Download as docx, pdf, or txt
Download as docx, pdf, or txt
You are on page 1of 16

1Filter Press Performance for Fast-Filtering Compressible

2Suspensions

3Anthony D Stickland1,1, Elizabeth H Irvin1, Samuel J Skinner1, Peter J Scales1,


4Andrew Hawkey2 and Francesco Kaswalder3
51Particulate Fluids Processing Centre, Department of Chemical and Biomolecular Engineering, The
6University of Melbourne, Parkville 3010, Australia
72Bilfinger Water Technologies, Auburn 3122, Australia
83Bilfinger Water Technologies, 48022 Lugo, Italy

9Abstract

10In some batch filtration processes, highly permeable suspensions dewater fast
11compared to the rest of the process. This work explores the impact of fast-filtering
12compressible materials on the throughput of fixed-chamber filter presses. The
13dewatering properties for a compressible yet highly permeable minerals processing
14slurry are used as inputs to a filter press model to explore the effects of operating
15and design parameters. For fast-filtering materials, maximum throughput is achieved
16with wide cavities and minimal handling time, while membrane resistance can be
17significant. Pressure has no effect beyond a small increase in maximum achievable
18concentration. Overall, this work demonstrates the combined use of material
19characterisation and device modelling for filter press optimisation.

20Keywords

21Filtration, Modeling, Permeability, Pressure filters, Suspensions

22 1. Introduction

23Ore grades are declining globally as higher-grade reserves are exploited and
24depleted. Lower ore grades not only increase the amount of waste, they also require
25increased milling and grinding for mineral liberation, leading to finer particle sizes,
26which affect downstream processes due to detrimental changes in rheological and
27dewatering properties. Finer particle suspensions are harder to pump at a given
28solids concentration due to increased shear yield stresses such that lower
29concentrations must be used, requiring more process water. Finer particles also
30settle slower and to a lower sediment concentration than larger particles such that
31more flocculant is used, less process water is recovered and the required tailings
32storage volume is increased. The poor dewaterability of fine mineral concentrates
33and tailings means that alternative dewatering techniques are required where
34pressures greater than those of gravity alone can reach high concentrations.
35Filtration under mechanical pressure is one way to process finer particles and reduce
36the tailings water content.

37Whilst recessed filter presses are certainly not new to the minerals processing
38industry, they are receiving increased attention. These filters are used throughout a
39number of industries and are capable of dewatering a range of slurry types such as
40water and wastewater treatment sludges, clays, sediments, etc. [1]. Filter presses
41consist of a series of vertical or horizontal plates [2]. Each plate is recessed such

11Email: stad@unimelb.edu.au
42that the plates form a series of cavities; commercial cavity widths vary from 25 to 45
43mm depending on the operation. A filter cloth is mounted to the face of each plate to
44provide a barrier to any solids in the suspension.

45Filter presses are operated in a batch process. Starting with the plate pack closed,
46the cavity is filled with suspension. Once all the cavities are full, the pumping
47pressure begins to rise and filtration begins. Slurry continues to be added and the
48filter cake builds up on the filter cloth. The pressure rises to a constant operating
49pressure, typically 7 to 12 bar. Filter designs can be fixed-cavity width (also called
50fixed-chamber or volume) where the only dewatering mechanism is the filling stage,
51or they can include variable widths (also called diaphragm or membrane filters). The
52latter allows the addition of hydraulic squeeze or pneumatic blow phases. Other
53presses incorporate cake washing features. At the end of the batch cycle, the press
54is opened and the filter cake is removed. Removal can be plate-by-plate or all at
55once. Depending on the design, the press may then have a cloth washing stage.

56Filtration performance is quantified as the throughput achieved for a given final filter
57cake concentration, although washing filters can also be rated in terms of recovery of
58valuable retentate or removal of contaminants. Prediction of filtration performance
59requires the dewatering properties of the suspension, the design of the press and the
60operating conditions [1, 3, 4]. Critical to accurate modelling is a representative
61constitutive model for the suspension behaviour. Fine minerals processing slurries
62are typically compressible [5] such that the achievable equilibrium concentration is
63pressure dependent or, alternatively, the strength of the material is concentration
64dependent. The rate of dewatering, usually given by a version of Darcy’s Law
65permeability, is also pressure (or concentration) dependent [6].

66Having a model of material behaviour and methods of characterisation that


67incorporate compressibility is vital for accurate predictions of filter performance. One
68such approach, coined ‘Compressional Rheology’, is based on the works of Buscall,
69White and Landman [7, 8]. Using solids volume fraction  as the measure of
70concentration, the compressive strength py() determines the extent of filtration. py()
71is zero at the gel point g, the concentration at which a networked suspension first
72forms, and equal to the applied pressure p in filtration at steady-state (py(∞) = p,
73where ∞ is the equilibrium cake concentration). The interphase hydrodynamic drag
74or hindered settling function R() determines the rate of filtration. R() approaches
75Stokes’ drag as   0, is inversely proportional to permeability and proportional to
76the specific cake resistance [9]. py() and R() can be characterised by laboratory-
77scale filtration [10], centrifugation [11] and gravity settling [12] experiments. Once
78measured, py() and R() can be used to predict the slurry behaviour in any
79dewatering device, provided a model of the device exists and there are no shear
80forces in the device. Numerous fine minerals processing slurries have been
81characterised using these methods (e.g. [13-15]) and successfully used to predict
82performance in thickeners [16] and filters [1, 3, 4, 17].

83Whilst material testing by filter press manufacturers is routine, comprehensive


84material characterisation combined with process modelling and optimisation are not.
85Instead, extensive pilot-scale testing and heuristics are usually used to inform filter
86press design and operation [18]. There are several reasons for this including the
87mathematical complexity of the problem, where complete solution in just one-
88dimension requires numerical methods, and the use of approximate analysis
89procedures to extract relevant properties from experimental tests. For many
90materials, characterisation procedures are not robust enough and suspension
91constitutive descriptions are too simple for models to be reliable. In contrast,
92validated models combined with accurately measured material properties based on a
93sound fundamental basis can be very useful, such as looking at the trends due to
94changes in process conditions and press design. This information can then be used
95to inform the direction of press optimisation and therefore limit the amount of
96expensive pilot-scale testing.

97Previous case studies of filter press optimisation using compressional rheology [1, 3,
984] have focussed on slow-dewatering suspensions where the majority of the batch
99cycle involves filtration or expression. For fixed-cavity width filter presses that
100process highly impermeable suspensions , optimisation is typically achieved by
101reducing the filtration time [1]. This requires small cavities and high feed
102concentrations so that high cake concentrations are reached quicker. Improvements
103to optimal material dewaterability through flocculants or other filter aids are highly
104beneficial. There are limited reasons for reducing the handling time beyond a certain
105point since it is not a major contribution to the overall cycle time. Since cake
106resistances are so high, membrane resistances can be fairly high without being
107detrimental to throughput, which reduces cloth washing requirements and permits
108membranes with improved particle capture and better filtrate quality.

109In contrast, filter press optimisation for fast-filtering compressible suspensions, where
110the handling time and membrane resistance are likely to be significant, has received
111limited attention [17]. The aim of this work was to measure the dewatering properties
112of a milled minerals processing slurry to predict full-scale filter press design and
113operation. The results from filtration and sedimentation tests are analysed to
114determine the dewatering properties of the suspension, which are used as inputs to a
115numerical model of a recessed filter press. The filtration parameters of feed
116pressure, cavity width, membrane resistance, handling time and initial concentration
117are varied in order to optimise the process. The limiting behaviours when the
118filtration time is negligible and membrane resistance overwhelming are explored.

119 2. Theory

120Filter Press Performance

121The specific slurry throughput Q (m3/(m2.s)) (or specific solids throughput 0Q, where
1220 is the feed solids volume fraction) of a batch filtration process is the volume of
123solids processed divided by the area and the total cycle time [1]:

V f +h 0
Q=
124 t f +t h …
125 (1)

126where Vf is the specific volume of filtrate at the end of filtration (m 3/m2), h0 is half the
127cavity width (m) (h0 is the full width for a single-sided cavity), tf is the fill time (s) and th
128is the handling time (s). Eq. 1 assumes that the cavity is a rectangular cuboid. The
129average solids volume fraction in the cavity f is given by a solid-phase volume
130balance over the filter [1]:

Vf

131
( )
φ f =φ0 1+
h0
…(2)

132The handling or technical time contains all the operations other than filtration.
133Starting from an open and clean press, the handling time can include closing the
134press, loading the suspension into the press (but not filtering), post-filtration cake
135manipulation such as squeezing, washing or blowing, press de-pressurisation, pipe
136flushing, press opening, cake removal and cloth washing. Most of these factors are
137determined by the press design; some indicative values for modern presses are
138given in Table 1. The time given for plate pack opening assumes that all cavities are
139opened simultaneously rather than one at a time and that there is good cake release
140(which is a function of cake dryness, cake weight and cake/membrane adhesion).

141Not included in Table 1 are other factors in the handling time include the loading
142time, which is given by the pump flowrate and press volume, and cake squeezing,
143washing and blowing, which depend on the material properties and the filter cake at
144the end of the fill stage.
145 Table 1: Typical durations of some of the different stages in the handling time of a fixed-cavity
146 filter press cycle (times are indicative only and will depend on press design and operation)

Cycle Stage Time (s)


Filtrate manifold drainage 30
Drip tray opening and closing 30
Simultaneous plate pack 60
opening
Plate shaking 30
Cloth rinsing 30
Plate pack closing 60
Total 240
147

148Vf for a given tf in Eq. 1 is determined by the material properties and the operating
149conditions. When the membrane resistance Rm is 0 and the operating pressure p is
150reached instantaneously, Vf is the square-root of tf [8] (this result is exact when 0 <
151g and approximate when 0  g [19]). In this case, by setting the derivative of Q with
152respect to tf to zero, the maximum throughput is when tf = th [19]. This rule of thumb
153gives an indication of how to optimise a filter. When a material is relatively slow to
154dewater such that tf >> th, changes to the material properties due to upstream
155processing or flocculation will have large effects on filter throughput whereas
156improvements to the handling time are not significant.

157When a material is relatively fast to dewater such that th >> tf, Eqs. 1 and 2 reduce to:
φ f h0
φ0 Q= t h >>t f
158 th …
159 (3)

160In this limit, improvements to handling time will bring large benefits but the material
161dewaterability is not important. The optimisation is more complicated when the
162handling time incorporates cake squeezing, blowing or washing due to their
163dependence on the preceding filtration stage and the cavity width, but Landman and
164White’s principle [19] of equating filtration and handling times remains (the additional
165‘cake manipulation’ stages add to the handling time).

166For fast-filtering materials, the filtration time may be dominated by the membrane
167resistance Rm rather than the cake resistance Rc, even though Rc increases as
168filtration proceeds. In the limit where Rm >> Rc, the specific filtrate rate is constant,
169i.e. dVf/dtf = p/Rm, where Rm is defined as Pa.s/m, which is proportional to the fluid
170viscosity . Upon substitution into Eqs. 1 and 2, this yields:
−1
φ h R h φ

171
φ0 t h [
Q= f 0 1+ m o f −1
Δ pt h φ0 ( )] Rm >> R c
…(4)

172One-Dimensional Filtration Modelling

173Cavity filtration is modelled by simplifying the process to a single dimension, z, with


174the origin at the membrane and h0 at the midline of a double-sided cavity (or the
175width of a single-sided cavity) [1]. The feed is assumed to enter the cavity at a plane
176at z = h0. The numerical model for fixed-width filter presses is based on the model for
177piston-driven filtration developed by Landman et al [8], where the cake formation
178stage is the build-up of the cake against the filter and the cake compression stage
179starts when the cake reaches the mid-plane. The governing equation to describe
180one-dimensional diffusion during compressional dewatering is a second-order
181hyperbolic partial differential equation:

∂φ ∂ ∂φ dV
182
=
∂t ∂z [
D(φ) +φ
∂z dt ] …(5)

183D() is the solids diffusivity, which describes the speed at which a concentration
184gradient propagates through the suspension. D() is a function of py() and R():

dp y (1−φ)2
D(φ )=
185 dφ R(φ ) …
186(6)

187The conditions at the start of filtration (once the cavities have been loaded with
188suspension) are:

φ( z ,0)=φ0
189 V (0)=0 …(7)
190The boundary condition at the membrane is:

dV
p y [ φ (0 ,t ) ] =Δp−Rm
191 dt …(8)

192where p is the applied pressure. If the cake hasn’t reached the mid-plane, then the
193top of the cake is at the maximum of 0 or g. Once the cake reaches the mid-plane,
194the iterative solution is given by the integrating the concentration to give the
195conservation of solids:
h0

∫ φ dz=φ0 (V +h 0 )
196 0 …(9)

197A numerical algorithm, in this case using a 4th-5th order adaptive Runge-Kutta
198method, uses the inputs of 0, py(), D(), p(t) and h0 to solve the governing
199equation subject to the initial and boundary conditions to give V(t).

200 3. Materials and Methods

201Material

202The sample characterised is a milled minerals processing slurry from the top of a
203flotation tank. The particle size distribution on a volume basis is d10 = 4.1 m, d50 =
20429.7 m and d90 = 99.8 m. The main components by mass as detected by x-ray
205fluorescence are silicon (76% as SiO2), aluminium (12% as Al2O3), iron (5% as Fe2O3)
206and potassium (4% as K2O), with the remainder trace components. The particle
207density, measured gravimetrically, is 2373 kg/m3.

208Material Characterisation

209Pressure filtration and batch settling tests are analysed to characterise the
210dewatering properties of the minerals processing slurry.

211Batch settling tests

212In the batch settling test, a measuring cylinder is filled with the slurry at an initial
213solids volume fraction of 0.138 and the height of the settling interface is recorded as
214it decreases over time. The results are used to determine R() and py() at low solids
215volume fractions near g [12].

216Filtration tests

217Laboratory-scale dead-end filtration tests are performed using the pressure rig
218described by de Kretser et al [10]. The rig applies a fixed pressure to a sample within
219a uniaxial filtration cell, causing a cake to build-up on the membrane and filtrate to be
220exuded. The pressure is applied by a pneumatic cylinder behind a piston and
221controlled via pressure transducer mounted in the face of the piston. The movement
222of the piston is measured using a linear encoder. The results from single pressure
223tests at pressures of 20, 50 and 100 kPa are used here to give rate information
224extracted from the slope of t versus V2, whilst extent information is taken from the
225final cake height. A stepped pressure test, in which each pressure is maintained until
226equilibrium is reached, is analysed to give the material compressibility at pressures of
22710, 20, 50, 100, 150, 225 and 300 kPa. The final cake concentrations are from oven
228drying at 80C.

229Filter Press Modelling

230The effects of altering various operating conditions on throughput and final cake
231solids concentration are examined using the model outlined in the theory. The
232parameters are chosen based on the range of values typically used in filter press
233operation and are outlined in Table 2. The initial solids volume fractions correspond
234to mass fractions from 0.15 to 0.45. The set points listed in the table show the value
235used when that particular parameter was not modified.
236 Table 2: Operating conditions used to model and optimise filter throughput. The set point is
237 the default value.

Parameter Values Set point


0 (v/v) 0.069, 0.153, 0.256 0.256
p (kPa) 700, 1100, 1500 700
h0 (mm) 25, 30, 35, 40, 45 25
th (s) 60, 120, 240, 480 240
Rm (Pa.m/s) 0, 108, 109, 5 x 109 0
238

239 4. Results and Discussion

240Material Characterisation

241The equilibrium cake concentrations from the filtration tests are plotted against the
242applied pressures to give py() at high concentrations while the settling test gives an
243estimate of the gel point as 0.254 along with some py() data above g (see Figure 1).
244The results show that the material is compressible since the equilibrium cake
245concentration is a function of the applied pressure. A power-law function that
246asymptotes to the gel point describes the data well.
1000
Compressive strength, py(f) (kPa)
py(f) = 0.225((f/0.254)17.46 – 1) kPa

100

10

Gel point fg

1 Pressure filtration

Gravity settling
0.1
0.2 0.25 0.3 0.35 0.4
Solids volume concentration, f (v/v)
247
248 Figure 1: Compressive strength py() for the minerals processing slurry from filtration and
249 gravity settling tests

250The slope of t versus V2 during the filter tests gives R() at high concentrations [20]
251while analysing the settling test using a Kynchian-type method gives low volume
252fraction data [12]. The R() results are shown in Figure 2. The data is fitted with a
2532nd order interpolating function to give a continuous function for R(). The values for
254R() are very low, indicating a material that is relatively fast to filter compared with
255many fine or clay-rich minerals processing slurries (e.g. [13, 21].
1.E+12
Hindered settling function, R(f)
Pressure filtration
1.E+11 Gravity settling
(Pa.s/m2)

1.E+10

Gel point fg
1.E+09

1.E+08
0.0 0.1 0.2 0.3 0.4
Solids volume fraction, f (v/v)
256
257 Figure 2: The hindered settling function, R(), from pressure filtration and gravity settling tests
258 of the mineral processing slurry

259Property Validation

260Since the extraction of material properties from the experimental data involves a
261number of approximations, it is important to verify that the properties are
262representative of the tests. The approximations used to analyse the cake formation
263region are not always valid [22, 23]. In this section of the paper, R() and py() are
264used as inputs to a numerical filtration model [8] along with the initial concentration,
265initial height, applied pressure and membrane resistance to predict the laboratory
266data. The predictions are reasonably accurate compared to the experimental data,
267as demonstrated by the results at 20 kPa in Figure 3. The values for the slope during
268the cake formation phase are very similar as are the final V2 values, such that the
269permeability and compressibility have been captured by py() and R(). The model
270predicts the onset of compression slightly later than the data and with steeper
271compression than the data. Despite these differences, the predictions validate the
272analysis procedure.
1200

20 kPa filtration test


1000
Model prediction

800
Time, t (s)

600

400

200

0
0 50 100 150 200 250 300
Specific volume squared, V2 î 106 (m2)
273
274 Figure 3: Laboratory filtration results at 20 kPa and re-prediction using extracted R() and py()
275 functions for the mineral processing slurry

276Filter Press Optimisation

277The fixed-cavity filter model is used with the inputs of py(), R(), 0, Rm, p, h0 and th
278to generate predictions of specific filtrate volume as a function of time. The specific
279solids throughput and average cake concentrations are calculated using Eqs 1 and 2.
280A parametric study of initial concentration, membrane resistance, applied pressure,
281cavity width and handling time is undertaken to investigate the optimisation of fixed-
282cavity filter performance for the minerals processing slurry.

283Eq 1 indicates that solids throughput versus concentration for a given set of
284conditions is expected to show low throughput at low cake concentrations since the
285amount of solids loaded per batch cycle is low. The throughput is also low at high
286concentrations since it takes an infinite amount of time to reach the compressibility
287limit of the material at that pressure, ∞. The position of the maximum between the
288low throughput limits depends on the handling and filtration times relative to each
289other. For a material that is fast to dewater, as in this case, the filtration time is small
290compared to the handling time and the material permeability is not important, in
291which case the maximum throughput is at high concentrations when the most solids
292are loaded per run and the throughput generally increases as a function of cake
293solids. This is in contrast to slow dewatering materials where the maximum
294throughput is at low concentrations since it takes a long time to filter to high
295concentrations and the throughput generally decreases as a function of cake
296concentration [1].

297The modelling results for feed volume fractions of 0.069, 0.153 and 0.256 (equivalent
298to mass fractions of 0.15, 0.30 and 0.45) are shown in Figure 4. For the highest feed
299concentration, the throughput is proportional to f and follows the th >> tf limit almost
300up to the compressibility limit at  = 0.397. The throughput deviates earlier from the
301th >> tf limit at lower feed concentrations as the filtration time begins to become
302significant such that lower feed concentrations have lower solids throughputs at high
303cake concentrations. In general, this follows the trend in most separations processes
304that higher solids throughputs are achieved through higher feed concentrations.

200
ɸ0 = 0.256 v/v
Specific solids throughput, f0Q

ɸ0 = 0.153 v/v
ɸ0 = 0.069 v/v
150 Q max
Limits
(kg/h.m2)

Compressibility limit
100

ff = f∞
50
th >> tf limit
f0Q = 3600r sffh0/th
0
0.0 0.1 0.2 0.3 0.4 0.5

305 Average cake solids volume fraction, ff (v/v)


306 Figure 4: Fixed-cavity model predictions for the minerals processing slurry at varying feed
307 concentrations (Rm = 0, p = 700 kPa, h0 = 0.0125 m, th = 240 s)

308The membrane resistance is important when it is large compared to the cake


309resistance and when the filtration time is at least comparable to the handling time.
310The limiting behaviour in this case is given by Eq. 4. The effects of membrane
311resistance on solids throughput are modelled here using values of Rm from 0 to 5 x
312109 Pa.s/m (see Figure 5). The results show that resistances below 10 8 Pa.s/m are
313insignificant whereas the membrane resistance begins to affect the filtration
314behaviour above 109 Pa.s/m. This is not a large value (values of 10 10 or 1011 Pa.s/m
315have been measured [1]), reflecting that membrane selection and cloth washing are
316important for fast-filtering suspensions such as this sample.

317The magnitude of the filtration impairment caused by the membrane resistance


318(effectively a scaled membrane resistance) is given by the right-hand side of Eq. 4.
319At f = 0.397, the values of Rm used in the predictions correspond to the following
320values in Eq. 4:

Rm ho φf

321
Δ pt h φ0 ( )
−1 ={0 , 0 . 004 , 0 . 04 , 0 . 2}
…(10)
322This shows that Rm = 5 x 10-9 Pa.s/m reduces the throughput by at least 20%.

200
Rm = 0
Specific solids throughput, f0Q

Rm = 10^8 Pa.s/m
180 Rm = 10^9 Pa.s/m
Rm = 5 x 10^9 Pa.s/m
Q max
160 Limits
(kg/h.m2)

Compressibility limit
140 th >> tf limit

ff = f∞
120
Rm >> Rc limits

100
0.25 0.3 0.35 0.4 0.45
Average cake solids volume fraction, ff (v/v)
323
324 Figure 5: Fixed-cavity model predictions for the minerals processing slurry at varying
325 membrane resistances (0 = 0.256, p = 700 kPa, h0 = 0.0125 m, th = 240 s). The Rm >> Rc limits
326 are given by Eq. 4.

327Another adjustable parameter in filtration is the applied pressure. Increasing the


328operating pressure increases the dewatering rate and the extent of dewatering for
329compressible materials. The model predictions for three feed pressures from 700
330kPa to 1,500 kPa are shown in Figure 6. The maximum achievable cake
331concentration shows a small increase from 0.397 to 0.410, but there is little
332difference between the throughputs for this material under these conditions. The
333benefits of increasing the pressure are minimal since the material is not very
334compressible and the filtration rate is already very fast compared to the handling
335time.

336Cavity width is a critical parameter in maximising throughput per unit filter area –
337increasing the cavity width increases the solids loading per cycle but also increases
338the filtration time required to reach a certain cake concentration. In general (within
339mechanical restrictions), the cavity width should be chosen to give a cake of the
340desired concentration in a filtration time roughly equal to the sum of all the other
341times in the batch cycle [19]. The model predictions for cavity widths from 25 to 45
342mm (0.0125 m < h0 < 0.0225 m) are shown in Figure 7. Note that the loading time, tL,
343which is part of the handling time, increases proportionally with the cavity width.

344Increasing the cavity width increased the throughput over the entire range of cake
345concentrations. This behaviour is in contrast to slower-filtering materials such as
346water treatment sludges [1] or fine iron ore tailings [17] where large cavity widths and
347long filtration times lead to low throughputs at high cake concentrations.
200
p = 700 kPa
Specific solids throughput, f0Q
p = 1100 kPa
180 p = 1500 kPa
Q max
Limits
160
(kg/h.m2)

Compressibility limit
140 th >> tf limit

ff = f∞
120

100
0.25 0.3 0.35 0.4 0.45
Average cake solids volume fraction, ff (v/v)
348
349 Figure 6: Fixed-cavity model predictions for the minerals processing sample at varying
350 operating pressures (0 = 0.256, Rm = 0, h0 = 0.0125 m, th = 240 s)

300
h0 = 0.0225 m, tL = 36 s
h0 = 0.02 m, tL = 32 s
Specfic solids throughput, f0Q

h0 = 0.0175 m, tL = 28 s
250 h0 = 0.015 m, tL = 24 s
h0 = 0.0125 m, tL = 20 s
Q max
(kg/h.m2)

Limits
Compressibility limit

200
ff = f∞

150

th >> tf limit
100
0.25 0.3 0.35 0.4 0.45
Average cake solids volume fraction, ff (v/v)
351
352 Figure 7: Fixed-cavity model predictions for the minerals processing sample with varying
353 cavity widths (0 = 0.256, Rm = 0, p = 700 kPa, th = tL + 220 s)
354The final modelled variable was the handling time. The results for handling times
355between 60 s and 480 s are shown in Figure 8. Due to the fast filtering material, the
356results were close to the th >> tf limit and the solids throughput was inversely
357proportional to the handling time.

300
h0 = 0.0225 m, tL = 36 s
h0 = 0.02 m, tL = 32 s
Specfic solids throughput, f0Q

h0 = 0.0175 m, tL = 28 s
250 h0 = 0.015 m, tL = 24 s
h0 = 0.0125 m, tL = 20 s
Q max
(kg/h.m2)

Limits

Compressibility limit
200

ff = f∞
150

th >> tf limit
100
0.25 0.3 0.35 0.4 0.45
Average cake solids volume fraction, ff (v/v)
358
359 Figure 8: Fixed-cavity model predictions for the minerals processing sample with varying
360 handling times (0 = 0.256, Rm = 0, p = 700 kPa, h0 = 0.0125 m)

361As mentioned in the introduction, some presses include hydraulic squeeze stages to
362remove additional water and homogenise the cake, pneumatic blow stages to
363desaturate the cake, and wash stages to remove valuable liquor or impurities. This
364work does not consider these additional stages since they are dependent on the
365preceding filtration stage, although incorporation of these stages provides an obvious
366extension to this work. However, since they all effectively add to the handling time, it
367is expected that the optimisation trends identified here will still hold.

368 5. Conclusions

369This work characterises the dewatering properties of a milled minerals processing


370slurry using laboratory-scale filtration and sedimentation tests. The properties
371indicate a compressible but highly permeable suspension. py() and R() are used in
372a numerical filtration models that is able to successfully re-predict the laboratory
373experimental data, thus validating the analysis.

374The limiting behaviours for fast-filtering materials in fixed-chamber filter presses


375identified here are general and can be applied to numerous industries and
376suspensions. The material properties for the milled concentrate are used to illustrate
377optimal filter performance when th >> tf and when Rm >> Rc. Maximising the solids
378loading per cycle through large cavity widths (using the widest physically allowable)
379and minimising the handling time has the greatest effects on filter press performance.
380Simultaneous rather than sequential plate pack opening is highly beneficial to reduce
381the handling time. Since the material is fast to dewater, the membrane resistance
382could be significant and consideration of cloth type and cloth washing protocols is
383essential. Other than marginally increasing the maximum achievable cake solids
384volume fraction, increasing the operating pressure does not have much effect.

385The limiting behaviour when th >> tf and when Rm >> Rc is very different to the limiting
386behaviour for slow-filtering suspensions identified in previous case studies [1, 17, 24].
387Combined, the different case studies cover a large range of scenarios for fixed-
388chamber presses, which highlights how understanding material properties can enable
389press operators and manufacturers to intelligently optimise design and operating
390conditions to maximise filter performance.

391Acknowledgements

392An earlier version of this paper [25] was presented by the lead author at
393FILTECH2015 in Cologne, Germany, 24-26 February 2015. Amar Sharma is
394thanked for performing the laboratory filtration and settling experiments. Bilfinger
395Water Technologies are thanked for the sample and student supervision; there was
396no commercial arrangement between Bilfinger Water Technologies and The
397University of Melbourne for this work. Samuel Skinner is supported by the Elizabeth
398and Vernon Puzey Scholarship. Infrastructure support was provided by the
399Particulate Fluids Processing Centre.

400Symbols Used

401Latin Symbols

402h0 [m] cavity half-width


403py() [Pa] compressive strength
404Q [m s-1] specific slurry throughput
405R() [Pa s m-2] hindered settling function
406Rc [Pa s m-1] cake resistance
407Rm [Pa s m-1] membrane resistance
408tf [s] fill time
409th [s] handling time
410tL [s] loading time
411Vf [m] specific volume of filtrate

412Greek symbols

4132 [m2 s-1] d(V2)/dt during cake formation


414p [Pa] operating pressure
415 [-] solids volume fraction
416f [-] average filter cake solids volume fraction
417g [-] gel point solids volume fraction
4180 [-] feed solids volume fraction
419∞ [-] equilibrium solids volume fraction
420References

4211. A.D. Stickland; R.G. de Kretser; P.J. Scales; S.P. Usher; P. Hillis; M.R.
422 Tillotson. Chem. Eng. Sci. 2006, 61 (12), 3818-3829.
4232. L. Svarovsky, Solid-liquid separation. 2nd ed, London: Butterworths. 1981.
4243. A.D. Stickland; R.G. de Kretser; A.R. Kilcullen; P.J. Scales; P. Hillis; M.R.
425 Tillotson. AIChE J. 2008, 54 (2), 464-474.
4264. M. Eberl; K.A. Landman; P.J. Scales. Colloids Surf., A. 1995, 103, 1-10.
4275. H.P. Grace. Chem. Eng. Progr. 1953, 49 (6), 303-318.
4286. F.M. Tiller; M. Shirato. AIChE J. 1964, 10 (1), 61-67.
4297. R. Buscall; L.R. White. J. Chem. Soc. Faraday Trans. I. 1987, 83 (3), 873-891.
4308. K.A. Landman; C. Sirakoff; L.R. White. Phys. Fluids A. 1991, 3 (6), 1495-1509.
4319. R.G. de Kretser; P.J. Scales. Trans. Filt. Soc. 2007, 7 (1), 60-66.
43210. R.G. de Kretser; S.P. Usher; P.J. Scales; D.V. Boger; K.A. Landman. AIChE
433 J. 2001, 47 (8), 1758-1769.
43411. R. Buscall. Colloids Surf. 1982, 5 (4), 269-283.
43512. D.R. Lester; S.P. Usher; P.J. Scales. AIChE J. 2005, 51 (4), 1158-1168.
43613. J. Hulston; R.G. de Kretser; P.J. Scales. Int. J. Miner. Proc. 2004, 73 (2-4),
437 269-279.
43814. R.G. de Kretser; P.J. Scales. J. Colloid Interface Sci. 2008, 328 (1), 187-193.
43915. S.P. Usher Ph.D. Thesis, The University of Melbourne, 2002.
44016. S.P. Usher; P.J. Scales. Chem. Eng. J. 2005, 111 (2-3), 253-261.
44117. R.G. de Kretser; H. Saha; P.J. Scales. Chem. Eng. Sci. 2010, 65, 2700-2706.
44218. R.J. Wakeman; E.S. Tarleton. Filt. Sep. 1999, 24-31.
44319. K.A. Landman; L.R. White. AIChE J. 1997, 43 (12), 3147-3160.
44420. K.A. Landman; J.M. Stankovich; L.R. White. AIChE J. 1999, 45 (9), 1875-
445 1882.
44621. S.P. Usher; R.G. de Kretser; P.J. Scales. AIChE J. 2001, 47 (7), 1561-1570.
44722. R.G. de Kretser; A.D. Stickland; S.P. Usher; P.J. Scales. FILTECH2011.
448 Wiesbaden, Germany, 22-24 March, 2011
44923. A.D. Stickland; C. Burgess; D.R. Dixon; P.J. Harbour; P.J. Scales; L.J. Studer;
450 S.P. Usher. Chem. Eng. Sci. 2008, 63 (21), 5283-5290.
45124. R.G. de Kretser; H. Saha; C. Biscombe; P.J. Scales. FILTECH2009.
452 Wiesbaden, Germany, 13-15 October, 2009
45325. A.D. Stickland; E.H. Irvin; S.J. Skinner; P.J. Scales; A. Hawkey; F. Kaswalder.
454 FILTECH2015. Cologne, Germany, 24-26 February, 2015
455

You might also like