Вы находитесь на странице: 1из 9

25

FUEL-INJECTION SYSTEM CALCULATIONS


By B. E. Knight, M.A.*
A simplified method of calculatinginjection system performance has been developed for the use of applications
engineers. It is particularly relevant to the problem of avoiding secondary injection and dribble.
The calculationof the complete behaviour of injection systems has been programmed for a digital computer.
Agreement with experiment is satisfactory and confirms that all important factors have been taken intok
account. These factors are quoted and the methods used for dealing with compression waves in the pipes are
described in detail. Viscous friction and voids in the fuel are taken into account. As an example of the use of
the programme, an approach to the problem of stroke to stroke variations of delivery in one line is described.
A longer term objective is the reduction of the time required to match injection equipment to engines by
using t h i s and other computer programmes instead of experimental methods, up to the final trial stage.

INTRODUCTION is an obvious approach but presents considerable dif€iculties


A SPECIFICATIONof fuel-injection equipment for a diesel in jerk-pump systems to which this discussion is restricted.
engine, particularly for automotive use, must meet many In such systems the delivery of fuel along the pipe con-
requirements simultaneously. Targets for fuel consumption necting the pump with the injector is by means of com-
and maximum torque against speed are set, together with pression waves. The delivery from the injector is often
standards of noise and smoke which may be implied rather delayed relative to the pump stroke by a time comparable
than stated. In addition to meeting these conditions over with the duration of the pump stroke. Moreover it is
a fuel delivery range of 5/1 and a working speed range now necessary to unload fuel from the pump end of the pipe
reaching 4/1 the specification must also give satisfactory in order to cause an abrupt end of the injection. This may
cold starting and regular idling at under half the lowest cause the fuel column to break leaving voids in the pipe
working speed. which close up and reform at different places in rapid
Many of these requirements tend to be incompatible, an succession.
improvement in one can only be obtained at the expense of The most detailed treatment so far published is that of
another. Rising standards accentuate this difficulty and GiEen and Rowe (r)t. Their method of dealing with pump
increase the attraction of devices giving entirely separate and nozzle-end phenomena w a s as detailed as was necessary
control of each aspect of fuel pump behaviouf. Nevertheless at the start of the present work. A major advance has been
such devices cannot be contemplated if they increase costs made, however, in the method of dealing with compression
against steady downward pressure and if the available com- waves, voids, and viscosity effects in the pipe. In order to
ponents can be manipulated to give a satisfactory s p e d - apply the method it was necessary to use a stored pro-
cation. gramme digital computer. The description of this computer
This is what frequently causes the fuel-injection equip- method of calculation forms the first part of the paper.
ment manufacturer to undertake long and expensive pro- The second topic of this paper is a system of simple cal-
grammes for development of injection equipment for new culations for use by application engineers with particular
engines. reference to avoiding secondary injection or dribble con-
This state of affairs leaves two clear targets for research. ditions. The framework argument for this calculation
One is to find means of reducing the effort and elapsed time scheme uses the DeJuhasz diagram (2) into which simple
required to tailor existing equipments to new engines. The void effects can be introduced if the effect of fuel spaces
other is to devise more readily adapted systems which cost at the pump and nozzle-end are omitted. This argument is
no more than the existing types. This paper deals with the too long to be reproduced here. The argument for the effects
first target. of fuel spaces at pump and nozzle-end does not follow the
The calculation of the behaviour of fuel-injection systems DeJuhasz method.
In this same context the paper of Wassenaar (3) is
The MS. of this paper was received at the Institution on SthJanuary valuable for the experimental evidence and the form of
1960. The report of the meeting, in London,at which this paper was presentation of the data.
presented i s on p . 98.
* Acting Chief Research Engineer, C . A . V . Ltd, Acton. t A numerical list of references is given in Appendix I .
Proc Instn Mech Engrs (A.D.) No I1960-6I
1
\
Downloaded from pad.sagepub.com at PENNSYLVANIA STATE UNIV on March 5, 2016
26 B. E. KNIGHT

The third topic is an example of the use of the computer (3) T o speed up the work of developing injection
programme to provide data not easily obtained otherwise. systems for new engines by substituting calculations on
possible designs for experimental work except in verifi-
Notation cation of the final specification.
a Pipe bore area, mm2. The first objective has been substantially achieved.
aN Nozzle orifice area, mm2. Agreement between calculation and experiment varies from
c 2/(1.01 x I O ~ K= /~ ) space wave velocity, cmlsec.
free very good, a confinnation of the description of the behaviour
K Bulk modulus of compressibility, 14 600 atm. in the pipe, to poor in some of the applications to distri-
M Compression wave pressure, defined at the end of butor-type pump systems with pintle nozzles. A more
p. 28, atm. elaborate but a more easily used version of the programme
Pi, j Pressure, am. is being prepared, which deals much more faithfully with
Qi, j Cumulative flow figure, mm3. distributor-type pumps and delivery valves.
t Time, sec. The existing programme is in use for calculating data for
2, Fuel kinematic viscosity, S. research purposes and an example is quoted below.
X Distance along the pipe, cm. The third objective requires a programme that is easy to
P Fuel density, 0.825 g/cm3. use, achieves a very high standard of reliability, and is
A1 Length of pipe section, mm. sufficiently cheap. The attainment of these objectives is in
At Time interval, sec. sight but will require some time yet. When that stage is
reached the combination of computer investigation of fuel
COMPUTER PROGRAMME system behaviour with the means of spec*ing engine
requirements foreshadowed by the other papers in this
The objects of the computer programme work are as group should enable the overall target to be reached.
follows: Fig. 1 shows diagrammatically a pump element, pipe,
(1) T o confirm that all the significant factors con- and nozzle. The factors taken into account in the computer
tributing to injection system performance are known by programme are listed.
obtainingagreement between calculation and experiment, Distributor' pump systems in which a single pump
notably for pressures in the pipe which are easy to element serves all the cylinders via a distributor connecting
measure. it in turn to each of a number of outlet parts will not be
(2) Subsequently to calculate data for research pur- discussed separately though some results will be given.
poses not readily obtainable by experiment. Examples are The pump and nozzle-end calculations involve con-
the rates of flow and the occurrence of voids in the fuel ventional steady-flow hydraulics, dynamics of moving
spaces in the system. masses, and dynamics of springs, and will not be described

PUMP INJECTOR
(I) TEST FOR OPENING I I (3) CALCULATE LIFT FROM
OR CLOSING OF I PIPE II FORCES O N NEEDLE
DELI~ERYVALVE IN TEN SECTIONS
CALCULATE FLOW
I I DOWNWARDS
SPRING STATIC FORCE
INTO OR OUT OF SPRING SURGE FORCE
PIPE USING DATA UPWARDS
BELOW m i (2) AT EACH OF THE NINE JUNCTIONS -fa)
HYDRAULIC FORCES O N
ANNULUS OUTSIDE SEAT
SELECT ONE OF FOUR ALTERNATIVE -ibj ANNULUS BETWEEN
I FORMULAE (ACCORDING TO PRESENCE
OR ABSENCE OF CAVITIES) FOR PRESSUPLE /(c)
SEAT A N D PINTLE
PINTLE
I WAVES ATTENUATED BY VISCOSITY
I CALCULATE BEHAVIOUR AT THE
I NINE JUNCTIONS

- 1 CALCULATE FLOW INTO NOZZLE


(4) CALCULATE PRESSURE RISE

PORT AREA
1
I
I
I
I N THE WHOLE CALCULATION

15W TESTS
THERE ARE

1350 WAVE CALCULATIONS


OUT OF
. DUE TO BALANCE OF FLOWS
INTO AND OUT OF FUEL SPACE
ALLOWING FOR NEEDLE
DISPLACEMENT AND
FUEL COMPRESSION

5400 POSSIBILITIES
I I CALCULATE PRESSURE LOSS
I I DEPENDING O N LIFT AT
I ALL THE OPERATIONS ARE CARRIED OUT I SEAT
I
I
FOR EACH OF IS0 TIME INTERVALS
MACHINE TIME FOR COMPLETE
CALCULATION I 2 MINUTES
I (Q)IFINAL ORIFICE
CALCULATE FLOW
A T FINAL ORIFICE
I I
I
I
Fig. I . Injection system calculation
R o c Imtn Mech Engrs (A.D.) No I 1960-61

Downloaded from pad.sagepub.com at PENNSYLVANIA STATE UNIV on March 5, 2016


FUEL-INJECTION SYSTEM CALCULATIONS 27

in detail. The most interesting part of the work concerns the pressures given by equation (1) can be said to apply.
the behaviour of the fuel in the pipe.
The biggest single difliculty in the whole programme was
to keep track of the volume and position of voids, which cm/sec2, the mass of fuel pd1/1000, g.
form and close up rapidly, and of the compression waves Thus mass x acceleration = force gives
reflected by them.

TIME INCREASING
NOZZLE END
= 1.01x ~ O ~ U ( Pj -~P-i~, ,j)--fi.iction forces. (2)
i=6 Equation (2) can be reduced in the absence of voids and
friction forces to the h i t e difference form of the usual wave
equation

by substituting for the P’s in terms of the Q s by means


of equation (1). For these units put AZ = 10 mm, A t = 1
sec.
Equations (1) and (2) are those uied whether there are
voids or not. For example, in the absence of friction forces
and when there are voids on both sides of station i then
Pi, = Pi-l, = 0 by equation (1) and consequently by
equatlon (2)
Qi, j+l-Qi, j = Qi, j - Q i , j-1

PUMP END
which says that the fuel between the voids continues in
motion at constant velocity.
Fig. 2. Network of points showing labelling with Trials of this system of calculation without the friction
cumulativeflms, Qi, forces soon showed its merit but since at that timeavoidance
of secondary injections was a major issue it was necessary
to add the friction forces because they reduce the amplitude
The method is illustrated in Fig. 2. The pipe length is of the secondary wave trains.
divided into sections of length AZ mm, often ten but six for The starting point for the analysis that gave the equation
illustration, giving five stations in the pipe excluding the used ever since is as follows: firstly, the injection period is
ends. The time from the start of the calculation is divided too short for the flow to become turbulent. Secondly, the
into intervals of time, A t sec, equal to the time of travel of a flow rate changes only slowly along the length of pipe when
compression wave through one of the sections (without expressed in terms of bore diameers. Thus the flow was
allowance for viscous drag). The volume of fuel passing treated as laminar but with a velocity distribution varying
the station in a time interval is added to a cumulative flow along the length of the pipe. The radial velocities resulting
figure Q,.,j , mm3, where i indicates the station, starting from the change of velocity distribution were treated as
with i = 0 at the pump end, a n d j the time interval. The s m a l l compared with the axial velocitiea.
initial values Qi,moare chosen when everything is at rest
and so that the Merence for adjacent stations i and i+l,
The result, in the units of a
t, x, and u already used, is
to introduce on the right of equation (3) the fiiction term
Q j , o-Qi+l, 0, is the fuel in the pipe section in excess of
just full. Then at any subsequent time Qi, j - Qj+l, is the
fuel in the section in excess of just f d . Thus the pressure
in this section at this time, Pi, atm, is and the integration is taken from the beginning of the
motion. The friction force term in equation (2) is a constant
factor multiplied by the h i t e difference form of expres-
sion (4).
unless Q,.,j - Q i + l , < 0, when Pi, = 0) The integrand as written involves the limit of the integral,
t. In this form the integral has to be recalculated afresh at
Qi, j - Q i + l ,< 0 indicates section not hll of fuel and each time hmal as t changes. This difficulty was overcome
Qi+l,j - Qi, is the volume of the voids.
Equation (1) is an equation of conservation of fuel. The by substituting as an approximation for
equation of motion to be combined with it to give the wave
equation is obtained by considering the a d e r a t i o n of the
fuel at a station as being the average acceleration of the fuel
between the c&tres of the two adjacent sections at which
Proc Zmtn Mech Engrs ( A D . ) N o 1 196M1

Downloaded from pad.sagepub.com at PENNSYLVANIA STATE UNIV on March 5, 2016


28 B. E. KNIGHT

which lies within f 2 8 per cent for ( t - T)/dt fiom 3 to 1000 representing the magnitude of the variable. Thus, for
and averages 86 per cent of the root for ( t - n/Atfiom 1 to example, a line drawn through all the 4’s (labelled cal-
1500. Subsequent experience has shown that the useful culated in Fig. 4) is the graph of calculated pipe line
range of (t- T)/At is 1 to 200 in which the approximation pressure against time. The full code is
is on average much better and can be improved when (1) Pressure drop across nozzle orifice.
necessary. In each of the two integrals that result from this (2) Rate of injection into engine.
substitution a factor containing the limit of the integral t (3) Cumulative delivery into engine.
can be taken outside the integral which can then be accumu- (4)Pipe line pressure at a station in the pipe.
lated step by step with time. (5) Rate of flow at this station.
The first check on this friction force formula was when (6) Nozzle needle lift.
it was found to give the same result for continuoussinusoidal
disturbances in a long pipe as a result quoted by Lamb (4). The computer prints also numerical values for these
Tests are diflicult to devise in which measurements are variables, the Q values and indicators of the presence of
made of the back pressure due to an intermittent and known voids in the pipes.
flow into a pipe. Nevertheless the calculated results of
Fig. 3 are of considerable interest. The curve of back SIMPLE CALCULATIONS FOR INJECTION
SYSTEMS
These calculations originated with the problem of avoiding
secondary injections and dribble, difficulties which at one
time occurred with many applications. The simplest ex-
planation is obtained by considering a pump stroke long
enough to establish steady flow in the system. The flow
rate Q,, mm3/sec, towards the nozzle is determined by the
pump. The pressure PI, a m , in the system is that required
to discharge the flow through the nozzle orifice.
If now a spill port opens at the pump, a wave of decom-
pression travels towards the nozzle reducing the pressure
to zero and reducing the rate of flow by 86-3aPI where a
is as before the pipe bore area. The flow rate behind the
wave front is
TIME-rnr. Q =Q1-86.3aP1 . . - (6)
Fig. 3. Effect of viscosity on back pressure due to The constant 86.3 a, mm3/sec.atm, relates the flow rate
compression wave in a long pipe induced by a compression wave to the pressure. In the
---- N o viscosity, 1.5 xmn bore pipe. simple case in which there is no back pressure it is con-
0.05 S, 1.5 mm bore pipe. venient to substitute for PI from the equation relating flow
--- 0.05 S, 3.0mm bore pipe, pressures multi- through an orifice to the pressure drop
plied by 4.
K = 13 600 am.
p = 0.825 g/cm3.
Q, = 10 950%dPI . . * (7)
where a discharge coe5aent 0.7 is used. Then
pressure against cam angle in the absence of the friction Q = 10950~,1/P,--86*3~P, . . (8)
force is the same shape as the curve of pumping rate against This resultant flow may be towards the nozzle (positive)
cam angle. When the friction term is included a back pres- if aN is big enough or away from the nozzle (negative) if
sure remains afier the pumping rate has f d e n to zero. This aN is small. It is convenient to divide the resultant flow Q,
is a well-known effect. A useful view is that friction in the by 8 6 . 3 ~to obtain the compression wave pressure M
pipe causes a partial reflection back to the pump. No matter developed by stopping the flow. It is convenient to have 11.1
how far the main pulse has travelled there is always some positive when Q is negative. Thus
reflection getting smaller at the pump end because of the
attenuation of the main pulse and because of the increasing
distance the reflection has to travel.
Concurrent work aimed at using the equation (2), without I n either case when the decompression due to the spill
the friction term, by measuring P at two stations an inch reaches the nozzle the needle seats. If M is negative (net
apart, in order to obtain the flow rate dQ/dt by an integra- flow towards the nozzle) the pressure M is developed im-
tion gave results similar to those shown in Fig. 3. mediately and must be below the nozzle-closing pressure
The best confirmation of the use of equations (1) and (2) to avoid a delay in the final closing of the nozzle. Negative
is the comparison with experiment of Fig. 4. The computer M is characteristic of dribble conditions.
output printer has been used to plot the graphs. Each of the If M is positive (net flow away from the nozzle) the
digits one to six represents a variable and is printed out needle seats cleanly and the pressure M is developed first
every time interval at a distance from the edge of the paper at the delivery valve. This pressure reverses the flow and
Proc Instn Mech Engrs (A.D.) No I 1960-6I

Downloaded from pad.sagepub.com at PENNSYLVANIA STATE UNIV on March 5, 2016


19Q961 I ON

6
6
6
6
6
6
5 6
5 6
6
6
6

6
3 1 6
I
6 6

Downloaded from pad.sagepub.com at PENNSYLVANIA STATE UNIV on March 5, 2016


30 B. E. KNIGHT

the pressure is subsequently redeveloped at the nozzle. In wave whatever the length of pipe. Thus the error, if any,
this case M must be below the nozzle-opening pressure to in treating intermediate cases as having the mean M of the
prevent a secondary injection. Fig. 5 is drawn from long and short pipe cases is nil if no nozzle-reopening
equation (9). pressure is forecast but may grow as the forecast reopening
pressure increases.
So far this argument takes no account of fuel spaces at
the pump and nozzle-end. These fuel spaces give rise to two
difficulties. Firstly, the injection pressure and rate are much
more difficultto forecast, and secondly,the nozzle-reopening
pressures become dependent on the unloading volume,
particularly in the secondary injection case. Considerable
further effort has been applied to remedy these deficiencies.
This has resulted in a calculation sheet containing 64 entries
of data, intermediate and final results. This is filled in with
the aid of 11 graphs. The number of graphs is minimized
by using the variables in dimensionless form.
The first part of the table consists of the data and the
conversion of these data to dimensionless form.
The next section of the table calculates the conditions at
the end of the pump stroke where this is long compared with
the wave travel time from end to end of the pipe and back.
In this case wave effects in the pipe are neglected. The pipe,
pump-end and nozzle-end fuel spaces are lumped together.
Then with the aid of the graphs it is easy to follow the rise
of pressure to the nozzle-opening pressure and the subse-
quent rise or fall until the time of the end of the pump
stroke. The pressure and flow at this time are the PI and
Qxto be inserted in the previous argument.
The fall of pressure from nozzle-opening pressure to the
time of spill occurs at low speeds where the pumping rate
cannot keep up with the nozzle discharge. A detailed analysis
of the peculiarities of the main injection under these con-
INJECTION PRESSURE DIFFERENCE-atm.
ditions can be made with the aid of the graphs.
Fig. 5. Dependence of nozzle-reopening pressure on injection The third section of the table is for the end of pump
pressure da#erme and ratio of pipe area to nozzle orifice stroke conditions where the pump stroke is short compared
area with the wave travel time from end to end of the pipe and
Pressure available to reopen nozzle is M+engine back pressure, back. At the nozzle-end the compression wave from the
taking account of signs. Positive pressure gives secondaries. pump and its reflection at the nozzle have to be considered
Negative pressure gives dribble; but without the complication caused by multiple reflections.
Wassenaar 175 a m nozzle-opening pressure: As in the previous section the graphs follow the pressure
x 35 m m 3 unloading. rise to nozzle-opening pressure and the subsequent rise to
+ 55 m m 3 unloading. the end of the pump stroke. No attempt has been made to
cover the variant of this case where the pressure continues
The argument leading to equation (9) holds in the to fall after the nozzle is opened.
absence of a back pressure. When there is a back pressure The fourth section of the table covers the use of Fig. 5
the same argument leads to the same equation with the for obtaining the basic nozzle-reopening pressure which is
back pressure added to each side if PI is treated as the the actual value for the dribble case (M < 0) as explained
pressure difference across the nozzle orifice. earlier. I n the case of secondary nozzle openings in the true
The assumptions made in this derivation are clearly very sense (M > 0) the nozzle-reopening pressure is modified
restrictive. However, a more elaborate argument gives the by the compression of fuel in the spaces at the pump and
same result for pump strokes short compared with the wave nozzle-ends. This is dealt with in the fifth section of the
travel time between pump and nozzle and back. In this table.
case as in the previous one PIis the injection pressure In this case the net flow in the pipe at zero pressure after
difference and Q, the rate of injection. the end of the pump stroke is toward the pump. The
The gap between long and short pump-stroke cases in ‘reflection’ of this flow does not begin until the unloading
this sense is in part closed by observing that if the nozzle volume is taken up, partly by the fuel stored in the pump-
orifice is such as to cause no reflection of the initial com- end volume, and partly by the flow. Then the pressure
pression wave from the pump then there is no secondary begins to build up exponentially sending a compression
Proc Instn Mech Engrs (A.D.) No I 196061

Downloaded from pad.sagepub.com at PENNSYLVANIA STATE UNIV on March 5, 2016


FUEL-INJECTION SYSTEM CALCULATIONS 31

wave to be reflected at the cavity formed at the nozzle by


the flow toward the pump. The exponential rise of pressure
at the pump is terminated by the arrival of the reflection
from the nozzle. This effect plays a major part in reducing
the nozzle-reopening pressure which is still further reduced
by nozzle volume. The effect of the nozzle volume is
enhanced by cavity formation there (if the unloading is
sufficient) because the early part of the flow that would
otherwise charge the volume and raise the pressure is lost
in the cavity. This completes the description of the
calculation.

Table I. Comparison of simple calculation and experiment


(C.A . V.experiments)

I Calc. Calc.
---
Exp. Calc. Exp.
Peak pressure at
nozzle end, atm . 557 402 400 314 370
Period to peak pres-
sure, degrees . 8.9 10.4 - 113 -
Secondarywavepres-
sure at pump end
P ~ , a t ~ n. . 202 104 110 41.5 40
Secondarywavepres-
sure at nozzle end
PN,atm . . 91 73 50 7 20
Dimensionlesspump
stroke (> 1 short
pipes, < 1 long
pipes) . . 2.44 1.89 - 2.1 1 -
Area of pipe bore/
area of nozzle
orifice. . . 8.85 6.55 6.55 7.35 7.35

A. 8 mm plunger, cam rate 0 3 mmldegree, 900 rev/&, 0.2 m m 2


nozzle orifice, 520 x 1.5 mm pipe.
B. 15 rnm plunger, cam rate 0.362 mmldegree, 625 rev/&,
1.08 mmz nozzle orifice, 1220x3 nun pipe, delivery
660 mm3lstroke.
C. 14 mm plunger, cam rate 0.376 -/degree, 625 rev]&,
0.96 m m z nozzle orifice, 1220x3 mm pipe, delivery
660 mm3/stroke.

Three comparisonswith experimentare shownin Table 1.


Two of these were done by an applications engineer on
proposed specifications both of which were shown by the
calculation to be satisfactory as was confirmed by subse-
quent tests. Such good agreement with experiment as is
shown in the table should be regarded as exceptional.
Nozzle reopening pressures to f20 a m should be regarded
as more normal. \OW W wwm
Comparison with Wassenaar's data (3) varies system-
atically from good at high nozzle-opening pressures and
higher unloadings to poor at low nozzle-opening pressures.
particularly with small bore pipes and lower unloadings.
This is in part shown in Table 2, case given by Wassenaar
in sufficientdetail to be calculated by the present method.
Columns 1-5 and 7-10 give the details of the injection P
systems used in Wassenaar's experiments. Column 6 gives fio
00) rcl tan .-
the lowest speed at which s e c o injections ~ were
observed for each system. Column 11 gives the calculated
Proc Insm Mech Engrs (A.D.) No 1 1 9 6 M I

Downloaded from pad.sagepub.com at PENNSYLVANIA STATE UNIV on March 5, 2016


32 B. E. KNIGHT

nozzle-reopening pressure and column 10 the nozzle-


opening pressure as set for the experiments and which must
be exceeded by the nozzle-reopening pressure before a
secondary injection is observed. Agreement with the experi-
ment is indicated when the pressures in columns 10 and 11
are equal. Columns 12, 13, and 14 are added to help in the
search for systematic effects. The Pischinger pressure is FUEL
the peak pressure to be expected at the pump when the
same pump with the same stroke as the injection system
under test is connected to a very long pipe.
Unfortunately most of Wassenaar’s high nozzle-opening
pressure cases are not given in sufficient detail. T o remedy
this deficiency his mean curves for 175-atm nozzle-reopen-
ing pressure (55 and 35 mm3 unloading) have been replotted
in Fig. 5 on an injection pressure basis by adding the
reflected wave pressure to the Pischinger wave pressure
used in his graphs. The Pischinger pressure is in any case
only a valid basis for correlation of long pipe cases. In the
present theory nozzle-reopening pressures of 175 a m
should lie to the right of the M = 161 atm, when the back
pressure is 14 a m as in Wassenaar’s experiments.
The correlation represented by the present method should
go a long way toward reducing the scatter of results in
Wassenaar’s correlation when the reasons for the very
interesting systematic effects are established. These are STEADY RUNNING

being investigated.

EXAMPLES OF APPLICATION DIFFICULTIES


RESIDUAL FUEL IN EXCESS OF JUST FULL -mm3
The case of secondary injections and dribble has already
been considered. It should be noticed that ability to fore- Fig. 6. Stable against unstabb pump chara.cteriS&s
cast these conditions has not led to widespread adoption of
modified parts. Rather the reaction where secondary in- delivery in one stroke, mm3, is plotted against residual fuel
jections are concerned is one of acceptance of the necessity in the system, mm3, in excess of just fill prior to that stroke.
of setting a lower limit on the ratio of the nozzle-hole area The horizontal line shows the pump displacement and
to pipe-bore area. In consequence high rates of injection are where this line intersects the pump characteristic is the
easier to achieve than when smaller hole areas were favoured. working point A for equal deliveries on successive strokes.
The application difficulty in which the computer pro- A small increase of delivery from this working point, that
gramme has proved particularly valuable is that in which is in excess of the pump displacement, causes the residual
deliveries from an injector vary from injection to injection. fuel at the beginning of the stroke shown at B to be below
The most widely known condition in which this arises is the steady delivery value. As drawn in Fig. 6a the loss of
with direct injection engines near idling speeds where the delivery more than compensates for this and the residual
pump displacement rate is insufficient to maintain the fuel at the beginning of the next stroke shown at C is
pressure at the nozzle above closing pressure. This occurs further removed from steady working value, but on the
more widely on small engines because nozzle hole areas are high residual fuel side. This alternation continues with
difficult to scale down. increasing amplitude. The critical slope of the characteristic
It must be emphasized that there are many causes of is 2, the higher slope shown in the upper diagram gives
irregular engine behaviour and each case must be properly unstable operation, lower slopes give convergence to the
diagnosed. steady running condition as shown in Fig. 6b. It should be
It is convenient to concentrate discussion on the exag- noted, however, that for slopes only slightly below 2 the
gerated form of irregular injection known as eight-stroking rate of decay of disturbances is small. Continuous sources
in which every alternate injection is completely suppressed. of disturbance in this case may give rise to irregular
When an injection is missed the residual fuel in the system behaviour that is difficult to diagnose.
rises, the next injection is large and the residual fuel falls. The characteristic of Fig. 6 can only be obtained experi-
The lower residual fuel can be said to be the reason for the mentally with elaborate apparatus, particularly for cases
subsequent mis-injection. where voids are left in the pipe at the end of injection.
As with servo-systems it is futile to ask where the dis- Consequently the computer programme was used. For one
turbance begins, the behaviour depends on the system as particular specification calculated slopes range from 1.5 to
a whole. The key question is illustrated in Fig. 6 where the 2.2 for known unstable conditions. Calculated slopes ranging
Proc Instn Mech Engrs (A.D.) No 1 1 9 6 M 1

Downloaded from pad.sagepub.com at PENNSYLVANIA STATE UNIV on March 5, 2016


FUEL-INJECTION SYSTEM CALCULATIONS 33

from 0.6 to 1-4 for known stable conditions have been versions of the programme so far completed, being repre-
obtained on other systems. sented respectively by Mr. L. Fantl and Mr. E. Cavanagh.
Mr. Cavanagh, now with C.A.V. Ltd, has contributed in
CONCLUSION other ways and is now responsible for writing the third
The computer programme and simple calculation scheme version of the programme. LEO Computers Ltd, operate
provide a range of calculation techniques adaptable to a the programmes on their service bureau.
wide range of circumstances. Their use is illustrated by Mr. K. Whiteman now with the A.E.A. (Atomic Energy
reference to two fuel-injection problems. The task ahead Authority) played an important part at the stage when the
is to learn how to apply these techniques quickly, effectively, methods of dealing with voids and viscosity in the pipes
and economically to the tailoring of injection equipments were evolved.
for new engines. Thanks are due to C.A.V. Ltd, for permission to publish
this paper.
ACKNOWLEDGEMENTS
This research has been going on for many years and thanks APPENDIX I
are due, not only to those who have worked directly on it,
but also to those who have contributed by discussion of REFERENCES
application diaculties and in other ways. In particular the (I) GIFFEN,E. and ROW, A. W. 1939 Proc. Znstn mech. Engrs,
author would like to mention the following: Lond., vol. 141, p. 519, ‘Pressure Calculations for Oil
Engine Fuel-injection Systems’.
Mr. H. C. Grigg who has laboured in this cause for four (2) DEJUHASZ, K. J. 1937 J . Franklin Znst., vol. 223, pp. 463,
years and has just completed the monumental task of 643,751, ‘Transient Phenomena in Linear Flow’.
rewriting the specification of the computer calculation, (3) WASSENAAR, H. 195455 Proc. Auto. Div. Znstn mech. Engrs,
including much more exact descriptions of the behaviour of p. 247, ‘Injection Phenomena in High-speed Diesel
Engines’.
delivery valves and distributor type pumps. (4) LAMB, H. 1945 ‘Hydrodynamics’, First American edition,
LEO Computers Ltd, as they now are, wrote the two p. 652 (Dover Publications, New York).

Proc Insrn Mech Engrs (A.D.) No 1 1960-61

Downloaded from pad.sagepub.com at PENNSYLVANIA STATE UNIV on March 5, 2016

Вам также может понравиться