Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
net/publication/245387210
Article in ARCHIVE Proceedings of the Institution of Mechanical Engineers Part C Journal of Mechanical Engineering Science 1989-1996 (vols 203-210) · February 2003
DOI: 10.1243/095440603762826558
CITATIONS READS
2 5,577
1 author:
Chris Lawn
University of London
63 PUBLICATIONS 1,114 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Chris Lawn on 16 January 2015.
by
C.J.Lawn
Department of Engineering
Mile End Rd
London E1 4NS
May, 2002
Revised: Sept 2002
Abstract:
The performance of a fuel jet issuing through a nozzle and entraining the air for
established theory for such ‘injectors’ based on momentum conservation for one-
dimensional flows. Earlier applications of this theory for injector calculations are
reviewed. It is shown that neglect of property variations introduces only small errors
for the cases that are likely to be considered in such combustion problems. Design
charts are generated in which the ratio of entrained air flow to fuel flow is related to
the area ratio of the two streams at the plane of the nozzle. The ratio is a function of
diffuser), the density ratio of the fluids, and the Mach number of the nozzle.
Predictions are compared with new experimental data for an injector with mixing tube
Reynolds number in the range 2000 to 12000 and shown to agree to within 10% when
2
NOTATION
M Mach number
R Gas constant
K2 +1
β Diffuser and burner loss coefficient parameter, β ≡ , where K2 is given
2
by eq. (5).
K 1′′ − 1
ε Inlet pressure loss parameter, ε ≡ , where K1′′ is given by eq. (2).
2
P1′ − Pa + ρ 1′U 1′ 2
ξ Fuel jet nozzle thrust parameter, ξ ≡
ρ1′U 1′ 2
ρ Density
3
Superscripts:
Subscripts:
0,1,2,3 At the planes of air inlet, fuel jet nozzle outlet, exit of mixing tube, exit of
diffuser
a Ambient
4
1. INTRODUCTION
Over the last sixty years, numerous papers have been published on the performance of
jet pumps, ejectors and injectors. These terms are variously used for devices to carry
out a similar process: the use of a high velocity fluid jet to entrain a larger quantity
from another stream and hence pump the fluid in that stream to higher pressure.
According to the definition adopted in a review by Sun and Eames [1], the term
gas to a higher pressure when that is required. Where the fluid is a liquid, ‘jet pump’
The above review paper concentrates on applications with supersonic primary fluid.
However, the context of the present study is the application of injectors to gas burners
in which the primary fluid is normally supplied at less than 0.1 bar. Some of the early
papers to be reviewed later treat this situation, but it was found that the results were
not easily applied to a design problem where the ratio of injector passage area to
burner flow area was to be optimised for maximum entrainment of air by a fuel jet.
Moreover, the effects of resistance to flow at inlet and outlet to such injectors do not
The following simple calculation based on the same physical principles as earlier
studies of constant area mixing in the injector passage was therefore developed. It
was then checked against an iterative procedure with more accurate treatment of fluid
properties, against calculations in the literature, and finally against experimental data
from a prototype burner. The result is that the method is shown to have sufficient
5
accuracy for all practical purposes, and can be applied with ease to new design
and these results give an indication (without further calculation) of the magnitude of
2. CALCULATION PROCEDURE
The system considered is shown in Fig. 1, where the numerical subscripts denoting
the stages of the injector are shown, and the problem is treated one-dimensionally. A
central fuel jet having stagnation pressure Pf exits from a jet nozzle of area A1′ with
static pressure P1′ , density ρ1′ and velocity U 1′ , and there it is surrounded by air at
static pressure P1′′ (and the other parameters are similarly labelled), having been
drawn from stagnation in the atmosphere at Pa through an orifice having area A0′′ into
an area surrounding the jet A1′′ . Mixing takes place in a duct of constant area
A2 = A1′ + A1′′ . The mixture then diffuses to an area A3 and is discharged through a
resistance coefficient K3
ρ 3U 32 ρ 3U 32
K3 ≡ P3 − Pa + (1)
2 2
The air inlet has a total pressure loss coefficient K 1′′ from stagnation at Pa to plane 1
where
the flow being treated as incompressible because the Mach number is<<1.
6
The fuel nozzle has an efficiency η1′ and the compressibility of the flow within it must
often be recognised, but when the pressure ratio is close to unity, it too can be
1 − η1′
where K 1′ = . The calculation of these air and fuel inlet parameters is described
η1′
in Appendix I.
P3 − P2
ηD ≡ , (4)
1 A
2
ρ 2U 22 1 − 2
2 A3
the overall loss coefficient of the diffuser and burner can be represented by
2
P − Pa A
K2 ≡ 2 + 1 = (1 − η D ) + (K 3 − 1 + η D ) 2 . (5)
1 A3
ρ 2U 22
2
Values of the diffuser efficiency can be obtained from diffuser performance charts,
The problem having been thus defined, conservation of mass and axial momentum
can now be applied to determine the relationship between the entrained air velocity
7
(P′ + ρ ′U ′ )A′ + (P ′′+ ρ ′′U ′′ )A′′ = (P
1 1 1
2
1 1 1 1
2
1 2 )
+ ρ 2U 22 A2 (7)
Solution of the above set of equations would be straightforward if the density ρ 2 were
not a function of the composition, temperature and ratio of the primary and secondary
(i) the steady flow energy equation, again in the low Mach number limit, assuming
The above equations can be solved iteratively (Pintossi, [3]) by taking the properties
at the outlet of the mixing tube to be the same as in the entrained air stream, as a first
approximation, on the basis that for combustion purposes the secondary flow will be
an order of magnitude greater than the primary. (A primary flow of methane has a
molecular weight only 55% of that of air, although this may be alleviated by up to
15% by temperature reduction in the nozzle. Thus if the air:fuel ratio is 10, the density
of the mixture may be about 6% in error.) Having calculated this first approximation
to a mass flow ratio, the properties can then be adjusted and the calculation repeated
8
However, as will be seen from the results, in practice the first approximation is
sufficiently accurate for most purposes, and moreover it allows an explicit solution.
P1′ − Pa + ρ 1′U 1′ 2
Dividing equation (7) by ρ1′U 1′ 2 A1′ , putting ξ ≡ , using equations (5)
ρ1′U 1′ 2
and (6) and noting that Pa is eliminated from the other terms,
2 2
1 − K 1′′ ρ 1′′ U 1′′ A1′′ 1 + K 2 ρ 1′U 1′ A1′ + ρ 1′′U 1′′A1′′ ρ 1′′ A1′ + A1′′
ξ + = (11)
2 ρ 1′ U 1′ A1′ 2 ρ 1′′U 1′ ( A1′ + A1′′) ρ ′
1 A ′
1
2
U ′′ U ′′ β
{ }
ζ α (α + 1)ε + α β 1 + 2αβ 1 + − ξ (α + 1) = 0
2
(12)
U 1′ U 1′ ζ
A1′′ ρ ′′ K +1 K ′′ − 1
where α ≡ ,ζ ≡ 1, β ≡ 2 , ε ≡ 1 . If the nozzle is unchoked, there is no
A1′ ρ1′ 2 2
further loss of accuracy in putting P1′ = P1′′≈ Pa , so that ξ = 1 . Solving the quadratic
m1′′ U ′′
= αζ 1 (13)
m1′ U 1′
property differences between the induced airflow and the fully mixed flow are
ignored,
=
2 2 2
{
m1′′ − αβ + α β + α (α + 1) + α β {ξζ (α + 1) − β } } (14)
m1′ {(α + 1)ε + αβ }
9
3. COMPARISON WITH VARIABLE PROPERTY SOLUTIONS
For a particular area ratio, α=277, an outlet resistance characterised by β=0.661, and
no inlet resistance, the effect of the pressure ratio Pf/Pa on the mass flow ratio was
examined for values of ζ and ξ corresponding to methane, air and propane, with the
There is a small loss of performance as the pressure ratio increases, with a distinct
change of gradient in the mass entrainment ratio at the pressure ratio where the
Also shown on Fig. 2 are the results of the iterative calculations using the spread-
sheet developed by Pintossi [3], which was for the choked condition only. With air as
the injection fluid, even at pressure ratios leading to temperatures some 50K lower
than ambient in the choked jet, the difference in the two sets of results is negligible.
However, the results for methane in the approximation of constant properties are
some 3% too high, while those for propane are some 3% low.
The first calculations similar to those in the previous section were carried out by
Keenan and Neumann [4]. They had no diffuser or outlet resistance and studied
constant area mixing with injection of air from a choked nozzle over a range of α up
to 1000. It was shown that optimum injection in their configuration occurred with the
10
nozzle outlet slightly upstream of the throat. However, the difference in entrainment
from the case with the injector at the throat was only a few percent.
Two area ratios were chosen for comparison of their data with the present
Essentially the calculations are in agreement, while their experimental data are some
10-15% lower. The asymptotic solution for large α and β=1 developed in Appendix II,
m1′′
= 2ξζα (15)
m1′
is seen to be only some 7% too high for an area ratio of 1024, and provides a useful
Von Elbe and Grumer [5] developed calculations which were specifically for low
pressure methane mixtures injected into burner tubes. They examined the effect of
having the injection plane upstream of the venturi entrance, as well as the effects of
tube length, gas flow rate, ignition and orientation of the flame. Their calculations
contained two empirical parameters to characterise the momentum fluxes, and these
were apparently sufficient to describe all the data. However, the significance of these
11
Another major contribution from over 50 years ago was that of Kastner and Spooner
association with area ratios up to 1110. Their interest was in the performance of their
was a variable inlet loss coefficient that is not easily translated to ε: in others there
was variable outlet loss. They showed, however, that the optimum mixing tube length
is about 7.5 diameters, and with this length, they confirmed that the optimum nozzle
In a very comprehensive set of calculations, Francis [7] developed the theory for gas
burners, and compared his predictions with the experimental data of Kastner and
Spooner [6] and Prigg [8]. Good agreement was found with the former, but Prigg’s
data for mass flow ratio for small area ratios tended to fall well below predictions.
However, these predictions were based on assumptions about the ratio of mixing tube
to burner port area, and the results were therefore inconclusive. Francis’ treatment is
identical in principle to the present one, but this one is easier to appreciate and apply.
The last classic contribution to this field was that of Kentfield and Barnes [9], but
their emphasis was on the performance of ejector pumps of small area ratio (<5).
Similar methods have been codified into design procedures in an ESDU data sheet
12
5. DESIGN RESULTS
Since the stoichiometric ratios of air to fuel for the gaseous fuels of interest are >14,
only mass flow ratios of this order are to be considered further. Figure 3 shows that,
for a density ratio corresponding to methane injected at low pressure (ξ=1, ζ=1.79), a
modest degree of loss in the diffuser (β=0.793) and no inlet loss (ε = -0.5), an area
ratio of about 50 is sufficient to achieve this mass ratio. Higher injector pressures in
the choked regime give slightly improved performance, but in practice, this is
seen from Fig. 4 to reduce the entrainment ratio. This mitigation arises because, for
the same momentum, the greater density corresponds to a greater mass flow through
the nozzle. The values of ζ chosen in Fig.4 correspond to the densities of methane, air
and propane in the jet when the driving pressure is Pf/Pa=1.05. It is the decrease in
density ratio as the nozzle injection temperature drops, and then the pressure rises
(after choking), that is the reason for the decrease in the performance for all three
Figure 5 illustrates the very dramatic effect of outlet resistance, either through
only a very small resistance and almost perfect diffusion, so that β is only a little
greater than 0.5, mass ratios of over 100 are possible with area ratios α>250. With
perfect diffusion to an area 4 times that of the mixing tube and then a loss coefficient
of 150, so that β=0.793, a mass ratio of 37 is predicted. But with a loss coefficient of
13
The outlet loss is referenced to the dynamic head in the full area of the mixing tube,
rather than the air inlet, but at large area ratios, ε and β are essentially equivalent, so
that the performance depends just on the sum of these parameters. With high inlet
loss, the device becomes an ‘ejector’ and simply operates at lower pressure than in the
‘injector’ mode when the outlet resistance is dominant. Thus Fig. 6 shows essentially
the same dependence as Fig. 5, with an increment of 2.5 in ε (from –0.5 to 2.0)
reducing the mass ratio to just below the levels calculated for an increment of 1.66 in
A problem of practical interest is one where the area of the burner nozzle, which
determines the fuel throughput, is held constant, and so is the burner outlet area,
which determines the velocity of the fuel-air mixture for a given equivalence ratio.
The objective is to optimise the mixing tube diameter. It has been seen from all the
preceding figures that increasing the mixing tube to nozzle flow area ratio increases
the mass ratio. However, as the mixing tube area approaches that of the outlet, the loss
of the dynamic head in the jet from the outlet becomes more significant, and so there
For the example chosen, a burner to nozzle area ratio of 1736, the minimum occurs
for mixing tube: nozzle area ratios of about 100, and with a burner loss coefficient of
150, equivalence ratios of less than 0.8 could not be achieved without a very efficient
diffuser.
14
6. COMPARISON WITH EXPERIMENTS
A small-scale laboratory burner was constructed with the essential features of the
injection equipment design shown in Fig. 1. Three configurations were tested having
mixing tube length:diameter ratios of 7.5 (‘long’) and 2.5 (‘short’), and no mixing
tube at all. The throat of the air inlet and the mixing tube (when fitted) had a diameter
of just 20 mm, and with a nozzle internal diameter of 1.2 mm, the area ratio was 277.
Diffusion after the mixing tube (or throat) was to a diameter of 50 mm over a length
In the first set of ‘calibration’ experiments, the actual burner section was omitted and
light-weight float. The top of the rotameter exhausted to the laboratory. The
calibration of this rotameter was checked against a Fisher TM-24X device with forced
flow. The purpose of the rotameter was to calibrate the suction pressure in the throat
of the air inlet against the air flow rate, so that this pressure could subsequently be
used to measure the flow when only the burner was in place. With a heavier float, the
resistance would have prevented the flow rates of interest from being obtained.
However, because the pressure across the rotameter had to support the weight of the
float in a passage of varying dimensions, it varied much less with flow rate than a
fixed resistance. Thus, this configuration was one where the loss coefficient K3 varied
with flow rate, over the range from 2500 to 700, so that β was between 32 and 10. In
all these cases, the diffuser efficiency and the exit pressure loss coefficient were
calculated from pressure measurements and the resulting value of β fed into the
15
Because of inefficiency in the nozzle, which was constructed from 40 mm of
hypodermic tube projecting from a larger diameter one, the nozzle did not quite choke
at the highest overall pressure ratio of 2.6, and so ξ=1 throughout this first set of
experiments. The measured flow rate through the nozzle was the primary independent
variable that was fed into the calculations of entrainment ratio, so as to predict the
entrained airflow.
All three mixing tube arrangements were tried with air injection through the nozzle,
with the results shown in Fig. 8. Also shown are data for the inlet restricted by a
circular plate to a 1 mm annulus which gave an inlet to throat area ratio of 0.5 and
ε=0. The pressure ratio range across the nozzle resulted in the density ratio ζ varying
from 1 down to 0.76. Finally, there are two results for methane injection, with
ζ =1.79.
Very satisfactory agreement of the predicted entrainment with that measured is seen
in all the data, apart from those with no mixing tube. The experimental results with air
injection and the optimum length of mixing tube (according to Kastner and Spooner
[6], 7.5 diameters) are actually better than predicted by about 10%, and this is a
mixing tube to just 2.5 diameters only degraded the performance by about 10%,
In the second set of experiments, the burner was in place and firing methane, the
nozzle flow being a mixture of methane and a diluent (usually CO2 to simulate a
16
biogas, but sometimes N2), except in one case when it was air and there was no
combustion. The density ratio ζ for the methane/diluent injection varied from 1.5 to
0.9 due to falling injection temperature (resulting from acceleration in the nozzle) and
composition. With the pressure ratio across the nozzle going up to 2.7, a few
conditions were choked, with ξ rising to 1.23. The Reynolds number in the mixing
tube ranged from 2000 to 12000, In some cases, burner and flame pressure losses
were fed into the calculations, together with the pressure rise up the diffuser. In other
cases, the averages of the results derived from the measurements (K3=150 and ηD=0.8)
were assumed, thus allowing predictions for the whole burner, rather than just for the
entrainment system. It is the predicted entrained airflows that are again compared
The experimental results from this set divide into four groups. The one with the long
mixing tube and air through the nozzle agrees precisely with the theory. Those with
the long mixing tube and methane/diluent lie some 10% lower than predicted, while
one set with the short mixing tube and the inlet restriction are 25% below. (This last
result is not consistent with the earlier one in the calibration experiments, possibly due
to uncertainty in the throat pressure calibration for the tight inlet restriction). With no
17
7. CONCLUSION
The indications of the comparison with experimental results in the range of Reynolds
number from 2000 to 12000 is that the mass entrainment of an injector, consisting of a
nozzle set with its outlet at the throat of a well-profiled air inlet and feeding a mixing
tube of 7.5 diameters, can be predicted to within 10% for a wide range of fluid density
ratios and Mach numbers in the nozzle. Failure to treat property variations accurately
The area ratio was 277 for these experiments and dependence on this parameter has
not been tested. However, this dependence is the same as that found theoretically and
experimentally by Keenan and Neumann [4], who tested area ratios up to 1100.
The outlet loss parameter was varied over a wide range in these experiments, although
it was ensured in the calculations that the exit of the mixing tube had approximately
the measured pressure relative to the injector inlet. There is therefore some confidence
in the validity of the results that have been generated by systematic variation of the
inlet and outlet loss parameters for preliminary design purposes. Moreover, design
calculations for a geometrically similar injector and burner of twice the scale of the
apparatus used in this study (with Reynolds numbers in the range 6000-20000) have
18
ACKNOWLEDGEMENT
iterative calculation to account for variable properties, and in surveying the literature,
is gratefully acknowledged.
REFERENCES
1. Sun, D-W and Eames, I.W, Recent developments in the design theories and
Press.
5. Von Elbe, G. and Grumer, J, Air entrainment in gas burners, Ind. and Eng,
design of the air ejector employing low pressure air as the driving fluid, Proc.
ejectors and jet pumps, Inst. Gas Engineers Journal, 1964, June, 399-414.
19
8. Prigg, J.A., Aerated burners, London and Southern-Junior Gas Association,
1955.
10. Engineering Science Data Sheet 84029, Ejectors and jet pumps – design and
20
Appendix I: Air and Fuel Inlet Loss Coefficients
Referring to the air entry in Fig. 1, and assuming incompressible flow, the static
1
P0′′ = Pa − ρ 1′′U 0′′ 2 (A1)
2
A0′′ A′′
P1′′− P0′′ = ρ1′′U 0′′ 2 1 − 0 (A2)
A1′′ A1′′
However, the fuel supply must in general be considered compressible and, with the
T01′ − T1′
η1′ ≡ , (A4)
T01′ − T1′s
γ
1 γ − 1 2
−1 γ −1
P1′ = Pf 1 − 1 − 1 + M 1′ (A5)
η1′ 2
ρ 1′U 1′ 2
Pf = P1′ + (A6)
2η1′
Pf − P1′ 1
so that K 1′ ≡ −1 = −1 (A7)
1
ρ1′U 1′ 2 η1′
2
21
Pa
1 −
P1′ − Pa P1′
ξ≡ + 1 = +1 (A8)
ρ 1′U 1′ 2 γM 1′ 2
In evaluating ξ and the Mach number for unchoked nozzle conditions, it is sufficiently
22
Appendix II: Asymptotic Solutions
The general solution of equation (12), which describes the injector velocity ratio if the
property differences between the induced airflow and the fully mixed flow are
=
2 2
{ 2
}
m1′′ − αβ + α β + α (α + 1) + α β {ξζ (α + 1) − β }
(A9)
m1′ {(α + 1)ε + αβ }
If ζ ≈ 1 and there are no inlet losses so that ε = -0.5, then in the limit of α → ∞ for
β ≠ 0 .5
m1′′ ξζα
= (A10)
m1′ β − 0 .5
This limit is further simplified if β=1, signifying that there is no diffusion and no
resistance at outlet, to
m1′′
= 2ξζα (A11)
m1′
On the other hand, if β=0.5, the case of perfect diffusion but no resistance,
m1′′
= ξζα (A12)
m1′
23
List of Figure Captions:
(α=277, β=0.661, ε=−0.5, ξ varying, ζ varying for a given gas only due to nozzle exit
after Choking
Pure Methane Jet with Mixing Tube Area (Burner: Jet Area Ratio=1736 fixed),
24
Fig 8: Air Entrainment by the Injector
with Simple Theory (CH4 with CO2 Dilution except where indicated)
25
Combustion
Chamber
Burner
A3
Diffuser
A2
Mixing Tube
′
A1
″
A1 Injector
″ Inlet
A0
Pf
26
27
28
29
30