You are on page 1of 9

SME Annual Meeting

Feb. 28-Mar. 1, Salt Lake City, Utah

Preprint 00-18

A NOVEL PROBE FOR INVESTIGATION OF SLURRY FILTRATION

F. S. Bourgeois, M. T. Haque and D. W. Wu


Dept of Mining, Mnrls and Mtrls Engrg; Univ of Queensland
Queensland, Australia

ABSTRACT It is worth noting that properties that are used to


A novel impedance probe has been commissioned characterize the filtration process can be spatial and
specifically for capturing fast dynamic information dynamic. For example, filter cake voidage varies
about filtration processes. It provides an attractive spatially and dynamically. Conventional tests only
alternative to more complex measurement tech- yield measurement of average cake voidage.
niques in situations where the slurry system is Limitations in our current level of understanding of
primarily one-dimensional. The probe can monitor the process dynamics can be partly attributed to the
mud-line, slurry densification and drainage. Dynamic absence of readily useable equipment for monitoring
variations in cake permeability are obtained by the process dynamically and spatially. Indeed, the
combining dynamic mud-line information obtained filtration process, as in the case of vacuum filtration
with the probe and filtrate rate data derived from a for example, is conventionally studied using vacuum
leaf test apparatus. The measurements are re- leaf tests. Such tests cannot yield information about
viewed in relation to microstructural measurements spatial properties, and are limited in their ability to
obtained on iron ore filter cakes. The practical quantify dynamic properties. Dynamic properties
significance of these measurements for optimization that are obtainable with filtration test rigs are nor-
of the filtration process is discussed. mally restricted to filtrate and drainage rates.
Although such tests are generally sufficient for
INTRODUCTION selecting filtration equipment, they do not contain
The filtration process is dynamically complex and sufficiently fundamental information to better our
not well understood. There are a great variety of understanding of the process to the point where
process properties that are required to draw a full current equipment design and plant practice can be
picture of the process macro- and microscopically. challenged.
From the characterization perspective of the filtration Basic dynamic properties are often not measurable
process, three types of properties of the filtration with conventional tests. One such property is filter
process can be considered: cake height. It is not possible to measure the
• Spatial properties are spatially distributed variation in filter cake height with time using conven-
within the filtration vessel, ie within the filter cake tional filtration test equipment. Although such
and/or the slurry. An example of such property is information can in some cases be recorded by visual
solids concentration. observation under slow settling conditions (Bhat-
tacharya, 1997), it is not possible to obtain such
• Dynamic properties are time dependent. An information manually when the filtration process is
important example is filter cake height. rapid and/or when the slurry does not permit visual
• Average properties are macroscopic in nature, location of the cake/slurry interface or mud-line. As
and are spatial and/or time averages of spatial a result, cake height can only be determined at the
and/or dynamic properties. Consequently, all end of the filtration process.
dynamic and/or spatial properties give rise to an It is also accepted practice to use the final cake
average property. An example of such property is height to determine the water permeability of satu-
filter cake moisture, which commonly refers to rated filter cakes. This is achieved by combining
average cake moisture as opposed to spatially dynamic filtration rate measured at the end of the
distributed moisture within the cake mass. cake formation time with the cake height measured
at the end of the filtration process. Strictly speaking,
1 Copyright © 2000 by SME
SME Annual Meeting
Feb. 28-Mar. 1, Salt Lake City, Utah
this is incorrect since shrinkage can cause cake The Impedance Multi-Probe
height to change during drying. Electrical techniques have been used by several
Furthermore, dynamic cake height monitoring is authors for measuring solids concentrations of
also necessary to quantify filter cake compression mixtures of conducting fluid and low concentrations
during both cake forming and drying. of small, non-conducting solid particles (Nasr-El-Din
et al., 1987, Turner, 1976). The apparatus devel-
Finally, the filtration process does not always lead oped by these authors were designed for measuring
to a distinct cake/slurry interface, in which case the local solids concentration in pipes and in mixing
notion of cake height and cake permeability remains tanks under quasi-steady conditions.
ill-defined, that is until the slurry level disappears
below the cake surface. The equipment that is used in this work (Wu, 1999)
utilises the two-electrode impedance measurement
Having shown the importance of dynamic monitor- principle. By maintaining a constant a.c. current
ing of the filtration process with the example of filter across a pair of platinum electrodes that is immersed
cake height, it is convenient to distinguish between in slurry, any change in slurry quality affects its
spatial properties that vary perpendicular to the electrical resistance, which is detected as a change
filtration septum and properties that vary in multiple in voltage across the electrodes. To obtain one-
directions. The following property attributes are dimensional information about the process, ten (10)
defined: pairs of electrodes are arranged in a vertical array.
• One-dimensional properties are spatial proper- The horizontal spacing between two electrodes from
ties that strictly speaking do only vary in the the same pair is 2.5mm. This spacing was chosen
direction perpendicular to the filtration septum. based on design criteria set by Nasr-El-Din et al
(1987) in order for the probe to be useable with
• Multi-dimensional properties are spatial prop- minus 500 microns coal particles. This top size is
erties that cannot be approximated as one- characteristic of fine coal slurries processed in
dimensional properties. vacuum filtration operations. Individual probes are
arbitrarily indexed from 1 to 10, probe 1 being the
All spatial properties that characterize the filtration
lowest and probe 10 the highest. Each pair of
process are intrinsically multi-dimensional. An
electrodes is sequentially activated to avoid cross-
interesting situation is with cake porosity, which
interference at a scan rate of three samples per
varies in all directions. Nevertheless, it has been
probe per second. A dedicated electronic circuit was
shown that it is a strongly anisotropic property that
designed to maintain the current constant at 250mA
exhibits the strongest gradient in the direction
perpendicular to the septum (Zhu Rui, 1998). Cake for slurry resistivity up to 100kΩ and to continuously
porosity can therefore be considered a one- switch between the electrode probes at the desired
dimensional property. scan rate. This relatively high sampling rate allows
even the most subtle and rapid changes in the
Properties that are purely multi-dimensional require filtration process to be captured. Each probe looks
3
complex techniques to be measured, such as three- at an approximately 40 mm hemispherical region of
dimensional tomography (Tarleton and Hancock, space and output a voltage signal (± 5 Volts) that
1996) or spatial geometry (Bourgeois and Lyman, relates directly to the solids concentration in the
1997). An example of such property is pore size measurement region. The vertical distance between
distribution. successive probes is 2.5mm, hence one-
This work is concerned with the measurement of dimensional information can be acquired every
one-dimensional dynamic properties. Although 2.5mm over a total height of approximately 30mm.
these properties cannot describe the entire filtration The electronic circuit acquires the analog output
process, they represent a very significant family of voltage from the probe and converts it into a digital
properties due to the inherent vertical anisotropy of signal. The latter signal is eventually passed on to a
the filtration process. Furthermore, one-dimensional dedicated portable computer for storage and analy-
properties are significantly simpler to measure sis. Figure 1 shows a schematics of the electronic
dynamically than multi-dimensional properties. set-up.

EXPERIMENTAL APPROACH
The following section introduces a unique probe
that was designed to capture dynamic one-
dimensional properties of the filtration process. The
reader is invited to refer to D.Wu (1999) for specific
details about the probe design and its performance.

2 Copyright © 2000 by SME


SME Annual Meeting
Feb. 28-Mar. 1, Salt Lake City, Utah

Multi-probe after
filtration test

Hole in cake surface


indicating position of
probe during test

Filtration vessel

Figure 1. Multi-probe and ancillary equipment. Figure 3. View of the probe and filtration vessel.

The platinum electrodes are soldered to a PCB Figure 4 illustrates the position of the electrode
board which is connected to the electronics circuit pairs relative to the filter cake and slurry. Probe 1
via a connecting cable. As shown in Figure 2, the sits closest to the filtration septum (exactly 4.5mm
ensemble is encased in an epoxide resin. The above the septum due to the probe configuration),
surface of the multi-probe is polished so that the and probe 10 is the highest probe in the vessel. It is
electrodes do not protrude inside the filtration vessel. emphasised that the probe is placed inside the
This prevents impeding the flow of particles and filtration vessel during testing, ie in direct contact
fluid. with the slurry and cake.

Figure 2. View of the multi-probe array. Electrode pairs can


be seen on the right-hand side of the probe.

Filtration Testing Figure 4. Schematics of probe position relative to filter


cake.
Filtration tests are conducted with the Single Leaf
Filter Test (SLFT) apparatus (Bourgeois et al.,
1995). The probe is held vertically against the Basic measurement principle
filtration medium in the filtration vessel in which top- In order to interpret results obtained from the probe
loading filtration tests are conducted. Figure 3 under actual filtration test conditions, it is important
shows the probe out of the test vessel after comple- to gain understanding of the probe output.
tion of a filtration experiment. The mark left in the
When a pair of electrodes is immersed in slurry,
cake after removal of the probe shows the location
the voltage across the electrodes is directly related
of the probe during the test.
to the electrical impedance of the slurry. Although
the probe output voltage can in principle be used to
quantify the properties of the slurry (Meredith and
Tobias, 1962), it will only be used qualitatively in this
paper. The probe output voltage V is related to the
solids concentration C via a relationship of the form:

V
= (1 − C) − n
Vo (1)

3 Copyright © 2000 by SME


SME Annual Meeting
Feb. 28-Mar. 1, Salt Lake City, Utah
where Vo is the voltage output that corresponds to probe. With the unflocculated filter cake, one
the liquid alone and n is a calibration exponent that observes that the higher the electrode pair in the
depends on the conductivity measurement appara- filtration vessel, the longer it takes for the bed/slurry
tus. Ongoing calibration work has indicated that the interface to reach it. Monotonically increasing time
multi-probe array is characterised by n ≈ 0.7. This intervals between successive voltage rises confirm
indicates that the output voltage from the probe rises this observation. This is clearly related to particles
as solids concentration increases. settling velocity, which decreases as the square of
particle diameter under laminar settling conditions,
Because probe output voltage increases with the finer particles being the last to settle.
solids concentration, a solid-liquid interface passing
in front of a given electrode pair gives rise to a sharp
stepwise increase in voltage. This permits mud-line
detection, the measurement principle being illus-
trated in Figure 5.

Figure 6. Mud-line monitoring with the multi-probe array.

The multi-probe was used to monitor the settling of


the same slurry after flocculation with a commercial
Figure 5. Detection of mud-line.
anionic flocculant. The corresponding output is
shown in Figure 7.
The voltage output signal starts rising when the
mud-line reaches the bottom of the hemispherical
measurement volume for a given electrode pair, and
stop increasing when the mud-line rises above the
measurement region. The mid-point of the voltage
step increase coincides with the mud-line being
precisely at the level of the electrodes. This behav-
iour permits dynamic and accurate monitoring of
filter cake height.
The following section reviews some of the prelimi-
nary tests conducted with the multi-probe under
filtration conditions. Figure 7. Mud-line monitoring with the multi-probe array.

EXPERIMENTAL FINDINGS The output signal reveals that the bed/slurry inter-
face moves upward at a steady rate, as indicated by
This section presents some of the preliminary the more regular spacing between the rising sections
measurements that have been obtained with the of the voltage curves. This is not surprising since
probe. The aim is to demonstrate the capability of flocculation renders the particle size distribution
the probe, hence experimental results are mainly more uniform, whereby eliminating the particle
discussed in a qualitative sense. segregation effect observed with the unflocculated
slurry.
Monitoring Mud-line
Figure 6 shows the multi-probe output signal ob- Figures 6 and 7 can be converted to actual bed
tained during a settling experiment with an height versus formation time using the mud-line
unflocculated fine coal slurry (d50= 200µm, d80= 425 detection principle presented earlier. The results are
shown in Figure 8.
µm). As explained earlier, a sharp voltage rise
corresponds to a drastic change in solids concentra-
tion, indicating the passage of the bed/slurry
interface in front of an electrode pair. The position
of the bed/slurry interface can therefore be moni-
tored accurately and dynamically with the multi-

4 Copyright © 2000 by SME


SME Annual Meeting
Feb. 28-Mar. 1, Salt Lake City, Utah
Further vibration of the bed allowed the latter to
compact further. This caused probe 8 (green curve)
to gradually move above the compacting bed, its
output voltage progressively reducing to that corre-
sponding to the supernatant liquid. Figure 9 also
shows that the signals from the probes that are
within the bed converge as bed compaction in-
creases. This is expected as the solids
concentration within the particle bed is becoming
more uniform with increased compaction.
Figure 9 demonstrates the capability and sensitivity
of the multi-probe for real-time monitoring of bed
height level variations.
Figure 8. Dynamic cake height measurement based on Figure 10 shows dynamic cake height measure-
multi-probe output from Figures 6 and 7. ments obtained with iron ore fines slurries (d50=
150µm, d80= 600 µm) tested in the SLFT apparatus.
In addition to providing information that will allow Besides confirming that the multi-probe can work
testing classical filtration models, routine access to with virtually any mineral slurry, it is interesting to
dynamic cake height measurement has significant note the short duration of the cake formation phase,
practical potential. Indeed, analysis of Figure 8 which stresses the ability of the multi-probe to
reveals that 70% of the unflocculated filter cake capture extremely fast changing conditions.
deposits in only 25% of the total form time. Such
information suggests that desliming might have
significant potential on process throughput with this
particular coal fines sample.
Figure 9 highlights the sensitivity of the multi-probe
for dynamic mud-line detection. Here, the multi-
probe array was used to monitor the formation of a
packed bed of monosize glass beads under gravity.
The initial voltage rises relate to the formation of the
bed as with the previous two Figures. The loosely
packed bed that formed was such that the top
probes (ie probes 9 and 10 in Figure 4) are sitting
above the bed surface at the start of the experiment. Figure 10. Dynamic variation of cake height obtained with
The signal from both probes 9 and 10 is that of a low iron ore fines cakes (minus 1mm).
voltage (Vo in Equation 1), which corresponds to the
supernatant liquid above the bed. In fact, probe 9 Monitoring Filter Cake Permeability
appears to be initially seeing a bed interface, as As mentioned earlier, measurements of cake per-
revealed by the initial voltage rise, then the bed meability with conventional filtration tests is always
compacts and the signal from probe 9 drops back to in error since cake height is measured at the end of
the level of the supernatant liquid as the mud-line the filtration cycle whereas filtration rate is measured
moves below probe 9. at the end of the filtration cycle. This problem can
be overcome with the multi-probe array. Indeed,
when combined with a leaf test apparatus, it is
possible to obtain filtrate rate and cake height
simultaneously, allowing instantaneous cake perme-
ability to be calculated at any time during cake
formation. Instantaneous cake permeability is
defined as the permeability of the saturated filter
cake to water at a specific instant during cake
growth.
The following section shows the steps that lead to
the calculation of instantaneous filter cake perme-
Figure 9. Mud-line detection and cake compaction. ability.

5 Copyright © 2000 by SME


SME Annual Meeting
Feb. 28-Mar. 1, Salt Lake City, Utah
Figures 11 and 12 show measurements of dynamic
cake height and instantaneous filtrate flow rate for
unflocculated and flocculated fine coal filter cakes.

Figure 13. Permeability versus cake height for unfloccu-


lated and flocculated fine coal filter cakes.

It appears that cake permeability remains relatively


Figure 11. Dynamic cake height and instantaneous filtrate uniform during cake growth. To the best of the
flow rate for an unflocculated fine coal filter cake. authors' knowledge, instantaneous variations of
cake permeability have not been measured or
reported elsewhere under vacuum filtration condi-
tions.
Similarly, Figures 14 shows instantaneous cake
height and filtrate flow rate for a 20mm iron ore fines
filter cake while Figure 15 gives the corresponding
instantaneous cake permeability.

Figure 12. dynamic cake height and instantaneous filtrate


flow rate for a flocculated fine coal filter cakes.

At any time during cake formation, the instantane-


ous permeability K(t) is the is calculated according
to Darcy's law:

Q(t) × µ L(t)
K(t) = Figure 14. Dynamic cake height and filtrate rate for a 20mm
A ∆P (2) iron ore fines filter cake.

where t is real time, Q(t) is the instantaneous filtrate


flow rate, L(t) the instantaneous cake height, µ the
filtrate viscosity, A the effective filtration area and ∆P
the applied vacuum pressure. Application of Equa-
tion 2 to the data of Figures 11 and 12 yields the
instantaneous permeability for the unflocculated and
flocculated fine coal filter cakes. The results are
shown in Figure 13.

Figure 15. Cake permeability versus height for a 20mm iron


ore fines filter cake.

It is interesting to note that the filter cake perme-


ability starts plunging once the cake is in excess of
approximately 16mm. This result suggests that the

6 Copyright © 2000 by SME


SME Annual Meeting
Feb. 28-Mar. 1, Salt Lake City, Utah
top of the cake is significantly more impervious than is expected that dynamic measurements of cake
the bottom part. Photomicrographs of the filter cake permeability will contribute to improving our under-
were taken using a cake resin impregnation tech- standing of the filtration process.
nique described elsewhere (Bourgeois and Lyman,
1998). These are shown hereafter. Monitoring Cake Drainage
Figure 17 shows an almost complete multi-probe
output signal for a filtration test with an unflocculated
fine coal slurry. Cake formation, which takes up to
approximately 700 seconds in this experiment, is
analogous to Figure 6. Beyond 700 seconds,
individual electrode pairs measure a significant
voltage increase. This voltage increase corresponds
to the moisture front going past individual electrode
pairs, which can be seen as the water front going
from point G to point A in Figure 5.
In Figure 17, the first probe to see the air-water
interface is the bottom probe (probe 1), then probe 2
directly above probe 1, and so on. In other words,
this test reveals that the filter cake is draining from
bottom to top.

(a) Cross-section 10mm above bottom of the filter cake,


showing a coarse microstructure.

Figure 17. Monitoring dewatering with the multi-probe


device for an unflocculated fine coal filter cake.

Figure 18 shows the multi-probe output for the


entire filtration cycle with a flocculated fine coal
slurry. The most significant difference with Figure 17
is the apparent randomness of the drainage pattern,
as opposed to the bottom-to-top drainage of the
unflocculated cake shown in Figure 17. This differ-
(b) Cross-section at the top of the filter cake, revealing very ence in drainage behaviour between flocculated and
fine microstructure. unflocculated fine coal filter cakes has been ob-
Figure 16. Photomicrographs of iron ore fines cross- served repeatedly. This is most likely due to
sections for various elevations in the cake.
variations in microstructural homogeneity between
these filter cakes.
Visually, it is apparent that the cake microstructure
is significantly finer at the top of the cake due to
particle size segregation. This result is in agreement
with the multi-probe measurements of Figure 15.
The constant permeability assumption used in many
filtration models (Holsten, 1993) is clearly inappro-
priate for filter cakes with such strong anisotropy. It
7 Copyright © 2000 by SME
SME Annual Meeting
Feb. 28-Mar. 1, Salt Lake City, Utah
In addition to the processes that were described in
this paper, it is planned to use the probe for meas-
uring densification processes that occur during cake
forming and drying, dynamic variations of cake
porosity and dynamic moisture distribution.

ACKNOWLEDGMENTS
This work was conducted with the financial assis-
tance of the Cooperative Research Centre for Mining
Technology and Equipment (CMTE).

REFERENCES
Bhattacharya, I.N., 1997, "Settling and Filtration
Characteristics of Fine Alumina Trihydrate Slurry",
International Journal of Minerals Processing, Vol. 49,
pp. 107-118.
Figure 18. Monitoring dewatering with the multi-probe
device for a flocculated fine coal filter cake. Bourgeois, F.S., Wightman, E., Clarkson, C.J., Rui,
Z. and Davis, J.J., 1995, "A New Benchmark Single
Leaf Filtration Test for Fine Coal Dewatering",
CONCLUSIONS AND RECOMMENDATIONS
Proceedings of the Seventh Australian Coal Prepa-
This paper discussed a unique impedance multi- ration Conference, pp. 177-199.
probe array that was designed specifically for
monitoring dynamic variations in one-dimensional Bourgeois, F.S. and Lyman, G.J., 1997, "Morpho-
properties during the filtration process. It was shown logical Analysis and Modelling of Fine Coal Filter
through a selection of examples that the multi-probe Cake Microstructure", Chemical Engineering Sci-
offers significant potential for unravelling dynamic ence, Vol. 52, No. 7, pp. 1151-1162.
aspects of the filtration process. Holdich, R.G. and Sinclair, I., 1992, "Measurement
Tests conducted under vacuum filtration conditions of Slurry Solids Contents by Electrical Conductivity",
demonstrated the sensitivity of the multi-probe Powder Technology, Vol. 72, pp. 77-87.
device for monitoring dynamic filter cake growth. Holsten, C., 1993. "Cake Filtration Rate Equa-
Examples of probe outputs were given for coal and tions: A Review of Classical and Modern
iron ore fines filter cakes. Approaches", Minerals Engineering, Vol. 6, No. 7,
Dynamic cake height measurements obtained with pp. 775-783.
the multi-probe were combined with filtrate rate Meredith, R.E. and Tobias, C.W., 1962, "Conduc-
measurement to yield instantaneous permeability tion in Heterogeneous Systems", Advances in
variations during cake formation. This information is Electrochemistry and Electrochemical Engineering,
novel, and holds considerable potential for improving Vol. 2, pp. 15-46.
our understanding of the filtration process.
Nasr-El-Din, H. A., Shook, A.C. and Colwell, J.,
It was also shown that the probe is capable of 1987, "A Conductivity Probe for Measuring Local
capturing fast dynamic information about filter cake Concentrations in Slurry Systems", International
drying. As expected, interpretation of the probe Journal of Multiphase Flow, Vol. 13, No. 3, pp. 365-
output during cake drying is more complex and will 378.
be the object of future work. Although not discussed
in this manuscript, the multi-probe can be used in Nasr-El-Din, H.A., Mac Taggart, R.S. and Masli-
conjunction with the SLFT apparatus to measure the yah, J.H., 1996, "Local Solids Concentration
air permeability of the filter cake as it dries. This is Measurement in a Slurry Mixing Tank", Chemical
done by combining cake height as measured by the Engineering Science, Vol. 51, No. 8, pp. 1209-1220.
multi-probe with air flow rate as monitored by the Tarleton, E. S. and Hancock, D.L., 1996, Imaging
SLFT apparatus. of Filter Cakes Through Electrical Impedance
Efforts are currently directed towards calibration of Tomography, Filtration and Separation, June, pp
the impedance probe, such that the probe output 491 - 494.
can be reliably interpreted in terms of solids concen- Turner, J.C.R., 1976, "Two-Phase Conductivity -
tration and particle size distribution. The Electrical Conductance of Liquid-Fluidized Beds

8 Copyright © 2000 by SME


SME Annual Meeting
Feb. 28-Mar. 1, Salt Lake City, Utah
of Spheres", Chemical Engineering Science, Vol. 31, Zhu, Rui, 1997, "Measuring the Structure of Fine
pp. 487-492. Coal Filter Cakes Using Image Analysis", Ph.D.
Dissertation, Department of Mining, Minerals and
Wu, D.J., 1999, "Dynamic On-Line Monitoring of
Materials Engineering, The University of Queen-
Fine Coal Dewatering Performance", Masters
sland.
Dissertation, Department of Mining, Minerals and
Materials Engineering, The University of Queen-
sland.

9 Copyright © 2000 by SME

You might also like