Application of Ecvt for Measurement of Gas-solids Flow

8/12/2019 Application of Ecvt for Measurement of Gas-solids Flow

1/12


2/12

INSTITUTE OFPHYSICSPUBLISHING MEASUREMENTSCIENCE ANDTECHNOLOGY

Meas. Sci. Technol.12(2001) 11091119 www.iop.org/Journals/mt PII: S0957-0233(01)22190-0

Application of electrical capacitancetomography for measurement ofgassolids flow characteristics in apneumatic conveying system

Artur J Jaworski1 and Tomasz Dyakowski2,3

1 School of Engineering, University of Manchester, Oxford Road, Manchester M13 9PL, UK2 Department of Chemical Engineering, UMIST, PO Box 88, Manchester M60 1QD, UK

E-mail: [email protected] and [email protected]

Received 22 February 2001, in final form 29 May 2001, accepted forpublication 13 June 2001

AbstractTransient three-dimensional multiphase flows are a characteristic feature ofmany industrial processes. The experimental observations andmeasurements of such flows are extremely difficult, and industrial processtomography has been developed over the last decade into a reliable methodfor investigating these complex phenomena. Gassolids flows, such as thosein pneumatic conveying systems, exhibit many interesting features and thesecan be successfully investigated by using electrical capacitance tomography.This paper discusses the current state of the art in this field, advantages andlimitations of the technique and required future developments. Variouslevels of visualization and processing of tomographic data obtained in apilot-plant-scale pneumatic conveying system are presented. A case studyoutlining the principles of measuring the mass flow rate of solids in avertical channel is shown.

Keywords:solids concentration, solids velocity, mass flow rate, tomography,correlation, pneumatic transport

1. Introduction

Transient three-dimensional multiphase flows are a character-istic feature of many industrial processes. The experimentalobservations and measurements of such flows are extremelydifficult, and industrial tomography has been developed overthe last decade into a reliable method for investigating thesecomplex phenomena [1, 2]. Various levels of visualization ofdata can be extracted from tomograms. Images can charac-terize the behaviour of the flow at a single level (or plane) asit varies with time and therefore allow distinguishing betweendifferent flow patterns. Individual images can reveal impor-tant cross-sectional information such as the concentration ofsolids in gassolids systems. A set of successive images, for

known velocities of solids, can be transformed into the three-dimensional distribution of solids along the direction of flowto provide additional body shape type information similar to

3 Author to whom correspondence should be addressed.

that obtained from conventional photography. However, there

are important differences between data obtained from the two

techniques and these are discussed in more detail in section 3.

On the micro-scale, the motion of a particle is governed

by various types of particleparticle and particlefluid

interactions. On the macro-scale, these are responsible for

the appearance of various types of macro-structures in the

flow with characteristic sizes much larger than the diameter

of the particles. Pneumatic conveying is usually classified into

two categories: dilute (or lean) and dense phase [3]. In

dilute-phase conveying the particles are usually transported

in the form of a suspension with the concentrations of solids

typically below 10%. On the other hand, dense-phase

transport is usually understood as the conveying of particles

along a pipe, which is filled with particles at one or more crosssections [4]. Flow patterns in a dense pneumatic conveying

system exhibit many interesting features and analogies with

gasliquid flows. Typically, materials conveyed in a dense-

0957-0233/01/081109+11$30.00 2001 IOP Publishing Ltd Printed in the UK 1109


3/12

A J Jaworski and T Dyakowski

Filter unit &

air reservoir

Removable

testsection

100 L

PT

Cooler

ELECTRONIC

CONTROLLER

4-20mA

AIR

COMPRESSOR

8 bar g

oil free air

Blowerunit

100 L

Sonic

nozzle

Approx. 7 m

Rotary

feeder

Upper tank

Lower tank

Approx.

3m

PNEUMATIC

CONVEYING

FLOW LOOP

AIR SUPPLY

Removable

test section

Removable

test section

Load

cell

Load

cell

PG PS

PS - Pressure overload switch

PG - Pressure gauge

PT - Pressure transducer

Bleed

valve

4-20mA

Figure 1.A schematic diagram of the UMIST dense-phase pneumatic conveying rig.

phase system are not very cohesive and exhibit a permeability

and de-aeration rate that are either both low (for a moving bed)

or both high (for slug or plug flow). High permeability and

de-aeration rates characterize polyamide chips, so the modes

of a dense transport of these chips are discussed in this paper.

When the gas velocity is reduced below the saltation

velocity, a settled layer of solids is formed at the bottom of

a horizontal pipe. The transport of solids occurs through the

propagation of flow instabilities referred to as slugs[57].

These pick up the solids from the settled layer and convey

them along the pipe for some distance. The slug propagation

velocity ishigherthan theaxialvelocityof solidsand thesolids,

after being mixed within the slug body, are dropped off to form

a settled layer behind the slug. The slope at the back of the

slug is usually steeper than the slug front.

In vertical pipelines, the behaviour of the solid particles is

somewhat different. Theflow instabilities responsible for thenet transport of solids take theformof nearly axis-symmetrical

discreteplugs, which move upwards along the pipe [8]. The

consecutive plugs are separated from one another by an air

gap. Flowvisualizationsreveal that the solidparticlescan raindownfrom the back of the preceding onto the following plug.

In both cases, the horizontal and thevertical transport, the flow

instabilities described exhibit a quasi-periodic behaviour with

a frequency typically in the region between fractions of a hertz

and a few hertz.

Granular materials conveyed in the pipelines are usually

relatively dry and can be assumed electrically non-conductive.

Moreover, their bulk dielectric constant is rather low, typically

between 1.5 and 5. Therefore, electrical capacitancetomography (ECT) seems to be ideally suited for investigating

this type offlow [913]. Typically, ECT systems operate at a

frame capture rate of up to 200 frames s1, which allowsone to

reproduce theflow patterns observed with sufficient accuracy,

the additional advantage, of course, being the possibility of

reconstructing the internal structure of the flow from the

distribution of dielectric permittivity in the cross-sectional

image.

The main problem with measuring multiphase flows is

associated with the fact that both the phases distribution

and the velocity profile vary widely both in the temporaland the spatial sense. The development of the so-called

twin-plane tomographic instruments, potentially offers an

excellent opportunity to develop techniques for measuring

the velocity field by cross-correlating, on a pixel-by-pixel

basis, the time series of tomographic images obtained. The

concept of such systems was describedfor example in [1,2, 14]

and was successfully implemented in the area of hydraulic

conveying by Loh et al [15]. However, attempts to apply

the cross-correlation for measuring the mass flow rates ofsolids in pneumatic conveying systems achieved only limited

success [16]. A more detailed discussion on the suitability

of twin-plane tomography and correlation techniques for

measuring massflow rates is given in [17].The objective of this work was to investigate the

complexities of flow morphology in dense pneumatic

conveying systems. This was done by using two

complementary techniques: a high-speed video camera (up

to 500 frames s1) and a twin-plane ECT system (with a

capture rate of 100 frames s1). A comparison between

the flow visualizations obtained using these two techniques

is provided. The prospects of using twin-plane ECT for

measuringmassflowratesof solids, arediscussed, a theoreticalanalysis is outlined and preliminary results concerned with

flow measurement within a vertical pipe to obtain estimates of

the mass flow rate of solidsare presented. Theresults obtained

1110


4/12


5/12


Figure 3.The design of the twin-plane ECT sensor.

Figure 4.High-speed camera visualization of the slugflow in the horizontal pipe.

3. Results and discussion

Qualitative and quantitative data were obtained during the

experimentalprogramme. The firstconfirmed theapplicability

of the ECT system for investigations of transient and three-

dimensional gassolids flows. Here we are presenting the

gassolids flow morphology in a vertical and a horizontal

channel. The latter is congruent with the earlier publications,

e.g. [1012]. Theflow structure in a vertical channel exhibits

many interesting features, in particular a bi-directional flow

sequence. Quantitative analysis was focused on applying the

ECT system for measurement of the mass-flow of solids. The

theoretical analysis is presented and simplifications made in

the current study are introduced and discussed in some detail.

3.1. Characterization of flow patterns

The flowpatterns in thepneumaticconveyingwereinvestigated

both for horizontal and for vertical sections (see figure 1).

Figure 4 shows a series of six photographs illustrating the

passage of two consecutive slugs through the viewing section

in the horizontal pipe. A few interesting features are worth

mentioning. Firstly, the structure of the two slugs is slightly

different. Thefirst slugfills the pipe completely and exhibits

rather clearly defined boundaries of the front and the tail.

This is not true for the following slug. A thin layer of gas

and suspended particles can be seen in the upper part of thepipe, above the main body of the slug. The front of the

slug appears less clearly defined and it looks as though the

material at the front of the slug is trying to catch up with the

preceding slug. Secondly, referring to the front of the first

slug, it can be seen that its slope can vary rather fast (compare

photographs (a) and (b)). Thirdly, the thickness of the settled

layer separating the two slugs evolves and becomes thinner

from one photograph to another (compare photographs (b)

(e)). The latter is accompanied by an increase in the distance

separating the two slugs. Finally, it is also worth noting that

clusters of particles are present immediately before the fronts

of both slugs described.

Figure 5 shows a time series of cross-sectional

tomographic images corresponding to the slugflow presented

in figure 4. The first seven images show the transition between

a half-filled pipe and a fullyfilled pipe that corresponds to the

passage of the slug front. Similarly, the last four images showthe passage of the slugs tail through the measurement plane.

All images in between correspond to the slug passing through

the sensing plane (images betweent = 0.12 s andt = 0.29 s

are omitted to save space).

Applying a twin-plane system allows reproducing the

shape of the slugs as presented in figure 6. Here, the pixels

lying along a vertical line passing through the centre are

selected from each frame. These are combined to give a

longitudinal cross section of the slug. Figure 6 illustrates

the problems encountered by the ECT measurement. These

relate to the limited spatial resolution of the images in the

cross-sectional plane, averaging of the concentration of solidsalong thefinite length of the electrodes (in our case 3 cm) and

smearing of the sharp boundaries between the phases. This is

why figure 6 showsshadesof grey, instead of sharp boundaries.

1112


6/12

Gassolidsflow characteristics

Figure 5.A series of ECT images corresponding to the passage of a slug through the sensor. (Images betweent = 0.12 s andt = 0.29 s areomitted since they show fullyfilled pipe (black).)

Figure 6.The longitudinal cross section of the slugflow obtainedfrom the tomograms. The data were obtained by extracting pixelsfrom the vertical axis of each tomogram and combining them as atime series.

Figure 7.The shape of slugs obtained by thresholding the data fromfigure 6.

Of course, one of the methods used to extract sharp

boundaries between the phases from the ECT images is the so-

calledthresholding. The areas of the normalized dielectric

permittivity above the threshold (e.g. 0.5) are represented inblack, while the areas of the permittivity below threshold are

shown in white. An illustration of this method is shown in

figure 7. It can be seen that the shape of the slug body can

(a) (b)

Figure 8.Still photographs showing plugs travelling upwards (a)and material dropping downwards in between the trains of plugs (b).

be determined, of particular interest being the slopes of the

front and tail of the slug. These exhibit many similarities

to the photographs shown in figure 4. Of course, the spatial

resolution does notallowimaging of thepresence of individual

particles or particle clusters as can be seen on a high-speed

video recording.

In the experiments conducted in the vertical pipe, theflow patterns observed were generally in agreement with those

described in the literature. The conveying of solids typically

consisted of two distinctive phases. Atrainof a few plugs,

1113


7/12


Figure 9.The cross-sectional distribution of dielectric permittivity for upward plugflow (a) and downward return of material (b).

eachapproximately 1020 cm long, appeared on average every

34 s. It could be seen that some particlesrain downfrom

the preceding onto the following plug. At the end of each

passage some granular material (most probably from the tail

of the train) dropped downwards under the gravity. This was

collected at the bottom of the vertical section and was picked

up by the next train of plugs. Figure 8 shows four photographs

taken from the high-speed video, which show a typical flow

pattern observed.

Again, the use of ECT allows looking into the internalstructure of the plugs, which would not be possible using

photographic techniques. Figure 9 givesexamplesof thecross-

sectional distribution of the normalized dielectric permittivity

due to material present in the sensing area. Figure 9(a)

corresponds to the plugs travelling upwards. Images 26 and

1012 correspond to the plugs present in the sensing area,

whereas images 1, 7, 8 and 9 show the spaces between the

two consecutive plugs. Figure 9(b) shows the permittivity

distribution for material dropping down between two trains

of plugs.

Figure 10 shows the axial cross section of the flow

reconstructed from the ECT measurements. It is worth noting

that the ECT reconstruction reflects the changes in porosityof the material within the plugs. It can be inferred that the

density increases towards the pipe wall, which most probably

corresponds to an increase in inter-particle stresses. The

centre of the plug, on the other hand, seems more porous,

probably due to the passage of the gas through the centre of

the plug. For the downward movement of the material the

highest concentration is usually close to the wall; however,

the location of the region of the highest concentration moves

around the circumference of the pipe.

Of course, on the fundamental level, the information

obtained from the flow visualizations (figures 4 and 8) is

differentfromthatobtainedby combiningtomographic images

such as those shown in figures 6, 7 and 10. Whereas thephotographs show thespatial information at a given instant, the

tomographic results represent the temporal changes at a given

spatial location (that of thesensor). The twoapproaches would

be equivalent only if the flow structures were frozenwhile

moving along the pipe. Although this is not necessarily true in

thepneumatic conveying system, it isworthnoting theapparent

similarities between thephotographs andthe tomographic data.

Of course, using a twin-plane tomographic system allows

obtaining the time delay between the appearance of slugs

or plugs in respective planes and therefore the propagation

velocity of slugs. The length of the slug can then be calculated

as the time that the slug was present in one of the planes

multiplied by the propagation velocity.It is apparent that, despite difficulties encountered in

the ECT measurements (reconstruction algorithm errors,

averaging along the electrodes), the technique can provide

1114


8/12


(a) (b)

Figure 10.Temporal changes in the permittivity distribution acrossthe diameter of the pipe: (a) upwards travelling plugs and

(b) material falling downwards.

unique information about the structure and the evolution of

three-dimensional and unsteady gassolids flow. In order to

provide more reliable information about theflow phenomena,

some more work is required in order to relate the measured

dielectric permittivity distribution to the concentration of

solids. This is of vital importance from the viewpoint both of

theoretical modelling and of the validation of computational

fluid dynamics (CFD) codes.

3.2. Flow measurements

As discussed in the introduction, one of the attractive ideas for

multiphaseflow measurement that still has to be investigated

is the use of twin-plane tomography systems and cross-

correlation techniques. The concept itself is not novel. In

fluid mechanics cross-correlation of pressure and velocity

fluctuations has been for many decades a standard technique

for investigating the flows within a boundary layer [21] and

for tracking the movement of coherent structures shed by

aerodynamic bodies [22]. In the area of multiphase flows

the velocity field can be measured by cross-correlating the

time-varying signal arising from one phase being dispersed

in another (e.g. gas bubbles in liquid) [23]. Some obvious (butoften tacit) assumptions made while measuring the velocity

field by cross-correlation techniques are that

(i) the sensorssize is small relative to their separation;

(ii) there are measurable disturbances in the flowfield being

investigated and

(iii) the velocityfield (convection velocity) can be associated

with the propagation velocity of these disturbances

It can easily be seen that the first assumption is not strictly

satisfied by the ECT sensor. The sensing electrodes have a

finite length (in our case 3 cm), which is often comparable tothe separation (in our case 13 cm, the centre-to-centre distance

between theplanes). In simpleterms, it couldbearguedthat the

ECT system detects the disturbance for thefirst time, while it

crosses the upstream end of plane 1. Similarly, it can detect its

presence for the last time when it crosses the downstream end

of plane 2. In these circumstances, it is not clear whether the

centre-to-centredistance is themostappropriate for calculating

propagation velocities of the disturbances. In general, this

problem is caused by the spatial averaging taking place along

the electrodes.

The detection of disturbances caused by gassolids

interaction is not straightforward. Two facts need to beremembered: firstly, that an ECT system has a spatial

resolution of about 10% of the pipe diameter; and secondly,

that the electrodes are rather long. Consequently, the sensing

volume within which detection takes place is large and there is

no way of detecting individual particles. Instead, an average

concentration of solids in a rather large control volume is

measured. This, of course, creates problems from the point

of view of correlation analysis. For example, it is not possible

to detect whether the settled layer at the bottom of the pipe is in

motion or stationary (tomography provides a constant signal).

Similarly, situations in which the system is nearly blocked (i.e.

the pipe isfilled completely with stationary, or slowly moving,

material) can be ambiguous for the purposes of correlation.Finally, the problems are compounded by the fact that

the gassolids structures do exhibit wavelike behaviour. The

classic example here is the propagation of slugs in the

horizontal pipeline. Here the velocity of material (plastic

pellets) cannot be directly linked to the propagation velocity

of the disturbances in theflow (slugs). Of course, it is possible

to measure the propagation velocity of the slugs (or, in other

words, how quickly the wavefronts are moving), but, strictly,

no inferences about theassociatedmass transportcan bedrawn.

The above mentioned issues should be borne in mind

while attempting to use the cross-correlation techniques for

measuring the mass flowrate of solids in pneumaticconveyors.It is not the intention of this paper to prove that such

measurements are not possible. It is simply to draw attention

to the fact that very often the results may be flow regime, or

indeedflow rig dependent and some serious caveats regarding

the underlying theoretical approach used by cross-correlation

techniques should be mentioned.

For the above mentioned reasons the experimental work

conducted at UMIST focused in thefirst instance on theflow

in the vertical section of the flow rig. It was thought that

theflow patterns could be assumed to a good approximation

axis-symmetrical and would be free of situations in which

a stationary settled layer of solids affects the correlation

analysis. The experiments focused on relatively lowconveyingvelocities (gas velocities between 1.5 and 2.0 m s 1 for an

empty pipe) and a feed of solids generally between 700 and

900 kg h1. This was a compromise between obtaining

1115


9/12


well-defined plug flow and ensuring that the ECT system

can still follow the passage of the flow structures with

the 100 frames s1 capture rate. The experimental results

presented are preliminary in character and illustrate the level

of analysis conducted.

Theinstantaneousmassflowratethrough thecross-section

of a pipe can be written as

m(t) =

A

(x,y,t)v(x,y,t) dxdy (1)

where (x,y,t) stands for the instantaneous density at

point (x,y) of cross section A and v(x,y,t) denotes

an instantaneous velocity at point (x,y) in the direction

perpendicular to the cross-sectional plane. It is then possible

to define averages of both density and velocity over the cross-

section of the pipe Aas follows:

A(t) = A

(x,y,t) dxdy (2)

vA(t) =

A

v(x,y,t) dxdy. (3)

Of course, the instantaneous density and velocity can now be

expressed in the following way:

(x,y,t) = A(t)+ (x,y,t) (4)

v(x,y,t) = vA(t)+v(x,y,t) (5)

where, by definition,

A

(x,y,t)dxdy = 0. (6)

A

v(x,y,t)dxdy = 0. (7)

Consequently,

m(t) =

A

[A(t)+ (x,y,t)][vA(t)+v(x,y,t)] dxdy

=

A

A(t)vA(t) dxdy+

A

(x,y,t)vA(t) dxdy

+

A

A(t)v(x,y,t) dxdy

+

A

(x,y,t)v(x,y,t)dxdy. (8)

Of course, the second and third terms on the right-hand side

are zero by definition. Thefirst term can be easily integrated

and therefore

m(t) = AA(t)vA(t)+

A

(x,y,t)v(x,y,t)dxdy. (9)

The above equation highlights the reasons why tomography is

required for highly non-uniform flows to correctly calculate

the mass flow rate of solids. It clearly shows that the more

spatially non-uniform theflow (both in the velocity sense and

in the density sense) the larger the second term on the right-

hand side of equation (9). Therefore, considering theflow on

a pixel-by-pixel basis becomes essential.

On the other hand for uniform flows the second term will

disappear, leading to a simplified equation:

m(t) = AA(t)vA(t). (10)

This level of simplification was applied in the study of the flow

in the vertical pipe presented here. Applying equation (10) in

discrete form for the tomographic measurements yields

m Ak=n

k=1 (tk)v(tk)t

tn t1(11)

where discrete times t1, . . . , t n correspond to consecutive

tomographicimages. To applyequation(11)someestimatesof

the instantaneousdensity andvelocitymustbe introduced. The

former could be found if the average permittivity distribution

in the sensor control volume could be linked to the average

density (orconcentration). Thisisnot straightforward, since no

appropriate mixing law is known a priori, so the concentration

of solids cannot be obtained in a reliable fashion. For the

work described here, a linear relationship between density and

permittivity hasbeenassumed. Moresophisticatedmodels can

be applied [18], but in the current work these did not seem to

provide too much improvement.

Finding the instantaneous velocity poses another type of

problem. In the vertical pipe, theflow is bi-directional and

a method to decide the instantaneous flow direction needs

to be devised. Here, it has been decided that a short-time

window cross-correlation could be a practicable candidate,although applying this concept may raise some questions of

a fundamental nature. At this stage, the spacing between the

sensors for the correlation analysis was assumed to be 13 cm.

A sample result showing the average concentration of

solids as a function of frame number is plotted in figure 11.

The upper graph shows its variation over 40 s. It can be seen

that theoccurrence of plugs is quasi-periodic. The lowergraph

shows the variation with time over a shorter time of 4 s. The

two phases of the flow, i.e. upward plug flow and return of

material, can be easily recognized.

Figure 12(a) shows a typical power spectrum obtained

for theflow in the vertical section. In this example, the dataset consisted of 32768 concentrations of solids (over 5 min

offlow rig operation). There are two maxima on the graph.

Thefirst, in the region of 0.3 Hz, most probably corresponds

to the frequency of the large scale flow patterns (i.e. a train

of plugs every few seconds). The second maximum, in the

region between 6 and 10 Hz, corresponds most probably to the

individual plugs within the train. It can be seen, however, that

the whole spectrum is of broad-band character, indicating a

quasi-periodic rather than regularflow character. Figure 12(b)

shows the cross-correlation of the signals obtained from two

planes of the sensor. Here again the plot contains low- and

high-frequency components. The maximum of the correlation

function for this example falls around +40 ms.However, the cross-correlation function shown in

figure 12(b) would indicate only the upward direction of

the transport of solids. In order to study this in more

1116


10/12


2000 2500 3000 3500 4000 4500 5000 5500 60000

20

40

60

80

100

frame #

averageconcentration[%]

2850 2900 2950 3000 3050 3100 3150 3200 32500

20

40

60

80

100

frame #

ave

rageconcentration[%]

upward plug flow material falling down

Figure 11.The average concentrations of solids obtained from the tomograms for a gas velocity of 2 m s1 and a feed of solids at 900 kg h1.

0.1 1 101E-011

1E-010

1E-009

1E-008

1E-007

1E-006

frequency [Hz]

powe

rdensity

(arbitraryunits)

time [s]

correlationcoefficient

-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 10

0.2

0.4

0.6

0.8

1

(a) (b)

Figure 12.The power spectrum (a) and correlation function (b) obtained for plugflow in the vertical pipe, for a gas velocity of 2 m s1 anda feed of solids at 900 kg h1.

-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 10

0.2

0.4

0.6

0.8

1

upward

downward

time [s]

correlationcoefficient

Figure 13.Two sample correlation results corresponding to theupward and downward transport of solids in the vertical pipe for agas velocity of 2 m s1 and a feed of solids at 900 kg h 1.

detail, the signal was divided into shorter sequences (typically

windows of the order of 128 samples). These were

subsequently correlated between two planes. Figure 13

presents the correlation function for the data taken for the

upward movement (filled circles) and material falling down

the vertical section (empty circles). It can be seen that thedirectionof theflowcannow be determined, a positive time lag

indicatingupward movement anda negative timelag indicating

the downward transport of solids.

The calculations of the mass flow rate of solids carried

out using formula (11) and estimates of density and velocity

as explained above typically underestimated the actual mass

flow rates by 2030%. For the two benchmark flow rates

of 700 and 900 kg h1, the typical results from calculations

would fall into the regions of 500600 and 700750 kg h1,

respectively. On the one hand, the results are encouraging,

because they prove that a formalized approach to calculating

the mass flow rate can give answers of the same order of

magnitude. On the other hand, it is apparent that there isstill a lot of scope for improvement in applying this approach.

In particular, the method described uses concentrations of

solids and therefore propagation velocities averaged over the

1117


11/12


cross-section of the pipe. However, in principle it is possible

that, while some solids travel upwards in one part of the pipe,

other particles travel downwards in another part. Therefore,

the cross-correlation analysis should be performed on a pixel-

by-pixel basis, rather than for the whole cross-section. This

could also include thecorrelation between pixels in two planes

which do not correspond to one another, to allow for lateralmovement of solids. Further work in this area is planned.

4. Conclusions

The paper presented is concerned with the application of

ECT for investigating the unsteady and three-dimensional

characteristics of gassolids flow associated with pneumatictransport both in horizontal and in vertical pipes. It has been

demonstratedthat ECTis able to image thedynamicsof macro-

structures (slugs and plugs) in a manner that is consistent

with high-speed photographic techniques. Moreover, unlike

photographic methods, ECT can give valuable insight into

the internal structure offlow instabilities such as slugs andplugs. This is important from the viewpoint offlow modellingand validating of CFD codes, which makes ECT an excellent

research tool.

Furthermore, the application of a twin-plane tomography

system to the measurement of multiphase flow is discussed.This is usually attempted by using well-known cross-

correlation techniques. Fundamental assumptions lying

behind these techniques are discussed in some detail and

these are related to the prospective application of ECT for

measurement offlow rates of solids in pneumatic conveyingsystems. Several potential weaknesses are identified, inparticular the finite length of electrodes, ability to detect theflow in a large control volume only and wavelike character ofdense gassolidsflow.

The theoretical analysis underlying the mass flowmeasurements is presented and appropriate simplifications arediscussed. This is applied to aflow in the vertical channelfor measuring theflow rate of solids in the presence of plugflow and preliminary results are presented. However, it isgenerally apparent that, before the tomographic techniques

can successfully be applied to multiphase measurement, a few

further developments need to take place. These should include

the following.

(i) The capture rate of the tomographic equipment should be

increased to the region of 5001000 frames s

1 to allowmore accurate estimation of the propagation velocity of

the disturbances in theflow. In the current work, theflowwas selected such that ECT could cope with representing

theflow sequences correctly, but it was felt that time lagestimates were not accurate enough.

(ii) Thedevelopment ofsuitablemodelsto relatethe measured

dielectric permittivity of the gassolids mixture to thedensity (or concentration). This is an issue that ECT

equipment manufacturers are well aware of, but it needs

to be addressed more strongly by the research community.

(iii) Theeffects of spatial filtering dueto finiteelectrode lengthon correlation analysis should be considered. At present,

a centre-to-centre distance between the sensor planes is

taken into account in calculations, but some more detailed

investigations are needed in order to justify the use of

correlation techniques for measurements.

Acknowledgments

We would like to gratefully acknowledge the support obtained

from the European Commission BRITE EURAM programme

(BRST-CT98-5402) and the UK Engineering and Physical

Sciences Research Council (GR/M31910). We would also

like to thank Process Tomography Ltd for providing a twin-

plane ECT systemandtheRutherford Appleton Laboratory for

access to the high-speed camera used forflow visualizations.

References

[1] Williams R A and Beck M S (eds) 1995Process Tomography:Principles, Techniques and Applications(London:Butterworth-Heinemann)

[2] Plaskowski A, Beck M S, Thorn R and Dyakowski T 1995Imaging Industrial Flows: Applications of ElectricalProcess Tomography(Bristol: IOP)

[3] Molerus O 1996 Overview: pneumatic transport of solidsPowder Technol.8830921

[4] Konrad K 1986 Dense-phase pneumatic conveying: a reviewPowder Technol.49135[5] Konrad K and Davidson J F 1984 The gasliquid analogy in

horizontal dense-phase pneumatic conveyingPowderTechnol.391918

[6] Konrad K 1986 Dense-phase pneumatic conveying throughlong pipelines: effect of significantly compressible airflowand pressure dropPowder Technol.48193203

[7] Konrad K 1988 Boundary element prediction of the freesurface shape between two particle plugs in a horizontalpneumatic transport pipelineCan. J. Chem. Eng.6617781

[8] Konrad K and Totah T S 1989 Vertical pneumatic conveying ofparticle plugsCan. J. Chem. Eng.6724552

[9] Huang S M, Xie C G, Salkeld J A, Plaskowski A, Thorn R,Williams R A, Hunt A and Beck M S 1992 Process

tomography for identification, design and measurement inindustrial systemsPowder Technol.698592[10] McKee S L, Dyakowski T, Williams R A, Bell T A and

Allen T 1995 Solidsflow imaging and attrition studies in apneumatic conveyorPowder Technol.8210513

[11] Ostrowski K L, Luke S P and Williams R A 1997 Simulationof the performance of electrical capacitance tomography formeasurement of dense phase pneumatic conveyingChem.Engng. J.68197205

[12] Dyakowski T, Luke S P, Ostrowski K L and Williams R A1999 On-line monitoring of dense phase flow using realtime dielectric imaging Powder Technol.10428795

[13] Dyakowski T, Jeanmeure L F C and Jaworski A J 2000Applications of electrical tomography for gassolids andliquidsolidsflows a reviewPowder Technol.11217492

[14] Dyakowski T 1996 Process tomography applied to multi-phaseflow measurementMeas. Sci. Technol.734353

[15] Loh W W, Waterfall R C, Cory J and Lucas G P 1999 UsingERT for multi-phaseflow monitoringProc. 1st WorldCongress on Industrial Process Tomography (Buxton)pp 4753

[16] Arko A, Waterfall R C, Beck M S, Dyakowski T, Sutcliffe Pand Byars M 1999 Development of electrical capacitancetomography for solids mass flow measurement and controlof pneumatic conveying systemsProc. 1st World Congresson Industrial Process Tomography (Buxton)pp 1406

[17] Jaworski A J and Dyakowski T 2001 Tomographicmeasurements of solids mass flow in dense pneumaticconveying. What do we need to know about theflowphysics? Proc. 2nd World Congress on Industrial Process

Tomography (Hannover 2931 August 2000)[18] 2000PTL 300 SystemUsers Manual(Wilmslow: Process

Tomography Ltd)[19] Yang W Q 1996 Hardware design of electrical capacitance

tomography systemsMeas. Sci. Technol.722532

1118


12/12


[20] Jaworski A J and Bolton G T 2000 The design of an electricalcapacitance tomography sensor for use with media of highdielectric permittivityMeas. Sci. Technol.1174357

[21] Kader B A 1996 The second moments, spectra and correlationfunctions of velocity and temperaturefluctuations in thegradient sublayer of a retarded boundary layer Int. J. HeatMass Transfer3933146

[22] Lee B H K and Marineau-Mes S 1996 Investigation of theunsteady pressurefluctuations on an F/A-18 wing at highincidenceJ. Aircraft3388894

[23] Hammer E A 1983 Three componentflow measurement inoil/gas/water mixtures using capacitance transducersPhD ThesisUniversity of Manchester

1119

Application of Ecvt for Measurement of Gas-solids Flow

Documents