Effect of increased fuel volatility on CDC operation in a light duty CIDI engine

Effect of increased fuel volatility on CDC operation in a light duty CIDI engine Fuel 194 (2017) 195–210 Contents lists available at ScienceDirect Fuel journal homepage www elsevier com/locate / fuel[.]

Full Length Article

Effect of increased fuel volatility on CDC operation in a light-duty CIDI

engine

University of Wisconsin – Madison, United States

DOE Great Lakes Bioenergy Research Center, United States

a r t i c l e i n f o

Article history:

Received 24 August 2016

Received in revised form 29 November 2016

Accepted 15 December 2016

Keywords:

Jet entrainment

Physical properties

Direct injection

Biofuel

Internal combustion engines

Volatility

a b s t r a c t Alternative diesel fuels derived from biomass can vary significantly in volatility compared to their petroleum-derived counterparts, and their appropriate utilization is contingent on their compatibility with existing engine infrastructure To investigate this compatibility, experiments were carried out to study the effect of fuel volatility on conventional diesel combustion (CDC) performance under a wide range of in-cylinder thermodynamic conditions at start of injection (SOI) Fuels of matched reactivity (i.e., cetane number (CN)) and varying volatility were produced by blending binary mixtures of 2,6,10-trimethyldodecane (farnesane) and 2,2,4,4,6,8,8-heptamethylnonane, octane number primary reference fuels (PRF), and cetane number secondary reference fuels (SRF) Nine fuel blends were tested in total, con-sisting of 3 volatility characteristics at 3 reactivity levels Five engine operating conditions were utilized, ranging from 14.7–29 kg/m3and 980–1120 K in-cylinder density and temperature at SOI Testing was performed in a single-cylinder GM 1.9 L diesel engine Only small differences in ignition delay (ID), in-cylinder pressure, and heat release rate (HRR) were observed between fuels of matched CN, regardless

of their volatility An analysis of the spray breakup and mixture formation process indicated that there were only small variations in ambient air entrainment and jet temperature between fuel blends, in agree-ment with the observed combustion behavior

Ó 2017 The Authors Published by Elsevier Ltd This is an open access article under the CC BY-NC-ND license

(http://creativecommons.org/licenses/by-nc-nd/4.0/)

1 Introduction

The increasing scarcity of petroleum resources coupled with the

deleterious ecological effects of continued carbon dioxide

emis-sions resulting from fossil fuel utilization present serious

engineer-ing challenges for industrialized nations Compoundengineer-ing this, the

demand for heavy-duty transportation and aviation fuels

tradition-ally provided by diesel-like middle-distillates is forecast to

increase steadily over the coming years[1] Alternative fuels,

par-ticularly those derived from biomass, offer a compelling means to

augment the existing petroleum fuel supply to meet this increase

in demand while simultaneously satisfying ever tightening vehicle

emissions standards[2–4] Hydrotreated renewable diesel (HRD)

fuels, derived from oils and fats, and hydrogenated isoprenoids

(such as farnesane), derived from fermentation, are of particular

interest as they offer superior physical characteristics (such as a

low cloud point) and emissions performance compared to fuels

derived from trans-esterified oils[2,4]

The introduction of a new fuel into the supply stream offers its own set of challenges Petroleum-derived fuels are by nature a complex blend of molecular species, each contributing individually

to the chemical and physical properties of the fuel as a whole In contrast, fuels derived from biomass are often composed of only

a few, perhaps even a single chemical species This compositional difference makes matching the properties of an alternative fuel

to its petroleum counterpart prohibitively difficult, and instead some allowance must be made for property differences Under-standing exactly how much, and in which properties variation can be tolerated without adverse effects will allow for the greatest utilization of alternative fuels In particular, this study focuses on properties of interest to compression-ignition direct-injection (CIDI) internal combustion engines, and the fuel most commonly utilized in such combustion applications, diesel fuel

This study is intended to expand upon work performed previously by the authors in a heavy-duty CIDI engine[5] In the previous study, the stock heavy-duty fuel injector was replaced

by one with a smaller injector orifice hole diameter in order to eliminate spay-wall interactions and piston bowl wetting While this was useful for eliminating potentially complicating factors associated with the injection event, it is not representative of http://dx.doi.org/10.1016/j.fuel.2016.12.064

0016-2361/Ó 2017 The Authors Published by Elsevier Ltd.

⇑ Corresponding author at: University of Wisconsin – Madison, United States.

E-mail address: groendyk@wisc.edu (M Groendyk).

Contents lists available atScienceDirect

Fuel

j o u r n a l h o m e p a g e : w w w e l s e v i e r c o m / l o c a t e / f u e l

typical diesel-engine operation In addition, the range of

thermo-dynamic in-cylinder conditions in the previous study was limited

and testing was only done at a single fuel reactivity level The

pre-sent study significantly expands the range of in-cylinder

tempera-ture and density conditions and performs tests at multiple fuel

reactivity levels In addition, the configuration of the engine used

is representative of a stock, small-bore diesel engine, which will

include the influence of spray-wall interactions under certain

in-cylinder conditions

1.1 Background

1.1.1 Diesel combustion

A discussion of the implications of varying the physical

proper-ties of a diesel fuel must begin with an understanding of the

phys-ical processes that diesel fuel must undergo to operate successfully

in a CIDI engine Conventional diesel combustion (CDC) is typified

by late direct fuel injection, whereby liquid fuel is injected at high

pressure into the combustion chamber near the end of

compres-sion Due to the compression event, the chamber contents will be

at elevated temperature and density when the fuel is introduced,

allowing it to rapidly evaporate, form an ignitable mixture with

the ambient oxygen, and begin to burn[6] The relevant physical

changes the fuel must undergo are evaporation and mixing, and

the fuel injection and resulting spray provide the mechanisms by

which these changes come about

Generally, CDC fuel injections can be thought of as mixing

lim-ited sprays, where the rate of ambient gas entrainment determines

both the extent of mixing and the thermal state at any position A

detailed discussion of the spray formation process will not be

included here, instead readers are directed to the works of Reitz

and Bracco[7], Naber and Siebers[8], and Siebers[9–11]for

addi-tional background on spray formation

1.1.2 The role of volatility in CDC

While the implications of physical properties for spray

phenom-ena have been well studied, the impact of spray behavior on the

overall combustion event is somewhat more complex The

varia-tions in the condivaria-tions produced by the dispersion and

vaporiza-tion of the spray are convolved with further variavaporiza-tions in the

chemical processes of ignition and combustion which follow

Using computer simulations, Ra et al.[12]were able to study

the combustion of a diesel fuel in which individual physical

prop-erties were systematically varied The effects of combustion

kinet-ics were isolated from the simulation by using the chemical

kinetics of n-tetradecane for all of the simulated fuels Their results

indicated that all of the 11 physical properties included in their

spray breakup and droplet collision/evaporation models had a

sig-nificant impact on the simulated combustion event In particular,

they found that the density, vapor pressure, and to a lesser extent

the heat capacities (both liquid and vapor) of the fuel exerted the

greatest influence on combustion performance[12]

A similar study by Kim et al.[13]using n-dodecane for the

base-line fuel and chemical kinetic mechanism, varied six fuel physical

properties within ranges found among commonly used diesel fuel

surrogates They found that at diesel-relevant conditions, varying

liquid density, viscosity, vapor pressure, and heat capacity had

sig-nificant impact on liquid penetration length, and that the liquid

density and heat capacity variations had the most significant effect

on ignition delay (i.e the length of time between injection and the

start of combustion)[13] Both the results of Ra et al.[12]and Kim

et al.[13]indicate that for kinetically identical fuels ignition can

still be impacted by changes in the local conditions resulting from

differences in fuel physical properties

Both of these studies included volatility properties: vapor

pres-sure (distillation curve), enthalpy of vaporization, and liquid and

vapor heat capacities Both studies also showed some dependence

of spray and or combustion behavior on these volatility properties For this study volatility can be defined as the thermal dependence

of a fuel’s vapor-liquid equilibrium fractions

According to both the mixing-limited jet vaporization model and to experimental observation, volatility plays a crucial role in determining the extent of liquid penetration into the cylinder [8,9,11,13] While this has obvious implications for cylinder-wall/ piston-bowl impingement, the effect of liquid length on subse-quent spray mixing and other important combustion parameters (such as ignition delay) is not fully understood[14–16] Ra et al [12] predicted a large change in both ignition timing and peak pressure for kinetically identical fuels differing only in vapor pres-sure; however, the study by Kim et al.[13]suggests that the effects

of volatility are limited mainly to liquid penetration, and that igni-tion timing is more strongly affected by a fuel’s liquid density and heat capacity

In addition to these computational studies, numerous experi-mental studies have been undertaken to assess volatility’s influ-ence on combustion directly [5,14–22] Generally, the ignition delay of a fuel has been observed to be insensitive to its liquid length (and therefore its volatility), and is instead correlated with its cetane number[5,17,21,22] This correlation is not perfect how-ever, and several investigations have noted deviations in fuel igni-tion behavior which can potentially be ascribed to volatility effects [14–16]

Alternative fuels with volatilities that differ significantly from petroleum diesel are being seriously considered Alcohols, such

as butanols for example, have received considerable attention as potential blend stocks, and are of significantly higher volatility than conventional diesel[23–25] Given the importance of ignition behavior for effective CIDI fuel utilization, the impact of volatility

on ignition should be resolved, so that appropriate recommenda-tions regarding the required volatility of alternative diesel fuels can be made

1.1.3 A word on the cetane number Ignition timing is of crucial importance for a fuel that is to be used in a compression-ignition engine[26,27] Engine load, emis-sions, and efficiency are all heavily dependent on combustion phasing, which in a CIDI engine is controlled by the injection tim-ing and the reactivity of the injected fuel Reactivity, betim-ing a key factor in combustion phasing, is of critical importance, and must

be quantified in some way

The cetane number (CN) is a practical reactivity metric that is widely used to characterize diesel-like fuels The CN of a fuel is defined by ASTM standard D-613[28], and is determined from tests performed in a specific engine by comparing the compression ratios necessary for fuels to achieve a specified ignition delay at a prescribed set of operating conditions The circumstances of the

CN test differ significantly from modern diesel engine operation, and despite its popularity, appropriate use of the CN has been the subject of repeated inquiry [10,15–17,27,29–32] Therefore, the use of the CN for the present work will be discussed, and its applicability established before proceeding

The present work is intended to provide insight into the physi-cal processes of autoignition in a modern CIDI engine This kind of autoignition event includes both chemical and physical processes

In order to accurately study variations in the physical processes only, the chemical portion of the auto-ignition should be held fixed

to as high a degree as possible to eliminate it as a variable One method to attempt to accomplish this is to match CNs of the fuels

to be tested The issue with this approach is that one would think that the CN is a characterization of the total auto-ignition process, and may include the influences of both chemical and physical

properties Before CN can be employed as a metric of strictly

chem-ical reactivity, more consideration is needed

A number of studies have been undertaken utilizing primary

reference fuel (PRF) blends, similar to those used in the current

work, in homogeneous-charge compression-ignition (HCCI)

engines For the current study, PRF refers to the reference fuels

used for octane number (ON) testing, i.e., n-heptane and isooctane

As the name suggests, HCCI utilizes a fully mixed and

pre-vaporized, homogeneous charge of fuel and oxidizer, and

compression-induced heating to initiate combustion Such a

strat-egy is entirely kinetically controlled and free of any influence of the

physical properties which govern mixture formation These studies

have shown that, for PRF blends and conventional diesel fuels,

combustion phasing is strongly correlated with CN [30–32] As

the combustion phasing is controlled entirely by the fuel chemistry

under these conditions, it follows that CN is correlated to fuel

chemistry for these types of fuels

Janecek et al.[33,34]have performed other HCCI studies using a

fuel substitution technique to relate the reactivities of diesel-like

fuels to blends of PRFs They report a strong correlation between

CN of the diesel-like fuel and the composition of the PRF blend

(the PRF number) that has matched combustion phasing - and

therefore equivalent reactivity The correlation is applicable over

a wide range of CN (20–75) and is insensitive to the particular HCCI

operating conditions used This offers additional compelling

evi-dence that CN primarily characterizes chemical reactivity, as PRF

number is based on ON testing, which does not directly involve

spray breakup and mixture preparation processes Based on these

findings, for the circumstances of this study, CN has been deemed

an appropriate metric of fuel chemical reactivity for PRFs and

diesel-like fuels, and is applied as such

1.2 Objective

The objective of this investigation is to ascertain the effect that

large variations in fuel volatility of will have on CDC in a typical

light-duty application indicative of current automotive diesel

engines Ignition delay is used as a key metric, as it has strong

influence on virtually all combustion parameters of importance

Volatility is isolated as completely as possible from other physical

properties of importance (such as liquid density) by utilizing fuels

composed of carefully selected pure chemical species To facilitate

the widest range of testing conditions, the test fuels were blended

to produce a wide range of chemical reactivity, and then run in the

engine under a wide range of intake conditions

2 Methods

2.1 Fuels

To investigate the effect of fuel volatility on ignition at various

levels of reactivity, both volatility, and reactivity must be

con-trolled as independently as is possible, across the widest feasible

range with minimal changes to other important fuel

characteris-tics The approach to do this used binary blends where each of

the two constituent compounds had approximately matched

volatility, but substantially different reactivity (as quantified by

CN) This allowed the reactivity of the blend to be adjusted by

altering the relative proportions of its constituents without

signif-icantly impacting the volatility of the blend as a whole Care was

also taken with respect to the chemical makeup of the blending

components to ensure similar reaction chemistry was also

achieved (e.g., similar stoichiometry, oxidation pathways, etc.)

We have discussed our metric for reactivity but have not yet

established our basis for the comparison of volatility Volatility is

a complex property, and it is difficult to fully quantify, but it can

be thought of as the thermal energy required to vaporize a fuel Vapor pressure, boiling point, enthalpy of vaporization, and heat capacity all influence evaporation, and so are all components of volatility Generally for fuels however, volatility is used to refer

to the temperature dependence of certain vapor-liquid equilibria (e.g boiling point, flash point, etc.) So, for the purposes of this study, the volatility of a fuel will be defined by its distillation curve

as measured by test method ASTM D-86[35] The distillation curve provides an approximate range of temperatures over which a sam-ple of fuel is vaporized, not a single number, so direct comparison

is not straightforward However, the use of binary blends whose components have very similar boiling points facilitates this com-parison by flattening the distillation curve to a near constant value PRFs are a natural choice for this application, due to their struc-tural similarity, availability, and the fact they are well studied in the literature In addition, their volatility is nearly identical, as shown inTable 1, meaning they can be blended in any proportion without affecting the distillation curve of the blend This volatility

is significantly higher than that of even the most volatile portion of

a typical middle-distillate fuel, which can be seen inFig 1 Finally, the constituent species of PRF-based blends, n-heptane and isooc-tane, are reported to have CN’s of 56 and 12.0–17.5 respectively [37] (as determined by ASTM D-613), which offers a reactivity range of practical interest to modern CIDI engines

An appropriate counterpart for the PRF-based fuels will have similarly matched chemical composition but much lower volatility, ideally in the range of currently available diesel-like fuel Farne-sane (FAR) or 2,6,10-trimethyldodecane, was selected as the high reactivity component of the low volatility binary blend; 2,2,4,4,6, 8,8-heptamethylnonane (HMN) was selected as the low reactivity component Both are saturated alkanes which differ primarily in their respective degrees of branching Both are also readily avail-able, and reasonably well studied with respect to combustion pro-cesses; HMN is used as a primary reference fuel in ASTM D-613 and farnesane is used as a bio-derived diesel/jet fuel substitute [2,4,26,28,48] These species, like the PRFs, offer an approximately matched volatility and a comparable range of reactivities (from CN

58 for pure farnesane[26]to CN 15 for pure HMN[28]), however,

as their higher molecular weight suggests, their volatilities are sig-nificantly lower and are within the range of typical diesel fuels The difference in distillation temperatures between PRF and FAR-HMN-based fuel blends (shown inFig 1) is uniform and approximately

150°C, indicating the magnitude of the volatility difference achieved

Utilizing binary components of approximately matched boiling point provides fuels with a narrow boiling ranges In contrast, con-ventional diesel fuels, by virtue of their diverse chemical makeup, provide wide boiling ranges (typically >100°C) The difference, for example, between a standard #2 diesel and the pure species blends

of PRF and FAR-HMN is readily apparent inFig 1 The width of the boiling range should also be taken into account in an assessment of the volatility of a fuel, and so its effect was also studied

Secondary reference fuels (SRF) for CN testing were selected to provide a third fuel blend which offered diesel-like volatility with respect to boiling range The SRFs are not pure chemical species, but instead consist of compositionally diverse petroleum distil-lates This allows SRF-based blends to vary in reactivity between

CN 75.2 and 19.4 while maintaining a wide boiling range Also, the SRF fuel blends are directly correlated to CN (as they are them-selves CN reference fuels) which offers a direct comparison to the

CN test for the pure component blends Given the CN range of the selected pure species components, 3 reactivity targets: CN 35, 45, and 55 were selected and appropriate splash blends of PRF, SRF, and FAR-HMN were produced for each PRF blending was done according to the ON-to-CN correlation of Kalghatgi [49]

FAR-HMN blending was done by volume according to CN weighted

average assuming ideal mixing as described in ASTM D-613, and

the author’s previous work[5,28] SRF blending was also done by

volume, per the manufacturer’s instructions For CN 45, emissions

certification #2 diesel fuel (Haltermann 2007 EPA tier II cert

die-sel) was also tested as a reference case

Unfortunately, it is impossible to alter the physical fuel

proper-ties completely independently of one another due to the strong

cor-relation between molecular weight and properties like volatility,

viscosity, density, and H/C ratio for a given molecular structure

Despite careful selection of the blending components, notable

dif-ferences in other properties of importance were unavoidable The

magnitude of the variation in these properties is shown relative to

the #2 diesel reference fuel inFig 2 In addition to the volatility

dif-ferences: density, heat capacity, surface tension, and viscosity have

significant deviations, and their effects on the spray breakup

pro-cess must be considered in the analysis of the experimental results

The three properties exhibiting the largest deviation from diesel fuel are viscosity, volatility, and molecular weight Our analysis will show that viscosity does not strongly impact the physical character of the sprays utilized in this study, as they are within the fully atomizing regime Similarly, molecular weight can have no direct influence on the spray on a macroscopic scale, and therefore is considered of secondary importance to the macroscopic physical properties to which it is related, such as liquid density This leaves volatility (with a minimum 48% differ-ence in distillation temperature for PRF-based fuels relative to diesel) as the physical property of largest variation As such, the combustion performance of the fuels selected should provide a clear indication of any volatility-driven phenomena, even in the presence of complicating factors A full summary of relevant phys-ical and chemphys-ical properties of the blend components can be found inTable 1

2.2 Engine setup The engine used in this study was an instrumented single-cylinder version of a General Motors/Fiat JTD 1.9-L light-duty die-sel engine In-cylinder pressure data were collected using a Kistler 6125A pressure transducer with a Kistler 5010 charge amplifier A 0.25-degree resolution crank angle encoder affixed to the crank-shaft was used to time data acquisition Three hundred cycles of pressure data were acquired at each test condition and used for analysis The dimensions and configuration of the engine are sum-marized inTable 2 Generally, the configuration was intended to be

as similar as possible to a stock engine

Fueling was provided by a Bosch CRI 2.2 direct injector equipped with a 7 hole 140-lm tip, pressurized with a common-rail (CR) fuel pump Return fuel from both the CR and the injector bleed was passed through a brazed-plate fuel-to-water heat exchanger and recycled to the high-pressure pump to limit fuel consumption Fuel was supplied to the CR system as needed by a

3 gal stainless steel canister, which was pressurized to 80–

100 kPa (gauge) with nitrogen Fueling rate was metered using a Coriolis mass-flow meter prior to introduction to the CR loop Injection duration was varied for each fuel to compensate for dif-ferences in energy density Injected energy was held constant for all fuels tested, however, the range of energy densities among the test fuels was small, so only small adjustments to injection duration were needed An injection pressure of 100 MPa was used for all fuels and conditions studied

Table 1

Summary of fuel properties for the blending components utilized [36,37,26,38–42,28,43–47].

Property Fuel component

n-Heptane Isooctane Farnesane HMN SRF-T SRF-U #2 Diesel

q[kg/m 3

m[cSt] (Temp [°C]) 0.496 (40) 0.547 (50) 2.95 (40) 3.08 (40) 2.238 (40) 1.145 (40) 2.70 (40)

MW [kg/kmol] 100.23 114.26 212.47 226.50 197.90à 148.70à 193.40à

à Average molecular weights determined from D-86 distillation data, see Ref [40].

y Density data at 293 K.

⁄ Density data at 288 K.

350

300

250

200

150

100

100 80

60 40

20 0

% Recovered [-]

Cert Diesel FAR-HMN PRF

CN 35 SRF

CN 45 SRF

CN 55 SRF

Fig 1 Volatility of representative fuel blends used in this study Note the variation

in the SRF curves due to compositional variation and the large difference between

PRF and FAR-HMN-based blends.

Fuel blends were run systematically by blend type and pseudo-randomly by CN The fuel system was designed to minimize trapped volume to facilitate drainage and purging when fuels were exchanged Great care was taken to minimize fuel cross-contamination by utilizing a 2-stage purge process First the low-pressure side of the fuel system (return fuel heat exchanger, fuel reservoir, and CR pump lines) are fully drained of fuel and purged with N2 Second, after the reservoir is charged with the new fuel blend, the high-pressure side (CR, injector, high pressure fuel lines) return lines are diverted from returning to the low-pressure side, and the CR pump is jogged This runs fuel from the low pressure side, through the high-pressure side and out of the system, where it is collected and its volume is measured Once 2 the trapped volume

of the high-pressure side has been collected, the system is closed by returning the high-pressure side return lines to their original flow pattern

Dry intake air was supplied to the engine through a surge tank from the dedicated building compressed-air line Intake temperature was maintained using electronically controlled hea-ters, and air mass flow rate was controlled using choked-flow ori-fices The exhaust back pressure was maintained at 10 kPa below intake pressure for all conditions investigated This was done to minimize the amount of exhaust residuals in-cylinder This also allowed a wider range of in-cylinder temperatures at start of injec-tion (SOI) to be achieved Exhaust gas emissions measurements were performed with a Horiba 5-gas emissions bench, Horiba flame ionization detector (FID) hydrocarbon analyzer, and an AVL-415 smoke meter Together the emissions bench and the FID measured the concentrations of CO, CO2, O2, NOx, and unburned hydrocarbons Emissions data were very similar for the different fuels except for the expected differences in CO2 concentrations due to the different H/C ratios[5] Only soot emissions showed sig-nificant differences and will be reported A schematic showing the major components of the air and fuel systems is shown inFig 3

2.3 Operating conditions Test conditions for the study were selected to provide a wide cross section of in-cylinder conditions that might be encountered

in automotive or heavy-duty diesel engines Temperature and den-sity of the cylinder contents at SOI characterize the operating con-dition, as these are the parameters which govern spray breakup and ignition phenomena Of particular interest to this study are

Percent Difference

MW [kg/kmol]

H/C ratio [-]

AFRst [-]

LHV [MJ/kg]

Cp (@298 k) [kJ/kg-k]

ρ [kg/m

3

]

σ [dyne/cm]

ν [mm

2

/s]

CN 35 PRF FAR-HMN SRF

(a)

Percent Difference

MW [kg/kmol]

H/C ratio [-]

AFRst [-]

LHV [MJ/kg]

Cp (@298 k) [kJ/kg-k]

ρ [kg/m

3

]

σ [dyne/cm]

ν [mm

2

/s]

CN 45 PRF FAR-HMN SRF

(b)

Percent Difference

MW [kg/kmol]

H/C ratio [-]

AFRst [-]

LHV [MJ/kg]

Cp (@298 k) [kJ/kg-k]

ρ [kg/m

3

]

σ [dyne/cm]

ν [mm

2

/s]

CN 55 PRF FAR-HMN SRF

(c)

Fig 2 Comparison of relevant physical properties of PRF, FAR-HMN, and SRF-based

fuel blends relative to the #2 diesel reference fuel Comparison is made for binary

mixtures of different reactivities (a) CN 35 (b) CN 45 (c) CN 55.

Table 2 Relevant engine geometry and operating parameters.

Engine geometry Base engine GM 1.9 L Diesel Compression ratio 16.5:1 Displacement [L] 0.477

Intake valve closure 132°

Exhaust valve opening 112°

Piston bowl type Stock, re-entrant

Injector geometry Number of holes 7 Orifice diameter [lm] 140 Included spray angle 144°

Injection timing 5°

Injection duration [ms] 0.6 Engine load [Bar IMEPg] 6.6–7.6 Injected fuel energy [J] 830 Injection pressure [MPa] 100

operating conditions resulting in low in-cylinder temperature and

density, where the volatility property differences are anticipated to

have the largest influence on the combustion process and engine

performance Intake air temperature and pressure were varied independently to achieve temperatures and densities in the range

of 980–1120 K and 14.6–29 kg/m3, respectively It was decided to test only the extremes of the possible conditions to assess the over-all magnitude of any global trends This sparse testing matrix con-sisted of 5 points, 4 at the extreme combinations of temperature and density, and one intermediate condition to provide informa-tion on the shape of the trend between them (see Fig 4 and Table 3)

In all cases, the engine was run at a speed of 2100 RPM and was fired for at least 1 h prior to any data collection to ensure that the engine had reached a stable thermal condition, and that any resid-ual fuel not eliminated during purging had been flushed out of the fuel system A portion of the conditions matrix was repeated mul-tiple times to determine data repeatability PRF and SRF-based fuels were inexpensive enough to allow for 2 and 3 measurements, respectively, taken on different days at each condition HMN was prohibitively expensive, even at the modest fuel flow rate required

of the light-duty engine; FAR-HMN data presented in the this paper represent only a single trial for each test condition The repeatability for the other fuels at equivalent conditions provides insight into the day-to-day variability and give an approximate measure of the uncertainty in the FAR-HMN results

Injection duration was selected such that ignition occurred prior to end of injection for the lowest CN at Condition 1 (lowest temperature and density) This criterion established the fueling rate for the entirety of the study, as fuel energy was maintained for all conditions and fuels These conditions resulted in a wide range of ignition delay (from 200 to 800ls), as well as a variety

of pressure and heat release characteristics This is exemplified in the pressure and heat release results shown inFig 5(a) for the SRF CN 45 blend at all 5 operating conditions, and inFig 5(b) for SRF blends of all 3 reactivities at Condition 3 Despite their differ-ences, Conditions 2 and 3 resulted in almost identical ignition delay and heat release, making their results difficult to distinguish

in the following analysis

2.4 Data analysis techniques Ignition delay (ID) is one of the characteristics of central interest for this study However, determining the ignition delay from in-cylinder pressure measurements is not always straightforward Numerous ID determination techniques have been described in the literature, each with merits and drawbacks For the present study, the ID determination technique of Groendyk and Rothamer [5]is used This approach determines the start of combustion (SOC) from raw in-cylinder pressure data by locating the onset of rapid pressure rise associated with pre-mixed burn Ignition delay is then derived from the difference between SOI and SOC timing Readers are directed to the work of Murphy and Rothamer

engines

In addition to ID data, cylinder pressure and heat release rate (HRR) data will be presented and discussed Pressure data are pre-sented in both the raw as acquired state and with low-pass

filter-Fig 3 Schematic diagram of the engine laboratory See Table 2 for the relevant

specifications.

450 400

350 300

Intake Temperature [K]

35

30

25

20

15

10

1200 1150 1100 1050 1000 950 900 850

SOI Density [kg/m3]

1

2

3

4 5

Fig 4 A map of the operating conditions utilized in this study The color scale

indicates temperature at SOI in Kelvin and the contours indicate density at SOI in

kg/m 3

The axes correspond to the intake conditions used to achieve them (For

interpretation of the references to color in this figure legend, the reader is referred

to the web version of this article.)

Table 3

Running conditions.

Temperature [K] Pressure [kPa] (absolute) Temperature [K] q[kg/m 3

]

ing The filter was designed to minimize passband ripple by

utiliz-ing a wide transition band with Gaussian roll off Passband and

stopband edges were fixed at 1650 Hz and 7500 Hz respectively,

and a stopband attenuation of 10 dB was specified for all pressure

data that was filtered However, analysis was generally carried out

on raw pressure data to avoid the inclusion of any filter artifacts

HRR was calculated from the pressure data (both raw and filtered)

and engine geometry using the 1st law balance given by Gatowski

et al.[50]

To detect any differences in mixture preparation as the result of

physical differences between the fuels, close scrutiny was applied

to pre-ignition phase (i.e., the period between SOI and main

igni-tion) of the cycle This proves problematic for the analysis of heat

release data, due to the low levels of HRR in this region, and the

susceptibility of HRR calculation to noise and artifacts introduced

by conventional noise-reduction techniques[5,21] However, the

specificity of region of interest allows for the application of an

alternative method of noise reduction A high-order polynomial

is fit to a small section of pressure data, offering good agreement with measurements and a noise free derivative for the calculation

of heat release Provided the resolution of the pressure data is high and the fitting window is carefully selected to eliminate the ringing that follows ignition, the technique is quite effective and provides a clear HRR trace for comparison free of any filtering artifacts, as shown inFig 6 This technique was applied to all HRR comparisons

in the pre-ignition region

3 Results and discussion 3.1 Ignition delay

Average ignition delays for each condition were determined from the mean of the ignition delays of the individual cycle pres-sure data.Fig 7(a) shows the average ignition delays for all fuels

at the 5 conditions tested as a function of fuel blend CN The igni-tion delay results show the expected sensitivity to cylinder tem-perature and density at SOI [5,51] ID was observed to decrease with increasing CN and the sensitivity of ID to CN at reduced CN

is increased for all conditions This is most apparent at conditions conducive to longer ID

As mentioned earlier, the similarity of the ignition behavior observed at Conditions 2 and 3 makes them difficult to discern in Fig 7(a) When the results are viewed individually by condition (Figs 7(b) and8) the agreement between the fuels can be seen more clearly Results for the 3 CN 45 binary fuel blends are also compared to the #2 diesel reference fuel at each condition Good agreement between the diesel and the blended fuels provides a check on the SRF composition (which serves as a known reactivity reference), and the ignition delay determination technique Uncertainty in the ID measurements are presented based on 95% confidence in the mean ID determined from repeated trials

at identical conditions The maximum uncertainty occurred at Con-dition 4 and was 51ls The average uncertainty in the ID data was 15.9ls Uncertainty in individual measurements was also investi-gated, but found to be small (2ls) relative to the variability in repeated trials Uncertainty is not presented for FAR-HMN data,

as no repeat trials were performed, as previously discussed

120

100

80

60

40

20

0

40 30 20 10

0 -10

Crank Angle [ º ]

400 350 300 250 200 150 100 50 0

Test Condition

1 2 3 4 5

SRF: CN45 SOI

100

80

60

40

20

0

40 30 20 10 0

-10

Crank Angle [ º ]

200 180 160 140 120 100 80 60 40 20 0

CN

35 45 55

SOI

SRF: Condition3

Fig 5 (a) CN 45 SRF blend pressure and HRR data for all test conditions, note the

closely matched HRR for Conditions 2 and 3 (b) Pressure and HRR data from

Condition 3 for SRF blends of each CN selected for testing.

50

48

46

44

42

40

Crank Angle [ º ]

40

30

20

10

0

-10

Raw Data

15th Order Fit

Fig 6 HRR data for CN 35 PRF at Condition 1 calculated using both raw pressure data, and pressure data that has been conditioned using the polynomial fit noise-reduction technique.

The shortest ignition delays were observed at Condition 5,

which had both the highest in-cylinder temperature and density

of all the conditions studied Ignition delays for Condition 5 also

showed the least sensitivity to CN, as was expected Energy

trans-fer to the fuel jet under these conditions is sufficiently rapid that

even fuels of widely varying activation energy will overcome their

energy barriers within a comparable interval following injection

This leads to the characteristic reduction in sensitivity to CN

In contrast, Condition 1 shows the longest ignition delays due to

its comparatively cool, low density, cylinder conditions at the SOI

Sensitivity of ID to CN is much greater, varying by more than

200ls from CN 35 to 55, compared to the roughly 50ls variation

seen at Condition 5 over an identical CN range Under these

condi-tions differences between the fuel blends become significant for

certain CNs It was observed that the PRF-based fuel blends for

both CN 35 and 45 experienced shorter ignition delay compared

to both FAR-HMN and SRF-based blends of comparable reactivity,

while at CN 55, ignition delay for the PRF blend was comparable

to the SRF-based blend FAR-HMN-based blends showed a

consis-tently longer ID compared to the PRF-based blends at Condition

1 Interestingly, agreement between the SRF and FAR-HMN-based

blends trends opposite of the SRF and PRF-based fuel blends, and

the best agreement is found at the lowest CN The reasons for this

behavior are not clear at present For all fuel reactivities tested

PRF-based blends ignited approximately 50ls in advance of

FAR-HMN-based blends at Condition 1, which is a small but potentially

statistically significant difference With the exception of Condition

1, no statistically significant differences in ID were observed for

any of the fuel blends at the conditions tested

3.2 Pressure and heat release

Cylinder pressure and apparent heat release rate data are

con-sistent with the ignition delay results The major features of the

pre-ignition, ignition, and mixing controlled combustion events

are well matched at each condition at a given cetane number, regardless of fuel blend.Fig 9(a) shows pressure data for the CN

45 fuels at Condition 3 as an example Onset of rapid pressure rise

is well matched between all fuels, as is peak pressure, ringing intensity, and chamber pressure through expansion Calculated heat release rates for this same condition show a corresponding degree of similarity, and, in addition, indicate that cylinder cooling during the pre-ignition phase due to fuel vaporization and heating

is comparable despite the increased volatility of the PRF blend Only small differences in HRR and pressure can be identified for any given condition, but some are present Most notably, there is

a slight increase in the rate of pressure rise for the PRF blends com-pared to either the SRF or FAR-HMN blends for most of the condi-tions tested This increase is reflected in the HRR as a slight increase in peak HRR during the premixed-burn spike This effect

is not observed at Condition 1 for fuels of CN 35 however, as shown

As previously indicated by the calculated ignition delay, Condi-tion 1 resulted in PRF-based fuels igniting in advance of both FAR-HMN and SRF-based blends of CN 35 Inspection of the pressure and HRR data for this condition (shown inFig 9(b)) supports this conclusion, clearly indicating the advanced onset of rapid pressure rise for the PRF-based blend relative to the other two However, this level of analysis offers no additional insight into how this advanced ignition came about.Fig 10offers a closer inspection

of the HRR data for Condition 1, focused specifically on the pre-ignition region Here again though, even under close scrutiny, HRR is comparable for all fuel blends regardless of CN

The lack of distinction among the fuel blends at a given condition is interesting given the expected influence of the physical properties on the spray breakup process and the large dif-ferences in fuel volatility Therefore, a detailed analysis of the physical processes relevant to ignition was carried out for the specific conditions used in this study to shed light on this interest-ing result

800

600

400

200

0

55 50

45 40

35

Cetane Number [-]

Condition1

Condition4

Conditions 2&3

Conditions 5

Condition 1 Condition 2 Condition 3 Condition 4 Condition 5

FAR-HMN

PRF

SRF

900

800

700

600

500

55 50

45 40

35

Cetane Number [-]

FAR-HMN PRF SRF Cert Diesel Condition 1

Fig 7 (a) Ignition delay for all fuels and conditions tested Error bars shown on SRF and PRF data are 95% confidence intervals based on 3 and 2 repeated measurements respectively (b) Detailed view of ID for all fuels at Condition 1, individual data points are shown for each replicated run, and the dashed lines connect the averages of the replicated data.

3.2.1 Physical considerations

Based on the computational models of Ra et al.[12], Kim et al

density, heat capacity, enthalpy of vaporization, volatility,

viscos-ity, and surface tension can be expected to impact the spray and

combustion in the following ways:

 Changes in volatility will lead to changes in the liquid

penetra-tion length

 Changes in viscosity and surface tension will lead to changes in

spray breakup

 Changes in liquid density will lead to changes in air

entrain-ment and mixture preparation

 Changes in heat capacity and heat of vaporization will lead

to changes in the axial temperature distribution within the

jet

Based on the observations made in the current study (that fuels having substantial differences in these properties but matched chemical ignition delay result in the same ignition delay and com-bustion performance) these effects must either be small in magni-tude, or have offsetting effects Each of these potential impacts are discussed in turn

Liquid Length: Using Siebers[11]liquid length scaling law, and available thermodynamic data, the liquid length of the pure spe-cies blends was estimated for each condition Liquid lengths are summarized inTable 4and are depicted relative to the piston bowl

FAR-HMN-based fuels Complete thermodynamic data for farnesane were unavailable, so the properties of pure HMN were used instead for these calculations

As expected, the lower volatility FAR-HMN-based fuel blends are consistently calculated to have a significantly longer liquid

500

450

400

350

300

250

200

55 50

45 40

35

Cetane Number [-]

FAR-HMN PRF SRF Cert Diesel Condition 2

500

450

400

350

300

250

200

55 50

45 40

35

Cetane Number [-]

FAR-HMN PRF SRF Cert Diesel Condition 3

600

550

500

450

400

350

300

55 50

45 40

35

Cetane Number [-]

FAR-HMN PRF SRF Cert Diesel Condition 4

400

350

300

250

200

150

100

55 50

45 40

35

Cetane Number [-]

FAR-HMN PRF SRF Cert Diesel Condition 5

Fig 8 Ignition delay for all fuels and conditions tested Individual data points are shown for each replicated run, and the dashed lines connect the average of the replicated data.

penetration length compared to the PRF-based fuel blends The

difference is greatest at Condition 1 (55%) and least at Condition 5

(48%) but appears to be relatively insensitive to the operating

condition The overall magnitude of the liquid penetration

length correlates strongly with ambient density, with the longest

penetrations predicted at Condition 1 and the shortest at

Condition 5

While this without question represents a substantial difference

between the PRF and FAR-HMN-based fuels with regards to spray

formation, Kook and Pickett, and Dumitrescu et al.[16,52] have

shown that liquid length itself does not impact ignition or

tion directly Liquid penetration only becomes relevant to

combus-tion in the event of wall impingement or piston bowl wetting Even

at Condition 1 the maximum liquid penetration of FAR-HMN-based

fuel is calculated to be well short of impingement These

differ-ences then, while significant, are not expected to have influenced

ignition or combustion for the conditions used in this study

Spray Breakup: Spray breakup regimes are generally delineated

by 2 dimensionless parameters which characterize the relative

influence of fluid-dynamic forces within the jet: the Weber number,

defined in reference to the ambient fluid into which the fuel is

injected (We) and the Ohnesorge number (Oh) For the purposes

of a fuel injection event, the We and Oh can be defined as:

We¼qaU2d

Oh¼

ffiffiffiffiffiffiffi

We

p

whereqais the ambient air density, U is the injection velocity of the

fuel, d is the mean droplet diameter (which was estimated to be on

the order of 10lm),ris the surface tension of the fuel, and Re is the

Reynolds number

Injection velocity was estimated using the Bernoulli equation

assuming a pressure drop of 100 MPa through the injector For

all conditions the We was of order 103and the Oh was of order

101which indicates a spray well within the atomization regime

[7] Simultaneously high We and low Oh indicate that inertial

forces dominate both surface tension and viscous forces in the spray Therefore, the effect of changes inrormwill not signifi-cantly alter the character of the fuel sprays used in this study, and will have limited influence of combustion as a result Experi-mental evidence of this is provided by Hiroyasu and Arai [53], who found that jet spreading angle is insensitive to fuel viscosity

in the atomization regime It is reasonable then that even the large differences in viscosity between PRF, FAR-HMN, and SRF-based fuels are not observed to have any significant impact on ignition

or combustion

Entrainment: For an atomizing spray the jet spreading angle has been shown to be dependent only on the liquid to ambient density ratio[7,9] Naber and Siebers[8]have shown that the spreading angle, together with principles of conservation of mass and momentum, is sufficient to calculate the rate at which air is entrained in a steady, non-vaporizing jet as it propagates Momen-tum flux was equivalent for all the fuel jets in this study, as it is dependent on injection pressure only, so any variation in jet spread-ing angle among the test blends directly correlates to variation in air mass entrainment This correlation is complicated somewhat

by the phase change of the fuel which can affect the jet boundaries independently of ambient gas entrainment, however, it still offers a reasonable metric for mixture preparation

The variation in fuel density between the blends utilized in this study is not especially large (PRF-based blends are on average 10% less dense than FAR-HMN-based counterparts) but is potentially significant In addition to liquid penetration length, Siebers[9]also provides a correlation for determining jet spreading angle as a function of fuel density and ambient density, the results of which are also summarized inTable 4 As with liquid penetration length, the difference in spreading angle between PRF and FAR-HMN-based fuel blends is consistent (0.34° on average) with the PRF-based fuel blends having the larger spreading angle Unlike liquid penetration however, the magnitude of the difference is quite small (2%) To assess the impact of this difference on charge preparation, the cross sectionally-averaged equivalence ratio was calculated for each fuel as a function of axial position An example case is shown inFig 11(a) for Condition 1

150

100

50

0

20 15 10 5 0 -5 -10

Crank Angle [ º ]

120

100

80

60

40

20

0

PRF FAR-HMN SRF Diesel SOI

80

60

40

20

0

20 15 10 5 0 -5 -10

Crank Angle [ º ]

400

300

200

100

0

PRF FAR-HMN SRF SOI

Fig 9 (a) Pressure data for the CN 45 blend of each fuel at Condition 3, representative of the global combustion trends Note that pressure rise, phasing, and magnitude of peak pressure are well matched for all three fuels Note also the slight increase in pressure rise rate for PRF relative to the other 3 fuels (b) Pressure data for the CN 35 blend of each fuel at Condition 1, the longest observed ID Note the slightly advanced onset of rapid pressure rise for PRF in agreement with the calculated ignition delay.