Part 5
1 INTRODUCTION
This part of the Manual is concerned with the design of gas atomised Venturi scrubbers
for the removal of fine particulates and mists from process gas streams. The design of
Venturi ejectors and scrubbers for gas absorption is not considered.
As noted in Part 2, Venturi scrubbers are wet collectors which rely on the high kinetic
energy created by accelerating a gas through a constriction to give good inertial
collection onto droplets distributed in the gas stream. This acceleration is achieved at the
expense of gas side pressure drop, which may be in excess of 1500 mm w.g. (14.7 kPa).
The liquid generally leaves the scrubber as a fine particulate laden spray which can be
removed by a low energy separator, such as a cyclone or wave plate demister. In this way
the Venturi can be seen merely as an agglomerator in which particles too fine for removal
by conventional means are attached to larger droplets in order to ease their subsequent
collection. A typical arrangement is shown in Figure 1. The efficiency of such a system is
therefore dependent both on the efficiency of the initial contact process between the
liquid and the particles and on the efficiency with which the demister operates. In a well
designed system the latter will be 100%. In severe cases where this is not the case, the
inclusion of a demister pad as a third stage will usually rectify the situation. Emissions,
therefore, are generally the result of penetration through the scrubber, and this is where
the problems of design are focused.
(1)
where d and dD are the respective sizes of the particle and of the droplet collector, vr their
relative velocity, p the particle density and g the gas viscosity. Inertial collection is
therefore favoured by large, dense particles, small droplet collectors, and high relative
velocities.
An overemphasis on the maximisation of this process is a major reason for the many and
varied designs of Venturi scrubber available on the market. It must be stated at the outset
that such attempts are largely futile. Inertial impaction is energy intensive: the efficiency
which can be achieved in a given situation is determined principally by the energy
expended in the contact process, not the fine detail of how this is brought about. For
practical scrubbers, in which the majority of the energy is supplied in the form of gas
phase pressure drop, this means that the Venturi geometry and liquid and gas flow rates
(on which so much emphasis is often laid) are largely secondary considerations: they
affect the Venturi`s performance only in that they determine the pressure drop.
The practical consequence of this for Venturi design is that it separates the design
problem into two distinct parts: prediction of the required pressure drop, and the
mechanical design of a unit to generate this required pressure drop. These aspects will be
considered separately.
A third aspect which requires consideration is the process design of the scrubbing system
as a whole. All scrubbers require a considerable amount of ancillary equipment,
principally for liquid handling and treatment, but also for gas pre-treatment. An extensive
discussion of such equipment is given in Part 2 Section 2 of this manual. The design of
such equipment can have a significant influence on the mechanical design of the
scrubber. For example, the effectiveness of the liquor recycle clarification may determine
the method of the liquid injection; pre-quenching certainly will change the gas volume
and hence the physical size of the Venturi.
The emphasis of this report is on the prediction of the required pressure drop. This
enables selection of the fan which, once installed, defines both the pressure drop range
over which the scrubber can operate (hence the scope for changing the Venturi`s
performance), and the operating cost. Concise procedures for predicting the pressure drop
are presented in Section 2.
The process design is considered in Section 3. Scrubber psychrometry is discussed to
enable specification of both the gas volume to be handled by the Venturi and the load on
the fan.
The mechanical design of the scrubber requires specification of the geometry of the
Venturi, the method of water injection, and the liquid flow rate required to achieve the
desired pressure drop. This is discussed in Section 4. Any number of specification
combinations are possible. Generally, these will work equally well provided adequate
liquid distribution is achieved in the scrubbing section. However, predictive models are
not sufficiently advanced to enable a unit to be designed to a specific pressure drop.
Venturi scrubbers are therefore usually designed with the aid of either past experience or
simple predictive models to give approximately the correct pressure drop. This enables
the liquid flow rate to be fixed, and hence the size of the liquid handling and treatment
plant to be established. Once installed the scrubber may be "fine tuned" to the required
pressure drop. For small changes this may be done by altering the liquid to gas ratio.
Larger changes may require alteration of the throat geometry to increase or decrease the
gas velocity. Such engineering considerations are generally the province of the equipment
vendor. However, some discussion of these is given to enable the reader to complete the
design themselves if this is desired.
2 CONCEPTUAL DESIGN
2.1 General
2.2 Scoping design
2.3 Design by scaling
2.1 General
The major consideration in the design of Venturi scrubbers is the required pressure drop.
This defines both the efficiency which can be achieved in a given application and, to a
large extent, the operating cost. From a practical standpoint it also defines the size of the
fan. The correct selection of this is crucial since, once installed, the performance
characteristics of the fan dictate the scope for altering the performance of the scrubber.
Whilst some "fine tuning" may be possible, subsequent changes from the design point of
the fan (eg. demanding a higher pressure rise) may be detrimental to both the efficiency
of the fan and the volume throughput which it can deliver.
The pressure drop required strongly depends on the fineness of the material to be
collected: the finer the material, the higher the pressure drop required. Indeed, the
relationship is essentially exponential. An unfortunate paradox results from this. It means
that the efficiency of a Venturi scrubber is limited by the size of the particles in the last
few percent of the fine tail of the size distribution. Since in practical situations it is
virtually impossible to measure this tail with any confidence, design techniques which
require precise size information are largely unworkable. The traditional grade efficiency
calculation approach is therefore not to be recommended for Venturi design. It is
included here solely for the purposes of scoping design and is aimed at providing an
approximate indication of the required pressure drop. The procedures are given in Section
2.2 below.
Accurate specification of the required pressure drop can only be achieved on the basis of
pilot plant or, following SPS research, miniscrubber tests. This latter approach is given in
Section 2.3. Clearly, such tests are only possible when dealing with an existing process.
Conceptual processes require the grade efficiency approach, and therefore demand a
significantly greater design margin.
by the equipment vendor and sevrral major suppliers operate mobile pilot facilities.
Similar trends are expected in Europe with the general tightening of control legislation.
2.3.1 Miniscrubber prediction of pressure drop
3 PROCESS DESIGN
3.1 General
3.2 Pre- and post quencher design
3.3 Physical and thermodynamic properties
3.1 General
An important consideration in the selection of any wet deduster is the need for a
comprehensive support system. An example of the complexity which may be involved is
given in Figure 4. In addition to the basic scrubber and entrainment separator there is
normally a slurry treatment device (most often a simple settling chamber), a final liquor
filtration plant to reduce the waste to a disposable form, and provision for recycle of the
treated scrubbing liquor to reduce the water usage. Whilst these ancillaries do not affect
the scrubber performance per se, poor liquor quality is often a cause of increased
maintenance, especially in spray irrigated scrubbers, and the overall size of the liquor
plant may limit the scope for controlling the scrubber pressure drop on liquid to gas ratio
(See Section 4.1). Wherever possible, it is advised that clear scrubbing liquor should be
used. This is to ensure that nozzles remain free from blockage and are protected against
corrosion. It is recognised however, that there are frequent cases where untreated liquor is
recycled to scrubbing plant. In these instances careful attention should be paid to the
selection of materials and sizing of equipment to ensure low maintenance costs and
consistent performance.
psychrometric chart for air-water is given in Figure 5; charts for other gas
systems and operating conditions can be obtained from SPS on request.
< 0.5
500 - 800
0.5 - 1.5
800 - 1200
1.0 - 2.0
These times are however, very subjective. For example, on some steel plants,
blast furnace gas is cooled from 1200C in a primary quench with virtually no
hold-up, using a liquor rate of approximately 10 times the evaporative loss.
PT
total system pressure (static pressure), kPa
PV
partial pressure of water vapour is given by Y PT/(Y + MV/Mg)
Y
gas humidity, kg vapour/kg dry gas (Figure 5).
For complex gas mixtures, such as combustion products, the gas density may be
determined using a tabular method described in Mini Manual 2 Section 7.2 (Bahu et al,
1986). However, for most cases this additional accuracy is not warranted.
(4)
where Tr is the reduced temperature relative to the critical temperature, Tc, (Tr =T/Tc) and
the superscript "*" refers to reference conditions. For water Tc = 647.3 K and @ 273 K
= 2501 kJ/kg.
(6)
4 MECHANICAL DESIGN
The mechanical design of Venturi scrubbers is the province of the equipment vendor and
relies heavily on past experience in similar applications. Different manufacturers tend to
concentrate on different parts of the market, for example: metallurgical fume;
incinerators; flue gas desulphurisation; and so on, thus care is required to select a vendor
with appropriate experience. Failure in this regard generally results in lengthy
commissioning and endless maintenance problems. The mechanical design embraces a
number of considerations:
Venturi geometry;
the required pressure drop and method of pressure drop control;
design of the entrainment separator and liquor handling facilities;
erosion and fouling (Venturi, entrainment separator, and liquor utility);
materials of construction;
performance monitoring facilities;
maintenance.
Tradition plays a part in the selection process: European practice is to use circular
Venturis on furnace applications, whereas North American practice often favours
rectangular throats. In general, higher liquid rates and lower gas velocities are
encountered in North American scrubbers. Such differences are ultimately immaterial
since it is the operating pressure drop which governs the collection efficiency. Claims
that units are designed for "optimal" or "enhanced" performance, should be viewed with
suspicion.
4.1 Selection of Venturi geometry
4.2 Sizing of Venturis for a given pressure drop
4.3 Entrainment Separation
4.4 Materials of Construction
This method is therefore generally restricted to wetted wall, fixed throat scrubbers. This
geometry is by far the most amenable to mathematical treatment, and hence pressure drop
models may be used with some confidence to predict required changes (see Section 4.2).
Excessive liquid turn-down, however, may result in inadequate throat coverage with a
resulting loss of performance. A practical minimum for scrubbers handling over 5 m3/s
gas is 0.3 m3 water per 1000 m3 gas. Lower values may be used with smaller scrubbers.
Demands on the liquid handling system generally restrict liquor rate changes to 25%.
Control on throat area.
Pressure drop, both dry and wet, generally vary with the square of the gas velocity.
Scrubbers are therefore inherently more sensitive to changes in gas velocity than changes
in liquid rate. This has two consequences: it is better to control pressure drop on gas
velocity by changing the throat area; inadvertent changes in gas velocity are difficult to
accommodate by changes in liquid rate. Scrubbers designed for varying gas rates are
therefore universally of the variable throat type. Several examples are shown in Figure 7.
Butterfly dampers are simple in construction and actuation. They are prone to
erosion and have severe non-linear sensitivity. They can also cause liquid
maldistribution in spray irrigated scrubbers. They may be used with both circular
and rectangular throats but require adequate duct length before the entrainment
separator to enable pressure recovery.
Hinged walls are used with rectangular (prismatic) Venturis. Again they are
simple in construction and actuation. Excessive throat opening can, however,
lead to poor liquid distribution in wall irrigated scrubbers.
Falling plug ("plumb bob") designs give proportional control and form a positive
seal when fully closed. However, they require upstream spray irrigation to
prevent fouling.
Rising plug designs again give proportional control together with increased
fouling resistance. They may be used with either spray or wetted wall irrigation.
They fail open.
The actuation may be manual or automatic. Manual adjustment is generally used where
set-point changes are required. Fine control of pressure drop can then be made on the
liquid rate. Automatic adjustment is preferable for non-steady processes. However,
independent monitoring of the liquor flow is required since falls in the flow (often due to
lime scaling) can result in decreased collection efficiency through inadequate atomisation
even though the overall pressure drop is maintained (see SPS RR 40 Section 3.3.3).
separation and all quantities can be described by local axial mean values. Clearly, this is
more applicable to circular than rectangular Venturis. Furthermore, rectangular units tend
to have steeper converging and diverging angles, so flow separation and two-dimensional
effects are exacerbated. For practical purposes this is a second order problem as
rectangular units tend to be of the adjustable throat type or, if not, are simple enough in
construction to allow inexpensive modification to give the desired pressure drop at the
commissioning stage. The approach adopted here is, therefore, to treat rectangular units
as circular on the basis of their hydraulic equivalent diameter. Once a geometry has been
established a more rigorous pressure drop prediction can be obtained using the Boll
model presented by Yung et al (1977). This is too lengthy to be reproduced here.
START
In p u t d e s ig n p r e s s u r e d r o p , Po a n d s y s t e m
p a ra m e te rs
S e le c t
L i q u i d / G a s r a t i o Q i/ Q g
D iffu s e r a n g le 2
D r o p le t v e lo c ity r a t io
E s tim a te d ry p re s s u re d r o p fa c to r R
E s tim a te d e s ig n w e t p r e s s u r e d r o p P
END
D t e r m i n e t h r o a t v e l o c i t y vg t
D e t e r m in e d r y p r e s s u r e d r o p p
C a lc u la t e o v e r a ll p r e r s s u r e d r o p P
= P
D e te r m in e
T h r o a t l e n g t h lt
T h jro a t d ia m e te r d t
No
R e c a lc u la te R
Yes
o
o
than 60 m/s)
gas at elevated temperature without pre-quench
gas absorption required (Q /Q up to 5) l g
large gas flows ( less than 7 m3/s) to ensure good throat
coverage, esp. wetted wall Venturis
o
spray irrigated Venturis (dictated by nozzle requirements)
Select
o
anticipate high turn-up for pressure drop control
low
o
once-through liquor circulation
3. Select diffuser angle,
o
o
o
Typical diffuser half angles are in the range 3 - 15 with typical choices for
conventional applications in the range 5 - 8. Classical long throat Venturi
scrubbers have a diffuser half angle of 3.5, selected as the critical value to
prevent flow separation. Prismatic scrubbers and scrubbers with plug type
dampers tend to have much steeper angles, typically to 15, and hence shorter
diffusers, in the latter case to minimise the problems of supporting the actuator
arm. Steeper angles can cause "jetting" of the scrubbing liquor, resulting in bypassing and impaired efficiency. Butterfly dampers can exacerbate this problem
by concentrating the flows to one side of the scrubber. Short diffusers should
therefore be followed by at least two duct diameters of straight ducting before the
entrainment separator. Shallower angles are encountered, but the suggestion is
that this merely increases the overall pressure drop by unproductive wall friction.
4. Select droplet to gas velocity ratio,
A significant proportion, if not the majority of the pressure drop in a scrubber
represents energy expended in atomising and accelerating the liquid. Clearly, if
the Venturi throat is sufficiently long, then the liquid would attain the gas throat
velocity. However, as inertial collection depends on the relative velocity between
the particles (ie. gas) and the liquid, there is little virtue in such designs. In the
author`s experience a practical ratio of droplet to gas velocity, when used in
conjunction with the above assumptions, is 80%. This gives a reasonable balance
between throat length and gas velocity.
With the constraining parameters set, the gas velocity for a given pressure drop
can be determined by iteration.
5. Estimate the dry pressure drop factor, R
Allowance must be made for the contribution of the dry pressure drop to the total.
The fraction which this represents depends on both the throat velocity and the
liquid to gas ratio, and thus must be determined iteratively once the throat velocity
has been established. The initial estimate should be in the range 0.1 - 0.3,
preferably between 0.15 and 0.2, with high values representing large divergent
angles ( > 12) and low liquid to gas ratios ( <0.5 l/m3).
(8)
o
where, from 4, = 0.8. Compare the calculated pressure drop Pe with the
design figure PR and, if necessary, iterate with a new velocity given by:
(9)
o
Continue until the pressure drop error is acceptable ( <5%). Then vgt = vte.
(11)
(14)
(15)
(16)
(17)
where x is a dimensionless throat length which, following Yung et al (1977), is
related to by:
(18)
For = 0.8 as used here, x = 1.3417.
Figure 12 Design Graph for Throat Length (air/water @ 20C and 1 bar with
= 0.8)
(19)
For rectangular scrubbers the throat area is given by Qg/vgt but the breadth and
width are arbitrary. Common practice is to use ratios of 1.5 to 2.5 for small
throats, rising to 3 and above as the width approaches 300 mm. The difficulties of
adequate liquid distribution generally restrict throat widths to 500 mm, but the
breadth may increase almost indefinitely provided liquid supply can be ensured
along the entire distance.
The remaining dimensions may now be determined to suit the requirements of the
application. Note that typical convergence half angles are between 8 and 15,
with suggested lower and upper limits of 5 and 30 respectively, with low values
generally giving low dry pressure drops. Higher values are recommended for
wetted wall Venturis where a steep angle ensures good liquid penetration across
the throat.
1. small droplets due to the inherent inability of the separator to remove the finest
droplets;
2. re-entrainment of collected liquid, the result of:
o transition from separated to separated-entrained flow,
o rupture of bubbles,
o creeping of liquid on entrainment surfaces,
o shattering of liquid droplets resulting from splashing.
With most separators the former may be reduced by increasing the gas velocity, the latter
by decreasing the gas velocity. Design is therefore largely a compromise.
The transition from separated flow to separated-entrained flow occurs at the reentrainment velocity. This is the overall gas velocity through the separator at which liquid
droplets will be stripped from wetted surfaces. The magnitude of this velocity depends on
the Reynolds number of the liquid film, the liquid properties, and the angle at which the
gas strikes the wetted surface. A correlation for cyclones is presented in Figure 13
(Calvert et al, 1975). Here the liquid Reynolds number is given by:
(20)
where a is the cyclone inlet height and l the liquor viscosity (water @ 20C = 1.02 x 10-3
kg/ms). Practical inlet velocities are between 20 and 50 m/s, a typical choice being
between 30 and 40 m/s, with values in the lower range being used to minimise erosion
with liquors of high solids concentrations. Values in the middle range generally result in
pressure drops of the order of 0.75 to 1.0 kPa. Calvert estimated that drop sizes resulting
from operation above the transition point are approximately 250 microns. These can
readily be removed with a vane baffle eliminator.
Vane eliminators are widely used as secondary eliminators. These have the advantage of
being directly insertable into the cyclone body as the gas spin merely enhances their
performance. These items are proprietary design and should be sized on the basis of the
manufacturer`s data. Zig-zag baffles and knitted mesh are also insertable, but generally
benefit from being separated from the cyclonic body, either by a vane or a baffle plate
(Figure 14b). Zig-zag baffles are again proprietary items and should be sized in
accordance with manufacturer`s specifications. Typical superficial velocities in vertical
flow are 3.5 to 5 m/s.
(21)
high velocity gas regions or changes of flow direction, especially the junction at
the start of the throat, dampers, and the elbow;
seals, especially casing seal with plug-type dampers when used with a flooded
elbow;
wet/dry interfaces;
spray nozzles;
welds on instrument pockets and generally post manufacture welds.
Rarely will a location be suitable if less than 2 duct diameters downstream or one
diameter upstream of a disturbance. Methods of improving a sampling location have been
discussed in SPS SAR 30 Section 2.2 (Wallin and van Santen, 1982).
Concentration fluctuations or differences across the duct are unimportant. However, the
miniscrubber is sensitive to small differences at the fine tail of the size distribution.
Sizing measurements should therefore be made using an in-stack cascade impactor fitted
with a suitable pre-cutter. At least four points, two each on two lines at right angles,
should be investigated. No variation should be found in the sub 2 micron tail. More
practically, where differences do exist, the likely effect on the miniscrubber prediction
can be estimated by running the size data through the scoping design procedure, Part 2
Section 2. The reader must then determine for himself whether this uncertainty is
acceptable. If an alternative (and better) sampling location is not available, then some
legitimacy can be gained by traversing the miniscrubber probe across the duct during the
sampling. Replicate measurements under identical conditions should be made on both
sampling lines in order to estimate the error. Note that this method biases the sampling to
the high concentration regions, which, because these tend to be of the coarser dust, results
in an underprediction of the required pressure drop.
The sampling probe should be as short as practically possible to minimise deposition
losses. If all sampling points are equal the sampling should be carried out at a single point
just far enough away from the duct wall to avoid flow distortions caused by the sampling
port. All lines should be lagged or trace heated if condensation is a possibility. The
conveying velocity should be of the order of 20 m/s to further minimise deposition losses.
Sampling should be isokinetic to within 5%.
Operating philosophy.
The notes below refer specifically to the SPS mini-scrubber operating as part of a
standard extractive dust sampling train. However, the general principles can be applied to
any miniature pilot plant scrubber. Indeed, there may be virtue in operating on a
somewhat larger scale, especially if analysis of the emitted fraction is required.
Given that the sampling is isokinetic, each "run" produces three key pieces of
information:
The operating velocity, liquid to gas ratio, sampling duration and sampled volume are
ultimately irrelevant to the analysis. However, the weight collected is determined from
the concentration of the collected liquor. The liquid rate should therefore be kept to a
practical minimum in order to improve the measure ment accuracy. Typical liquid to gas
ratios are 0.8 - 1.3 l/m3 (1 x 10-3 m3/m3), although satisfactory results can be obtained with
ratios as low as 0.6 l/m3 (0.6 x 10-3 m3/m3) or as high as 1.9 l/m3 (1.9 x 10-3 m3/m3). As
sampling proceeds, deposition on the back-up filter causes the sample rate to fall, thus
reducing the scrubber pressure drop. Compensation should be made by increasing the
suction pressure to maintain isokinetic sampling rather than by increasing the liquid rate.
Preparation.
The following items and preparation are required:
exhaust filters. Glass fibre filters are suitable, but should be baked for at least 2
hours at 105C to remove trace organic components. Weighing when cool
should be to 0.05 mg. The filter housing should be heated during the
measurements to prevent blockage through condensation.
liquor collection. A dry 5l capacity water bottle weighed to 0.1 g and fitted with
a means of sampling a representative aliquot. (In the laboratory tests this was a
20 ml syringe with a wide bore needle fitted through a bung in the cap).
concentration determination. Liquor concentration may be determined either by
evaporating to dryness and weighing, or by filtering, drying and weighing. The
former is less accurate because the tare weight of the evaporating dish reduces
the sensitivity (small changes in large numbers). The latter is only suitable for
non-soluble particles but overcomes the tare weight problem. Hence each run
requires either: four 50 ml capacity beakers suitable for oven drying pre-weighed
to 0.1 mg; or four 50 ml containers pre-weighed to 0.1 mg plus four 0.22 micron
acetate filters pre-baked and weighed as per the exhaust filters. (Handling and
identification of these fragile filters is simplified if placed individually in light
weight containers. Aluminium foil baking dishes are ideal).
scrubbing liquid. 2 to 4 litres of fresh demineralised water placed in a constant
head tank with at least 3m elevation. Alternatively pressurise or use a small
centrifugal pump to supply head.
Operating procedure.
Typically six data points are required for an acceptable accuracy. These should represent
replicate sets at each of three pressure drops; at approximately the desired pressure drop
(estimated via the impactor size distribution), 50% higher and 50% lower. The standard
procedure adopted in the research programme (SPS RR 40) was then to:
1. Determine the gas velocity at the sampling point by pitot measurement (see Mini
Manual 1 Part 2) and select a sampling nozzle for isokinetic sampling at 2 l/s.
2. Select probe to give conveying velocity approximately 20 m/s. 12 mm polished
stainless steel is usual.
3. Rinse collection bucket and assemble parts. No need to dry. Assemble back-up
filter.
4. Leak test the system: Close water supply valve. With the probe outside the duct
switch on suction. No flow should be recorded with the sampling nozzle capped.
5. Adjust water and air rates to give desired pressure drop at isokinetic flow. If
liquid to gas ratio outside normal operating range select a different Venturi
throat piece and repeat set-up procedure from 4.
6. With the miniscrubber running place sampling probe in the duct at the required
sampling location. If this proves difficult detach probe and reconnect to
miniscrubber when in correct location.
7. Continue sampling, adjusting the sample rate to maintain isokinetic conditions at
a constant scrubber pressure drop. Halt measurement when no further adjustment
is available. With the miniscrubber still running detach probe to allow a
complete air change in the de-entrainment bucket. Simultaneously switch off
water.
8. Carefully remove dust from outside of probe, then rinse out internal deposits and
add washings to the bucket.
9. Remove, dry and weigh filter to 0.05 mg. Note the weight gain, Wo gm.
10. Determine the liquor concentration. Note: in the SPS research this was done
entirely by weighing as volume determinations using measuring flasks were
found to be inaccurate for these purposes. Throughout, the weight contribution
of the collected particulate matter was ignored.
Transfer collected scrubbing liquor to the 5l bottle. Rinse bucket and add
washings to the bottle. Re-weigh bottle to 0.1 g and hence note volume of liquid
collected, Vc ml. Fit the sampling syringe to the bottle and shake well to give a
uniform suspension. Invert bottle and remove an approximately 20 ml aliquot.
Repeat to give a total of 8 samples. Pair together into the four 50 ml containers
and re-weigh to 0.1 mg to determine their volumes V ml. Evapori ate to dryness
or filter as desired to determine the weight of particulate matter in each, Wi gms.
11. Determine the miniscrubber efficiency, as:
(23)
12. where Wb is the total weight of particulate collected in the bucket
(24)
13. Plot efficiency in transfer units against pressure drop on log-log axes as per
Figure 2(b) where Number of Transfer units, Nt = -ln(1-/100).
14. Repeat entire procedure at least six times at three different pressure drops.
Analysis and use of data.
The data plot will inevitably be scattered. The scatter may have two causes: poor
experimental technique, in which case a best fit straight line should be taken through all
the data; or genuine variations due to fluctuations in the particulate size distribution, in
which case a best straight line should be taken through the tops of the data pairs. This can
result in two markedly different lines and consequently markedly different predictions of
the required pressure drop. The correct approach can be inferred from the cascade
impactor data. If this shows no variation in size distribution at the various sampling
locations then the scatter taken can be assumed to be random. If the impactor data is
scattered then the latter course would be more valid. However, with only three useful data
points the accuracy of the method must be in question. Further replicate cascade impactor
and miniscrubber measurements should therefore be made. If the confidence interval
cannot be reduced below 30%, or if the predicted pressure drop for the required
efficiency is greater than 6.0 kPa, then further pilot testing should be considered.