Title Page
 Table of Contents
 List of Symbols
 Literature Survey
 Experimental investigation
 Fortran listing of the combustion...
 Measured pressure versus time listings...
 Biographical sketch

Title: A study of the combustion phenomena related to "knock" for a spark ignition engine operating on hydrogen and air
Full Citation
Permanent Link: http://ufdc.ufl.edu/UF00085805/00001
 Material Information
Title: A study of the combustion phenomena related to "knock" for a spark ignition engine operating on hydrogen and air
Physical Description: x, 176 leaves. : illus. ; 28 cm.
Language: English
Creator: Lenertz, James Edward, 1945-
Publication Date: 1974
Subject: Hydrogen as fuel   ( lcsh )
Combustion   ( lcsh )
Mechanical Engineering thesis Ph. D
Dissertations, Academic -- Mechanical Engineering -- UF
Genre: bibliography   ( marcgt )
non-fiction   ( marcgt )
Thesis: Thesis--University of Florida.
Bibliography: Bibliography: leaves 170-175.
General Note: Typescript.
General Note: Vita.
 Record Information
Bibliographic ID: UF00085805
Volume ID: VID00001
Source Institution: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
Resource Identifier: aleph - 000869357
oclc - 14267278
notis - AEG6382

Table of Contents
    Title Page
        Page i
        Page ii
    Table of Contents
        Page iii
        Page iv
    List of Symbols
        Page v
        Page vi
        Page vii
        Page viii
        Page ix
        Page x
        Page 1
        Page 2
        Page 3
    Literature Survey
        Page 4
        Page 5
        Page 6
        Page 7
        Page 8
        Page 9
        Page 10
        Page 11
        Page 12
        Page 13
        Page 14
        Page 15
        Page 16
        Page 17
    Experimental investigation
        Page 18
        Page 19
        Page 20
        Page 21
        Page 22
        Page 23
        Page 24
        Page 25
        Page 26
        Page 27
        Page 28
        Page 29
        Page 30
        Page 31
        Page 32
        Page 33
        Page 34
        Page 35
        Page 36
        Page 37
        Page 38
        Page 39
        Page 40
        Page 41
        Page 42
        Page 43
        Page 44
        Page 45
        Page 46
        Page 47
        Page 48
        Page 49
        Page 50
        Page 51
        Page 52
        Page 53
        Page 54
        Page 55
        Page 56
        Page 57
        Page 58
        Page 59
        Page 60
        Page 61
        Page 62
        Page 63
        Page 64
        Page 65
        Page 66
        Page 67
        Page 68
        Page 69
        Page 70
        Page 71
        Page 72
        Page 73
        Page 74
        Page 75
        Page 76
        Page 77
        Page 78
        Page 79
        Page 80
        Page 81
        Page 82
        Page 83
        Page 84
        Page 85
        Page 86
        Page 87
        Page 88
        Page 89
        Page 90
        Page 91
        Page 92
        Page 93
        Page 94
        Page 95
        Page 96
        Page 97
        Page 98
        Page 99
        Page 100
        Page 101
        Page 102
        Page 103
        Page 104
        Page 105
        Page 106
        Page 107
        Page 108
        Page 109
        Page 110
        Page 111
        Page 112
        Page 113
        Page 114
        Page 115
        Page 116
        Page 117
        Page 118
        Page 119
        Page 120
        Page 121
        Page 122
        Page 123
        Page 124
        Page 125
        Page 126
        Page 127
        Page 128
        Page 129
        Page 130
        Page 131
        Page 132
        Page 133
        Page 134
        Page 135
        Page 136
        Page 137
        Page 138
        Page 139
        Page 140
        Page 141
    Fortran listing of the combustion model computer program used for this study
        Page 142
        Page 143
        Page 144
        Page 145
        Page 146
        Page 147
        Page 148
        Page 149
        Page 150
        Page 151
        Page 152
        Page 153
        Page 154
        Page 155
        Page 156
        Page 157
        Page 158
        Page 159
        Page 160
        Page 161
        Page 162
        Page 163
    Measured pressure versus time listings for all experimental runs
        Page 164
        Page 165
        Page 166
        Page 167
        Page 168
        Page 169
        Page 170
        Page 171
        Page 172
        Page 173
        Page 174
        Page 175
    Biographical sketch
        Page 176
        Page 177
        Page 178
Full Text







I wish to thank all the members of my supervisory committee,

especially the co-chairman, Dr. V. P. Roan, for the assistance and

guidance they provided. The Mechanical Research Laboratory provided

excellent support for the experimental portion of this study. The

high priority on short notice which was received during this study

was invaluable. My thanks go to all the technicians, especially to

the shop foreman, Mr. R. Tomlinson, and the laboratory director,

Professor E. P. Patterson, for the interest they showed and also for

the many helpful suggestions which they offered.

Most of all, I thank my wife, Edwina, who supported this effort

financially as well as through constant encouragement. Her understand-

ing and support made this study possible.



ACKNOWLEDGMENTS . . . . ... . ii

LIST OF SYMBOLS . . . . ... . v

ABSTRACT . . . . ... . . viii


I INTRODUCTION . . . . ... 1

II LITERATURE SURVEY . . . .... .. 3

Historical Review ............... .. 4
Recent Investigations Concerning Hydrogen Engines . 6
Knock Research for Conventional Fuels . . .. 11
Nonequilibrium Effects . . . ... 13
Effect of Engine Parameters on Combustion . ... 16


Introduction . . . .... .. . 18
Experimental Objectives . . . ... 18
Experimental Apparatus . . . ... 19
Instrumentation . . . . ... ... 23

Pressure Transducers . . . ... 23
Ionization Gauges . . . ... 24--
Oscilloscope . . . .... 26
Air Flow Meter . . . ... 27
Hydrogen Flowmeter . . . ... 28
Water Flow Meter .. . . . ... 28
Engine Speed Measurement . . ... 29
Procedure . . . . ... ..... .29
Experimental Accuracy . . . ... 32

Load Cell . . . ... .. 32
Pressure Transducers . . . ... 33
Air Flow Meter . . . . ... 34
Hydrogen Flowmeter . ... ....... 35
Ionization Gauges .......... . . 35
Oscilloscope . . . .... 36
Water Injection Carburetor . . ... 37
Engine Speed . . . .... 37
Relative Humidity . . . ... 37
Ignition Timing . . . ... 38






Introduction . . . .
Combustion Simulation . . .
Reaction Calculations . . .

Adiabatic Flame Temperature .
Pressure Rise Simulation . .
Residual Gases . . .
Reaction Products Equilibrium .

Calculated Flame Front Position . . . .

Combustion Chamber Geometry . . .
Volume Calculations . . . .
Apparent Quench Distance . . . .

V RESULTS . . . . . .

Performance Data .
Reaction Progress Data
Flame Speed Data .
Unburned Reactants .
Knocking Combustion .
Flame Front Model Results
Oscillograms ..

VI SUMMARY . . . .


FIGURES . . . . . . . .




REFERENCES . . . . . . .





. . 39

. . 39
. . 40
. . 42

. . 42
. . 48
. . 51
. . 52

. . . 83


Cl Constant, ratio of unburned volume to unburned moles of
reactants in the combustion chamber

C2 Constant, ratio of products volume to moles of products
for segment k

CM Constant, equal to the number of moles of reactants unburned
plus those reacting for each segment k

Cp(j) Heat capacity at constant pressure for species j

Cp Average value for mixture heat capacity at constant pressure

CV Average value for mixture heat capacity at constant volume

CV Constant, equal to the combustion chamber unburned volume
plus the volume of segment k products of combustion

Hj(T) Enthalpy for species j at temperature T

AHF(j) Heat of formation for species j referenced to 298K

AH (T ) Total enthalpy change for the combustion process with
final temperature (Tf) for. the products

Kpj Equilibrium constant based on partial pressures for reaction

NB Moles of products which underwent reaction in segment k

NUB Moles of unburned reactants in the combustion chamber

NTOT Total moles in a mixture

NP(i) Moles of combustion products for species i

NR(i) Moles of reactants for species i
Species i, i = 1,8 (Gram-moles)
(1) 02 Diatomic oxygen
(2) N2 Diatomic nitrogen
(3) H2 Diatomic hydrogen
(4) H20- Water vapor



(5) 0 Monotomic oxygen

(6) H Monotomic hydrogen

(7) OH Hydroxyl

(8) NO Nitric oxide

P. Partial pressure of species i

Pk Total pressure in combustion chamber at time k

PON Defined as total pressure divided by total moles

P TOT Total pressure for combustion products

R Universal gas constant

R Ratio of moles of reactants to moles of products for
segment k

r Mole fraction of initial reactants in segment k

T Final temperature for products at equilibrium

T. Initial temperature for reactants mixture

Tk Temperature of products for segment k

T Products temperature
TR Reactants temperature

TUB Temperature of unburned gas

tk Time from spark ignition for reaction of segment k

VB Volume occupied by the products of combustion for segment
undergoing reaction

Vk Volume for previously burned segment k

VUB Volume of unburned reactants

VTOT Total combustion chamber volume at time tk

y Specific heat ratio

Hydrogen fuel equivalence ratio


Flame Front Geometry

a x coordinate for cylindrical segments centerline

b y coordinate for cylindrical segments centerline

AB Burned area approximation for the intermediate zone between
spherical and cylindrical assumed flame front shape

k1 Minimum value for z coordinate of cylindrical segments

k2 Maximum value for z coordinate of cylindrical segments

R Flame travel distance from spark origin

RAD Cylindrical segments radii

r,6,q Spherical coordinates

r,e,z cylindrical coordinates

v Cylindrical segments volume

x,y,z Cartesian coordinates

Abstract of Dissertation Presented to the Graduate Council
of the University of Florida in Partial Fulfillment of the Requirements
for the Degree of Doctor of Philosophy



James Edward Lenertz

August, 1974

Chairman: Dr. R. B. Gaither
Co-Chairman: Dr. V. P. Roan
Major Department: Mechanical Engineering

"Knock" in hydrogen fueled spark ignition internal combustion

engines does occur at high compression ratios and water induction does

control the "knocking" with no appreciable loss in maximum power output.

Combustion "knock" is caused by the rapid rate of combustion chamber

pressure rise (> 50 atm/msec) associated with flame speeds on the order

of 300 feet per second. "Knock" was most severe with fuel rich condi-

tions, at high compression ratios. Compression ratios up to 10.9:1

were investigated and the fuel was supplied via high pressure tanks and

metered into the intake manifold.

The single cylinder experimental engine used in this study was

operated at full throttle with an engine speed of approximately 1800 RPM

for a range of fuel flow rates, water induction rates, and compression

ratios. The fuel flow rates were those corresponding to maximum power

output and compression ratios were varied from 8.2 to 10.9:1. Water

induction decreased knock by decreasing flame speed in all cases inves-

tigated, although some evidence of incomplete combustion was found for


fuel lean mixtures. Water flow rates were varied up to maximum

values approximating fuel flow rates for gasoline operation and

"knocking" was reduced to nonaudible levels for all cases when the water

flow rate approached this maximum. The experimental engine had pressure

transducers and ionization gauges mounted in the combustion chamber to

measure pressure rise and flame spread rates. A computer program was

developed to predict flame front position based on measured combustion

chamber pressure rise data. The types of combustion phenomena which

caused "knock" were found to be high flame speeds as determined from

flame front position versus time calculations and measurements. No

evidence was discovered of a true detonation wave being established in

the combustion chamber or of autoignition of the last portion of the

burning charge.

An additional combustion phenomenon was discovered during the

study. Incomplete combustion occurred for slightly fuel lean mixtures

as determined from pressure rise data at times corresponding to reaction

completion, i.e., complete inflammation of the combustion space. Wall

quenching was the probable explanation and wall quench distances at

reaction completion were predicted. An increase in predicted incomplete

combustion also occurred for high water induction rates in fuel lean

runs. This could possibly be explained partially by nonuniform water

distribution which would tend to cause incomplete combustion in the

local areas of high water concentration. However, no trend of increas-

ing amounts of incomplete combustion was noted for fuel rich runs as

water induction was increased. Therefore, it was believed that wall

quenching was responsible for the predicted incomplete combustion at

the time of reaction completion.

Excellent correlations were obtained in all cases for predicted

reaction completion times from pressure rise data and ionization gauge

output from the last portion of the charge to burn. Good agreement with

previous hydrogen fueled engine performance data was also obtained.




The spark ignition internal combustion engine fueled by hydrogen

has the potential to become one of the major portable sources of power

in the future. Hydrocarbon fuel reserves are rapidly being depleted and

new sources of power must be developed. The amount of hydrogen which

could be made available by the electrolysis of water is practically


The recent emphasis on ecology will also undoubtedly lead to

consideration of new, "cleaner" fuels. Hydrogen-oxygen combustion

produces only H20 as the product of combustion, and when hydrogen is

burned in air the only combustion product which is considered harmful

is nitric oxide. When hydrogen is used as fuel for a conventional

internal combustion engine, the levels of NO which are emitted can be

controlled to meet the 1975-1976 Environmental Protection Agency st.an-

dards and no carbon monoxide or unburned hydrocarbons are emitted [1].

The probable source of the energy needed for the electrolysis

process would be nuclear power plants [1]. The majority of the hydrogen

used would be transported via pipeline but some could be liquified and

stored cryogenically since the space program has developed the capabil-

ity for handling liquid hydrogen safely during storage and transport [2].

The problems involved in using cryogenic fuel for motor vehicles consist

of both safety and economic considerations [3]. The low density of

gaseous storage can be improved by absorbing hydrogen in metal hydrides

where gaseous hydrogen is then released when the metal hydride is

heated and the storage volume is approximately the same as for liquid

hydrogen [4]. Present technology appears capable of obtaining solutions

to the above problems in the future. The problems of production, trans-

portation, and storage of hydrogen must be solved further before hydrogen

can be used extensively as an automotive fuel.

Another major area for study is concerned with the combustion

characteristics of hydrogen when used in an internal combustion engine.

The problems of preignition and "knock" at high compression ratios

require more investigation in order to achieve the full potential for

hydrogen fueled internal combustion engines [4]. This study is con-

cerned with determining the way in which "knock" develops at high com-

pression ratios and what effect water injection has on "knock" and pre-

ignition. The following section presents some of the previous work

done in the area of hydrogen engines, "knock" research, and hydrogen-

air reactions.

The problems associated with building an experimental apparatus

to model an internal combustion engine cycle are varied and numerous.

The nonuniform geometry of the typical combustion space is normally

difficult to describe mathematically. Piston movement yields a contin-

uously changing pressure and temperature environment as well as an

unknown level of turbulence. Therefore, using an actual engine is the

best way to satisfy the problems of simulating an I.C. engine cycle.


The use of an actual engine makes the instrumentation problems more

difficult since vibration, heat, and electrical interference may cause

problems as well as the mechanical problems normally associated with

machinery. The additional problem of water leaks was of concern in

this study since most of the instrumentation was mounted through a

water jacket.



Historical Review

The first interest in using hydrogen as a fuel for internal

combustion engines developed in the 1920s due to the development in

Germany and England of the rigid airship as a passenger carrier.

The huge airships, filled with hydrogen for buoyancy, used large quan-

tities of liquid fuel in their internal combustion engines during their

three- to five-day-long nonstop cruises. Therefore, large quantities

of hydrogen were vented to maintain the optimum altitude as the fuel

load lightened. In 1927, the Zeppelin Company began to utilize the

previously vented hydrogen by adding 5 to 30 percent hydrogen to the

regular fuel used [5]. Operating at a compression ratio of 7:1,

these spark ignition engines ran without knocking and the hydrogen

decreased liquid fuel consumption by as much as 14 percent, thereby

increasing the payload an equivalent amount. In 1935, the Zeppelin

Company and the Royal Airship Works experimented with the use of

hydrogen booster fuels in diesel engines [5]. There was no preignition

noted for compression ratios as high as 16:1 for the hydrogen enriched


In England in the 1920s, Ricardo and Burstall performed similar

experiments on single cylinder spark ignition engines utilizing hydrogen

as the sole fuel. They both encountered combustion knock but attributed

it to different phenomena. Ricardo felt that preignition was the

problem and Burstall felt that the combustion knock was due to detona-

tion [6].

In Germany, Rudolf Erren developed many schemes for operating

internal combustion engines on hydrogen in mixtures with air or oxygen

in the late 1920s and early 1930s. The novel feature was the admission

of pressurized hydrogen during the compression stroke to eliminate back-

firing in spark ignition engines. One of the main motivations for

Erren's work was the need for a fuel that would produce no exhaust

bubbles during submerged operation in a submarine. Hydrogen-oxygen fuel

was the answer. The problem of knock was circumvented by using a

diesel engine and injecting high pressure hydrogen in such a way as to

result in essentially constant pressure combustion [5]. The Erren

method of injecting hydrogen during the compression stroke for spark

ignition engines was investigated by Manfred Oemichen on behalf of the

German Ministry of Transport in the early 1940s. A specially built Otto

cycle engine was operated at various speeds for compression ratios

varying from 7.5 to 12.0:1. Knocking combustion always occurred as the

mixture strength approached the stoichiometric value or richer [6].

World War II ended these tests and allied bombing in Berlin destroyed

the complete records of these tests in Germany. After World War II,

testing began in Canada and the United States concerning the develop-

ment of Otto cycle engines powered by hydrogen fuel.

Recent Investigations Concerning
Hydrogen Engines

Since 1950, a considerable amount of work has been done in the

field of hydrogen fueled spark ignition engines. Anzilotti investigated

the response of hydrogen-air mixtures in a single cylinder engine to the

addition of tetraethyllead [7]. He concluded that hydrogen knocked

readily in the engine and that the addition of tetraethyllead increased

the knock limited compression ratio. Since tetraethyllead also increased

the ignition temperature for hydrogen-oxygen mixtures at atmospheric

pressure, Anzilotti felt the knock must have originated from an auto-

ignition of the end gas. The effect of tetraethyllead on flame speed

was not investigated.

R. O. King did a long-term study on hydrogen fueled engines at

the University of Toronto in the 1950s [6,8,9]. He discovered that

autoignition would occur when prevaporized oil was added to the hydrogen-

air mixture at a compression ratio of 10:1 [9]. This allowed the engine

to develop power when the ignition spark was absent and eventually led

to engine destruction due to the extremely heavy knocking which was

present. He discovered that a clean combustion chamber, cold range

spark plug, low water jacket coolant temperature, and a sodium cooled

exhaust valve allowed operation at compression ratios as high as 12:1

with little preignition or backfiring for any fuel mixture strength.

The presence of combustion knock was noted, however, and King attributed

it to the high reaction rates which occurred without any evidence of an

end gas explosion as determined by pressure indicator diagrams [8,6].

No flame front position measurements were attempted. Very lean mixtures

were tested for compression ratios as high as 20:1 and performance data

were collected for various compression ratios, engine speeds, and fuel

mixtures [8,6]. Average flame velocities were predicted by determining

the burning time to maximum pressure from pressure indicator diagrams;

the estimated values were 160 to 170 feet per second for maximum power

output at 1500 RPM with compression ratios of 12 and 14:1 [6].

A recent study conducted by Shell Oil Company utilized ioniza-

tion gaps to determine the type of reaction which caused knock in

gasoline fueled engines [10]. Their study consisted of using an ion-

ization gauge to trigger oscilloscope sweep so that the reaction in the

last portion of the charge could be studied by using an expanded time

scale. For gasoline fuel, they concluded that knock was caused by

essentially instantaneous inflammation of the end gas by multiple flame

fronts. However, they operated the engine briefly by using hydrogen

as a fuel with a compression ratio of 10 to 1, and they suggested that -

for hydrogen fuels knock might be caused by the establishment of true

detonation waves in the combustion chamber [10]. This was based on the

observance of strong pressure waves which preceded the flame front.

They concluded that the knock process for hydrogen differed radically

from that for hydrocarbon fuels.

The search for an "ideal" way in which to utilize hydrogen as

a fuel for internal combustion engines is presently under way. Murray

and Schoeppel at Oklahome State University have developed a hydrogen

fueled single cylinder spark ignition engine with a compression ratio

of 6.5:1 [11,1]. They avoided the problems of preignition and combus-

tion knock by using a hydrogen injection system to inject the hodrogen

slowly during the combustion process. They reported knock-free oper-

ation and low emissions but excessive fuel consumption due to poor

cylinder-air utilization as well as frictional losses in the fuel injec-

tion mechanism [12].

Swain and Adt at the University of Miami have developed a

similar engine modification but the hydrogen is inducted at the intake

valve rather than injected late in the compression cycle [13]. A 1971

four-cylinder Toyota station wagon was converted to operate on hydrogen

fuel. The throttling was accomplished by varying only the hydrogen flow

rate. Hydrogen gas at low pressure was inducted each time the intake

valve opened since the hydrogen supply tube was in the valve seat.

Uneven cylinder-to-cylinder hydrogen distribution was discovered so

that little emission or performance data were reported. Preliminary

results indicated that lower nitric oxide emissions and higher thermal -

efficiencies were achieved than would be expected for a similar gasoline

fueled car.

Harold Sorensen of International Materials Corporation has

developed a method for producing hydrogen from a hydrocarbon fuel via

an onboard reformer [14]. Unleaded gasoline is used as fuel and a

reformed gas mixture consisting of carbon dioxide and hydrogen was used

as the fuel for a 1971 Ford V-8 engine. Air is mixed with the reformed

gas in the intake manifold and the car is throttled in the regular way.

Preliminary work indicated satisfactory engine power and exhaust

emissions which were within the 1976 Federal Emissions Standards.

The compression ratio was increased from 9:5 to 10.5:1 but no data were

given for efficiency, power output, or mixture stoichiometry.

Underwood and Dieges have developed several engines that

operated on hydrogen and oxygen [15]. The first was a six-cylinder,

1950 Studebaker. This engine was first operated on pure oxygen in order

to standardize the flow with the intention of adding hydrogen gradually.

The hydrocarbon residuals ignited in the oxygen rich atmosphere and

caused superficial damage before any hydrogen was admitted. Therefore,

hydrogen rich operation was attempted first with air and later with

oxygen. The very rich hydrogen-oxygen mixture produced a quiet and

smooth operation until the oxygen supply was accidentally increased

and the engine was tested to destruction [15]. Next a recirculation

system was used on a 1930 Model A Ford to provide very hydrogen rich

operation while the incoming mixture was stochiometric. The exhaust

water vapor was condensed so that only hydrogen was recirculated to

dilute the stoichiometric intake mixture which originated from two high

pressure cylinders. A more modern vehicle, a 1960 Ford pickup, was

then converted to operate on cryogenic supplies of hydrogen and oxygen.

Specific performance data werenot developed due to a general lack of

instrumentation but the power output and efficiency were very low due to

experimental limitations, especially the limited oxygen supply rate.

A computer study of a hydrogen-oxygen engine operating fuel

rich with exhaust gas recycling was performed by Karim and Taylor

[16,17]. Constant volume combustion to equilibrium conditions was

assumed to occur at top dead center. The compression was nonadiabatic

and the expansion process was considered a nonadiabatic but frozen con-

centration process. Also an independent prediction of possible auto-

ignition during compression was made by considering the chemical reac-

tion kinetics during compression. Results indicated that autoignition

would occur at compression ratios greater than 11:1 for hydrogen-oxygen

system. Nitrogen or steam diluents suppressed autoignition more effec-

tively than argon or helium due to the high specific heats. Auto-

ignition was predicted for very low compression ratios if the hydrogen-

oxygen mixture was diluted with either argon or helium. The maximum

indicated thermal efficiency was predicted to be approximately 35 per-

cent for fuel rich hydrogen-oxygen combustion similar to that employed

by Underwood and Dieges [15].

Starkman et al. compared various fuels to determine predicted

peak temperatures, pressures and exhaust emissions [18]. Hydrogen was

considered as one of the fuels for the computer study. An assumed 9:1

compression ratio was used and the cycle process approximation con-

sisted of an isentropic compression of a nonreacting ideal gas mix-

ture with variable specific heats followed by an adiabatic constant

volume combustion which proceeded to equilibrium conditions. The

amounts of NO and CO which were predicted at peak temperature were

assumed to be indicative of the comparitive values which would be

present after expansion was completed. Isooctane was used as the

baseline fuel and all other fuels were compared to it. Hydrogen was

predicted to yield a higher temperature but lower pressure for all

air-fuel ratios investigated. For stoichiometric proportions, hydrogen

was predicted to yield a 7 percent higher temperature and a 13 percent

lower pressure than isooctane. The predicted nitric oxide concentration

was a few percent less than the isooctane level at stoichiometric pro-

portions but had a maximum level slightly higher than isooctane for

20 percent lean fuel-air mixtures [18]. However, since hydrogen can be

consumed in very lean proportions for part load operation, the emis-

sions would normally be much less than for a gasoline fueled engine.

Knock Research for Conventional Fuels

Two predominant theories concerning knock have prevailed in the

past. These are the autoignition theory and the detonation theory

[19-22]. The autoignition theory postulated that knocking was produced

by local pressure imbalances produced by virtually simultaneous auto-

ignition of the last part of the charge to react. The detonation

wave theory predicted that the pressure imbalance was caused by a

supersonic detonation wave which was established in the end gas.

Studies by Curry in 1963 proposed yet another combustion mode which

could cause knock [23,24]. He conducted a three-dimensional study of

flame propagation with the aid of multiple ionization gaps mounted on

the piston top and cylinder head of a single cylinder engine. Results

indicated that knock could occur from an acceleration of the primary

flame front to speeds of from 300 to 1200 feet per second. These

measured flame speeds were faster than normal but much less than would

be observed for a true detonation wave. He also found that the flame

speeds were higher throughout the combustion period for knocking

combustion and that the addition of tetraethyl lead decreased the flame

speed for knocking cycles. Ellison et al. also investigated the effect

of tetraethyllead on flame speed for 18 different hydrocarbon fuels [25].

They also measured a flame front acceleration prior to knock but attrib-

uted it to the effects of preflame reactions in the end gas. They did

note that for a given operating condition, the combustion cycles which

had the fastest combustion knocked, whereas the slow burning cycles

produced no knock. The cycle-to-cycle variation was responsible for the

measured range of flame speeds.

Preflame reactions are very important for hydrocarbon fuels

where two-stage ignition is prevalent [26,27]. Up to 10 percent of the

chemical energy of the fuel may be released during the preflame reac-

tion [27]. Autoignition information can be obtained through the use

of rapid compression devices. Delay times and ignition temperatures

can therefore be determined. Hydrogen-air mixtures and many hydrocarbon

fuels were tested by Jost [28]. Hydrogen-air mixtures were determined -

to have a high ignition temperature and an extremely high pressure rise

rate after autoignition as compared to isooctane fuel [28]. An excel-

lent summary of early, prior to 1962, literature on spark ignition

engine combustion phenomena was presented by Starkman [29].

Some recent work in Italy by Cornetti et al. established that

a load washer (load cell) could be mounted external to the combustion

chamber to detect knock [30]. Results of endurance tests for various

detonation intensities were also presented to allow predictions for the

length of time that a given knock level could be tolerated before

piston failure occurred.

Karim and Watson developed a method to predict autoignition in

a compression apparatus [31]. Reaction kinetics were utilized;

therefore, the first fuel mixture considered was hydrogen and oxygen

which have relatively well-known kinetics. A thirteen reaction step

kinetic model was used to describe the reaction. Autoignition was

"predicted" only after six of the rate constants were "adjusted" to

match experimental ignition delay times [31]. The authors are presently

working on modifying the model to allow predictions for hydrogen-air

and hydrocarbon-air fuel mixtures.

The rate of pressure rise in the combustion chamber must be

limited to a reasonably low value for both spark ignition and diesel

engines to prevent structural damage. The effects of various param-

eters on pressure rise for diesel and Otto cycle engines are presented

by Brewster and Kerley of Ethyl Corporation [32]. They also have a

good discussion of preignition effects for Otto engines. Diesel knock-

ing characteristics are further described by U. W. P. Anders [33]. He

stated that the value for dP/dt "has proven to be a suitable criterion

for assessing the diesel knocking characteristic and the consequent

production of noise."

Nonequilibrium Effects

It is well known that chemical equilibrium is not instantan-

eously attained in the spark ignition combustion process for all the

species of interest. Nitric oxide in particular has been shown to

ignore the dictates of equilibrium during the expansion process [34,35].

Starkman and Newhall at the University of California, Berkeley, compared

the effects of frozen and equilibrium expansion and determined that

a frozen composition could result in an up to 10 percent loss in expan-

sion work [34]. They also found that the measured exhaust values for

NO were close to those predicted by equilibrium conditions at peak

cycle temperature. Spadaccini and Chinitz assumed equilibrium condi-

tions existed at combustion completion and by using those composition

values for initial conditions, they predicted the exhaust levels of NO

by considering chemical kinetics during the expansion process [35].

Thirty-four reactions were used for the C-H-O-N system which included

fourteen species, the rate constants which were used are listed [36].

The predicted exhaust concentrations of NO agreed closely with the

measured values. Muzio et al. [37] considered the effects of temper-

ature variation within the combustion chamber due to compression of

the first burned gases to higher temperatures than the last burned.

They assumed the C-H-O system was in equilibrium behind the flame but

assumed that the NO species was not present immediately following com-

bustion. They then used chemical kinetics for the NO formation behind

the flame front. The results indicated higher NO levels in the first

burned gases but the predicted levels were lower than the measured

values. The effect of engine parameters on NO was investigated by Huls

and Nickol as well as Ohigashi et al. [38,36].

In spite of these limitations for the assumed equilibrium con-

centrations, the predicted temperature is quite accurate [39]. Also

the reaction kinetics experiments have not yet provided enough data

to simulate the entire flame front reaction kinetics. Therefore, if

accurate concentrations of exhaust gases are needed, it is better to

calculate temperatures and pressures from equilibrium considerations

and then use these values for initial conditions for the reaction

kinetics calculations [39].

The above results for hydrocarbon fuels are similar to those

obtained for hydrogen-air reactions. Shahed and Newhall [40] pre-

sented results that indicated NO was produced chiefly by post flame

reactions and that the time required for low temperature reactions, at

three atmospheres'pressure, to proceed to equilibrium was appreciable.

They also concluded that the effect of diluents such as water or nitro-

gen reduced the rate of formation of NO as well as lowering the equi-

librium level. Homer and Sutton recently tried to improve on the NO

prediction model for hydrogen-air flames at one atmosphere pressure by

considering the effect of radical overshoot [41]. Atomic oxygen and

hydrogen and the hydroxyl radical all are present in proportions

exceeding equilibrium values during the reaction within a flame front.

Radical recombination to equilibrium values occurs very quickly but

has an effect on the initial rate of nitric oxide formation [41].

The recombination reaction for dilute hydrogen-oxygen-argon reactions

was studied with the aid of shock tube experiments by Blair and

Getzinger [42]. The recombinations of the radicals to equilibrium con-

ditions were found to have occurred within 100microseconds for very

dilute mixtures at an initial pressure of 10 centimeters of mercury [42].

This study was conducted to provide more information about the rate

constants for termolecular recombination reactions. All of the above

results for hydrogen flames assumed equilibrium was achieved in the

flame zone for the bimolecular chain reactions involved in the

hydrogen-oxygen reaction [40-42]. The width of the flame zone is very

thin, on the order of a millimeter, for hydrogen-oxygen reactions [43].

Dixon-Lewis and Williams measured temperature profiles for hydrogen-

oxygen-nitrogen flames which were very dilute and hydrogen rich [44].

For approximately twice the stoichiometric amount of hydrogen, 75 per-

cent nitrogen, 1 atmosphere pressure, and an initial temperature of

3350 Kelvin, the mixture had a flame thickness such that the temper-

ature variation occurred within approximately 2 millimeters. Higher

pressures and flame speeds would tend to decrease the flame thickness

even further [45].

Effect of Engine Parameters on Combustion

The effects of many parameters on combustion in a spark igni-

tion engine have been investigated in the past [19,20,21]. The effect

and cause of cycle-to-cycle variations have been studied recently by

several investigators using hot wire anemometers to determine combus-

tion cylinder velocity fluctuations [46,47,48]. Results indicated

that these mixture velocity fluctuations at the spark plug could cause

the cycle-to-cycle variations for a given operating condition. A very

approximate average eddy size of 0.1 inch was predicted [46]. The

effect of induced swirl in the combustion chamber was determined by

Johnson [49]. Robert Siewert of General Motors Corporation examined

the effects of variations in valve timing [50]. R. C. Lee of Phillips

Petroleum Company investigated the interactions of many engine

parameters on exhaust emission and power output [51]. Sukai et al.

of Nissan Motor Company examined the effects of combustion chamber

shape and the spark plug placement [52].

Obviously, much additional research has been done on internal

combustion engine research in the past fifty years. However, the more

important aspects are adequately presented in the above-mentioned

papers. Since relatively little work has been done previously in the

area of hydrogen fueled engines, much of the hydrogen combustion

research has been conducted at lower temperatures and pressures than

those present in an I.C. engine. The results of such research must be

considered not directly applicable even though certain trends can be

expected at the higher temperature and pressure levels present in

this study.




An experimental study of the combustion phenomena related to

"knock" was undertaken to gain more knowledge on how "knock" develops

in hydrogen fueled engines and how variations in engine parameters

affect "knock" or detonation. The combustion chamber data were dis-

played on an oscilloscope where the sweep was externally triggered by

the ignition spark. Therefore, the data which were collected were

displayed from time of spark until after the combustion reaction was

completed. This enabled the use of an expanded time scale during the

combustion period of the cycle so that precise time measurements

could be obtained.

Experimental Objectives

The effects of variations in compression ratio, air fuel ratio,

and water injection rates on pressure development and flame travel

times were investigated. Knowledge of the pressure rise and flame

propagation histories during combustion of hydrogen-air mixtures were

needed to determine how "knock" originated. The range of variation in

engine parameters allowed operation at "knocking" conditions and also

at conditions where "knock" did not occur. This permitted cycle

comparisons to determine how the pressure waves characteristic of

"knock" developed. The pressure waves which caused the "knock" were

strong enough to cause audible vibrations at some of the experimental


The occurrence of "knock" was detected from measured combus-

tion chamber pressure oscillations. The pressure versus time trace

obtained from the pressure transducer was also a measure of the rate

of heat release during the pressure rise associated with combustion.

Flame front position was determined by mounting ionization gauges in

the combustion chamber and monitoring their output. As the flame

reached each ionization gauge, it produced a small current flow which

was displayed on an oscilloscope to indicate when the flame front

reached the gauge position. The type of combustion phenomena which

apparently caused "knock" was investigated by analyzing the data on

pressure and flame development.

Experimental Apparatus

The four stroke cycle, single cylinder, spark ignition engine

used in this study was a Fairbanks-Morse Diesel Engine converted to

spark ignition operation. Table 1 gives some of the more important

engine details. All runs were made at full throttle and the power

generated was absorbed by an electric dynamometer (AC generator)

which could also be used as a starting motor to turn the engine over

when it was not developing power. The dynamometer was a specially

wound electric motor which would provide the power to motor the engine

at approximately 1775 RPM. When the engine developed power and the RPM

increased to values greater than 1800 RPM, the electric motor began

to dissipate power by generating alternating current and transmitting

it to the existing power lines. The dynamometer-starting motor

description is given in Table 2. When developing power, the hydrogen

fueled engine operated at approximately 1835 RPM for the tests conducted.

Engine torque was measured with an electronic load cell made

by the Electronics and Instrumentation Division of Baldwin-Lima-

Hamilton Corporation. Load cell output was displayed on a Leeds-Northrup

voltmeter with a 10-millivolt maximum range which corresponded to a

50-pound load on the load cell. The load cell was attached to the

dynamometer via a 12.6-inch loading arm. The engine-dynamometer instal-

lation is shown in Figure 1 and the apparatus details are listed in

Table 2.

The hydrogen fuel supply consisted of two high pressure hydrogen

cylinders containing approximately 200 standard cubic feet of hydrogen

per cylinder. This allowed approximately 2 hours of engine operation

per set of tanks. The hydrogen gas was inducted into the intake mani-

fold approximately 8 inches from the intake valve. This allowed suffi-

cient time for hydrogen-air mixing while keeping the volume of the

explosive mixture as small as possible for the sake of safety in case

of preignition, causing an intake manifold explosion.

The ignition timing was varied while each test was in progress

to produce maximum power output for each test run condition. The

ignition system was a standard six-volt automotive system consisting of

a six-volt battery, induction coil, ignition contact points, condenser,

and a low heat range spark plug. Ignition timing was varied by rotat-

ing the ignition points set around the drive shaft which had a slot

machined in it. Ignition timing could be read directly in degrees

from the engraved mounting for the ignition point set shown in Figure 2.

The compression ratio of the engine was varied by inserting

shims of different thickness between the cylinder head and block. Two

head gaskets were used, one gasket on each side of the shim, to prevent

compression leaks. Three shims were used with thicknesses of 0.213,

0.130, and 0.055 inches which produced compression ratios of 8.16:1,

9:30;1, and 10.93:1, respectively. The shims and a head gasket.are

shown in Figure 3.

The cylinder head on the Fairbanks-Morse engine was modified

to allow mounting the required instrumentation in the combustion chamber.

The pressure transducers and the ionization gauges had the same size

threads (14-millimeter) and could be interchanged at the desired loca-

tions. As shown in Figure 4, two holes were drilled through the water

jacket from the top of the head, one hole on each side of the valves.

These 1-inch.diameter holes were tapped to accept a watertight sleeve

after being centered over the locations picked for mounting instrumen-

tation in the combustion chamber. The 14-millimeter instrumentation

mounting holes were drilled and tapped at locations shown in Figure 5.

The lower valve shown in Figure 5 is the exhaust valve and it is

recessed approximately nine-tenths of an inch from the cylinder head

surface. A spark plug and an ionization gauge were mounted on opposite

sides of the recess with the spark plug on the left and the ionization

gauge on the right side in the photograph. These were installed in

mountings drilled and tapped in existing passageways in the cylinder

head which had been used for diesel operation. The spark plug was

mounted in the energy cell location and the ionization gauge was

mounted in the fuel injecter passageway. The cylinder head combustion

chamber was cleaned and sanded and the valves were ground before begin-

ning the test runs.

Crankcase ventilation was provided to prevent possible explo-

sive concentration of hydrogen in the crankcase due to piston blow by.

Plentiful ventilation was achieved by connecting a compressed air line

to one side of the crankcase. The crankcase vapors were allowed to

exit from the cylinder head after leaving the crankcase via the valve

push rod channels.

Engine cooling was accomplished by circulating water from a

water storage tank through the engine block and water jacket. This

circulation was caused by free convection effects only since no water

pump was used. Cool water entered near the bottom of the engine and

heated water exited at the top of the engine and was returned to the

water storage tank. The storage tank had an overflow so that storage

tank temperature was kept low by adding cool water from the laboratory

water line while allowing the warmer water to drain through the overflow.

An overhead exhaust fan provided sufficient ventilation to

prevent any large exhaust or crankcase vapor concentration around the

engine. It also kept the air temperature around the engine at approx-

imately ambient levels. An overall view of the experimental apparatus

is shown in Figure 6.


Pressure Transducers

The pressure transducers used in this investigation were of the

quartz crystal type. The basis for their effective operation is the

piezoelectric effect. Pressure applied to the transducer diaphragm is

transferred as a force acting on the transducer crystals; this generates

an electrical charge output proportional to the pressure input. The

sensitivity of these instruments is expressed as picocoulombs per PSI,

or units charge per unit pressure.

The two pressure transducers used were Kistler 601H miniature

quartz piezoelectric crystals as listed in Table 2. The transducers

were mounted in Kistler model 628C water cooled adaptors designed for

flush mounting in high temperature environments. They are capable of

measuring pressures up to 15,000 pounds per square inch and have a

rise time of 3 microseconds. The operating temperature range for the

transducers is from -450 to +5000F. The intermittent gas temperature

on the diaphragm can exceed 3000F. The crystal has a resonant fre-

quency of 130,000 hertz, linearity within 1 percent and a temperature

sensitivity of 0.01 percent per degree Fahrenheit. The pressure trans-

ducer without the water cooled adaptor measures six-tenths of an inch

in length and has a one-quarter-inch diameter.

The transducer-adaptor assembly shown in Figure 7 was mounted

in the water jacket of the cylinder head. The steel sleeve also shown

in the photograph was installed around the transducer to provide a

watertight seal. The sleeve was tightened down on the neoprene gasket

to ensure no leaks at the bottom of the sleeve and the threads at the

top were sealed with Permatex gasket cement. The pressure transducer

was tightened in place by using the long slotted socket also shown in

Figure 7 and the copper spacer on the threads ensured flush mounting

plus a leak-free seat in the combustion chamber.

The pressure transducer signals were amplified by using

model 504 Kistler Charge Amplifiers as listed in Table 2. The pressure

transducer sensitivity in picocoulombs per PSI was specified and the

selected output range was 100PSI per volt for both amplifiers. This

voltage trace was displayed on an oscilloscope as is described later.

Ionization Gauges

Ionization gauge operation is based on the principle that the

presence of ionized gases in a gas mixture increases the electrical

conductivity of that mixture. When a potential difference is applied

across two electrodes immersed in a gas, no discharge is produced if

the electrode gap is wide enough or the potential difference is small

enough. If an ionized gas then passes between the electrodes, the

increased conductivity will allow a current flow across the gap which

may be used to produce a suitable signal. Therefore, since ionization

occurs in flame fronts, the arrival of a flame front at a given point

can be determined by installing an ionization gauge at the given point

and monitoring its output.

The ionization gauges used in this investigation were cold

range AC-43 spark plugs. These plugs were gapped to 0.030 inch and

had a potential difference of 200 volts applied across the electrodes.

Figure 8 shows one of the ionization gauges along with the tool used to

install it and the watertight sleeve which was used to protect the

gauge when it was mounted in the water jacket. The modifications which

were made to the spark plug ionization gauge were required due to space

limitations in the water jacket portion of the cylinder head. The

external steel casing on each ionization gauge was turned down on a

lathe and slots were machined in the remaining casing to provide a method

for installing the gauges in a very limited space. The installation

tool fitted over the gauge and had prongs which were inserted in the cor-

responding slots in the casing to provide a means of tightening the ion-

ization gauges.

These modifications provided just enough clearance for the

watertight sleeve which was installed over the ionization gauge and

tightened onto a neoprene gasket at the base of the sleeve to provide

a watertight seal.

The ionization gauge circuitry is shown in Figure 9. The

circuitry is similar to that used successfully in shock tube studies

to determine the velocity of shock waves [53,54]. Since hydrogen flame

fronts have relatively low levels of ionization compared to hydrocarbon

fuels, some experimentation was needed to get satisfactory ionization

gauge output [45]. A 200-volt DC power supply was used to provide the

desired potential difference. Two ionization gauges were used in all

test runs, one was installed in the water jacket and one was installed

in the diesel fuel injector passageway. The output from each gauge was

displayed as a separate oscilloscope trace. Flexible plastic tubing

was clamped to the two watertight sleeves to provide an oil-free

environment for the instrumentation electrical connections. The two

plastic tubes were each approximately 1 foot long and both extended

outside the cylinder head valve cover to give complete protection from

oil contamination. All instrumentation leads were shielded to minimize

the effects of electrical interference from the engine ignition system

and associated apparatus.


A Tektronix Type 564 storage oscilloscope was used to display

pressure and ionization gauge traces. The vertical amplifier unit used

was a Type 3A74 four-trace amplifier and each trace had gain adjust-

ments varying from 0.02 to 10.0 volts per division. A 3B3 time base

was used to provide external triggering capability. The time base had

an adjustable sweep rate in the range between 1.0 seconds per division

and 0.5 microseconds per division. The basic oscilloscope unit also

contained an internal calibration module. This module produced a

60-cycle square wave with adjustable amplitude and could therefore be

used to calibrate the vertical amplifiers and also the time base sweep

rate. The sweep rate usually used in this investigation was 5 milli-

seconds per division. This allowed three complete cycles of the 60-

cycle square wave to be displayed on the 10-division wide oscilloscope

screen for purposes of calibration.

The oscilloscope was fitted with a camera attachment. The

camera used was a Tektronix Oscilloscope Camera C-12 and had a Polaroid

back for quick development of photographs. The camera allowed the

capability for variations in shutter speeds and aperture. Photographs

taken with 1/25-second shutter speeds showed only one cycle of the

engine operation. Shutter speeds of 1/5 second were used to determine

the cycle-to-cycle variations in the combustion chamber since multiple

cycles were displayed in these photographs. Details of oscilloscope

equipment are listed in Table 2.

Air Flow Meter

Air flow to the engine was measured with a Meriam Laminar Flow

Element as listed in Table 2. This flow element has an air flow capac-

ity of 100 cubic feet per minute for an 8-inches-of-water-pressure

difference across the sensing element. The sensing element consisted

of a fine matrix enclosed in a cylindrical case with two pressure taps

located within the matrix section to determine the pressure difference

between the pressure tap locations. A slant manometer was used to

measure the pressure difference between the two pressure taps. This

allowed flow rates to be determined from calibration curves furnished

with the element. An air filter and straightener section were fitted

upstream of the element to ensure a clean uniform air supply at the

metering element.

Intake air temperature was noted as well as absolute pressure

at the upstream pressure tap; these data were needed to correct the

volumetric flow rate to standard conditions. Intake air temperature

was measured with a mercury thermometer and absolute pressure was

measured with a mercury manometer.

Intake air relative humidity was measured by using a Bacharach

sling psychrometer. The partial pressure of water vapor in the air was

calculated and a pressure correction was made to determine the volu-

metric flow rate of dry air.

Hydrogen Flowmeter

Hydrogen was furnished via 2200 PSI Airco high purity hydrogen

cylinders. Two cylinders were connected with high pressure pigtails and

fed through a pressure regulator where the pressure was reduced to

30 PSIG. An Airco Hydrogen Dual Range Flowmeter was used to measure the

hydrogen flow rate. Flowmeter and pressure regulator details are listed

in Table 2. The flowmeter inlet pressure was maintained at 30 PSIG and

the flow rate could be adjusted with a valve located downstream of the

flowmeter for flow rates up to 300 standard cubic feet per hour. The

hydrogen was then introduced into the intake manifold close to the

cylinder head to minimize the volume of combustible hydrogen-air mixture

in the intake manifold.

Water Flow Meter

An Octa-Gane Water injection carburetor was used to mix water

with the intake air. The flow rate was controlled by adjusting a

needle valve in the carburetor venturi. The water was therefore intro-

duced into the combustion chamber as a mist which was mixed into the

intake charge.

The volumetric flow rate was determined by using an electric

stop clock to measure the time needed for a given volume of water to

be consumed. A burette was used to measure the volume of water injected.

The measurement was accomplished by closing the valve to the water tank

and allowing the water in the 25-milliliter burette to be consumed.

The flow rate measured in this way was recorded in milliliters per


Engine Speed Measurement

Engine speed was determined by counting the number of engine

revolutions which occurred in a measured amount of time. A stop clock

and an electronic counter were connected to the same switch to allow

simultaneous starting and stopping for both. A signal generator was

used to produce one electronic pulse per revolution.

The signal generator and electronic counter are described in

Table 2. The signal generator shaft was attached to the engine drive

shaft so that one pulse per engine revolution was output to the counter.

Since the electric stop clock was activated by the same switch that

activated the counter, the number of engine revolutions and the elapsed

time were determined simultaneously when the switch was turned on.


The basic experimental program which was completed consisted

of varying compression ratio, air fuel ratio, and water injection rates

with ignition timing set for maximum power output. All runs were made

at full throttle and the engine speed was approximately constant at

1835 RPM.

Preliminary runs were made which indicated there was no pressure

imbalance in the combustion chamber during nonknocking operation.

This was determined by mounting two pressure transducers on opposite

sides of the combustion chamber and comparing their output. Therefore,

only one pressure transducer was used for the basic experimental

program listed in Table 2. This allowed two ionization gauges to be

used in the combustion chamber for the experimental runs listed in

Table 2. The second pressure transducer was used to check the pressure

imbalance in the combustion chamber for knocking conditions at the

highest compression ratio used. The results are presented later.

The compression ratio was varied by changing the thickness of

shims which were installed between the cylinder head and block. Each

compression ratio was determined by calculating the combustion chamber

volume with the piston at top dead center. The cylinder head volume was

determined by measuring the volume of water needed to completely fill

the cylinder head recesses. The thickness of each shim was recorded

after being measured with a micrometer. The average diameter of the

combustion chamber was also recorded for each shim. Each shim was then --

installed between the cylinder head and block with a head gasket on each

side of it to prevent any compression leak. The cylinder head bolts

were tightened to 95 foot pounds' torque andthe total thickness of the two

head gaskets, gasket cement, and shim was determined by measuring the

distance between two sets of 3/16-inch pins. The two sets of pins were

located on opposite sides of the piston, and one pin from each set was

mounted in the cylinder head while the other pin in the set was mounted

directly below the first in the engine block. The average distance

between the two sets of pins allowed the total gasket plus shim thickness

to be calculated while the cylinder head gaskets were compressed.

This allowed the combustion chamber volume to be calculated for each

of the shims used.

The instrumentation was allowed to warm up for at least 30 min-

utes. This allowed the oscilloscope, charge amplifiers, voltmeter,

and power supply ample time to reach a steady state operating condi-

tion. Cooling water inlet temperature was recorded, the air flow meter

slant manometer was zeroed, and the compressed air flow used to venti-

late the crankcase started.

After the instrumentation warm-up period was over, the pressure

transducer charge amplifiers were zeroed. The load cell zero reading

was also checked. The oscilloscope was calibrated by using the internal

calibrator circuit mounted on the scope. The time base sweep rate and

the vertical amplifier gain were checked at the same time by examining

the trace of the 60-cycle square wave of chosen amplitude which was

produced by the calibrator. The amplifier and time base were corrected

if any calibration was needed.

The overhead exhaust fan was started to ensure no build-up of

hydrogen near the engine. The two high pressure hydrogen tanks were

moved just outside the building and the hydrogen flow meter was leveled;

the pressure regulator was adjusted to give 30 PSI pressure at the flow

meter inlet. Cooling water for the engine and the pressure transducer

adaptor was turned on and the engine was brought to approximately 1775

RPM with the use of the starting electric motor. The ignition was then

turned on and when the hydrogen flow valve was opened the engine began

to develop power which was absorbed by the electric dynamometer.

The engine was warmed up until the water jacket exit temper-

ature reached approximately 1150F. The hydrogen flow rate was set at

the desired level as was the water flow rate. The ignition timing was

then adjusted to give maximum torque. Two Polaroid oscilloscope pic-

tures were taken, one with a shutter speed of 1/25 of a second and

one at 1/5 of a second. Engine RPM, torque, and ignition timing were

recorded. Water flow rates were measured by recording the time required

to allow a measured volume of water to be inducted into the intake


Air flow was determined by recording the following: the slant

manometer pressure difference reading for the laminar flow element, the

intake air temperature, the intake air relative humidity, and the intake

air absolute pressure after entering the sensing element. Due to vis-

cous damping plugs in the pressure taps of the laminar flow element,

air flow reading did not stabilize quickly. The damping plugs were

necessary to prevent large fluctuations in the slant manometer fluid

when backfiring occurred. At several operation conditions, preignition

took place before the air flow reading stabilized. In these cases, the

air flow was approximated by extrapolating data from similar runs since

all runs were conducted at full throttle.

Experimental Accuracy

Load Cell

Torque developed by the engine was measured by using a fifty-

pound capacity electronic load cell. The load cell was calibrated by

loading it with standard weights. The millivolt meter which was used

during the experimental runs was also used during calibration. The

results yielded a slope of 5.0 pounds per millivolt output and an accur-

acy of 0.05 my or 0.25 pounds in the range of interest. This indi-

cated an accuracy of approximately 1.7 percent for the torque range


Pressure Transducers

Cylinder pressure development was measured with two piezoelec-

tric crystal pressure transducers. They were mounted in special adap-

tors with serial numbers 17819 and 17820 and were calibrated by Kistler

Instrument Company in Redmond, Washington. The calibration factors in

picocoulombs output per pound per square inch input were 1.09 and 1.10,

respectively. The linearity of the calibration factor is listed as

within 1 percent of nominal. At 750 PSI, which is approximately the

maximum pressure measured during experimental runs, the error in mea-

surement from the calibration curves furnished by Kistler is approx-

imately 0.5 percent. Static tests using the dead weight tester were

used to further check the calibration. The calibration checked to

within the oscilloscope resolution of 5 PSI for pressures up to

290 PSIA. The pressure transducers were also dynamically checked by

comparing their output during engine motoring. The two transducers

gave identical results within the accuracy of the oscilloscope resolu-

tion ( 5 PSI) at pressures up to 250 pounds per square inch.

Air Flow Meter

A Meriam Laminar Flow Element was used for air flow measurement.

A calibration curve was furnished for the flow element by the manufac-

turer. Each individual flow element is calibrated to within 0.35 per-

cent for air flowing at a pressure of 29.92 inches of mercury absolute

and a temperature of 700F. This flow element is particularly well

suited for measuring a pulsating flow because of its linear character-

istics. Viscous pulsation dampers were installed in the differential

pressure lines to insure an arithmetical average of the differential

pressure across the element. Tables for correction factors for temper-

atures and pressures other than the reference values are included with

the calibration curve.

The temperature correction factor is 1.0 percent for 3F

temperature change at 700F. The air temperature was measured with a

mercury thermometer accurate to approximately 10F; therefore, the error

due to temperature measurement is approximately 0.33 percent of the

air flow.

The pressure correction factor is approximately 1.0 percent

for 0.3 inch of mercury pressure change. The absolute pressure was

measured with a mercury barometer and was corrected for temperature and

latitude. Therefore the pressure was known to within 0.1 inch of

mercury which gave approximately a 0.33 percent error in air flow.

The pressure differential was measured with a slant manometer

containing red oil, specific gravity = 0.827 at 600F. The specific

gravity variation with temperature was approximately 0.05 percent for

10F. The pressure difference was corrected for the change in specific

gravity. The differential pressure was determined within 0.005 inch

of water which led to an error of 0.064 cubic feet per minute. For

the minimum air flow in the range of experimental runs, 6.8 cubic feet

per minute, this led to a possible error of up to 0.94 percent in the

air flow measurement.

Since all these error sources are independent, the root mean

square of the individual error sources is a reasonable estimate of the

total error. Therefore, the total error in air flow measurement is

less than 1.1 percent of the total air flow.

Hydrogen Flowmeter

An Airco Dual Range Hydrogen Flowmeter was used to measure the

hydrogen flow rate from the two high pressure cylinders which were used

as a fuel source. The accuracy was rated at 5 percent for the low

flow range and 10 percent for the high flow range. The total intake

flow of the engine was approximately constant for the experimental runs

since they were all performed at full throttle and essentially constant

RPM. Therefore, the variation in the hydrogen flow rate could be esti-

mated by examining the variation in total intake flow for the engine.

By this means, the flow variation for the hydrogen flow meter for these

tests was estimated to be Less than +5 percent of the nominal value.

Ionization Gauges

Two ionization gauges were used to determine the presence of

a flame front at the gauge locations. The ionization gauges'output was

displayed on an oscilloscope. The desired result was a detectable

signal variation with a fast response time when the flame front reached

the gauge. The time constant for the RC circuit which was used was

approximately 25 microseconds. The ionization gauge trace which was

displayed on the oscilloscope had sufficient amplitude to be detected

for all the experimental runs. The amplitude of the trace variation for

hydrogen fuel was approximately 10 percent of the value which was

observed using gasoline fuel. This is due to a lack of the primary ion

producing process which is present for hydrocarbon flames but not for

hydrogen-air flames [45,55],

CH + 0 CHO + e


The Tektronix Oscilloscope used had an internal calibration

circuit for checking sweep time and vertical amplifier calibration.

The calibration unit produced a 60-cycle square wave with variable

amplitude. The published accuracy of the internal calibration circuit

output was 3 percent for both amplitude and sweep time. The ampli-

tude of the square wave was checked by using a digital voltmeter and

was within the published error tolerance. The sweep time calibration

was checked at slow sweep rates with a stop clock and was also within

the published accuracy limits. The oscilloscope vertical amplifiers

and time base were calibrated frequently by using the internal cali-

bration circuit as a standard.

Water Injection Carburetor

The rate of water injection was varied manually by adjusting

a needle valve in the carburetor venturi. The rate was calculated by

measuring the time needed for a given volume of water to be inducted.

The timing clock had divisions of 0.1 second and the volume was measured

in a burette with divisions of 0.1 milliliter. Several measurements

were made to obtain an average value for the water induction rate.

The given values for water induction are accurate within approximately

7 percent as determined from measurement variations. A 7 percent

error in water induction leads to an error of less than 1 percent for

the total calculated moles in the combustion chamber.

Engine Speed

The engine speed was determined by counting the number of

revolutions which occurred in a given time period. An electronic counter

counted the number of sine wave cycles output by a sine wave generator

which was attached to the crankshaft. The counter was digitized and

accurate to 1 revolution while elapsed time was measured on an elec-

tric stop clock accurate to within 0.1 second. The counter and clock

were activated with the same switch. The engine speed was accurate

to within approximately 4 RPM or 0.2 percent for the speed range

around 1835 RPM.

Relative Humidity

A sling psychrometerwas used to measure relative humidity.

The psychrometer used was compared with another sling psychrometer

and also a battery powered psychrometer. The results were repeatable

and accurate to within approximately 5 percent relative humidity.

Ignition Timing

Ignition timing was adjusted manually by rotating the ignition

points. The timing bracket had divisions marked for one-degree incre-

ments when adjusting the timing. Ignition timing accuracy was checked

with a timing light and found to be within 1 degree of the desired

value. The spark timing was further checked by measuring the time

from spark to top dead center on a pressure-time diagram. Again the

results indicated the ignition timing was within one degree of the

desired value.




The pressure versus time diagrams, which were obtained by the

experimental methods described in Chapter III, are also a measure of

the rate of chemical heat release versus time. Therefore, a computer

program was developed which calculated the volume of fuel-air mixture

burned which would produce the measured pressure at each time increment.

The basic input for the program consisted of a listing of selected time

points after spark ignition and the corresponding cylinder pressures.

As many as twenty time and pressure data points could be input for each

run condition. The additional data needed for initial calculations of

fual mixture, combustion chamber geometry and piston position as func-

tions of time were also furnished as input data. A Fortran listing of

the source deck used can be found in Appendix I. The output from the

program consisted of predicted extent of reaction completion and pre-

dicted flame front position at each time point for an assumed spherical

flame front. Additional output of interest was the calculated temper-

ature and chemical composition variation in the burned gases.

Combustion Simulation

The actual combustion in an internal combustion engine is

extremely complex. Temperature gradients exist at the walls of the

cylinder and near the flame front as well as in the burned gases.

The highly turbulent gases throughout the cylinder cause mixing in the

burned gases and uneven flame fronts. Some preflame reactions occur in

the unburned gases and chemical equilibrium is not instantaneously

obtained-in the flame front. The flame front has an irregular shape

and a finite thickness. Thermal radiation from the flame front is

absorbed in varying amounts by the burned and unburned gases. There-

fore, in order to model mathematically the combustion process, certain

simplifying assumptions must be made.

The following combustion model was used to predict the amount

of reactants burned as a function of time to produce the measured

pressure-time history. The flame front in the hydrogen-air mixture

was considered to be a discontinuity which defined the boundary between

the unreacted gas and the reacted gas. The unburned gas was assumed to

have a fixed composition and the burned gases were assumed always to be

in chemical equilibrium. These assumptions are reasonable approxima-

tions since the hydrogen-air reaction rates are very fast and the heat

release takes place over a very short distance even at much lower

initial temperatures and pressures than those present in an internal

combustion engine [44,43]. There is also assumed to be no heat trans-

fer across the flame front from the burned to the unburned gas as well

as no heat transfer between the gases and the combustion chamber walls.

These simplifying assumptions are reasonable due to the fact that heat

transfer via conduction and convection from the cylinder walls is not

appreciable due to the short time needed for the flame to sweep through

the combustion chamber [19,20,21]. Less thermal radiation is emitted

by hydrogen flames than from hydrocarbon flames and therefore heat

transfer to the combustion chamber walls via radiation is also negli-

gible [56,19], since the effect of radiation from hydrocarbon fuels

is small. Since the unburned gases are quite transparent to the radia-

tion, little error is involved in neglecting the heat transfer across

the flame front via radiation [57,58]. The calculated total heat loss

to the combustion chamber walls during the flame travel time is less

than 2 percent of the total heat release during combustion and is there-

fore assumed to be unimportant [19,21]. Each of the small segments of

gas which are burned can be considered to react at constant pressure

since each reaction produces only a small change in pressure [43,39].

The physical model for this study consisted of a homogeneous

mixture of perfect gases at the time of spark ignition. The moles of

each component present were calculated from air, fuel and water flow

rates and pressure was measured directly. Therefore, an initial temper-

ature was calculated by using the perfect gas equation,

T .

The total pressure, P, is equal to the sum of the partial pressures of

the component gases and the number of moles, N, is equal to the sum of

the moles of the components. Changes in cylinder pressure occur because

of piston movement and the heat release due to chemical reaction.

The combustion chamber volume was determined as a function of piston


The unburned portions of the cylinder gases is adiabatically

compressed by piston movement and by the combustion which has occurred

in the gas already reacted. No heat transfer is considered since the

time between spark ignition and reaction completion is very short and

the cylinder pressure is assumed to be equalized at all points in the

combustion chamber. Chemical equilibrium is assumed for all combus-

tion products, and the products are assumed to be ideal gases. Burned

gases are also adiabatically compressed by the latter portions of the

charge which are reacting. Chemical compositions in the burned gases

are constantly corrected for equilibrium changes due to the temper-

ature rise from compression.

Reaction Calculations

Adiabatic Flame Temperature

The adiabatic flame temperature was calculated for reactants

at given initial conditions. The products were assumed to be in chem-

ical equilibrium and the reaction was assumed to occur at constant

pressure. The small size of reacting segments allowed the constant

pressure approximation since the gases were free'to expand and their

reaction created only a small pressure rise in the combustion chamber.

The moles of reactants NR(i) were known and the moles of products NP(i),

and final temperature were calculated. The species present in the

equilibrium mixture of products were assumed to be 02, N2, H2, H20, 0,

H, OH, and NO. The amounts of NO2, N, NH3, HO2 and other species were

negligible at the temperatures which were obtained. The overall reac-

tion which was considered is the following one:

NR(1)O2 + NR(2)N2 + NR(3)H2 + NR(4)H20

NP(1)O2 + NP(2)N2 + NP(3)H2 + NP(4)H2 0

+ NP(5)O + NP(6)H + NP(7)OH + NP(8)NO.

Therefore, for a given final temperature and pressure the values for

NP(i) are determined as follows. Five equilibrium equations are used:

1/2 0 + 1/2 N NO
2 2

H2 + 1/2 02 1 H20

1/2 02 0

1/2 H2 H

1/2 02 + 1/2 H2 OH.

Three atom-balance equations are also needed to ensure conservation of

mass throughout the reaction for hydrogen, oxygen, and nitrogen atoms.

The atom balance equations are given below for the three species.

Hydrogen: 2NR(3) + 2NR(4) = 2NP(3) + 2NP(4) + NP(6) + NP(7)

Oxygen: 2NR(1) + NR(4) = 2NP(1) + NP(4) +.NP(5) + NP(7) + NP(8)

Nitrogen: 2NR(2) = 2NP(2) + NP(8).

Equilibrium constants, Kp, for NO, H20, O, H, and OH were input in the

form of tables which were a function of temperature in increments of

100 Kelvin. The values for the equilibrium constants were obtained

from the JANAF tables [59]. The equilibrium constants, Kp, are based

on partial pressures and are defined as the following definition for

KpNO indicates. Nitric Oxide: KpNO = PN (P N2) The values for

Kp were all based on the formation of one mole of the product species

as shown above in the five equilibrium equations which were considered.

For mixtures of perfect gases, the partial pressure of one

species is proportional to the mole fraction of that species; i.e.,

8 8
NO = NP(8)(P /N ) where POT= Pi and NTOT = NP(i).
i=l i=l

Therefore there are nine independent equations and nine unknowns.

After some algebra the equations can be combined in terms of NP(3),

NP(2), and (PTOT/NTOT) PON to yield three nonlinear equations in the

three unknowns, i.e.,

f(NP(3), NP(2), PON) = 0

g(NP(3), NP(2), PON) = 0

h(NP(3), NP(2), PON) = 0.

The numerical iteration method used to obtain a solution was similar

to Newton's Method for a single equation. The method required the

use of improved values for each component as they become available.

Only one partial derivative evaluation for each component was necessary

[60,61]. The recursion formulas used to generate succeeding aproxima-

tions are:

NP(3)i = NP(3)i f(NP(3)i, NP(2)i, PONi)/fN
i+1 1 i i. P NP(3)

NP(2) = NP(2)i g(NP(3)i, NP(2)i, PONi)/gN
i+l i+1' NP(2), PON.) 5Np(2)

PONi = PONi h(NP(3) i+lNP(2)i+, PONi)/hpO

The functions fNP(3)' gNP(2)' hpON' are the partial derivatives of

f, g, and h, respectively, and are evaluated by using the same com-

ponent values that were used for the original function evaluation.

The solution is unique but there is no criterion which will

guarantee convergence of the iteration [60,61]. However, convergence

was obtained in all cases by using the following initial values for the


NP(3)o = NR(3) 2(NR(1)), ( > 1.0 (rich)

NP(3) = 0.1(NR(3)), < < 1.0 (lean)
NP(2)o = 0.9(NR(2))

PON = 1.5 P I/ NR(i)

= [H2/ 2[021 Hydrogen Equivalence Ratio.

The iteration proceeded until the following convergence criterion

was satisfied.

(IPON i+ PON. /PON) < 1.0 x 107.

At that point, the eight product mole fractions were used to calculate

the mixture enthalpies. The method of determining burned gas composi-

tion as described above is shown in detail in Subroutine FLAME as listed

in Appendix I.

The assumed adiabatic constant pressure combustion of each

gas segment permitted a determination of the adiabatic flame temper-

ature for each segment. Enthalpy data for the eight chemical species

considered were obtained from JANAF tables [59]. The values were

referenced to 2980K and had the units of kcal/mole. A least squares

polynomial curve fit was used to match the enthalpy data over the tem-

perature range from 500 to 40000K. Two curve fits were needed for each

species to obtain a good fit over the complete temperature range. For

temperatures greater than 15000K, the best least squares polynomial

curve fit was a fifth order polynomial in temperature,

2 3 4 5
H = A + BT + CT + DT + ET + FT

The coefficients varied for each species. The polynomials matched the

data points within 1 percent for each of the species. For temperatures

less than 15000K, the best least squares polynomial curve fits were

obtained with fourth order polynomials. They matched all data points

with a maximum of less than 1 percent for temperatures greater than

6000K. These curve fits are entered in the program as external func-

tions and are included near the end of the Fortran listing in Appendix I.

The standard heat of formation for each species at 2980K was

obtained from the JANAF tables where the units are kcal/mole. The

overall enthalpy change for an adiabatic constant pressure process is

zero. In order to simplify the programming, the conversion for reac-

tants to products was assumed to take place at the reference temper-

ature of 2980K so that the heat of formation data was needed for only

one temperature. The enthalpies of the reactants and products at their

respective initial and final temperatures were also referenced to

2980K. For a given initial reactants temperature, T., and pressure,

the final temperature of the products, Tf, was determined to be the

value for Tf which satisfied the following equation:

8 8
AH (T ) = 0 = E NP(j)H.(Tf) + Z NP(j)AHF(j)
j=l j=l

4 4
E NR(j)H (T.) E NR(j)AHF(j).
j=l j=l

The values for NP(j) were determined from chemical equilibrium con-

siderations as described earlier. The values for the enthalpies of

the components come from their respective least squares curve fits.

The first two estimates for the final temperature, Tf, were 2000 and

39000K. Further estimates for the final temperature were calculated

from the following recursion formulas for each iteration:

Tf (i+l) = Tf(i) DELT

DELT = (Tf(i)- Tf(i-l)/(AH (Tf(i)) -AH (Tf(i-l)) )]AH (Tf(i))

Rapid convergence to the desired accuracy was obtained. Details are

shown in the last portion of Subroutine FLAME as listed in Appendix I.

For each new estimate for Tf, the chemical composition was recalcu-

lated to satisfy the chemical equilibrium equations. The final out-

put from Subroutine FLAME was the calculated burned gas temperature,

Tf, and chemical composition for initial inputs of pressure, temperature

and chemical composition of the reactants.

Pressure Rise Simulation

One of the intents of this analysis was to predict the rate

at which reactants were burned during the hydrogen-air combustion

reaction. The burned and unburned gases were assumed to be at the same

pressure and separated by a "thin" flame front. The unburned gases

underwent no reaction while they were being compressed but the burned

gases continually reacted to remain in equilibrium as the temperature

increased due to compression. The rate at which the reactants were

consumed was calculated by analytically matching a measured pressure-

time history. The measured pressure-time history was approximated by

a series of reactions which occurred at given pressures and times.

The segment of reactants which burned to produce each desired pressure

rise was treated independently. The identity of each segment was pre-

served, although the temperature, volume, pressure, and chemical compo-

sition of each segment was continuously changing. The first burned

segments reached a higher maximum temperature than the segments which

reacted later because the compression of the burned gases created

a greater temperature rise than the compression of the reactants.

This occurs because a temperature ratio is calculated due to compres-

sion and the temperature change is therefore greater for the higher

temperature gases.

The moles of reactants which were burned to produce a pressure

rise from P. to Pi+ in the time period from t. to ti+1 were calculated

in the following way. The temperature rise in the unburned gas, TUB'

due to adiabatic compression from P. to Pi+1 is calculated as follows


TUB = T UB(P i+/Pi)
UB. UB. 1+1 i
i+l 1

The specific heat ratio, y, was calculated from the least squares curve

fit enthalpy data. The value for Cp at a given temperature for each

species was taken as the average slope of the enthalpy rise over a

one-hundred-degree temperature span. The derivative of the least

squares polynomial was not used since the instantaneous slope for a

least squares curve fit may be quite different than the average slope.

The average value for Cp of the mixture, Cp, was then

4 4
Cp = E NR(j)Cp(j)/ Z NR(j) (Unburned Gases)
j=l j=l

CV = Cp R y = Cp/CV

Similarly, the values for CV and y in the burned products were found

by calculating Cp for the eight component gases. The reacted gas was

also adiabatically compressed to a higher temperature which could be

determined from known pressure rise and specific heat ratio calcula-

tions as shown above for the unburned gases. The reacted gas then

underwent dissociation to remain in chemical equilibrium at the higher

temperature. Since total moles, pressure, and temperature were then

known for the burned segments, the volume occupied by each segment was

determined by the perfect gas equation of state.

The total volume in the combustion chamber was determined from

the calculated piston position. The adiabatic flame temperature and

the ratio of moles of reactants to moles of products were calculated in

Subroutine FLAME for a reaction occurring in the unburned gases.

The following four equations determine the number of moles of reactants

which must react to create the desired pressure rise from P. to P i+

The unburned portion of the combustion chamber reactants are

defined as those present after the reaction occurs.

(1) = -- C (Unburned Reactants).
UB i+l

The previously burned segments that have again reacted to

remain in chemical equilibrium have volumes which can be calculated

from known values for temperature, pressure, and number of moles,

Vk = (for segment k). Therefore, the total combustion chamber

volume, VTOT, is equal to the sum of the volumes for the burned seg-

ments, unburned volume, and the volume occupied by the products of the

segment undergoing reaction, VB.

(2) VB + V = VOT E V CV

For m previously burned segments, CV can be calculated. The products in

volume VB were assumed to be perfect gases.

(3) P C2 (Reaction Products).
B i+l

The moles of reactants unburned plus the moles of reactants

being burned for this segment are equal to the total moles of reactants

initially available minus the moles of reactants burned in previous


(4) NUB + N Rm = NRTOT NBk Rmk CM

The ratio of moles of reactants to moles of products for the reaction

under consideration is R The subscript, k, denotes the segments which

burned prior to the segment presently reacting.

From equations 1 through 4, the following expression was

derived for NB:

N = (CM(C1) CV)/(C1(R ) C2)
B m

The value obtained for N was the number of moles of burned products

that were needed to produce the desired pressure rise. The amount of

reactants needed to give NB moles of products is simply equal to NB

multiplied by R which is the ratio of moles of reactants to moles of

products for the reaction of interest. Therefore, the total moles

remaining unburned at a given cylinder pressure was determined by

subtracting the moles burned for each segment from the total initial


Residual Gases

The amount of exhaust residual gases in the combustion chamber

was calculated for an estimated average exhaust temperature. For full

throttle operation, the amount of residual gases is quite small and the

effect of the residual gases is slight [19,20]. The estimated mole

fraction of exhaust gases which remained in the combustion chamber

were 0.050, 0.043 and 0.036 for the compression ratios 8.2, 9.3, and

10.9, respectively, and were input values. The residual mole fraction

is defined as the ratio of moles of exhaust gas in the fresh charge to

total moles of exhaust gas after combustion. The residual gas composi-

tion was approximated by assuming the complete consumption of hydrogen

for fuel lean operation or oxygen for fuel rich operation. Also, since

no appreciable dissociation occurs at the exhaust temperatures, the

amounts of 0, H, OH, and NO were assumed to be negligible. This allowed

the estimated moles of H2, N2, 02, and H20 in the residual gases to be

added to the known intake mixture composition to obtain the total moles

of reactants in the combustion chamber. The fuel equivalence ratio, q,

does not include the effects of residual gases since 4 is a measure of

the intake composition only. The chemical composition of the residuals

was dependent on p even though the residual fraction was assumed to be

constant for a given compression ratio. Therefore, the actual composi-

tion of the combustion charge was slightly more fuel lean or rich than

the corresponding lean or rich intake mixtures, since any unburned

hydrogen or oxygen in the residual gas was added to the intake mixture.

Reaction Products Equilibrium

The products of combustion can be reduced to an equivalent set

of reactants with the same total enthalpy and mass. The reactants could

then be used as initial conditions to obtain the adiabatic flame temper-

ature and determine the equilibrium chemical composition of the products.

This was done for the previously burned segments of reaction products

which were compressed to conditions of temperature and pressure such

that the chemical composition did not satisfy equilibrium. The conver-

sion of products to reactants with the same enthalpy was done to

facilitate the use of Subroutine FLAME which was written for four

reactants. The Fortran subroutine which accomplished the above reduc-

tion to reactants is presented in Subroutine INIT as listed in

Appendix I. The input for the subroutine consisted of an array of

products which were compressed to a given temperature which was dif-

ferent than the temperature before compression. The product gases were

not in chemical equilibrium at the new temperature.

The array of reactants which satisfied the mass balance

requirements were found in the following way:

NR(2) = NP(2) + NP(8)/2 (Nitrogen).

The value, r is defined as the mole fraction of the total reactants

which are present in the segment of interest. Therefore, r can be
determined by comparing the mass of nitrogen in the segment with the

total amount of nitrogen in the charge,

r = NR(2)/NR(2)TOT

The moles of 02, H2, and H20 are then found by multiplying the original

moles of each reactant species by ro. This resulted in the same molar

concentrations for the segment reactants as were present in the original

reactants. The total enthalpy of the products was calculated since the

moles of each product species and the temperature were known, and the

heat of formation, AHF(j), for each species j was given for a temper-

ature of 2980K. Therefore, the enthalpy equation which must be satis-

fied was the following one:

8 8
E H.(T )NP(j) + Z NP(j) AIF(j)
j=1 j j=l

4 4
SE H (T )NR(j) + E AHF(j)NR(j)
j=l j=l

The value for T the products temperature, was known and the value for
TR, the reactants temperature, was solved for by iteration. The correct

reactants temperature was the value which reduced the difference in

total enthalpies between reactants and products to zero.

Calculated Flame Front Position

The flame front location was calculated as a function of piston

position and calculated volume burned for a given compression ratio.

The piston position was calculated for each time point as a function of

crank angle, therefore the geometry was continuously changing. For a

given calculated burned volume, the flame front position could be

established for the combustion chamber geometry which was modeled at

that time point. The calculated burned volume was determined from mea-

sured pressure-time history in the combustion chamber. The Fortran

listing for the subroutine which performed these calculations,

Subroutine CALVOL, is in Appendix I.

Combustion Chamber Geometry

The combustion chamber geometry model was represented by three

cylindrical shapes. The first described the exhaust valve recess, the

second defined the piston cylinder and combustion chamber except for

the valve recesses, and the third cylindrical volume was the intake

valve recess.

Cartesian coordinates were used to describe the combustion

chamber, the axis origin was located at the spark plug position in the

exhaust valve recess as shown in Figure 10. Each of the three cylin-

drical volume segments had boundaries which were described by equa-

tions of the following type:

2 2 2
(x-a) + (y-b) = (RAD)2, k1 < z < k2.

The centerline x, y, coordinates for each of the vertical cylinders,

were defined by a and b, while the cylinder radius was given by RAD

and the top and bottom of each cylinder was located at k2 amd kl, respec-

tively, for the z coordinate. The exhaust valve recess had a depth of

0.888 inch and the intake valve recess had a depth of 0.145 inch.

The flame front was assumed to have a spherical shape and to

have originated at the spark plug location. The spherical flame front

location was described in sperical coordinates with the origin at the

spark plug. The spherical coordinates r, 0, (, are related to x, y,

and z in the conventional manner as shown in Figure 11. A volume incre-

ment was defined for an incremental change in r, 0, and 4,
(5) dV = r sin e dr dO dq.

A volume increment may also be defined in cylindrical coordinates

as shown in Figure 11,

(6) dV = r de dr dz

The reasons for using cylindrical coordinates are explained in the

following section.

Volume Calculations

The flame front was assumed to expand spherically throughout

the combustion chamber from the initial starting point at the spark plug

as shown in Figure 12. Therefore, spherical coordinates were used

where possible to predict the burned volume behind the flame front for

a given flame travel distance -R. Ignition took place in the exhaust

valve recess; therefore, some portions of the combustion chamber were

not in a line-of-sight position with respect to the ignition as shown

by the shaded area in Figure 10-B. For those portions of the combus-

tion chamber, the spherical flame front was approximated by a cylindri-

cal flame front with an adjusted origin. A transition region where

part of the flame front was spherical and part of it was neither

spherical nor cylindrical was needed when burning was occurring both in

and out of the line-of-sight volume. The transition region was accounted-

for by means of a curve fit polynomial which added a correction to the

calculated spherical volume.

The volume burned during the first 0.75 inch of flame travel

was accounted for by using only the spherical flame front model. The

volume beyond a flame travel distance of 1.63 inches was accounted for

by using only the cylindrical flame front model. The region of flame

travel between 0.75 and 1.63 inches was the transition region. The

burned volume in that region was approximated by calculating the total

volume in the spherical region for a given flame travel distance and

adding to it the volume which burned outside the line-of-sight volume.

The volume which burned outside the line-of-sight volume, or exhaust

valve recess, was calculated in the following way. The flame front was

assumed to expand spherically throughout the exhaust valve recess until

it reached the clearance volume between the piston and the cylinder

head. When the flame reached the clearance volume, it was assumed to

begin expanding spherically into the unburned gases with the origin for

the expansion located at the point where the original flame front

reached the clearance volume. The distance the flame traveled to reach

a position at the edge of the clearance volume depended on the location

of that position. The first point at which the flames began spreading

into the clearance volume was located directly below the spark plug with

the Cartesian coordinates (0, 0, -0.75). The last point at which the

flames reached the clearance volume was located directly across the ex-

haust valve recess at the coordinates (1.45, 0, -0.75) that locate

point F, as shown in Figure 10. The value for R, therefore, varied

from 0.75 to 1.63 inches during the transition region. The value for

the z coordinate, -0.75 inch, approximately corresponds to the midpoint

of the clearance volume. This value was used to approximate the actual

expansion into the clearance volume. The actual expansion was simulated

by assuming that a cylindrical wave front with no vertical curvature

began expanding cylindrically into the clearance volume when the flame

front reached the midpoint of the clearance volume, as shown in Fig-

ure 10-C. Flame front position, R, was assumed to be equal to the sum

of two reference distances for all points outside the exhaust valve

recess. The first distance was the distance from the spark plug to

a position at the edge of the exhaust valve recess and clearance volume

midpoint, similar to point F in Figure 10-C, and the second distance was

an arc from that position to the flame front. Therefore, by determin-

ing the distance to a series of points around the perimeter of the

exhaust valve and at the clearance midpoint (z = -0.75) the flame front

position R could be drawn as the curve that was an envelope for the

arcs drawn from the series of points. The arcs from each point were of

such a length that when the arc length was added to the distance to the

spark plug, the result was equal to the given flame travel distance R,

as shown in Figure 10-C. Curves were drawn for the values of R rang-

ing from 0.75 to 1.63 inches; the area under those curves was equal to

the area burned beyond the exhaust valve recess. A planimeter was used

to measure the burned area for the various values of R. A curve fit

for area burned AB, versus R was determined,

(7) AB = 0.0017 (0.8859)R + (1.1178)R2 0.75 < R < 1.63.

The volume burned was equal to AB multiplied by the piston-cylinder

head clearance distance, which was a function of piston position and

compression ratio.

When R was equal to 1.63 inches, the shape of the model flame

front was almost a perfect cylinder, as shown in Figure 13. The calcu-

lated flame front position was closely approximated ( 0.05 inch), by

using a cylindrical flame front with axis origin located at x = 0.3

and y = 0.04 with a radius of 1.15 inches. The flame front location

and volume burned were then determined by the cylindrically expanding

flame front in the clearance volume. The cylindrical flame front was

a reasonable approximation for the assumed spherical expansion since

the effect of vertical curvature was slight in the small clearance

between the piston and cylinder head.

The problem of volume calculation was posed in the following

manner. The volume burned was known and the flame front position had

to be calculated. Since the geometry was constantly changing due to

piston movement, the numerical integration which was used for volume

calculation usually had to be repeated for each segment burned. The

spherical volume was determined by numerically integrating the spherical

volume increment for each value of r, for values of ( and 0 ranging

from 0-90 and 0-180 degrees, respectively, and multiplying by two since

the spherical portion was symmetrical with respect to (. The geomet-

rical boundaries were checked to ensure that no volume was calculated

when the boundaries were exceeded during the integration. The values

for volume versus R were only calculated once for R less than 0.75 since-

the changing geometry did not affect that portion of the combustion chamber.

Therefore, integration started at R equal to 0.75 inch for the cal-

culations which occurred after the flame front distance had exceeded R

equal to 0.75 inch.

After the flame front location exceeded 1.63 inches, the inte-

gration consisted only of cylindrical integration with 0 varying from

0 360 degrees for each value of r. The volume behind the flame front

for R equal to 1.63 inches was calculated analytically as the sum of

the exhaust value recess volume plus the volume from the curve fit for

area burned outside the exhaust valve recess. This calculated volume

was used as an initial condition for volume calculations for R greater

than 1.63 inches. Considerable computer time was saved by minimizing

the number of volume integration performed. The incremental values

used for all the integration were 5 degrees for angles and 0.1 inch

for radius increments. For all cases, the value of R, or r, was

increased until the volume burned for position R. was greater than the

calculated burned volume. Then the correct value for R was determined

by linear interpolation between R. and Ri_-. (The volume for R. was

greater than the calculated burned volume and the volume for R._ was

less than the calculated burned volume.)

The flame travel distance R was measured along the approx-

imated spherical expansion path, as shown in Figure 10-C. Flame speed

was calculated as the average speed required to move from RK to Rk+1 in

the time from tk to tk+l. The subscript k stands for values before

a particular segment was burned and k+l represents values after the

segment was burned. If the calculated burned volume exceeded the total

volume in the cylinder, a warning message was printed as output from

computer Subroutine CALVOL. This could occur since the program was not

limited in the amount of reactants which could be consumed to satisfy

the pressure criteria. The calculated burned volume exceeded the total

combustion chamber volume whenever the calculated number of moles which

reacted exceeded the measured initial moles of reactants. This phys-

ically impossible result was allowed to occur without causing calculation

termination since it gave an estimate of the small amounts of additional

reactants needed to match the given pressure-time criteria. The amount

of reactants consumed in excess of those available will be discussed

more completely in the following chapter dealing with results.

Apparent Quench Distance

The presence of an apparent quench distance was suggested from

certain experimental runs that consistently had calculated burned

volumes slightly less than the total cylinder volume at reaction com-

pletion. The unburned gas was assumed to be near the cylinder walls

and of uniform thickness throughout the combustion chamber.

An apparent quench distance was input in Subroutine CALVOL to

determine how large the wall quenching effect would have to be to

account for the calculated unburned volume. The effect was simulated

in Subroutine CALVOL by replacing the combustion chamber boundaries

with artificial boundaries which were separated from the geometrical

boundaries by a distance equal to the quench distance. The flame front

position was calculated such that the burned volume behind it did not

include any of the volume which was within the assumed quench layer.

The correct input value for the apparent quench distance gave maximum

flame travel distance at reaction completion. This simulated the case

of a flame front sweeping the entire combustion chamber but leaving a

thin layer of unburned gas along the combustion chamber boundaries at

reaction completion.


Flame quenching does occur in internal combustion engines

[62,63]. Daniel observed quenching photographically for propane fuel

[62], and the presence of unburned hydrocarbons in the exhaust of

internal combustion engines has been credited to wall quenching by

Gottenberg et al. [63]. The magnitude of the flame quenching effect

is discussed in the following chapter where results are presented.



Performance Data

The experimental tests in this study were conducted at full

throttle and the spark timing was set for maximum power output.

Compression ratios were expressed in the results section as single

numbers and were abbreviated C.R., while engine speeds in revolutions

per minute were abbreviated RPM. The fuel mixture was described by

the fuel equivalence ratio, C, which is the ratio of actual fuel-air

ratio to stoichiometric fuel-air ratio. Therefore, ( greater than

1.0 corresponds to a fuel rich mixture and p less than 1.0 corresponds

to a fuel lean mixture. For hydrogen, } is defined in the following

way for a mixture where the hydrogen and oxygen concentrations are

expressed as [H2] and [02], respectively,

S= [H2]/2[02].

The engine power output for operation on hydrogen fuel was

plotted along with the power output obtained from Gulf regular commer-

cial gasoline in Figure 14. Brake horsepower is a measure of the work

output of the engine and was determined from measured engine speed and

dynamometer load. Indicated horsepower was calculated by measuring the

engine friction and pumping work losses and adding that value to the

brake horsepower. Friction and pumping losses were determined by

motoring the engine and measuring the torque needed to maintain the

engine speed. Friction and pumping losses were found to correspond to

approximately 2.94 horsepower at 1835 RPM. As seen from Figure 14,

when operating on hydrogen the engine developed approximately 75 per-

cent of the power which was developed with gasoline. This is due to

the fact that the hydrogen displaces a significant volume of the intake

air in the intake manifold and therefore decreases the available oxygen.

The stoichiometric air to fuel ratio is 2.38 for hydrogen on a mole

basis; therefore, 29.6 percent of the intake manifold mixture would be

hydrogen for a stoichiometric mixture in dry air. Based on lower heat-

ing values of 19,080 btu/lb. for gasoline and 51,608 btu/lb for hydro-

gen [18], the hydrogen fueled engine used only 70 percent as much fuel

energy at maximum power as it did when fueled with gasoline. Therefore,

the indicated thermal efficiency must be higher for hydrogen fuel since

it yields 75 percent as much indicated horsepower as gasoline for 70

percent of the energy input. This is shown in Figure 15 where the indi-

cated thermal efficiency for hydrogen is 33.7 percent and for gasoline

31.5 percent for maximum power output. The brake and indicated thermal

efficiencies are based on the lower heating values given above for

hydrogen and gasoline. Thermal efficiencies are defined as power out-

put divided by energy input; therefore, the indicated and brake thermal

efficiencies are based on indicated and brake horsepower output as

shown in Figure 14. The indicated mean effective pressure, defined in

the usual way as a measure of the work done per cycle, was 120 pounds

per square inch for the maximum indicated horsepower from Figure 14.

The volumetric efficiency was defined for this study as the actual

volumetric flow rate of intake mixture divided by the ideal volumetric

pumping capacity of the engine which was based on the piston displace-

ment volume and engine speed. Measured values for volumetric efficiency

were nearly constant, varying from 74.2 to 77.9 per cent. The perfor-

mance data agreed well with those of R. O. King [8]. His data were

extrapolated to the engine operating conditions used for Figures 14

and 15 for hydrogen fuel and gave agreement within the estimated accur-

acy of this study. Indicated horsepower was estimated to be accurate

to within 0.4 horsepower and indicated thermal efficiency was estimated

to be accurate within 2.3 percent. Data on power output are listed

in Table 4.

As seen from Figure 14, the experimental runs were all conducted

in the range of fuel equivalence ratios which produced high power out-

puts, since this was the range where "knocking" combustion was found -

to occur. The indicated thermal efficiency continues to increase as the

mixture is leaned to a maximum which occurs for mixtures 50 to 60 per-

cent lean (q = 0.5 0.4) as shown by King [8]. This trend was indi-

cated in Figure 15 along with the decreasing brake thermal efficiency

as the mixture was made more lean. The increasing proportion of power

which is used to overcome friction for lean mixtures causes the

decrease in brake thermal efficiency for > less than 1.0.

Reaction Progress Data

The combustion pressure rise model (Chapter IV) was checked by

calculating the predicted pressure rise for a stoichiometric mixture of

air and hydrogen and comparing it to data collected by Fenning in exper-

iments with a constant volume vessel as reported by Bone and Townsend [56].

The predicted value for maximum pressure was corrected to within 1 percent

for a maximum pressure of 709 psia. This suggests that the computer

program should predict the pressure rise in the combustion chamber quite


Figures 16 through 24 present the calculated reaction progress

as a function of time from spark ignition. The time corresponding to

reaction completion was the time at which the calculated rate of heat

release went to zero as determined from the pressure-time traces input

for each run from spark ignition to reaction completion. The total

reaction times as determined by these figures are listed in Table 5 for

each experimental run. The value for nH20, the water vapor mole frac-

tion which is displayed on each figure is an average value for the set

of runs on the figure; the actual value for nH20 for each experimental

run is listed in Table 4. In all cases, the richer mixtures burned the

fastest, which was to be expected since the maximum laminar flame speed

for hydrogen-air mixtures at atmospheric pressure occur at approximately

= 1.6, or 40 percent hydrogen by volume [43]. Increasing water induc-

tion tended to increase the total reaction time for all three compression

ratios, and increasing the compression ratio tended to reduce the

total reaction time for a given rate of water induction. Ionization

gauge number one (I.G. No. 1) was located directly across the exhaust

valve recess from the spark plug at a distance of 1.45 inches. Ioniza-

tion gauge number two (I.G. No. 2) was located in the last part of the

combustion chamber to be inflamed and the flame travel distance from

the spark plug to I.G. No. 2 approximately 2.90 inches, as shown in

Figure 5. The time of peak output from I.G. No. 2 is listed in Table 5

for each experimental run along with the calculated total reaction time

as determined from Figures 16 through 24. The time of peak output from

I.G. No. 2 matches the calculated total reaction time quite closely for

all runs.

The calculated reaction completion increments for each test

included one point after reaction completion as shown in Figures 16-24.

That point corresponded to the last time that heat release was predicted

from the input cylinder pressure-time data. Zero heat release was

predicted for all test runs when time values were greater than the pre-

dicted time of reaction completion; i.e., no detectable late burning


The maximum rate of reaction completion, as determined from

Figures 16-24, occurred well before the end of the reaction. The

inflection points in the curves of reaction completion versus time

defined the maximum rate, and the inflection point invariably occurred

when the reaction was approximately half completed. This was true in

particular for experimental runs 2C and 3C as shown in Figure 22, which

produced very heavy knocking. There was no indication in Figure 22 of

rapid combustion of the last part of the charge for the knocking runs

2C and 3C as might be expected if autoignition had occurred.

Reaction completion was plotted versus crankshaft angle for

representative slow and fast burning mixtures in Figure 25. The igni-

tion timing for all tests was set for maximum power output; this led to

early spark ignition for slow burning mixtures and later spark ignition

for faster burning mixtures. For run 1A, the combustion period covered

28 degrees of crankshaft rotation; the faster burning rich fuel mixture

used for run 3A yielded combustion completion within approximately

16 degrees of crankshaft rotation. At 1835 RPM engine speed, the crank

angle changes 11.0 degrees per millisecond. Therefore, the combustion

period covered a range of crankshaft rotation varying from 14 to 40

degrees for the full range of experimental runs conducted. The ignition

timing used for each run is listed in Table 3. The amount of time it

takes from spark ignition for a noticeable pressure rise to occur was

consistently longer for the slower burning mixtures, as shown in

Figures 16-25. This tended to produce more cycle-to-cycle pressure

rise variations for the slow burning mixtures since variations in the -

initial flame front development are related to cycle-to-cycle varia-

tions [55,47,52].

Flame Speed Data

All flame speeds were measured relative to the cylinder head

not relative to the unburned gases. The flame speed thus defined is

equal to the sum of the flame .speed relative to the unburned gas and

the speed of the unburned gas normal to the flame front due to expan-

sion of the burned gases [39].

Figures 26-34 present the calculated average flame speeds

predicted by the input pressure rise data for all the experimental

runs as well as the measured average flame speed. The measured aver-

age flame speed was determined by measuring how long after spark

ignition the number two ionization gauge was triggered by the arrival

of the flame front and then dividing the flame travel distance

(R = 2.90 inches) by the time taken for the flame front to reach

I.G. No. 2. The calculated average flame speeds plotted versus

time in Figures 26-34 were the values predicted by the computer program

to be needed in order to match the pressure rise data for the assumed

spherical flame front model. The flame speed value at each time point

is the average flame speed needed to produce the change in flame front

position calculated for the volume segment which was burned. All the

calculated flame speed values are therefore dependent on the accuracy of

the time and pressure changes for each reaction segment. For fast pres-

sure rise, the time step for each reaction segment was necessarily smalL_

so that the pressure rise for that segment was small. The combustion

chamber pressure as a function of time was listed for each of the exper-

imental runs in Appendix II. The errors tend to average out so that if

one segment had a higher than actual flame speed the next segment should

have a lower than actual flame speed. The first data points for each

run had a relatively large possible error in flame speed because the

burned volume was so small that a slight error in pressure rise yielded

a large percentage error in flame front position immediately after

spark ignition. The spherical flame front model was not exact but was

a reasonable model to use for a combustion chamber with no induced

swirl [64,23,24]. Since in general each portion of the actual flame

front would have a slightly different flame speed and the cycle-to-

cycle variation also affects flame speed, the predicted values for a

spherical flame front can be viewed as indicative of average flame speeds.

Curves were not drawn through the data points since trends are more

obvious from examination of the individual data points. The rich mix-

tures invariably had the maximum flame speeds and the flame speeds

always decreased as the reaction progressed even for the heaviest

knocking condition runs, shown in Figure 32. Estimated maximum flame

speeds as obtained from Figures 26-34 were 225, 275, and 300 feet per

second for compression ratios of 8.2, 9.3 and 10.9, respectively.

Time of calculated reaction completion for each experimental run was

denoted by zero flame speed in Figures 26-34. The dashed lines in

Figures 26-34 represent the measured average flame speeds needed to

cross the cylinder in the time span from spark to I.G. No. 2 triggering.

The effect of water induction was shown to decrease consistently the -

flame speed and increase the total reaction time. The lean (4 < 1.0)

mixtures have much lower maximum flame speeds than the richer mixtures;

i.e., approximately 125, 140, and 150 feet per second for compression

ratios 8.2, 9.3, and 10.9, respectively, for no water induction.

The measured average flame speeds were consistently less than the max-

imum calculated flame speeds due to the presence of low flame speeds

during the initial and final portions of the reaction.

Unburned Reactants

The previous reaction data presented in Figures 16-24 were

based on calculated reaction completion which implied that the reac-

tion was completed when the predicted heat release rate was zero.

At the time corresponding to predicted reaction completion, the reac-

tants in general were not completely burned. This could occur for the

computer model since the amount of reactants consumed was the sum of

all the reactants consumed for the input series of pressure increases.

The combustion model was checked to determine how sensitive the calcu-

lated amount of reactants burned was to changes in input data. The

possible input variations considered were: one degree in ignition

timing, 5 percent in hydrogen flow, 1 per cent in air flow, 4 RPM

engine speed, 5 percent relative humidity, 0.05 cubic inches volume

at TDC, 2 degrees Fahrenheit in intake air temperature, 10 PSI in

maximum pressure, 3 percent for input pressure rise data, 10 per-

cent for water induction, and 0.02 for residual mole fraction. The

results indicated that the largest effects were from ignition timing,

pressure data, and hydrogen flow rate variations. The RMS error for all

the considered error sources yielded a 7 percent variation in pre-

dicted moles of reactants that were burned to match the input pressure

rise data.

As shown in Figure 35, the lean mixtures consistently had

significant amounts of unburned reactants predicted, whereas the richer

mixtures had very little unburned reactants predicted at the time of

predicted reaction completion. In fact, for some of the cases, the

predicted amounts of reactants burned for the rich mixtures exceeded

the available amount of reactants. This overprediction of the amount

of reactants consumed was quite small, less than 5 percent of the moles

of reactants, and was due to random variations in input data. As

explained above, the variations in input data can produce up to 7 per-

cent variations for the predicted moles of reactants burned. The mole

fraction of the total charge which reacted was listed for each exper-

imental run in Table 4. Increasing water induction tended to decrease

the mole fraction which reacts, especially for lean mixtures. The total

reaction time used for Figure 35 was the value predicted from the com-

bustion model. As shown in Figure 35, the effect of increasing total

reaction time results in decreasing the predicted mole fraction which

reacts. Since flame speeds are directly related to reaction time, the

experimental runs with the higher flame speeds had the more complete

combustion. This might be expected since the gas movement due to com-

bustion pressure equalization is more rapid for high flame speeds which

would tend to decrease the amount of unburned gas along the combustion

chamber walls. This trend was also shown in Figure 36, the unburned

volume was seen to decrease as the predicted total reaction time was

decreased. As seen by comparing Figures 35 and 36, at predicted reac-

tion completion approximately 20 percent of the initial reactants which

may remain unburned occupy only 10 percent of the volume at the higher

temperature and pressure conditions existing at reaction completion.

Since the gas near the combustion chamber wall also would be close to

the wall temperature which was a lower temperature than the predicted

compression temperature, the actual volume occupied by the unburned gas

would be somewhat less than that shown in Figure 36 so that an even

larger percentage of the volume was actually burned.

The combustion chamber geometry which was modeled had a pre-

dicted burned volume at a given flame front position that varied for

every .experimental run since the geometry was a function of piston

position. Results for two widely differing experimental conditions

are shown in Figure 37. The percentage of volume burned can be seen to

be relatively independent of the run conditions for flame front travel

greater than approximately 2.5 inches (which corresponded to approx-

imately 90 percent of the total volume being behind the flame front).

The maximum flame travel distance was approximately 2.93 inches for all

cases and corresponded to complete combustion.

As shown in Figures 35 and 36, the fuel lean mixtures tended

to have incomplete combustion predicted for all cases with values vary-

ing from 12 to 23 percent of the total moles of reactants. Since this

was clearly greater than the predicted error bounds, further study was -

made of the phenomena. A wall quench distance of 0.010 inch was input

in the computer program for all fuel lean runs and the results are

shown in Figure 38. The quench distance was defined as a uniform layer

of unburned gas along the walls of the combustion chamber. As shown in

Figure 38, the quench distance of 0.010 inches allowed approximately

5 percent of the total volume to be unburned at reaction completion or

time of complete inflammation. From Figure 36, wall quenching of less

than 0.010 inch was needed to yield the predicted results for all

experimental runs which were fuel rich, i.e., P greater than 1.0.

Also shown in Figure 36 is confirmation of the fact that wall quenching

at a distance somewhat greater than 0.010 inch was needed to explain

the predicted incomplete combustion results for fuel lean mixtures.

The "inflammed" volume plotted in Figure 38 is the volume behind the

flame front which included the volume of the unburned gases along the

walls of the combustion chamber for the portion of the combustion

chamber traversed by the flame front.

The presence of a quench zone in internal combustion engines,

as discussed above, has been verified by other researchers [62,63].

The quench zone 0.010 inch thick led to a small change in the calculated

flame speeds. In Figure 39, the changes in calculated flame speeds are

shown for one of the experimental runs. The change in flame speed

arose due to the increased flame travel needed to produce a "burned"

volume which does not include the volume in the quench layer. The 0.010

inch quench layer increased the calculated flame speed only 5 feet per

second for the average datum point, which was a relatively insignificant-

effect. Figures 26 through 34 do not include the flame speed increase

due to the quench layer effect.

The magnitude of the apparent quench zone was estimated for each

fuel lean experimental run and is displayed in Figure 40. The thickness

of the quench zone for each run was determined as the distance needed

to yield the predicted unburned volume in the quench zone at the combus-

tion chamber walls. The quench zone was assumed to be at the temper-

ature predicted by compression alone and no heat transfer or temperature

gradient corrections were considered. The effect of heat transfer would

tend to reduce chamber pressure; therefore, if heat transfer corrections

were considered the predicted quench distance would be somewhat smaller.

However, the short time available for heat transfer during combustion

and the nonluminous character of hydrogen flames limited the amount of

heat transfer which was possible to about 2 percent of the heat release.

The approximate effect of 2 percent of the heat of combustion heat loss

during combustion would decrease the quench distance only about 0.005

inch. Therefore, an appreciable quench distance was predicted for the

fuel lean experimental runs which yielded low flame speeds which could

not be wholly explained by the effects of heat transfer independent of

the quenching process. The quench distance for stoichiometric quiescent

hydrogen-air mixtures at atomspheric pressure and room temperature is

approximately 0.025 inch between parallel plates [43]. This value com-

pares very well with the predicted values although the effects of high

pressure,turbulence, and high temperature gradients which are present

in combustion chamber do not allow direct comparison of the two condi-

tions with any assurance of accuracy. The order of magnitude is com-

parable, however, just as it was for the hydrocarbon fuel experimental

quench distance determinations discussed in Chapter IV [62,63].

Daniel presented measured values, for a propane fueled engine, of

quench distances up to 0.015 inch. Earlier results for isooctane and

benzene fueled engines were also presented and these showed only 81 to

84 percent reaction completion at time of complete inflammation, as

recorded photographically. Gottenberg predicted quench distances of

up to 0.010 inch for an isooctane fuel from an experimental pressure

vessel where the quench distance calculation was based on unburned

hydrocarbons present after the products were exhausted to a reservoir.

The predicted effect of incomplete combustion increased as the

amount of water vapor inducted increased; these increases in incomplete

combustion led to larger values for apparent quench distance, as shown

in Figure 41. The quench distance was apparently larger, mainly due to

the decreased flame speed at the high water induction rates which led to

less turbulence in the combustion chamber. This was apparent from

inspection of the data, since little effect on quench distance or un-

burned volume was predicted for the increased water injection rates

when the flame speed was higher, as for the fuel rich runs (( > 1.0)

listed in Table 4. Part of the predicted quench distance increase for

high water flow rates could also be due to some local areas of incom-

plete combustion due to the nonuniform distribution of water.

"Knocking" Combustion

Pressure oscillations in the combustion chamber of an internal

combustion engine are widely recognized as the cause of the phenomenon

called combustion knock. The pressure imbalance may be achieved by dif-

ferent means and the intensity of knock varies from almost indetectable

pressure fluctuations to oscillations so large that structural damage

occurs. The larger pressure oscillations cause audible vibrations but

the smaller pressure oscillations are not severe enough to cause audible


The conditions encountered in this study produced a wide range

of pressure oscillations in the combustion chamber. The oscillations

varied from essentially nondetectable to pressure fluctuations which

caused severe audible knocking. The data corresponding to no water

induction are presented in Figure 42. The maximum rate of pressure

rise in the cylinder (dP/dt) was found to affect the amount of pressure

oscillation present as predicted by Anders, Brewster, Curry and others

[33,32,24]. As shown in Figure 42, audible knock occurred for rates of

pressure rise greater than approximately 45 atmospheres per millisecond

for these tests. Since the values for (dP/dt)max were obtained by deter-
mining the slope from pressure-time oscillograms, the values have an

estimated accuracy of approximately 3 atm per millisecond for the

highest values and 1 for the lowest values of (dP/dt). The calculated

values for (dP/dt) for all experimental runs are tabulated in Table 5.

The effect of increasing water induction rate on pressure rise

was shown for each compression ratio in Figures 43-45. The addition of

water to the unburned charge tended to lower flame speeds and therefore

decreased the maximum rate of pressure rise. The curves shown in

Figures 43-45 are dotted to indicate that the slopes of the linear

curves are tentative, since the data quantity and accuracy do not allow

a precise cycle-to-cycle mean determination of slope. This is especially

important if the curves are extrapolated to higher values of water induc-

tion, since it could possibly lead to rather large errors. The effect

of water injection was more pronounced for higher compression ratios and

also for the fuel rich mixtures. This was fortuitous since the high

compression ratio and fuel rich mixture define the worst case for com-

bustion knock. Therefore, water injection appeared to be most effective

at the conditions where it was most needed. In all cases, the water

injection proved to decrease the pressure oscillations which cause knock.

The worst case tested was shown as the top curve in Figure 45, and

even there injection of water at moderate levels stopped the audible

knock as the maximum pressure rise rate dropped below approximately

45 atmospheres per millisecond.

Flame Front Model Results

The calculated reaction completion time from the pressure versus

time oscillograms agreed closely with the output from the ionization

gauge located in the last part of the combustion charge (I.G. No. 2).

The data are tabulated in Tables 5 and 6. Table 5 lists calculated

reaction times and I.G. No. 2 peak output times which corresponded to

the flame front reaching the gauge location. Table 6 lists the time

that I.G. No. 2 first began to respond (triggered) and also the time at

which the signal had decayed to near zero values, which would indicate

reaction completion. All the calculated reaction completion times, as

determined from Figures 16-24, occurred between the triggering and decay

times obtained from I.G. No. 2 and most of the reaction completion times

agreed very closely with the time of peak output for I.G. No. 2 as shown

in Table 5.

The time at which the first ionization gauge (I.G. No. 1) was

triggered is also listed in Table 6. That ionization gauge was located

1.45 inches away from the spark plug and directly across the exhaust

valve recess. The predicted flame front position at the time of

I.G. No. 1 triggering is shown in Figure 46 as a function of total

reaction time. As can be seen there, the assumed spherical flame front

had a larger radius predicted from pressure development than was mea-

sured by I.G. No. 1 triggering data. Therefore, the assumed spherical

flame front shape was not exact, especially for the fast burning mix-

tures. As predicted by Curry, the three-dimensional flame front tends

to expand fastest in the directions corresponding to less containment

by the geometrical boundaries [23]. Figure 46 shows that the effect of

high flame speeds which produced short total reaction times tended to

increase the nonspherical shape of the flame front for the initial

development. Figure 46 indicates that the variations due to cycle-to-

cycle variations and operating conditions gave appreciable data scatter

for flame front shape deviation from the assumed spherical shape. Fig-

ure 46 indicates that the predicted flame speeds must be considered to -

be average values, since the velocity of the flame front was actually

dependent on the position of the flame front segment under consideration.

As shown by Curry [23], the assumed spherical geometry was a better

approximation for the last part of the charge which underwent reaction

than it was for the initial flame development.

Figure 47 displays several predicted temperature profiles across

the combustion chamber as they were calculated at the time of predicted

reaction completion. The temperature difference from first burned to

last burned segments reached values as high as 1700K, where the first

burned segments were the hottest. The cylinder pressure at which

a segment initially reacted is shown in Figure 47 as a function of the

final temperature attained by the segment at reaction completion.

The temperature for each segment is equal to a temperature rise due to

combustion plus additional temperature rise due to compression of the

segment as the reaction proceeds. The predicted effect of water injec-

tion is shown in Figure 47 (Runs 1A and 7A) to cause a decrease of

more than 2000K in the maximum cycle temperatures predicted. Higher

compression ratios led to higher cylinder pressures, higher peak temper-

atures and larger temperature gradients from first to last burned seg-

ments as shown in Figure 47 for representative cases. The temperature

gradient would be needed for accurate calculations of the initial rates

of formation of NO during the expansion stroke. The existence of a

temperature gradient was confirmed in experiments conducted by Muzio

et al. [37] for hydrocarbon fueled engines.


The polaroid photographs of the oscilloscope screen are

referred to as oscillograms and show combustion chamber pressure versus

time traces and also usually show ionization gauge output. Representa-

tive oscillograms are included in this report to illustrate the observed

combustion chamber phenomena noted for hydrogen fueled engines. Figures

48 and 49 show the cycle-to-cycle variations in combustion chamber pres-

sure rise. The slower flame speeds consistently show larger cycle-to-

cycle variations in pressure rise than the fast burning cycles do.

The ionization gauge output traces in Figures 48 and 49 show the corre-

sponding cycle-to-cycle variation in average flame speed. Higher than

average flame speeds corresponded to fast pressure rise rates as was

expected. Figures 48 and 49 both have vertical pressure trace sensitiv-

ities of 100 PSI per division and sweep time rates of 0.5 milliseconds

per division. The oscilloscope sweeps were triggered by the ignition

spark so that all traces began at the time of spark ignition. The

effects obtained by varying water induction rates are shown in Figures

50 and 51 for single cycles at the given conditions. Figure 50 shows

the pressure rise which was obtained for operation at the given fuel

rich, high compression ratio operating condition with no water induc-

tion, and Figure 51 shows the pressure trace at the same operating con-

ditions, with the exception of water induction rate, which was increased.

The increased water induction lowered peak pressures and extended the

flame travel time. The pressure oscillations which cause knock are

present in Figure 50 throughout the expansion stroke, whereas no pres- -

sure oscillations are evident in Figure 51. This graphically shows the

effectiveness of water induction on suppressing hydrogen engine knock.

Figures 50 and 51 both have vertical pressure sensitivities of 100 PSI

per division and horizontal sensitivities of 0.5 milliseconds per divi-

sion for sweep rate. Figure 52 is an oscillogram of a single cycle for

maximum power operation on gasoline fuel at a C.R. of 8.2 with both

ionization gauges output displayed. The results showed high magnitude

ionization gauge response and essentially no pressure oscillation.

The delay time after spark ignition and before significant pressure

rise occurred was much longer than for hydrogen fuel, the flame speed

was much slower, and the pressure rise rate was lower than the compar-

able values for hydrogen fuel. The pressure sensitivity in the ver-

tical direction was the usual 100 PSI per division but the time scale

was expanded for Figure 52 so that one horizontal division was equal

to 1 millisecond. Figure 53 shows a dual trace for the pressure develop-

ment during a single cycle for hydrogen fuel at knocking conditions.

Two pressure transducers were mounted at opposite sides of the combus-

tion chamber to determine if any significant pressure imbalance occurred

at any time during the cycle. The time scale for Figure 53 was 0.2

millisecond per division horizontally and the vertical pressure sensi-

tivity was 200 PSI per division. The No. 1 ionization gauge trace out-

put is also shown in Figure 53 to give an indication of the flame front

development during the initial travel of the flame front. No signif-

icant pressure imbalance was noted in Figure 53 which was obtained at

extreme knocking conditions. This reinforced the predicted flame speed--

data which showed that knocking on hydrogen fuel did not originate due

to detonation of the last part.of the charge to burn.



Knock in hydrogen fueled internal combustion engines does occur

at high compression ratios and water induction does control the knock-

ing with only a slight loss in maximum power. Combustion knock was

caused by the rapid rate of pressure rise (>50 atm/msec) associated

with flame speeds on the order of 300 feet per second. The encountered

knocking was severe and was of a high enough intensity level to cause

eventual structural failure [30]. Some piston pitting occurred on the

test engine, as shown in Figure 54, but was relatively minor since the

engine was not operated at knocking conditions for long periods of time.

The pitting occurred mainly below the exhaust valve port and also to a

lesser degree in the last burned portion of the combustion chamber.

The predicted engine life at the worst knocking conditions encountered

was approximately 16 hours based on the magnitude of the pressure oscil-

lation and engine RPM and the results of recent endurance tests for

"knocking" engines [30].

The mathematical model of the combustion process predicted the

complete combustion of the charge for hydrogen rich mixtures as deter-

mined by pressure rise data. However, for fuel lean mixtures the model

predicted significant amounts of the charge were not burned in the flame

front's first passage. Water injection increased the predicted amount

of incomplete combustion for the fuel lean case. The unburned portion

was assumed to be near the combustion chamber walls in a quenched reac-

tion layer which was estimated to have a thickness of up to 0.030 inch

for high water induction rates.

The results for this particular engine are probably somewhat

conservative since it was operated with a low turbulence and noncompact

combustion chamber. Knocking would no doubt be more severe for a more

compact geometry or a swirl producing inlet since both would lead to

higher pressure rise rates which leads to knock for hydrogen fueled

engines. High speed knock could also be a problem since the higher

engine speed leads to greater turbulence which again leads to high pres-

sure rise rates. Therefore, additional work with different engines is

needed as well as quench layer measurements at low power settings to

extend the results obtained in this study.

The results of this study agreed with the combustion research

work done by Curry and King [24, 6 ]. Knock in this study was found to --

be caused by high flame speeds as the above authors suggested and was

not due to autoignition of the end gas or the establishment of a true

supersonic detonation wave as suggested by other investigators [7,10,22].

Therefore, the methods used to control knock for hydrogen fueled engines

should be based on limiting the rate of pressure rise rather than

minimizing the combustion time. Water induction proved to be very

effective in this study for reducing knock intensity and should ser-

iously be considered for use in engines operating at high power levels.


Additional investigation is needed in the area of low power

setting operation and its effect on reaction completion. Part throttle

and lean fuel mixtures should have relatively large quench distances.

By determining the heat release history for the charge, an approach

might be found to increase reaction completion and thereby improve

engine thermal efficiency.


Table 1. Details of Fairbanks-Morse-Model 45B3-1/8
Four-Stroke Cycle Engine


Fuel delivery

No. of cylinders



Swept volume

Compression ratio range

Engine speed.


Ignition timing

Water injection

Valve timing

Inlet valve opens

Inlet valve closes

Exhaust valve opens

Exhaust valve closes

Diesel engine converted to
spark ignition

Hydrogen gas inducted into
intake manifold

3-1/8 inches

4 inches

30.8 cubic inches

8.2 10.9:1

Approximately 1835 RPM

Water cooled



10-150 BTDC

35-400 ABDC

40-450 BBDC

0-100 ATDC

Table 2. List of Experimental Equipment

DYNAMOMETER STARTING MOTOR General Electric three-phase induction
motor, 220 volt, 60 cycle AC
Model 67A18 Serial Number 4677002

LOAD CELL Electronics and Instrumentation Division of Baldwin-Lima-
Hamilton Corp. SR4, Type UlB,
Serial Number Z-3883 Fifty pound capacity

PRESSURE TRANSDUCERS Kistler 601H Quartz Crystal
Serial Number 17820, 17819, Mounted in
Kistler Model 628C water cooled adaptors

CHARGE AMPLIFIERS Model 504 Kistler Charge Amplifiers
Serial Numbers 858 and 860

IONIZATION GAUGES AC-43 Spark Plugs, Cold Range, 0.030 inch electrode

OSCILLOSCOPE- Tektronix Type 564 Storage Oscilloscope
Serial Number 004161

OSCILLOSCOPE TIME BASE Type 3B3 with external triggering
Serial Number 008037

CAMERA Tektronix Oscilloscope Camera c-12 with polaroid back
Serial Number 007907

AIR FLOW METER Meriam Laminar Flow Element
Model 50MC2-2SF Serial Number 6-13141

VACUUM PUMP Denco Vacuum Pump Serial Number 35870


HYDROGEN FLOWMETER Airco Hydrogen Dual Range Flowmeter
Style Number 805-1604

WATER INDUCTION CARBURETOR Octa-Gane Water Injection Carburetor
Model H-44

SIGNAL GENERATOR Sine Wave Generator manufactured by Electric
Indicator Co., Inc. Serial Number 208510

ELECTRONIC COUNTER Potter Aeronautical Corporation Pottermeter
Model 21B Serial Number 2859

Table 3. Experimental Test Conditions

Run Hydrogen Spark Water Compression
Number Flow Timing Flow Ratio
(SCFH) (Deg. BTDC) (ml/sec)













Table 4. Calculated Reaction Completion, Water Induction
Mole Fraction and Power Output

Mole Fraction Percent Volume Brake
of Total of Total Charge Power Thermal
Run nH20 Charge Reacted Reacted Output Efficiency
(%) (BHP) (%)













University of Florida Home Page
© 2004 - 2010 University of Florida George A. Smathers Libraries.
All rights reserved.

Acceptable Use, Copyright, and Disclaimer Statement
Last updated October 10, 2010 - - mvs