI Combustion Burn Rates: The Role of Thermal Stratification

Homogeneous charge compression ignition (HCCI) was originally demonstrated by Onishi et al. [1] and Najt and Foster [2]. The hallmarks of HCCI are lean, unthrottled operation. The lack of throttling coupled with lean operation increases engine thermal efficiency (the former through less pumping work and the latter through higher gamma) while offering low NO_x and particulate emissions.

While HCCI combustion has multiple advantages over traditional spark ignited engines as well as diesel engines, it is not without drawbacks. HCCI combustion is limited to moderate loads due to excessive combustion noise at high load, and low load operation is limited due to misfire. Further difficulties are encountered with regard to combustion control and transient operation.

While the majority of HCCI research has focused on gasoline fuels, HCCI combustion may in fact benefit from a fuel with an octane rating in between that of typical gasoline and diesel fuels, as has been suggested by Hildingsson et al. [3]. The use of low octane naptha fuels offers potential environmental benefits, as they require less processing (cracking) at the refinery, producing less CO₂. A lower octane fuel combined with high-efficiency HCCI operation may emit less greenhouse gasses from a well-to-wheels perspective.

In addition to environmental benefits, multiple studies have demonstrated attainment of higher load for low octane fuels, relative to gasoline, under HCCI conditions. Shibata and Urushihara [4,5] compared different blends of pure hydrocarbons with n-heptane and noted, for toluene blends, two-phase burning that enabled lower peak pressure rise rate and longer burn duration, which extended the load limit. Yang et al. [6,7] built upon the work of Dec and Yang [8]—boosted gasoline HCCI combustion with partial fuel stratification—with a low octane distillate fuel termed Hydrobate. The Hydrobate fuel exhibited strong $ϕ$ sensitivity and under naturally aspirated conditions offered higher thermal efficiency and load compared to gasoline.

In a previous study by the authors [9], higher load was attained with a low octane gasoline/n-heptane blended fuel compared with regular gasoline (RON = 90.5). Related to the higher load operation, faster burn rates and shorter burn durations were observed for the low octane n-heptane blended fuel compared with gasoline at fixed combustion phasing. For that study, the amount of negative valve overlap (NVO) was varied to maintain constant combustion phasing. Varying the NVO also varied the amount of total EGR (tEGR) and thus the chemical and thermal composition, obscuring the effects of the fuel chemistry. A recent study by Kuboyama et al. [10] studied low octane naptha blends under HCCI conditions on a multicylinder and found a slight decrease in maximum load (due to pressure rise rate) for a 78 RON naptha blended fuel compared with gasoline. In fact, the loss in load was due to applying thermal efficiency constraints to late combustion phasing limits. For the naptha blended fuel, burn rates were faster (and burn duration shorter) for equivalent load and CA50, which was in agreement with the authors' previous work.

The purpose of this work was to isolate the fuel chemistry effects of low octane fuel, if any, as they affect the burn characteristics of HCCI combustion. To this end, experiments were conducted to compare gasoline and PRF40 under constant composition conditions. For this study, the charge composition was defined as the total mass of fresh air, fuel, and tEGR. The effects of changing NVO on charge composition are discussed in detail in the Experimental Procedure section. While these new experiments were able to isolate compositional effects, the changes in IVC temperature necessary to maintain fixed combustion phasing possibly introduced thermal gradient changes between the fuels which could have potentially influenced HCCI heat release rates [11,12]. To further interpret the experimental results and separate thermal effects from fuel chemistry effects, simulation work was performed with a multiple zone balloon model (no mixing between zones) utilizing a simplified chemical kinetics mechanism. The model was developed by Kodavasal et al. [13] for HCCI combustion.

This work was conducted at the University of Michigan on a single-cylinder Ricard Ltd. (Shoreham-by-Sea, UK) Hydra engine equipped with a fully flexible valve actuation (FFVA) system manufactured by Sturman Industries (Woodland Park, CO). General characteristics of the FFVA engine can be found in Table 1, and a schematic illustration of the engine test cell is given in Fig. 1. A more detailed description of the FFVA hardware has previously been documented by Manofsky et al. [14]. While valve events are fully adjustable, for this study symmetric NVO was maintained for all test points. The valve lash was set at 0.1 mm, and this was the criteria for determining valve opening and closing events. Each valve had a linear position encoder, and this measurement was recorded during operation.

Air was supplied to the engine from compressed shop air after being dried and filtered and was metered through both critical flow orifices and a Fox Thermal Instruments (Marina, CA) hot-wire anemometer for measurement redundancy. The entire intake plenum and runners were wrapped with insulation to enable inlet temperatures up to 200 °C in conjunction with a 5 kW inline heater. The high inlet temperatures were required to operate HCCI with small levels of NVO and high levels of external EGR (eEGR). The eEGR was introduced into the intake plenum from the exhaust (as shown in Fig. 1), and flow was controlled by an electrically actuated needle valve. Fuel was direct injected into the cylinder at 100 bar with a start of injection at 330 deg before top dead center (BTDC). Fuel flow was measured with a Max Machinery (Healdsburg, CA) Model 213 piston-driven flow meter.

Exhaust gas emissions were measured with a Horiba Instruments (Ann Arbor, MI) MEXA 7500D-EGR and included total hydrocarbons (THC) on a C1 basis, oxides of nitrogen (NO + NO₂), oxygen, carbon monoxide, carbon dioxide, and methane; a secondary CO₂ analyzer was used to measure intake CO₂ for calculating eEGR fraction.

High-speed (crank angle resolved) engine data were collected at a resolution of 0.1 deg crank angle (CA) using a Kistler Instruments (Amherst, NY) crank angle encoder and logged by an AVL Test Systems (Plymouth, MI) combustion analysis system. Cylinder pressure was measured with a Kistler 6125 A piezoelectric transducer. For each data point, 200 cycles were logged, and the average pressure trace was analyzed. The modified Woschni heat transfer correlation as developed by Chang et al. [15] was used in postprocessing cylinder pressure-based heat release rates.

The use of NVO trapped hot residuals in cylinder to provide the temperatures at top dead center (TDC) to facilitate auto-ignition. The amount of internal EGR (iEGR) was not directly measured in cylinder; the iEGR was inferred using the method developed by Fitzgerald et al. [16] and examined by Ortiz-Soto et al. [17], where it was found to provide an accurate estimation of residual gas for HCCI conditions.

The simulation approach used here was developed by Kodavasal et al. [13] to investigate the thermal and compositional stratification in HCCI combustion. The multizone model provides a platform to study the effects and the interaction with detailed kinetics without employing full computational fluid dynamics (CFD) tools with high computational cost. Accordingly, the model represents the charge as a set of 40 nonmixing zones, interacting only through pressure work. The temperature of each reacting zone was allowed to evolve during compression according to heat transfer weighting factors determined by a one-time calibration against a 3D CFD motoring calculation. As used here, the detailed kinetics calculations were carried out for each zone using the 33 species skeletal mechanism of Tsurushima [18] for primary reference fuels (PRFs) modified to include the extended Zeldovich mechanism for NO. Once the temperature in the hottest zone reaches 1100 K, ignition was assumed and the heat transfer from each zone was subsequently determined by a modified formula based on each zone's individual temperature. The details can be found in Ref. [13].

The low octane number (ON) fuel for this work was PRF40. A pure PRF was desirable for simulation studies, as the kinetic mechanisms are well developed. The gasoline used for the study was a research-grade gasoline supplied by Haltermann Solutions (Houston, TX), with an antiknock index (AKI, equal to the average of the RON and motor octane number (MON)) rating of 87. The RON and MON of the gasoline are controlled by the supplier such that both the fuel sensitivity and AKI are the same from batch to batch. The specific fuel properties are listed in Table 2.

With the knowledge that under HCCI operating conditions, low ON PRFs will demonstrate low-temperature heat release (LTHR) as engine speed is decreased [20], the experimental operating space included engine speeds of 2000, 1500, and 1000 RPM to explore the impact of LTHR on combustion burn rates. The remaining engine conditions for the experiments are summarized in Table 1.

For otherwise equivalent NVO HCCI operating conditions, maintaining constant combustion phasing between two fuels with a large difference in ON requires changing the duration of NVO to adjust the TDC temperature. Changing the amount of NVO also changes the iEGR and has the potential to change HCCI burn rates due to changes in thermal stratification. Thermal stratification in this context refers to the mass–temperature distribution in cylinder. Consider low and high iEGR cases that have an equal mean temperature at intake valve closing (IVC). The high iEGR case would have a larger distribution of the mass–temperature profile compared with a low iEGR case. This is a consequence of mixing cooler intake and external EGR (eEGR) with a larger quantity of hot trapped internal residual, which increases the temperature distribution of the mass in cylinder relative to the low iEGR case. An example of the impact of thermal stratification on auto-ignition burn rates can be found in the work by Middleton et al. [21].

With potential differences in thermal stratification when varying iEGR, quantifying the effect of iEGR on burn rates is necessary before comparing burn rates between fuels with varying levels of iEGR. To quantify the effects of varying iEGR requires changing NVO while keeping all other conditions the same: combustion phasing, fueling rate, equivalence ratio, and composition. Here, composition is the sum of the global masses of all constituents, as illustrated conceptually in Fig. 2. The injected fuel combined with inducted air sets the exhaust equivalence ratio, $ϕ$⁠. The combination of eEGR and iEGR equals the tEGR. The injected fuel mass as a fraction of the total charge mass is the fuel/charge equivalence ratio, $ϕ′ ≈ϕ·(1−tEGR)$⁠. This is essentially the traditional equivalence ratio recast as a measure of the energy density of the charge [22,23].

For the charge composition to remain constant while changing NVO, any reduction in iEGR must be met by a commensurate increase in eEGR to maintain constant tEGR. This is implemented experimentally by adding eEGR as NVO (iEGR) is reduced. As iEGR is reduced and relatively cool eEGR is added, the temperature at IVC timing is reduced. For combustion phasing to remain constant, the intake temperature must be increased to maintain a constant temperature at IVC; with a fixed composition charge, the IVC temperature determines combustion phasing.

To experimentally quantify the effect of iEGR on HCCI burn rates, two experiments were performed with gasoline at 2000 RPM as shown in Fig. 3. The curve with diamond markers represents the first attempt, in which eEGR was added as iEGR was removed (less NVO) at constant manifold pressure and constant IVC timing; eEGR was added to maintain constant exhaust $ϕ$ as iEGR was reduced. In the second attempt, represented with the curve with round markers in Fig. 3, the IVC was not held constant. With the first approach, as the NVO (iEGR) was reduced the intake manifold temperature was increased to maintain constant combustion phasing. Increasing the intake manifold temperature (with less iEGR) resulted in changes in runner dynamics which had the effect of reducing pressure in the runner at the time of IVC. The lower pressure at IVC reduced the total charge mass (and tEGR mass fraction) as the iEGR was reduced, and the fuel/charge ratio (⁠$ϕ′$⁠) changed accordingly.

To compensate for the fact that tEGR was decreasing as iEGR decreased due to IVC pressure changes, the iEGR study for gasoline at 2000 RPM was repeated, varying IVC timing as NVO was changed (curve with round markers, Fig. 3). For the cases where the duration of NVO was greater than ∼100 deg, the IVC was retarded, and for the case where NVO was less than 100 deg, the IVC was advanced. This changed the pressure in the runner at IVC and provided essentially constant tEGR, or constant total composition, for all NVO settings studied. Corresponding to the constant composition, $ϕ′$ remained constant as well, compared with the case with a fixed IVC. For reference, the required change in IVC timing for the data in Fig. 3 was from IVC = 118 deg BTDC at 151 deg NVO (43% iEGR) to IVC = 168 deg BTDC at 69 deg NVO (19% iEGR).

The above method of varying IVC to maintain constant composition was sufficient at 2000 RPM. The same procedure at lower engine speeds, especially for PRF40, was observed to yield higher tEGR fractions (and lower $ϕ′$⁠) as speed was reduced. At 90 deg of NVO for PRF40, reducing speed from 2000 to 1000 RPM led to a total increase in tEGR from 42% to 52%, respectively. There were only two methods feasible to maintain a constant $ϕ′$ for all engine speeds: throttle the lower engine speed cases or increase fueling at the lower engine speed cases. The latter would change the exhaust $ϕ$ while the former would change the pressure. Considering the relative impact of pressure compared with $ϕ$ on the ignition delay correlation proposed by He et al. [24], the exponent for pressure has a smaller power than that for $ϕ$⁠. Furthermore, it is standard practice to scale for pressure differences in chemical kinetics, but not equivalence ratio. For this reason, intake throttling was the approach taken to reducing the tEGR at lower speeds. The result was that the intake air was throttled to 0.95 bar at 1500 RPM and to 0.9 bar at 1000 RPM. With this mild throttling, the tEGR was held to 44% (±2 points) for all three engine speeds. For the experiments that follow, constant composition was maintained by adjusting IVC as a function of NVO for a given engine speed and throttling the intake as the speed was reduced.

With a method in place to fix charge composition for a wide range of iEGR, the effect of iEGR on HCCI combustion was examined with gasoline at three engine speeds: 1000, 1500, and 2000 RPM. The range of iEGR was from 19% to 46% of the total EGR (note that all reported EGR fractions are mass fractions) depending on engine speed. The crank angle-based rate of heat release (HRR) for each iEGR at 1000 and 2000 RPM is plotted in Figs. 4 and 5, respectively. There is no discernible difference between engine speeds in terms of the distribution of maximum HRR between iEGR levels when evaluated on a crank angle basis.

As mentioned above, as iEGR is reduced and eEGR increased at a constant tEGR, it is necessary to increase the intake temperature to maintain a constant IVC temperature and combustion phasing. For gasoline at 2000 RPM, reducing iEGR from 43% to 19% required an increase in intake port temperature from 50 to 200 °C, respectively, to maintain an IVC temperature of 505 K. At 1000 RPM, the reduction in iEGR from 46% to 32% required an increase in intake port from 100 to 200 °C, respectively, to maintain an IVC temperature of 465 K. The maximum intake temperature of the setup was 200 °C, and this was what limited the minimum amount of iEGR for both engine speeds. The lower IVC temperature at 1000 RPM was necessary to lengthen the ignition delay, as there was a longer time period before ignition when holding a fixed 6 deg ATDC combustion phasing.

Examining the maximum HRR across the different iEGR values at 2000 RPM, it can be noted that at the lowest iEGR level, 19%, the maximum HRR was 80 J/deg CA. Increasing iEGR corresponded to reduced maximum HRR: At the highest iEGR (43%), the maximum HRR was 69 J/deg CA corresponding to a decrease of 11 J/deg CA compared with the lowest iEGR case.

The difference in 10–90 burn duration (CA10–CA90) for gasoline at all three engine speeds as a function of iEGR is shown in the upper region of Fig. 6. The burn duration trend with iEGR was nearly identical for all engine speeds, which was consistent with the HRR having no crank angle-based engine speed dependence as shown in Figs. 4 and 5. Regarding iEGR dependence, as iEGR was increased, there was for all speeds a small but consistent increase in the burn duration above an iEGR of ∼30%. At 2000 RPM at the lowest iEGR level (19%), the 10–90 burn duration was ∼8 deg, whereas at the highest iEGR (43%), it was ∼10 deg.

The iEGR studies conducted for gasoline were repeated for PRF40. The HRR curves for PRF40 at 1000 and 2000 RPM are shown in Figs. 7 and 8, respectively. As expected of a low ON PRF, the PRF40 exhibited LTHR which increased in magnitude with decreasing engine speed. The LTHR can be observed as the small “bump” in the HRR curve at ∼15 deg BTDC (Fig. 7) prior to the main ignition and is especially pronounced at 1000 RPM. At both engine speeds, the PRF40 showed a slight change in the maximum HRR as iEGR increased, though the sensitivity of burn rates to iEGR was much less pronounced than for gasoline and was in different directions for the two speeds. Despite the presence of LTHR at 1000 RPM, note the magnitude of the main HRR was essentially identical between 1000 and 2000 RPM, demonstrating that the presence of LTHR did not significantly impact the main HRR.

As the PRF40 had a lower ON than gasoline, the required IVC temperature for a given phasing was less than that of gasoline. At 2000 RPM, the required IVC temperature for PRF40 was 480 K, compared with 505 K for gasoline. The intake port temperature was increased from 50 to 125 °C as the iEGR was reduced from 38% to 17%, respectively, for the 2000 RPM cases in Fig. 8. At 1000 RPM, the required IVC temperature for PRF40 was 405 K compared with 465 K for gasoline. The intake port temperature was increased from 35 to 60 °C as iEGR was reduced from 23% to 14%, respectively, for the 1000 RPM cases in Fig. 7.

Comparing the 10–90 burn duration of PRF40 for different iEGR values and speeds was complicated by the LTHR portion of the mass fraction burned (MFB) curve. The LTHR, especially at 1000 and 1500 RPM, significantly advanced the location of CA10. While strictly correct in terms of the actual cumulative heat release, what was of primary interest was the 10–90 burn duration of the main combustion event. To examine the 10–90 burn duration of the main event, the MFB curve was truncated and rescaled; an example of which is demonstrated in Fig. 9 for PRF40 at 1000 RPM, iEGR = 23%. The dashed line is the original MFB curve, and the solid line is the rescaled MFB curve with the LTHR portion removed. To remove the LTHR from the MFB curve, the portion of lowest slope after the onset of LTHR was determined and anything prior to this portion of the curve was removed; the remaining MFB curve was then rescaled from zero to one.

The effect of the rescaling of the MFB curve on the CA burn locations is shown in Fig. 9, where it is observed that only the location of the 10% burn location was significantly altered; the CA50 and CA90 locations remain practically unchanged (plotted essentially on top of each other), and the difference of 0.3 deg being within the error bars. For clarity and completeness in presentation when the LTHR portion was removed, the CA10 of the main event is denoted CA10_M, and the corresponding burn duration is designated 10 _M–90 with the “M” designating the 10% burn of the rescaled “main” combustion event.

Examining the 10 _M, 50, and 90 burn locations in Fig. 10 for PRF40 at 1000 RPM shows that as with HRR, there was no significant dependence of burn location on iEGR level. While not shown, the results at 1500 and 2000 RPM were essentially identical at equal iEGR levels. Corresponding to the 10_M, 50, and 90 burn locations, the 10_M–90 burn duration was nearly identical for all iEGR and engine speeds for PRF40, as shown in the lower portion of Fig. 6. The main conclusion from Fig. 6 is that over the range of engine speeds tested, the effect of iEGR on burn rates was small, especially in comparison with the changes observed between gasoline and PRF40 at equivalent iEGR levels.

To examine the fuel effects on HRR, comparisons between PRF40 and gasoline at the same iEGR are made in this section. While differences between iEGR levels were small for the PRF40, points of different fuels at the same iEGR are compared to remove any question of iEGR effects on burn rates when comparing different fuels. Due to inlet temperature limitations, there was no overlap between the two fuels at equal iEGR levels at 1000 RPM, so experimental comparison between the fuels was made at 2000 RPM. The HRR curves for PRF40 and gasoline at 38% iEGR are shown in Fig. 11. The PRF40 had a maximum HRR considerably higher than that of gasoline: 95 J/deg CA compared with 70 J/deg CA. This change in HRR was larger than any observed by changing iEGR levels.

Moving from comparison of the maximum HRR to comparison of the overall burn duration, Fig. 6 shows the 10_M–90 burn duration as a function of iEGR for both PRF40 and gasoline. For all iEGR levels, the PRF40 demonstrated a shorter burn duration compared with gasoline. At lower iEGR levels, between 10% and 30%, where gasoline maintained a relatively constant burn duration, the PRF40 had a burn duration ∼2.5 deg shorter than gasoline. This difference increased beyond 30% iEGR as the gasoline began to exhibit increased burn duration with increasing iEGR. This was consistent with the observed sensitivity of HRR to iEGR: For gasoline, there was a more pronounced decrease in maximum HRR as iEGR increased, and this corresponded to increasing burn duration for gasoline in Fig. 6. Conversely, the PRF40 showed little variation in maximum HRR with varying iEGR, and this was supported by the comparatively constant burn duration as a function of iEGR.

The results highlight that for a given iEGR, and for all engine speeds, PRF40 demonstrated a notable increase in maximum HRR and a corresponding decrease in 10 _M–90 burn duration relative to gasoline. Factors which in general could contribute to the observed differences in burn rates include:

1. (1)

   equivalence ratio (⁠$ϕ$⁠)

2. (2)

   energy addition per cycle

3. (3)

   intake pressure

4. (4)

   combustion phasing

5. (5)

   oxygen concentration

6. (6)

   dilution/tEGR level

7. (7)

   turbulence

8. (8)

   compositional (local) stratification from NVO

9. (9)

   thermal (local) inhomogeneities from NVO (incomplete mixing of fresh and residual charge)

10. (10)

    thermal gradient from bulk average temperature to wall temperature

11. (11)

    fuel chemistry differences

In the list, items (1)–(6) were fixed explicitly in the design of the experiments. For item (3) (intake pressure), we have made comparisons between fuels at the same engine speed, and the inlet pressure was the same. For item (7) (turbulence), the engine speed and valve timing were the same between the two cases, so there was no reason to believe there was any significant difference in turbulence. For item (8) (composition), the cases selected had the same level of iEGR so any compositional stratification that was due to iEGR would have be the same between the cases; the same is true for item (9) (local thermal stratification). This only leaves items (10) (thermal inhomogeneities due to bulk thermal gradients) and (11) (fuel chemistry).

Consider Fig. 12, where the mean estimated TDC temperature is shown as a function of the estimated IVC temperature from the experimental data. The temperatures are grouped by engine speed, in this case 2000 and 1500 RPM as there were data for both fuels at iEGR = 38% at these speeds. At 2000 RPM, gasoline had an estimated average IVC temperature of 505 K, while the PRF40 had an estimated average IVC temperature of 489 K (a difference of 16 K). At 1500 RPM, the gasoline had an estimated average IVC temperature of 481 K, while the PRF40 had an estimated average IVC temperature of 427 K (a difference of 54 K). For both speeds, a lower IVC temperature for PRF40 would be consistent with the observed (fixed) combustion phasing, as PRF40 was the more reactive fuel. As engine speed was reduced from 2000 to 1500 RPM, more LTHR occurred for the PRF40, and this was responsible for the larger decrease in required IVC temperature; in order to align combustion phasing, the IVC temperature for PRF40 had to be reduced further as speed decreased to account for the increase in temperature that occurred from the LTHR prior to TDC (the LTHR occurred between 10 and 20 deg BTDC at 1000 RPM, as shown in Fig. 7).

At both speeds, the PRF40 had a lower TDC temperature than the gasoline. At 1500 RPM, the gasoline had a TDC temperature of 1059 K versus 1040 K for PRF40, a difference of 19 K. At 2000 RPM, the gasoline had a TDC temperature of 1098 K versus 1090 K for PRF40, a difference of 8 K. At both speeds, a lower bulk gas temperature will reduce the thermal stratification as the temperature gradient between the bulk gas and cylinder wall is reduced (a flatter gas temperature profile from the center of the cylinder to the wall). At each engine speed, a reduction in thermal gradients for the PRF40 could have been responsible for the increase in maximum HRR relative to gasoline, as it has been shown by Sjöberg and Dec [11] that larger thermal gradients will reduce HCCI burn rates.

The data suggest that changes in the thermal stratification, as a function of IVC temperature requirements for combustion phasing, likely affected the HCCI burn rates. The difference in required IVC temperatures between the fuels is a function of the fuel chemistry so items (10) and (11) in the above “factors” list cannot be completely separated experimentally.

As conceptualized by Maria et al. [12], the heat release rate in HCCI combustion occurs as a cascading ignition process and can be affected by three factors:

1. (1)

   the temperature gradient (local and bulk)

2. (2)

   the sensitivity of the ignition delay to temperature—a parameter related to the slope of the ignition delay time on an Arrhenius diagram

3. (3)

   the intrinsic burn rate of each small portion of the charge (the burn rate of a small homogenous fraction of the charge)

The balloon model that was available to the authors and described earlier represents each of these factors and was used to further investigate these effects.

For this study, the following conditions were matched to the experiments: fueling rate (9 bar EMEP), total EGR (43%), $ϕ$ (0.7), and CA50 (6 deg ATDC). The two fuels investigated were PRF40 and PRF87; the latter was chosen to represent gasoline in the framework of the simplified PRF model. For each fuel, the IVC temperature was adjusted to match CA50.

Simulated heat release results at 1000 RPM are shown in Fig. 13. PRF40 showed significant early heat release followed by a short duration main heat release with a high peak. In contrast, PRF87 showed minimal early heat release and a longer burn event with lower peak heat release rate. This trend is the same as observed experimentally (Figs. 4 and 7); however, the difference between the simulations of the two fuels was greater than in the experiments.

For two of the simulations, the wall temperature was identical and set at 400 K. Since the PRF40 required a significantly lower IVC temperature to maintain constant CA50, the temperature gradient across the charge was lower than for PRF87 and may provide a possible explanation of the experimental results. This is shown in Fig. 14, where the temperature distributions of the charge are displayed at 700 deg ATDC as a function of the cumulative mass fraction from hottest to coldest. The hottest zone is represented at a cumulative mass fraction of zero, and the remaining zones are plotted in decreasing temperature against the cumulative mass fraction. The coldest zone is near the cylinder wall.

To test the hypothesis that the temperature gradient was a key factor in determining the heat release rate, a third simulation was performed reducing the wall temperature to 240 K with PRF40 in order to approximate the gradient for PRF87. The resulting temperature distribution is shown in Fig. 14 labeled PRF40*. The results clearly show that the average temperatures for the two PRF40 cases were predicted to be significantly lower than for PRF87. Correspondingly, the predicted IVC temperatures for the three cases in the figure were 384, 491, and 415 K for PRF40, PRF87, and PRF40*, respectively.

Referring back to Fig. 13, increasing the temperature gradient for PRF40* produced a heat release rate for the PRF40* fuel which was very similar to that of PRF87. Comparing PRF40 to PRF87 for the fixed wall temperature, the PRF40 had a significantly higher HRR, and this corresponded to the much more uniform temperature profile for the PRF40. The results support the idea that sequential ignition across a thermal gradient can be a key factor in determining the heat release rate.

To visualize how the temperature profile affects the burn rates, the normalized heat release curves for selected zones are plotted in Figs. 15, 16, and 17 for PRF40, PRF87, and PRF40*, respectively, each at 1000 RPM. The zones shown in the figures are at cumulative mass fractions: 0.0, 0.1, 0.3, 0.7, 0.9, and 1.0 (ordered from hotter to colder). Shown as the dashed line are the average normalized heat release curves corresponding to the absolute heat release results in Fig. 13.

In Fig. 15, PRF40 had a comparatively tight grouping with all zones igniting within a ∼5 deg CA window. For PRF87 in Fig. 16, the crank angle window for ignition was almost doubled. Although the intrinsic rate of heat release for each zone was higher, the ignition of the zones was spread farther apart in time. From this, the steeper temperature gradient for PRF87 appeared to generate a lower overall heat release rate. When the wall temperature was reduced for PRF40* and the IVC temperature was increased to maintain constant CA50, the resulting zone heat release rates, as shown in Fig. 17, had a spread in ignition timing similar to (but slightly greater than) that of PRF87. In addition, the heat release rate of the individual zones appeared to have decreased. Also, the zones near the wall demonstrated slower ignition and combustion, and contributed potentially to the decrease in overall burn rate. All effects combined to reduce the overall burn rate close to that of the PRF87 case.

The model was also used to simulate PRF40 and PRF87 at 2000 RPM corresponding to the experimental data compared in Fig. 11. While the experiments clearly showed higher heat release rate for PRF40 compared with gasoline, the model showed almost identical behavior for PRF40 and PRF87 at the higher speed. The IVC temperatures required for the same CA50 were within 4 K of each other, and results for temperature throughout the cycle were almost identical, as were the average and individual zone predictions for heat release. The predicted heat release rates for individual zones at 2000 RPM were roughly twice as long (in crank angles) and half as high as the PRF87 curves at 1000 RPM. This was consistent with the idea that the intrinsic heat release rate should be approximately constant in the time domain. Inspection of the ignition delay correlation from the reaction mechanism suggests that 2000 RPM was well away from the negative temperature coefficent (NTC) heat release regime, so this result was consistent with the basic premises of the model. Nevertheless, the model disagreed with the experimental data. The cause of the discrepancy is unknown, but possibly is due to the relatively simple kinetic mechanism used, which may not accurately represent the location of the NTC regime.

The experimental methods, and corresponding results, presented in this work allowed for the collection of data which could largely rule out differences in fuel, air, and EGR charge composition as impacting burn rates when comparing different fuels. Three effects were identified which could contribute to observed differences in burn rates between PRF40 and gasoline, which are listed again below:

1. (1)

   the temperature gradient

2. (2)

   the sensitivity of the ignition delay to temperature

3. (3)

   the intrinsic burn rate of each small portion of the charge

These factors cannot be easily separated with regard to their impact on HCCI burn rates. Items (2) and (3) are fuel properties, and ultimately the result of different fuel chemistry. Item (1) is a direct consequence of the latter two factors affecting burn rates.

PRF40, at typical TDC temperatures, will have a shorter ignition delay time and less sensitivity of ignition delay time to temperature compared with PRF87 based on published ignition delay data [25]. The consequence of this for a fixed combustion phasing is that at a given engine speed, the IVC temperature must be lower for PRF40 compared with gasoline.

As was shown with the simulation, a lower IVC temperature at a fixed wall temperature (which is representative of an engine at fixed load, speed, phasing, and jacket water temperature) produced a thermal gradient that was shallower for the PRF40 relative to PRF87, and this increased the average burn rate for PRF40. In the cases where the wall temperatures were the same for both PRF40 (Fig. 15) and PRF87 (Fig. 16), the intrinsic burn rate for PRF87 was faster, despite the average heat release being lower. However, the time of ignition for PRF87 was distributed over a longer time in crank angle degrees, due to the larger thermal gradient. When the thermal gradients were matched by reducing the wall temperature for PRF40* (Fig. 17), the average heat release for PRF40* matched that of PRF87, despite the fact that the intrinsic heat release rates were still lower for PRF40*.

Based on the simulation results presented, the intrinsic burn rate was considered a secondary effect on the overall burn rate in relation to the effect of the thermal gradient. Yet, the intrinsic burn rate, insomuch as it is a function of the fuel chemistry, dictates to a degree the thermal gradient by imposing an average IVC temperature for correct combustion phasing. However, the model has limitations, and agreement between the simulated cases at 2000 RPM did not match the experimental trends. A validated full CFD model with detailed chemistry may capture the differences in experimental burn rates between gasoline and PRF40 at 2000 RPM, and provide further insight on the relative importance of intrinsic burn rate compared with the sensitivity of ignition delay to temperature as the thermal gradient is changed.

The goal of this study was to understand how low octane primary reference fuels impact combustion rates in an NVO-controlled HCCI engine equipped with hydraulically actuated intake and exhaust valves. By adjusting the valve timing and the external EGR level, it was possible to control internal EGR (and its potential effects on local inhomogeneities) while maintaining constant tEGR and composition. In this way, a low octane fuel, PRF40, was compared to 87 AKI gasoline under constant composition conditions with equal levels of iEGR. The effects of varying iEGR at constant composition were also investigated for each fuel. To support the experimental results and gain further insight, a multizone balloon model was used with a reduced reaction mechanism for isooctane and n-heptane fuel blends to investigate the effect of changing in-cylinder thermal gradients on burn rates for PRF40 and PRF87. Significant findings of this work are as follows:

1. (1)

   HCCI burn rates demonstrated significant sensitivity to fuel type when total charge composition, internal EGR, and combustion phasing were held constant. PRF 40 showed shorter burn durations and higher peak heat release rates—approximately 35% higher than gasoline. Little or no sensitivity to engine speed was observed for either fuel under these conditions.

2. (2)

   Burn durations for both fuels showed little sensitivity to internal EGR up to about 35%. At higher levels (up to 45%), gasoline burn durations increased by approximately 20%. Internal EGR levels higher than 35% for PRF40 were not tested.

3. (3)

   Simulation results suggest that the increase in burn rates for PRF40 relative to PRF87 may have been due primarily to decreased thermal gradient between the bulk or average temperature and the wall temperature for PRF40 (because the PRF40 required significantly lower charge temperature), and secondarily due to changes in the intrinsic burn rate of the fuels. Using the simulation to artificially apply thermal gradients to PRF40 equal to that of PRF87 produced lower burn rates similar to that of PRF87.

Questions that remain unanswered include the relative importance of thermal versus compositional stratification, as iEGR is varied. In this study, the fuel was direct injected, so some change in compositional stratification of fuel may occur as iEGR is varied. For comparison between fuels, iEGR was held constant, but the changes in burn rates observed for the gasoline as a function of iEGR could be due to compositional as well as thermal stratification. Another unknown is the role of fuel chemistry in the different trends between PRF40 and gasoline iEGR studies: Is the constant burn duration for PRF40 due primarily to the reduced thermal gradients relative to gasoline, or is the fuel chemistry less sensitive to the changes in thermal gradients as iEGR is changed?

The model used a reduced mechanism, and significant assumptions were made with regard to turbulence and heat transfer. A more detailed chemical kinetic mechanism, coupled with the multizone model, could be used for more refined results. The results presented here demonstrate that at low speed (1000 RPM), the majority of the differences in heat release rates are likely due to thermal stratification and that differences in fuel chemistry are less important. At 2000 RPM, the experiments showed significantly higher heat release rates for the PRF40 fuel, but the model was unable to explain the results since all thermal conditions were the same. It is possible that under these conditions, a more refined model is needed to improve understanding of the interplay between thermal and chemistry sensitivities of the fuels as they affect HCCI combustion.

The authors would like to thank Dr. Janardhan Kodavasal (Argonne National Laboratory) for providing expertise and access to the simulation tools he developed for HCCI combustion while at the University of Michigan.

This work was supported by the Department of Energy under the University Consortium on Efficient and Clean High Pressure Lean Burn (HPLB) Engines, directed by the University of Michigan under Department of Energy Contract No. DE-EE0000203.

AKI =

antiknock index—average of RON and MON

CA =

crank angle (deg)

CA10 =

10% mass fraction burn location (deg ATDC)

CA50 =

50% mass fraction burn location (deg ATDC)

CA90 =

90% mass fraction burn location (deg ATDC)

CA10_M =

CA10 with LTHR portion removed (deg ATDC)

CFD =

computational fluid dynamics

deg ATDC =

degrees after top dead center, combustion

deg BTDC =

degrees before top dead center, combustion

eEGR =

external EGR—recirculated to intake (% mass)

EGR =

exhaust gas recirculation, or residual gas (% mass)

EMEP =

energy mean effective pressure (bar)

HCCI =

homogenous charge compression ignition

HRR =

net heat release rate (J/CAD)

iEGR =

internal EGR—in-cylinder residual gas (% mass)

IMEP_g =

gross indicated mean effective pressure (bar)

IVC =

intake valve closing location (deg BTDC)

LHV =

lower heating value (kJ/kg)

LTHR =

low-temperature heat release

MFB =

mass fraction burned (%)

MON =

motor octane number

NTC =

negative temperature coefficient

NVO =

negative valve overlap

ON =

octane number

PRF =

primary reference fuel

PRF40 =

PRF, 40% isooctane and 60% n-heptane (mol basis)

PRF87 =

PRF, 87% isoocatne and 13% n-heptane (mol basis)

RON =

research octane number

RPM =

revolutions per minute

TDC =

top dead center

tEGR =

total EGR—total exhaust gas diluent (% mass)

$ϕ$ =

fuel/air equivalence ratio

$ϕ′$ =

fuel/charge equivalence ratio