University
stringclasses 19
values | Text
stringlengths 458
20.7k
|
|---|---|
Virginia Tech | aqueous SO oxidation by dissolved O was controlled by adsorption of SO onto the
2 2 2(aq)
catalyst, activated carbon. With respect to the oxidized species, this conclusion is disputed in
a more recent study also using activated carbon as a catalyst for the reaction. Of the detected
S(IV) species, HSO - and SO , it was reported that HSO - was the oxidized species and
3 2(aq) 3
SO was non-reactive (Govindarao and Gopalakrishna, 1995). Furthermore, it was
2(aq)
observed that the SO had a deactivating effect on the catalyst due to its competition with
2(aq)
HSO - for adsorption sites and its retention on the catalyst surface after adsorption. This
3
blinding effect was also observed in a similar study using poly-4-vinylpyridine-Cu as a
catalyst (Kumar et al., 1996).
Yet another study of aqueous SO oxidation by O , again in the presence of activated
2 2
carbon, examined reaction kinetics and found that the reaction was limited by solution
acidity, which limited aqueous oxidant concentration since with increased acid concentration,
O becomes less soluble (Komiyama and Smith, 1975).
2
The specific limitations and catalyst blinding experienced in the previous SO
2
oxidation studies are not expected to be problematic in the Fe(II) regeneration, although
similar phenomena could occur. First, as mentioned above, at high acid concentrations – pH
below two – SO not HSO -, is the predominant species when SO is dissolved
2(aq), 3 2(g)
(Govindarao and Gopalakrishna, 1995), (Garcia et al., 1998) and (Krissman et al., 1998).
Since typical electrolyte pH encountered in Fe(II) regeneration is well below two, SO will
2(aq)
be the species oxidized. SO is not easily oxidized by O , but it is by Fe(III). Next, while
2(aq) 2
the oxidant, Fe(III), is fully soluble in the acidity and temperature ranges typical for EW
electrolytes, the reducer, S(IV), may be difficult to keep dissolved, due to its partial pressure,
if reactors are not kept airtight. Finally, any catalyst deactivation by non-reactive species
would have to be caused by Fe(II) adsorption, which will be non-reactive in the absence of
O ; however, since published diffusivity data indicates that SO reaches the catalyst
2 2(aq)
surface much faster than iron ions, deactivation is not expected to occur (Han, 1990),
(Freiberg and Schwartz, 1981) and (Ramsing and Gundersen, no year given).
10 |
Virginia Tech | To further increase the reaction rate of aqueous SO oxidation by O , electro-catalysts
2 2
have been investigated, including the use of transition metals as mediators to provide an
alternate reaction route between the catalyst and reacting species (Garcia et al., 1998)
(Berglund et al, 1993). The Fe(II)/Fe(III) reduction-oxidation couple has been the subject of
experimentation for this purpose, and so associated literature may be reversibly applicable to
the Fe(II) regeneration process, accounting for both relevant half-reactions (Fe(III) reduction
and SO oxidation).
2
A particular study conducted by Garcia et al. in 1998 examined aqueous SO
2
oxidation with an Fe(II)/Fe(III) iron cycle in the presence of dissolved O and graphite, a
2
catalyst similar to activated carbon but of much lower surface area, whereby Fe(III) was
generated by Fe(II) oxidation and then used to oxidize the SO . It should be noted that while
2
the above study bears some similarities to the process of current interest, it was conducted
under different conditions (i.e., slightly lower molar acidity, lower temperature, differing
Fe(III), Fe(II), SO ratios, and batch-type tests rather than continuous flow) than the Fe(II)
2
regeneration must be and with different goals in mind.
At 3M H SO , all of the S(IV) was assumed to be of the form SO – as will be the
2 4 2(aq)
case in Fe(II) regeneration – which behaved differently in the absence and presence of
Fe(III). With only Fe(II) and SO in solution with dissolved O , the SO was almost
2(aq) 2 2(aq)
non-reactive and the Fe(II) oxidation occurred rapidly. This result is in agreement with other
studies on aqueous S(IV) oxidation, which concluded that the SO species was not
2(aq)
oxidized by O , but that HSO - was. For the Fe(II) regeneration process, the absence of
2 3
SO oxidation gives some insight into consequences of disproportional ratios of Fe(III) to
2(aq)
SO ratios: if ample Fe(III) is not available for reaction, SO may not be reacted,
2 2(aq)
producing problems in the EW electrolyte circuit (e.g., SO escape or depleted acid
2(g)
concentration). Also, if any O is present in the electrolyte, Fe(II) will rapidly oxidize and
2
counteract the desired process results.
As for SO behavior when Fe(III) was introduced, in batch tests of solutions
2(aq)
containing SO , Fe(II) and Fe(III) with dissolved O , SO was readily oxidized.
2(aq) 2 2(aq)
11 |
Virginia Tech | Additionally, it was observed that the oxidation was inhibited somewhat by Fe(II), due in
part to both the initial Fe(II) concentration and that produced by Fe(III) reduction. This
inhibition, which was explained by the competitive adsorption of Fe(II) on the catalyst, will
be to a lesser extent for the Fe(II) regeneration process, if present at all. The regeneration
process is continuous flow and the activated carbon catalyst should provide a larger number
of surface sites for adsorption, as opposed to the graphite used in the above study.
Although catalysts like carbon may provide two mechanisms of catalysis for
reduction-oxidation reactions, surface adsorption and electron transfer, this was not found to
be the case for the Fe(III)-SO reaction in the above study. Instead, it was proposed that
2(aq)
under the conditions studied, iron ions adsorbed onto the catalyst surface, inhibiting
adsorption of SO molecules, and surface-bound Fe(III) reacted with SO in solution,
2(aq) 2(aq)
meaning the reaction was only catalyzed by adsorption of one species, and not by electron
transfer through the solid. However, that mechanism does not seem plausible for the Fe(II)
regeneration process considering published diffusion coefficients. It is more probable that
the SO will quickly adsorb onto the carbon surface, followed by iron ions; Fe(III) will
2(aq)
immediately react with the SO and Fe(II) will be non-reactive so long as dissolved O is
2(aq) 2
negligible.
Regarding other major findings during review of the pertinent literature, one
similarity between studies involving SO was the problematic nature of SO measurement.
2 2
Iodimetry was used to directly measure SO in solution; although several authors noted
2
issues with loss of concentration during sample measurement and/or transport, and also with
separation if multiple species were present that might react with the triiodide used during
titration (Kumar et al., 1996), (Govindarao and Gopalakrishna, 1995) and (Garcia et al.,
1998). As well, several attempts were made in the current work to convert aqueous SO to
2
gaseous phase for measurement. This method proved difficult and inaccurate due to the
complexity of procedures required to drive all SO out of solution and capture it within a
2
known volume. In the 1998 Garcia et al. study, indirect measurements - determining changes
in SO concentration by stoichiometry with other measurable quantities – were reported to be
2
more accurate than direct methods. An SO -specific probe was found to be available during
2
12 |
Virginia Tech | a literature search on aqueous SO measurement; however, the probe was observed to be
2
unreliable during tests (Cook, 2004).
The Fe(II) regeneration process in AART using FFS is certainly complex and various
issues must be considered to study it. Combining general knowledge about solution and
catalysis chemistry with given process conditions and a review of reactions similar to those
of interest, several major points become clear:
• There are a number of variables that might affect the process, which may or may not
be dependent on each other;
• The kinetics of the process are likely governed by the rate of surface adsorption of
reactant(s) onto the catalyst;
• Measurement of SO concentrations may prove problematic.
2
Objectives
A better understanding of the Fe(II) regeneration process may prove beneficial in
numerous areas, including design and improvement of the process, its place within the EW
circuit, equipment and materials. Lessons learned with respect to the Fe(II) regeneration
might also shed light on other similar processes. With the above points in mind, it is possible
to study the regeneration from at least two perspectives, the first being a fundamental look at
the regeneration reaction and the second an industrial type optimization. For optimization, it
may only be necessary to relate the process outputs to the inputs, and then determine the
combination of input variables producing the most desirable results. This type of study is
quite common industrially, and can provide information for both improving a current process
and designing a new process or operation. For a fundamental study, it is still important to
relate process outputs to inputs, however a closer look might also be taken at the mechanics
of the relationship. With respect to the Fe(II) regeneration, process factors might be related
to results by the chemical reaction mechanisms occurring within the process. Through an
internal understanding of a process, a more fundamental perspective can also provide insight
into potential improvements.
13 |
Virginia Tech | Scope
To examine what factors affect the Fe(II) regeneration process, and how and why
they affect it, the two studies below were designed and conducted, and the details are
contained in the following papers. The studies were completed concurrently, so the design of
one was not typically dependent on the other.
Fundamental Kinetics
For a better internal understanding of the Fe(II) regeneration process, a model of the
fundamental kinetics was developed. The goal of the modeling process was to capture the
major reaction mechanism(s) and quantify the specific rate constant(s). General knowledge
was combined with observations of comparable reviewed reactions and a basic kinetic model
was hypothesized for the Fe(II) regeneration. Based on the model, variable process factors
expected to be most influential to the reaction rate were carbon surface area, initial Fe(III)
concentration and flow rate. The model was validated experimentally by collecting data
relating process variables to the reaction rate. The final model provides insight into both the
current process and other processes with analogous reaction mechanisms.
Empirical Optimization
For the purpose of improving the current Fe(II) regeneration process, an empirical
optimization was completed. The goal was to maximize throughput at an acceptable level of
Fe(III) reduction (Fe(II) regeneration) and without “breakthrough” (un-reacted) SO in the
2
effluent, which is potentially hazardous. Again, general knowledge and the literature defined
influential factors, which were varied over operable ranges during experimentation, to
examine their effects on process responses. Factors considered most important and included
in the optimization were carbon surface area, initial Fe(III) and SO concentrations, and
2
temperature. Design Expert, a software package by Stat-Ease, was utilized in experimental
design, to analyze and model experimental data, and, then, to optimize the process. Both the
model and optimization results may be useful in Fe(II) regeneration process design and
improvement.
14 |
Virginia Tech | Chapter 2: Fundamental Kinetics of the Ferrous
Regeneration for Alternate Anode Reaction Technology
Emily Allyn Sarver
Abstract
The Fe(II) regeneration process is an important aspect of Alternate Anode Reaction
Technology (AART) using a Fe(II)/Fe(III)-SO (FFS) for copper hydrometallurgy. The
2
process is basically Fe(III) reduction by SO , which is catalyzed by activated carbon
2(aq)
particles. For the current work, experiments have been conducted to examine the
fundamental kinetics of the process, including the primary reaction mechanism and the
effects of four variable factors – carbon particle size, flow rate, and initial Fe(III) and SO
2
concentrations – on the Fe(II) regeneration rate. As expected, the regeneration reaction is
mass transfer-controlled, and the rate is limited by the diffusivity of Fe(III). Carbon particle
size and initial Fe(III) are the most influential factors under the conditions tested.
Additionally, initial SO concentration has been determined to be insignificant to the reaction
2
rate, and flow rate affects the reaction rate via its affects on diffusivity. The following rate
expression has been hypothesized and validated by experimental data, with the diffusivity of
the Fe(III) ions being observed as 1.1x10-7 cm2/s.
⎡ 1 2 ⎤
dFe2+ 6M 2D V 2D 3
= ⎢ f +0.6 t f ⎥[C ]
dt ρdV ⎢ d 1 1 ⎥ Fe3+
d 2v 6
⎣ ⎦
15 |
Virginia Tech | Introduction
Copper is currently the third most used metal in the world, primarily because of
its exceptional electrical conductivity and resistance to corrosion. In 2003, about 17
million metric tons of copper were produced to supply the growing global demand
(ICGS, 2003). In addition to increasing demand, copper prices have also been on the
rise, skyrocketing to over $1.50 per pound in April 2005, up from just $0.68 only two
years earlier (LME, 2005). Rising prices are due in part to growing production costs,
which are not expected to decline to previous levels when energy was less expensive.
High prices present a real advantage to low-cost producers, and new technologies that are
capable of cutting costs are under serious consideration by the copper industry. One such
technology – alternate anode reaction technology (AART) – has shown potential for a
major reduction in power consumption, a primary cost center for electrowinning (EW) in
hydrometallurgical production of copper (Sandoval and Lei, 1993; Sandoval et al., 1995;
Sandoval and Dolinar, 1996).
AART basically changes the conventional EW anode reaction from water oxidation
to ferrous (Fe(II)) to ferric (Fe(III)) iron oxidation, while the cathode reaction remains the
same. This change results in two major advantages over the conventional process:
substantial reduction in overall energy (power) requirements for EW and elimination of
hazardous acid misting. The power reduction can be attributed to the reduced equilibrium
potential of the alternate anode reaction, and the elimination of acid misting to the absence of
O production at the anode, which is present with the conventional anode reaction. One
2
major aspect of the above AART is the necessity of Fe(II) regeneration – reducing the Fe(III)
produced at the anode back to Fe(II) – to maintain recyclable EW electrolyte streams.
The method of regeneration that has been most successful during testing is Fe(III)
reduction by SO . This method is attractive because SO can be potentially obtained from
2 2
nearby copper smelters, relatively easily and inexpensively. Also, the use of SO as the
2
reducer produces recoverable sulfuric acid as a by-product, which is beneficial because the
acid may be utilized in leaching operations or sold (Sandoval and Dolinar, 1996). The
16 |
Virginia Tech | combination of the Fe(II)-Fe(III) anode reaction and the subsequent Fe(II) regeneration by
SO is abbreviated FFS. The FFS electrode reactions for copper EW and the Fe(II)
2
regeneration reaction are shown below in Equations 1-4.
Anode reaction: 2×(Fe2+ → Fe3+ +e−) (1)
Cathode reaction: CuSO +2e− →Cu0 +SO 2− (2)
4 4
Overall EW reaction: CuSO +2Fe2+ →Cu0 +SO 2− +2Fe3+ (3)
4 4
Fe(II) regeneration reaction: SO +2Fe3+ +2H O → 2Fe2+ +SO 2− +4H+ (4)
2 2 4
As can be seen above, for every mole of plated copper (Cu0), two moles of Fe(III) are
generated at the anode and must be reduced back to Fe(II) by a mole of SO . In practice, the
2
SO is injected into the Fe(III)-rich electrolyte after it leaves the EW cells, and then the
2
solution is passed through a bed of activated carbon where the regeneration reaction occurs,
shown in Figure 2.1. The role of the activated carbon is as a catalyst, which is discussed
below. The recoverable acid by-product is produced when the excess H+ in Equation 4 reacts
with the copper sulfate (CuSO ) already contained in the electrolyte.
4
Figure 2.1 – Fe(II) Regeneration Reactor for AART Using FFS
Although the Fe(II) regeneration process has been successfully managed during both
bench and pilot scale testing of AART using FFS, little has been done to study it specifically.
A basic understanding of the reaction mechanism(s) and kinetics surrounding the
17 |
Virginia Tech | regeneration may provide insight into the current process or other similar processes. Such
understanding may be helpful in design or optimization of a process, a circuit, or equipment,
and potentially improve process results or efficiency or save costs. As such, the objective of
the current work was to develop a fundamental rate model for the Fe(II) regeneration. To do
so, it was first necessary to determine factors potentially affecting the regeneration and the
most probable reaction mechanism, such that a model could be hypothesized. Experimental
data were then collected and used to fit unknown model terms.
Fundamental Rate Model Development
Following is the development of a fundamental rate model for the Fe(II) regeneration
process, supported by a review of relevant literature and principles of reaction mechanics and
kinetics.
First, the regeneration reaction given in Equation 4 accounts for the overall reaction,
whereby the end products of the reaction between SO and Fe(III) in aqueous solution are
2
S(VI) (as sulfuric acid) and Fe(II); however, Equation 4 does not address any intermediate
reactions, the specific S(IV) species oxidized or the implications of such details. For the
Fe(II) regeneration, SO is introduced as a gas into the electrolyte before it enters the reactor.
2
A review of SO solution chemistry provides the following series of reactions (Equations 5-
2
8) that may occur upon dissolution of gaseous SO in aqueous solution:
2
SO ⇔ SO (5)
2(g) 2(aq)
SO + H 0 ⇔ H SO (6)
2(aq) 2 2 3
H SO ⇔ HSO − +H+............K =1.4×10−2 (7)
2 3 3
HSO − ⇔ SO 2− +H+..............K =6.3×10−8 (8)
3 3
When dissolved, SO may be of the form SO , or form sulfurous acid (H SO ) and
2 2(aq) 2 3
immediately dissociate to bisulfite (HSO -) or sulfite (SO 2-) as shown above. H SO is so
3 3 2 3
unstable that it is virtually undetectable spectroscopically in solution and will, therefore, not
be significant in the Fe(II) regeneration (Shriver and Atkins, 2003). Also, due to the relative
dissociation constants in Equations 7 and 8, SO 2- concentration will also be negligible or
3
nonexistent, leaving SO and HSO - as possible species. The acidity of the electrolyte will
2(aq) 3
18 |
Virginia Tech | play a major role in determining the extent of each. It has been established that in highly
acidic solutions– below pH of about 2 – SO is the predominant species upon dissolution
2(aq)
of SO (Krissman et al., 1998; Govindarao and Gopalakrishna, 1995). Due to the extreme
2(g)
acid concentration – nearly 4M – of the electrolyte in the Fe(II) regeneration process, it can
be safely assumed that essentially all of the dissolved SO will be of the form SO . As
2 2(aq)
such, intermediate reactions were not considered in the rate model development.
Unfortunately, it has also been observed that the homogeneous reaction between
S(IV) and Fe(III) is much slower when SO is the reacting species rather than HSO -
2(aq) 3
(Garcia, et al., 1998). Specifically under the conditions encountered in the Fe(II)
regeneration, the slowness of the Fe(III)-SO reaction was reported separately by Sandoval
2(aq)
and Dolinar in 1996. While the reaction rate can be increased by catalysis as discussed
below, electrolyte acid concentration cannot be changed without significantly affecting other
operations within the scheme of EW, and the entire hydrometallurgical process including
leaching and solvent extraction. Therefore, the rate model developed in the current work is
specific to a Fe(III)-SO reaction in solutions of high acidity.
2(aq)
For the Fe(II) regeneration, the reaction rate can first be expressed generically and
then expanded to include the mechanism(s) and/or known process factors, like flow rate.
Rate expressions for chemical processes are typically written as a function of the reactants
and products of the process – irreversible processes only require known concentrations of
either reactants or products. The Fe(II) regeneration reaction (Equation 4) may be assumed
irreversible, and if the reactants, Fe(III) and SO , react only with each other (i.e., the
2(aq)
consumption ratio of each reactant is dictated by the reaction stoichiometry), a rate
expression of the simplest form may be applied. Such an expression is shown in Equation 9,
dFe3+
whereby the rate of depletion of Fe(III), − , is related to the initial concentrations of
dt
both reactants, C and C , by a rate constant, k.
Fe3+ SO
2
dFe3+
− = k[C ]x[C ]y (9)
dt Fe3+ SO 2
19 |
Virginia Tech | The orders of the reaction, x and y, represent the relative influence of each reactant, Fe(III)
and SO , respectively. Since the amount of Fe(III) consumed in the reaction is equal to
2(aq)
the amount of Fe(II) produced, the rate of the regeneration can be expressed as follows in
Equation 10.
dFe2+
= k[C ]x[C ]y (10)
dt Fe3+ SO 2
As written, k accounts for the two to one, Fe(III)-SO consumption ratio (i.e., for an equation to
2
express the rate of change in SO the rate constant would be one-half of the value of k). It
2(aq),
also encompasses all process factors other than the initial reactant concentrations and the
reaction time, dt. For the Fe(II) regeneration, dt is the incremental time in which the
concentration of Fe(II) changes. In this continuous flowing system, dt is also assumed to be
the same as the retention time of the solution in the column of activated carbon. The retention
time,τ, can be easily determined if the electrolyte volume, V, and flow rate, F, through the
carbon are known, Equation 11.
V
τ= (11)
F
If the concentration of one reactant can be held constant during the reaction, the rate
limiting reactant and reaction order can be established. It was not possible to do this
experimentally in the case of SO because high residual SO concentration would have
2 2
impacted analysis of Fe(II). However, it may be assumed that the reaction is controlled by
the diffusion of Fe(III) through the boundary layer, since the diffusivity of Fe(III) is 3.5x10-6
cm2/s (Han, 1990) and that of SO is 3.0x10-5 cm2/s (Ramsing and Gundersen, no date
2(aq)
given). As such, the order of SO should be much lower than that of Fe(III) (i.e., x>>y) and
2
Equation 10 can be re-written as Equation 12:
dFe2+
= k'[C ]x (12)
dt Fe3+
For mass transfer-controlled reactions, x = 1. Also, since the concentration of SO should
2(aq)
not influence the reaction rate, [C ]ycan now be included in the apparent rate constant, k’.
SO
2
To further define k’, a more in depth understanding of the Fe(II) regeneration reaction is
required, particularly with respect to the reaction mechanism.
20 |
Virginia Tech | As stated above, since the homogeneous reaction between Fe(III) and SO has been
2(aq)
observed to be very slow, catalysts are needed for this reaction, and activated carbon has
been successfully utilized (Sandoval and Dolinar, 1996; Sandoval et al, 1995). The carbon
provides a surface on which the reaction can occur, and its success as a catalyst, like that of
other similar materials, has been well documented for reduction-oxidation reactions. In fact,
surface catalysts have been shown to increase reaction rates by up to three orders of
magnitude for both SO oxidation and Fe(III) reduction in aqueous solution (Thomas and
2
Ingraham, 1963; Komiyama and Smith,1975; Govindarao and Gopalakrishna, 1995). The
primary mechanism of catalysis is generally supposed to be the provision of adsorption sites
for reactants, although electron transfer through the solid catalyst is also proposed as a
secondary mechanism for reduction-oxidation reactions (Garcia et al., 1998). Since
adsorption is highly dependent on surface area, that of the activated carbon was presumed to
be of major importance to the regeneration rate model. Additionally, because the rate of the
reaction has been observed to be so much faster than that of the homogeneous reaction
(Sandoval and Dolinar, 1996), the Fe(II) regeneration rate model was developed for a purely
mass transfer-controlled reaction, meaning that the products of collision-controlled reactions
in solution were assumed negligible. Evidence supporting this assumption is also found in a
study by Garcia et al. (1998), where the extent of the homogeneous reaction between Fe(III)
and SO in the presence of a surface catalyst proved insignificant.
2
For purely mass transfer-controlled Fe(II) regeneration, complete reaction requires
that one SO ion and two Fe(III) ions must adsorb on the carbon surface as depicted in
2(aq)
Figure 2.2.
Fe3+
Carbon
SO
2(aq)
particle
Fe3+
Figure 2.2 – Depiction of Fe(II) Regeneration Reaction Mechanism
21 |
Virginia Tech | The Fe(II) regeneration can be described as a particulate system of carbon particles within an
electrolyte matrix. The rate constant, k , for a mass transfer reaction in a particulate system
m
is of the form of Equation 13 (Han, 2002).
1 2
2D V 2D 3
k = f +0.6 t f (13)
m d 1 1
d 2v 6
D is the diffusion coefficient of an adsorbing species, d is the mean diameter of carbon
f
particles, v is the kinematic viscosity of the electrolyte solution, and V is slip velocity, which
t
for the Fe(II) regeneration system is given by Equation 14.
F
V = (14)
t A
s
Slip velocity represents the difference in flow rates of materials or fluids; for the Fe(II)
regeneration system, since the carbon particles are stationary, V gives the flow of the
t
electrolyte around them. A is then the cross-sectional area of the reactor through which
s
electrolyte solution can flow (i.e., the area not occupied by the carbon particles).
Additionally, v can be calculated by Equation 15 as the quotient of the dynamic viscosity of
water at the reaction temperature,μ , and the density of the electrolyte, ρ .
w s
μ
v = w (15)
ρ
s
D is expressed in units of area per time, representing the rate at which a species can
f
diffuse through a boundary layer. For the Fe(II) regeneration, this quantity can be
understood as the rate at which a species diffuses and adsorbs onto the carbon surface. Since,
per the above developments adsorption of all reactants is required prior to reaction, the rate
limiting reactant will be the one with the slower adsorption rate (i.e., lower D value.).
f
Factors known to influence diffusivity include temperature, size of the diffusing species,
molar concentration, and acidity of the solute (Han, 1990). Diffusivities of the reactants for
the current work are expected to be somewhat different from the published values due to
differing process conditions, including acidity, temperature and the continuous flow nature of
the current experiments, as opposed to the batch tests from which the above diffusivity data
22 |
Virginia Tech | were determined. However, there is still expected to be a large degree of separation between
the Fe(III) and SO diffusion coefficients, since both reactants will be subject to the same
2(aq)
process conditions during the Fe(II) regeneration and should, consequently, experience the
same types of changes in diffusivity. As such, Equation 12, which assumes that Fe(III) is the
rate limiting reactant, is expected to be valid for the regeneration. Analysis was conducted
for verification as shown in the Results and Discussion section below.
In order to adapt the general rate expression of Equation 12 to apply to the mass
transfer-controlled Fe(II) regeneration, both k and a term representing the potential contact
m
between the reactants and the carbon surface must be incorporated into k’. The contact can
be expressed in terms of the fixed volume of electrolyte (containing the reactants), V, around
the surface area of carbon particles, A, in the reactor. The rate expression then becomes
Equation 16, where the order of reaction, x, is excluded since it should be equal to one for the
mass transfer reaction.
⎡ 1 2 ⎤
dFe2+ A 2D V 2D 3
= ⎢ f +0.6 t f ⎥[C ] (16)
dt V ⎢ d 1 1 ⎥ Fe3+
d 2v 6
⎣ ⎦
Finally, if the carbon particles are assumed to be spheres, A can be expressed in terms of the
carbon mass, M, density,ρ, and mean diameter, d, and Equation 16 becomes Equation 17,
the final hypothesized rate model.
⎡ 1 2 ⎤
dFe2+ 6M 2D V 2D 3
= ⎢ f +0.6 t f ⎥[C ] (17)
dt ρdV ⎢ d 1 1 ⎥ Fe3+
d 2v 6
⎣ ⎦
By examining Equation 17, the factors likely to influence the Fe(II) regeneration were
identified. First, carbon particle size was expected to have a large effect on the reaction rate
as was mentioned previously. If the particle size is reduced, the reaction rate should increase
significantly. It should also be noted that carbon density varies slightly with particle size
(due to the decrease in porosity with decreasing size), so this had to be considered when
fitting the model to experimental data. Flow rate through the fixed volume reactor will affect
Fe(II) regeneration in two ways: by dictating the retention time and the slip velocity term.
23 |
Virginia Tech | Since the reaction is expected to be first-order, which should be verified experimentally, the
only unknown quantity is the Fe(III) diffusion coefficient.
Acidity of the Fe(II) regeneration cannot be changed for the reasons stated
previously, which also apply to the solute concentrations, since various other copper
hydrometallurgical operations require specific concentrations. Regarding temperature,
increased temperature positively affects reaction rates via a number of phenomena, most
commonly through the provision of activation energy, or reduction of a thermal reaction
boundary. However, separate experiments surrounding the Fe(II) regeneration process
(associated with the optimization problem in Chapter 3) yielded complex results with respect
to the significance of reaction temperature on rate. These results and some potential
explanations are discussed in Chapter 3. An optimal temperature for the reaction, 120 °F,
was observed during these experiments. This temperature is common for electrolytes in the
Fe(II) regeneration area of EW circuits. Given the complexity of the results associated with
the separate experiments, the observed success at an optimal temperature and the time
necessary to gather sufficient data to conclude temperature dependence of the Fe(II)
regeneration, the rate model developed in the current work is for Fe(II) regeneration at 120
°F.
Experimental
To collect data capturing the effects of carbon particle size, flow rate and the initial
concentrations of Fe(III) and SO on rate of regeneration, a bench-scale Fe(II)
2(aq)
regeneration reactor was constructed. The reactor was operated under conditions similar to a
larger-scale process (e.g., temperature, acidity, Fe(II) to Fe(III) ratio). All test work was
completed at the SX/EW Test Facility, operated by the Phelps Dodge Process Technology
Center in Morenci, AZ.
Apparatus and Materials
The apparatus for Fe(II) regeneration experiments consisted of the following parts:
24 |
Virginia Tech | • Electrolyte ingredients: reagent-grade sulfuric acid (97%), ferrous (100%) and ferric
(76%) sulfates, technical-grade copper (II) sulfate (98%), and de-ionized water;
reagent-grade chemicals were obtained from chemical the suppliers, Cole-Parmer and
VWR, and the copper (II) sulfate was obtained from a Phelps Dodge refinery.
• Gaseous SO (from pressurized tank)
2
• Activated carbon (of bituminous coal type) of three mesh class sizes: 12x40 (mean
particle size 843 μm), 16x45 (mean particle size 649 μm) and 20x50 (mean particle
size 499 μm); carbon was obtained from TIGG, but as 16x45 mesh carbon was not
commercially available, larger carbon was crushed and screened for use during
testing
• De-ionized rinse water used to wash carbon between experiments
Basically, known reactant concentrations were passed through the reactor, within which
the carbon particle size could be varied, and the amount of Fe(II) regenerated was measured
in the effluent. From the known flow rate through the reactor, the solution volume within the
reactor, and the measured change in Fe(II), the rate of regeneration was calculated. Desired
test conditions (i.e., carbon particle size, flow rate and reactant concentrations) were
determined from the testing protocol as described in the Design section below. A general
schematic of the experimental process flow is shown in Figure 2.4 and a description is as
follows:
1. electrolyte (mixed and heated to desired acid, iron and copper concentrations and
temperature) was pumped from the mix tank at a known flow rate;
2. SO was injected into the electrolyte flow at a known rate satisfying desired feed
2
concentration, dissolving to aqueous form, before entering the Fe(II) regeneration
reactor; and
3. feed flowed through the reactor, where Fe(III) was reduced to Fe(II), and the effluent
was sampled for comparison to the feed.
26 |
Virginia Tech | Temperature and acid concentration were kept constant for the reasons discussed previously
in the model development section. Total iron concentration was kept constant at the stated
concentration, since this is what would be experienced in commercial EW electrolytes for
AART using FFS; the Fe(II) and Fe(III) ratio is varied by how much iron is reacted at the
anode and how much is converted during the regeneration process. Similarly, copper was
included in the electrolyte solution – also at a typical commercial concentration – since it is
inherent to the EW process, and, hence, the Fe(II) regeneration aspect of AART using FFS.
Although specific studies have not been conducted with respect to the effects of copper
concentration, it has not been observed, nor is it expected, to significantly influence the
regeneration.
Procedure
The experimental procedure should be prefaced with the following notes. First, the
carbon used during experimentation was soaked in de-ionized water for at least one day prior
to its use. Additionally, the same carbon was used for each experiment for which it was the
correct size because the activity was assumed not to decrease significantly during
experimentation, as has been observed in previous studies (Thomas and Ingraham, 1963).
Since there were three carbon sizes tested, there were three volumes of carbon used during
experimentation. After the carbon had been soaked, solution volume within the carbon was
determined experimentally for use in calculating residence times. Finally, the density was
determined for each volume of carbon (one for each of the three tested size classes) to be
used during experimentation.
Since the experimental apparatus allowed for direct changes in electrolyte and SO flow
2
rates, flow rates were varied during testing and, then, converted to residence time values
since the carbon volume and, hence, the solution volume within it, was fixed. Just before
each experiment, reagent grade chemicals were measured to make electrolyte to
concentrations per the test design protocol, and allowed to heat and mix for approximately
one hour. When completely mixed and at the desired temperature, 120 °F, the reactor was
operated, and data were collected:
28 |
Virginia Tech | 1. An initial sample of the electrolyte was taken from the mix tank just before each
experiment to use as a baseline for comparison with the effluent sample(s).
Comparison with this sample negated concentration errors due to evaporation or
measurement mistakes when mixing electrolyte.
2. About 2 L of de-ionized water was pumped through the carbon column to rinse any
residual electrolyte or SO from previous experiments. Then, about 2 L of the freshly
2
mixed electrolyte was pumped through the column, or more, until the temperature of
the column reached the desired temperature.
3. The electrolyte pump was set to the flow rate corresponding to the residence time
called for by the test protocol, allowed to reach equilibrium, and verified by manual
measurement.
4. Once the electrolyte had reached a desired constant flow, the SO flow was started
2
and set at a rate calculated to give the desired SO concentration for a given
2
experiment.
5. Based on the feed flow to the column, after enough time had elapsed for the reaction
to come to equilibrium (about five residence times), a sample was taken from the
effluent and allowed to cool. Particularly if the ratio of SO to Fe(III) tested was
2
high, the samples were capped and shaken, then uncapped and exposed to air several
times to ensure that most of any SO in solution was released.
2
The cooled initial and effluent electrolyte samples were analyzed for Fe(II) and total iron
content. Measurements on each sample were completed in duplicate, and results were
averaged to better ensure the accuracy of the data. Corrections, if necessary, were made for
evaporation, and the initial Fe(III) concentration was determined to verify that it reasonably
matched the test protocol. The amount of Fe(II) produced by the regeneration process was
then determined as the difference between the concentrations in the initial and effluent
samples, and converted to a rate value given the residence time for the experiment. The
titration and AA methods described below were utilized for all samples, and, periodically,
some sample duplicates were sent to another lab for verification.
29 |
Virginia Tech | most importantly carbon particle size and flow rate (via slip velocity). In Figure 2.7, both the
slope of the trend-lines and closeness of data points to them indicate the relative model fits.
0.0600
0.0500
0.0400
0.0300
0.0200
0.0100
0.0000
0.0000 0.0100 0.0200 0.0300 0.0400 0.0500 0.0600
dFe(II)
dt
)II(eFd
td
mass transfer
general
nim-L/lom
,
Predicted , mol/L-min
Figure 2.7 – Experimental Fe(II) Regeneration Rate Values vs. Rate Models
The general model has a slope of 0.52 and an R2 value of 0.26, compared to the mass transfer
model slope of 0.75 and an R2 value of 0.82. Inclusion of the mass transfer factors, carbon
particle size, diffusivity, flow rate and kinematic viscosity, results in both improved accuracy
and precision of the Fe(II) regeneration rate model.
Conclusions and Recommendations
The Fe(II) regeneration process is an important aspect of AART using FFS for copper
hydrometallurgy; however little has been done to study it specifically. SO has been chosen
2
as the most attractive oxidant for use in the process and, because the homogenous reaction
between Fe(III) and SO in solution is very slow, activated carbon is utilized as a surface
2
catalyst. Although several factors may affect the regeneration reaction, few may be varied
due to its place within the scheme of the entire hydrometallurgical process. The current work
examined the effects of four variable factors – carbon particle size (i.e., surface area), flow
34 |
Virginia Tech | rate and the initial concentrations of the reactants, Fe(III) and SO – with the purpose of
2
developing a fundamental kinetic model for the Fe(II) regeneration.
A fundamental rate expression was derived from principles of chemical reaction rates
and mass transfer phenomena, and a series of experiments were conducted to collect data that
captured the effects of above factors. The following statements can be made about the Fe(II)
regeneration under the tested conditions:
• A mass transfer rate model predicts the Fe(II) regeneration rate more accurately and
precisely than a general model, which does not account for the reaction mechanism.
• The reaction rate is first-order, and is limited by the rate at which Fe(III) diffuses
onto the carbon particle surfaces.
• Initial concentration of SO does not significantly affect the reaction rate.
2(aq)
• The reaction rate is significantly increased by increased initial Fe(III) concentration
or decreased carbon particle size.
• Flow rate significantly affects the diffusivity of a species; as a result, reaction rate is
inhibited at very high flow rates.
The experimental data supported the hypothesized mass transfer rate model very well,
especially within the currently operable ranges of initial Fe(III) concentration and flow rate.
The finalized kinetic model for the Fe(II) regeneration rate (in mol/L-min) is of the following
form, where D has a value of 1.1x10-7 cm2/s:
f
⎡ 1 2 ⎤
dFe2+ 6M 2D V 2D 3
= ⎢ f +0.6 t f ⎥[C ]
dt ρdV ⎢ d 1 1 ⎥ Fe3+
d 2v 6
⎣ ⎦
It should be noted that the order, x, has been dropped from the above model equation since
the regeneration was found to be first-order.
Furthermore, several recommendations can be made regarding additional test-work
pertinent to the Fe(II) regeneration or similar reactions. First, the model might be validated
within a wider range of variable factors by performing more tests designed with the results of
the current work in mind. Such designs might quantify the effects of some process factors on
35 |
Virginia Tech | Chapter 3: An Empirical Optimization of the Current
Ferrous Regeneration Process for Alternate Anode
Reaction Technology
Emily Allyn Sarver
Abstract
The Fe(II) regeneration process is an important aspect of Alternate Anode Reaction
Technology (AART) using a Fe(II)/Fe(III)-SO (FFS) for copper hydrometallurgy. The
2
process is basically Fe(III) reduction by SO , which is catalyzed by activated carbon
2(aq)
particles. The current work examines the effects of four variable factors – carbon particle
size, temperature, and initial Fe(III) and SO concentrations – on the Fe(II) regeneration rate,
2
followed by an optimization to maximize the rate. A requirement of negligible or no SO in
2
the process effluent is an added constraint on the rate, imposed by concerns for consequences
of SO in processes subsequent to the regeneration. Using Design Expert software to analyze
2
and model the experimental results, carbon particle size and initial Fe(III) are the most
influential of the tested factors, related to the rate linearly. Temperature is related to the rate
by a squared term. Also, it is included in the model expression as a two-factor interaction
with initial SO concentration, although this term is the least significant. Optimization of the
2
Fe(II) regeneration rate model results in the following combination factors over their tested
ranges: minimum carbon particle size, maximum initial Fe(III) concentration, and moderate
values of temperature and initial SO concentration.
2
37 |
Virginia Tech | Introduction
Copper is currently the third most used metal in the world, primarily because of its
exceptional electrical conductivity and resistance to corrosion. In 2003, about 17 million
metric tons of copper were produced to supply the growing global demand (ICGS, 2003). In
addition to increasing demand, copper prices have also been on the rise, skyrocketing to over
$1.50 per pound in April 2005, up from just $0.68 only two years earlier (LME, 2005).
Rising prices are due in part to growing production costs, which are not expected to decline
to previous levels when energy was less expensive. High prices present a real advantage to
low-cost producers, and new technologies that are capable of cutting costs are under serious
consideration by the copper industry. One such technology – alternate anode reaction
technology (AART) – has shown potential for a major reduction in power consumption, a
primary cost center for electrowinning (EW) in hydrometallurgical production of copper
(Sandoval and Lei, 1993; Sandoval et al., 1995; Sandoval and Dolinar, 1996).
AART basically changes the conventional EW anode reaction from water oxidation
to ferrous [Fe(II)] to ferric [Fe(III)] iron oxidation, while the cathode reaction remains the
same. This change results in two major advantages over the conventional process:
substantial reduction in overall energy (power) requirements for EW and elimination of
hazardous acid misting. The power reduction can be attributed to the reduced equilibrium
potential of the alternate anode reaction, and the elimination of acid misting to the absence of
O production at the anode, which is present with the conventional anode reaction. One
2
major aspect of the above AART is the necessity of Fe(II) regeneration – reducing the Fe(III)
produced at the anode back to Fe(II) – to maintain recyclable EW electrolyte streams.
The method of regeneration that has been most successful during testing is Fe(III)
reduction by SO . This method is attractive because SO can be potentially obtained from
2 2
nearby copper smelters, relatively easily and inexpensively. Also, the use of SO as the
2
reducing agent produces recoverable sulfuric acid as a by-product, which is beneficial
because the acid may be utilized in leaching operations or sold (Sandoval and Dolinar, 1996).
The combination of the Fe(II)-Fe(III) anode reaction and the subsequent Fe(II) regeneration
38 |
Virginia Tech | by SO is abbreviated FFS. The FFS electrode reactions for copper EW and the Fe(II)
2
regeneration reaction are shown below in Equations 1-4.
Anode reaction: 2×(Fe2+ → Fe3+ +e−) (1)
Cathode reaction: CuSO +2e− →Cu0 +SO 2− (2)
4 4
Overall EW reaction: CuSO +2Fe2+ →Cu0 +SO 2− +2Fe3+ (3)
4 4
Fe(II) regeneration reaction: SO +2Fe3+ +2H O → 2Fe2+ +SO 2− +4H+ (4)
2 2 4
As can be seen above, for every mole of plated copper (Cu0), two moles of Fe(III) are
generated at the anode and must be reduced back to Fe(II) by a mole of SO . In practice, the
2
SO is injected into the Fe(III)-rich electrolyte after it leaves the EW cells, and then the
2
solution is passed through a bed of activated carbon where the regeneration reaction occurs,
shown in Figure 3.1. The role of the activated carbon is as a catalyst, which is discussed
below. The recoverable acid by-product is produced when the excess H+ in Equation 4 reacts
with the copper sulfate (CuSO ) already contained in the electrolyte.
4
Figure 3.1 – Fe(II) Regeneration Reactor for AART Using FFS
Although the Fe(II) regeneration process has been successfully managed during both
bench and pilot scale testing of AART using FFS, little has been done to study it specifically.
For commercial operation, a better understanding of which process factors affect the
regeneration, and in what ways, may assist in process design or improvement with respect to
39 |
Virginia Tech | the process parameters, circuitry, equipment or materials. To relate process factors to results,
an optimization problem was defined and solved. First, process constants, variable factors
and their operable ranges, and desirable responses were defined. Then, a series of
experiments were conducted to determine the combination of variable factors that, under
fixed conditions, will produce the most desirable results.
Literature Review
An understanding of the Fe(II) regeneration process parameters is essential to
developing and solving the optimization problem. Due to its place within the scheme of EW,
and the entire hydrometallurgical process including leaching and solvent extraction, many of
the regeneration process parameters are fairly fixed. Acidity and copper concentration, for
instance, cannot be varied without changing many other processes, while parameters specific
to the regeneration process, like temperature and catalyst properties, might be adjusted rather
easily. Other factors, such as iron and SO concentrations, might be varied within certain
2
ranges such that subsequent operations will experience small or negligible effects. For
example, the Fe(III) concentration might be varied somewhat, but it must be kept in certain
proportion with Fe(II) in the EW cells to maintain high current efficiency. Likewise, SO
2
concentration might be varied within its potential range of consumption by the Fe(II)
regeneration reaction, but if it is varied considerably outside of this range the regeneration
will not be efficient or un-reacted SO could escape during EW, a hazardous consequence.
2
A review of the regeneration and similar processes provides some insight into
reaction mechanisms and influential factors. The homogeneous Fe(II) regeneration reaction
has been observed to be slow occurring in solution, a fact documented by Sandoval and
Dolinar in a 1996 study of AART using FFS. Similar observations have been made of both
of the half-reactions associated with the regeneration, Fe(III) reduction and SO oxidation,
2
when studied separately (Kumar et al., 1996; Govindarao and Gopalakrishna, 1995; Garcia et
al., 1998; Thomas and Ingraham, 1963; Seaburn and Engel, 1993; Komiyama and Smith,
1975; Berglund et al., 1993). When catalyzed by a solid complex, such as activated carbon
or graphite, the rates of such reduction-oxidation reactions have been shown to increase by
up to three orders of magnitude (Thomas and Ingraham, 1963). Sandoval and Dolinar noted
40 |
Virginia Tech | a dramatic increase in the Fe(III)-SO reaction rate when they tried catalyzing the reaction by
2
passing the electrolyte containing the reactants through a bed of activated carbon, although
the enhanced performance was not quantified. Several mechanisms have been suggested for
reduction-oxidation reactions catalyzed by solid surfaces, most commonly surface adsorption
and electron transfer phenomena; however, the behavior of each reactant, and other non-
reactive species, and the reaction responses to various factors are typically quite specific to
the process in question (Kumar et al., 1996; Govindarao and Gopalakrishna, 1995; Garcia et
al., 1998; Thomas and Ingraham, 1963; Komiyama and Smith, 1975; Berglund et al., 1993).
In a 1998 study by Garcia et al., the proposed mechanism included a heterogeneous reaction,
whereby the oxidant adsorbed onto the catalyst surface and reacted with the reducing agent
still in solution. Despite differing degrees of catalysis, a connection can generally be made
between surface area and reaction responses, so catalyst surface area will clearly be a factor
in the Fe(II) regeneration process.
Additionally, solid-surface catalyzed reduction-oxidation reactions involving iron
ions and/or S(IV) species have been observed to be affected by several other variables
(Sandoval and Lei, 1993; Kumar et al., 1996; Govindarao and Gopalakrishna, 1995; Garcia
et al., 1998; Thomas and Ingraham, 1963; Seaburn and Engel, 1993; Komiyama and Smith,
1975; Krissman et al., 1998; Berglund et al., 1993). Typically, increased temperature has
been observed to have a positive effect on reaction rate, while things like acidity and
adsorption of non-reactive species have tended to inhibit reactions. In batch tests, non-
reactive species have even been determined to deactivate the catalyst by blinding potential
adsorption sites from reactive species (Kumar et al., 1996; Govindarao et al., 1995).
However, in continuous flow tests using activated carbon as a catalyst, the activity has been
observed to be long-lasting in the presence of similar iron and acid concentrations as are
common for EW electrolytes (Thomas and Ingraham, 1963).
Concentrations of reactants and non-reactive species have been observed to influence
solid-surface catalyzed reactions differently, specific to the reactants, products and reaction
mechanism(s). As stated above, if the non-reactive species adsorb onto the catalyst and are
retained, they can have a mild to severe deactivating effect. They might also assist or be
41 |
Virginia Tech | insignificant in the reaction, as was observed of Cu(II) ions by Kumar et al. in a 1996 study
of SO oxidation by dissolved O . As well, one reactant may dominate the reaction if, for
2 2
example, it adsorbs onto the catalyst surface much faster than another reactant or is present in
relatively large concentrations. Reaction products can also affect reaction progression if they
aid or inhibit it in any way. For instance, in the aforementioned Garcia et al. study, the rate
of Fe(III) reduction was slowed by the production of Fe(II) when the Fe(II) adsorbed onto the
graphite catalyst and prevented further adsorption of the reacting species. The affects of
solution constituent concentrations for the Fe(II) regeneration will depend on reaction
stoichiometry, mechanism(s) and the continuous flow nature of the process.
Optimization Problem Development
To set up the optimization problem, process constants, variable factors and their
operable ranges, and desirable responses were defined.
Particular to the current regeneration process, several fixed conditions are present
which may certainly affect it: electrolyte acid, total iron and copper concentrations. Due to
the relatively high acidity, when SO is dissolved into the electrolyte the predominant species
2
will be SO , which is known to react slower than bisulfite (HSO -), the predominant
2(aq) 3
species at lower acidity (Kumar et al., 1996; Govindarao and Gopalakrishna, 1995; Garcia et
al., 1998; Krissman et al., 1998). Also, Fe(II), either initially in the electrolyte or produced
by the regeneration reaction, could inhibit the progression of the regeneration if it prevents
adsorption of reactants onto the catalyst. In the range of concentration for the regeneration
process, such inhibition does not appear likely considering published diffusivity data that
suggests SO will adsorb onto the catalyst faster than iron ions (Han, 1990; Freiberg and
2
Schwartz, 1981; Ramsing and Gundersen, no date given). Conversely, if the copper
concentration has any effect it will likely interfere as a catalyst for the homogeneous reaction
rather than as an inhibitor, as has been proposed in previous studies of similar reactions
(Kumar et al., 1996). Although these conditions may play a role in the regeneration, they are
fixed within the current process and will be maintained as constant parameters for the
purpose of optimization.
42 |
Virginia Tech | Considering these constants, the regeneration reaction and the findings of previous
studies surrounding similar reactions, the (variable) influential process factors and their
operable ranges were identified. The factors are activated carbon surface area, temperature,
and reactant – Fe(III) and SO – concentrations. Carbon surface area can be easily varied
2
without affecting other operations, the most problematic issues being carbon containment
within the reactor (e.g., if particles are too small, they will either wash through containment
mesh at the bottom of the reactor or the mesh will be so tight that it inhibits flow) and
obtaining odd sized carbons (i.e., sizes not commercially available). Temperature can also be
varied fairly easily, although excessive temperature changes could present large costs in time
and/or money. Fe(III) concentration can be varied somewhat; since total iron is constant,
varying Fe(III) means varying Fe(II) equally but oppositely. In order to increase Fe(III)
concentration, more Fe(II) must be oxidized at the anode and this is limited by concern for
EW efficiency (i.e., Fe(II) is depleted in the EW cells, so increased oxidation requires more
time or voltage and thus reduces efficiency). Finally, SO can be varied with Fe(III)
2
concentration such that the two are within the proper proportions for the regeneration
reaction, which depend on both stoichiometry and the extent of the reaction. It is important
to achieve a complete reaction (i.e., prevent un-reacted SO ) while regenerating as much
2
Fe(II) as possible, precisely the conditions of the optimization, as discussed below.
Considering how each factor might be varied and their values during previous successful
Fe(II) regeneration within AART using FFS testing, an operable range was defined for each
process factor. Ranges are shown in Table 3.1 along with the constant process parameters
and their respective values for the optimization.
43 |
Virginia Tech | Table 3.1 – Experimental Parameters
Varied Process Factors Low-Value Mid-Value High-Value
Fe(III) Concentration 2g/L 4g/L 6g/L
SO Concentration 0.50g/L 0.75g/L 1.00g/L
2
Temperature 110°F 120°F 130°F
Mean Carbon Particle Size 499μm 649μm 843μm
Constant-value Parameters Value
H SO Concentration 160g/L
2 4
Total Fe Concentration 30g/L
Cu(II) Concentration 38g/L
As is often the case, the main goal of the optimization problem for the current Fe(II)
regeneration process was to maximize throughput. Since the process reaction is essentially
contained within fixed volume reactors, maximum throughput occurs at the minimum
residence time of electrolyte within the reactor(s) because the two quantities are inversely
related, as shown in Equation 5, where F is electrolyte throughput or flow, V is electrolyte
volume in the reactor, and τ is residence time.
V
F = (5)
τ
Additionally, two conditions on this goal had to be considered for regeneration efficiency and
safety: Fe(II) regeneration should be as high as possible in order to supply as much Fe(II) for
the EW anode reaction as possible, and break-through (un-reacted) SO in the reactor effluent
2
should be as low as possible to avoid SO escape when electrolyte is exposed to air (i.e., in
2
the EW cells). These conditions are individually satisfied at process extremes. With
excessive SO supply to the regeneration reactor, the highest amount of Fe(II) will be
2
regenerated but break-through SO will occur; with minimal SO supply, break-through SO
2 2 2
will be easily prevented but less Fe(II) will be regenerated.
Considering the above goal and conditions, criteria were established to measure the
success of the Fe(II) regeneration process as the factors were varied over their respective
ranges. First, process success only occurs if there is no (or negligible) break-through SO .
2
Second, the optimal combination of process factors is the one that results in the minimum
successful residence time at a maximum level of Fe(II) regeneration. For experimentation,
44 |
Virginia Tech | the criteria could be quantified by a single response to a particular combination of the process
factors; the Fe(II) regeneration rate, R , which could be calculated using Equation 6 where
Fe2+
dFe2+is the amount of Fe(II) produced at the minimum residence time, τ . By
min
maximizingR , the optimal ratio between Fe(II) regeneration and minimum residence time
Fe2+
could be determined. The solution may be used to improve current operations or to design
reactors and process conditions for future operations. R may be easily used in scale-up
Fe2+
problems since the minimum residence time may be related to any number of reactor sizes
(i.e., solution volumes) and feed flow rates.
dFe2+
R = (6)
Fe2+ τ
min
During the experimentation described below, the process factors were empirically related to
the responses and an optimization was completed.
Experimental
To collect data capturing the effects of carbon particle size, temperature and the initial
concentrations of Fe(III) and SO on rate of regeneration, a bench-scale Fe(II) regeneration
2
reactor was constructed. The reactor was operated under similar conditions to a larger-scale
process (e.g., temperature, acidity, Fe(II) to Fe(III) ratio). All test work was completed at the
SX/EW Test Facility, operated by the Phelps Dodge Process Technology Center in Morenci,
AZ.
Apparatus and Materials
The apparatus for Fe(II) regeneration experiments consisted of the following parts:
• 12” Plexiglas column from Waters Equipment (1.45” internal diameter) with top and
bottom mesh-covered screw-caps fitted for electrolyte and effluent flows; when filled
with activated carbon, the column functioned as the Fe(II) regeneration reactor
• 150 lb pressurized SO tank with a digitally-set Porter Instruments mass flow
2
controller rated for 0-150 mL/min flow
• 20 L electrolyte mix tank with variable speed mixer, and coil immersion heater with
Cole Parmer Dyna-Sense controller
45 |
Virginia Tech | size 499 μm); carbon was obtained from TIGG, but as 16x45 mesh carbon was not
commercially available, larger carbon was crushed and screened for use during
testing
• De-ionized rinse water used to wash carbon between experiments
Basically, known reactant concentrations were passed through the reactor, within which
the carbon particle size could be varied, and the amount of Fe(II) regenerated was measured
in the effluent. From the known flow rate through the reactor, the solution volume within the
reactor, and the measured change in Fe(II), the rate of regeneration was calculated. Desired
test conditions (i.e., carbon particle size, flow rate and reactant concentrations) were
determined from the testing protocol as described in the Design section below. A general
schematic of the experimental process flow is shown in Figure 3.3 and a description is as
follows:
1. electrolyte (mixed and heated to desired acid, iron and copper concentrations and
temperature) was pumped from the mix tank at a known flow rate;
2. SO was injected into the electrolyte flow at a known rate satisfying desired feed
2
concentration, dissolving to aqueous form, before entering the Fe(II) regeneration
reactor; and
3. the feed flowed through the reactor, where Fe(III) was reduced to Fe(II), and the
effluent was sampled for comparison to the feed.
47 |
Virginia Tech | Figure 3.3 – Process Flow for Experimental Fe(II) Regeneration
Design
Experimental design, results analysis and optimization were completed using Design
Expert 6.0, a software package by Stat-Ease with capabilities to relate multiple process
factors and responses. Using a response-surface method and (face-centered) central
composite type design, the four process factors (independent variables) were varied over
three levels each in a set of 30 experiments, a sufficient number for creating a statistically
significant model. This method and design type are ideal for optimization problems because
experiments capture points on the interior and boundaries of the design space such that the
model(s) generated are representative of the entire space. Subsequent optimization
determines the combination of factor values within the design space that produces the desired
response.
The testing protocol included six replicates on the center point of the design space,
eight axial points and 16 factorial points in the 30-experiment set. The four numeric process
factors and their respective operable ranges were as defined in Table 3.1. Particle size was
characterized by the geometric mean size within a given class. Carbon particle size was
modeled as a numeric factor so that during optimization the size could be indicated between
48 |
Virginia Tech | actually available sizes if preferred. The process response was Fe(II) regeneration rate at
minimum residence time, as defined above. Each experiment was conducted at its particular
combination of process factors, per the test protocol, and the response was measured per the
procedure below.
Procedure
The experimental procedure should be prefaced with the following notes. First, the
carbon used during experimentation was soaked in de-ionized water for at least one day prior
to its use. Additionally, the same carbon was used for each experiment for which it was the
correct size because the activity was assumed not to decrease significantly during
experimentation, as has been observed in previous studies (Thomas and Ingraham, 1963).
Since there were three tested carbon sizes, there were three volumes of carbon used during
experimentation; after the carbon had been soaked, solution volume within the carbon was
experimentally determined for use in calculating residence times. Finally, approximate flow
rates were determined during a series of preliminary tests at various Fe(III) and SO
2
concentrations, such that maximum flow rate (minimum residence time) could be determined
relatively quickly during each of the 30 experiments, as to avoid evaporation or critical
depletion of the electrolyte feed.
For each experiment, a two-part procedure was conducted: determination of
minimum residence time followed by determination of Fe(II) regeneration at the minimum
residence time. Since the experimental apparatus allowed for direct changes in electrolyte
and SO flow rates, flow rates were varied during testing and then converted to residence
2
time values since the carbon volume, and hence the solution volume within it, was fixed.
The experimental procedure is described below.
Determination of Minimum Residence Time
Reagent grade chemicals were measured to make electrolyte to concentrations required
by the test protocol just before each experiment, and allowed to heat and mix for
approximately an hour. When completely mixed and at the desired temperature for a given
experiment, the minimum residence time was determined:
49 |
Virginia Tech | 1. An initial sample of the electrolyte was taken from the mix tank just before each
experiment to use as a baseline for comparison with the effluent sample(s).
Comparison with this sample negated concentration errors due to evaporation or
measurement mistakes when mixing electrolyte.
2. About 2L of de-ionized water was pumped through the carbon column to rinse any
residual electrolyte or SO from previous experiments. Then about 2L of the fresh
2
electrolyte was pumped through the column, or more until the temperature of the
column reached the desired temperature for a given experiment.
3. The electrolyte pump was set to a conservative flow, one which allowed for a nearly
complete reaction of initial Fe(III) and SO concentrations required for a given
2
experiment, and the flow rate was allowed to reach equilibrium, verified by manual
measurement.
4. Once the electrolyte had reached a desired constant flow, the SO flow was started
2
and set at a rate calculated to give the desired SO concentration for a given
2
experiment.
5. Based on the feed flow to the column, after enough time had elapsed for the reaction
to come to equilibrium (about five residence times), a sample was taken from the
effluent and immediately capped. All samples were taken and analyzed for break-
through SO as stated below in the Analytical Methods section, and SO readings and
2 2
corresponding flow rates were recorded. If break-through SO was determined to be
2
present in the sample, the experiment was run at a slower flow rate; if SO was not
2
present, the flow rate was increased.
This procedure was repeated incrementally until the maximum flow rate could be
determined within 10-20mL/min increments; minimum residence time was calculated from
the maximum flow rate and solution volume within the carbon using Equation 5.
Additionally, all SO -flow rate data were analyzed for each experiment with multiple data
2
points to determine if a general trend could be established (i.e., how does break-through SO
2
change with flow rate?)
50 |
Virginia Tech | Determination of Fe(II) Regeneration
Upon determination of the minimum residence time for each experiment, the initial
electrolyte sample and the effluent sample from the reactor when operated at the minimum
residence time were analyzed for Fe(II), total iron and Cu(II) content. Measurements on each
sample were completed in duplicate and results were averaged to better ensure the accuracy
of the data. Corrections, if necessary, were made for evaporation, the initial Fe(III)
concentration was determine to verify it reasonably matched the test protocol, and the
amount of Fe(II) produced by the regeneration process was determined as the difference
between the concentrations in the initial and effluent samples. The titration and AAS
methods described above were utilized for all samples, and periodically some sample
duplicates were sent to another lab for verification.
Analytical Methods
For determining minimum residence time, it was necessary to determine whether or not
the reactor effluent contained break-through SO . Measurement of SO in solution is
2 2
possible via iodimetry, however this method was not feasible for the Fe(II) regeneration
experiments for several reasons (e.g., lack of required materials, time constraints per the
experimental procedure, and/or reliability of analysis by an outside party due to loss of SO
2
concentration). Since, per the two-part experimental procedure below, it was only necessary
to find if SO was present in the effluent, not an exact quantity, a hand-held gas monitor was
2
used. The monitor was an Industrial Scientific T82 single gas monitor with an SO sensor,
2
which detected SO gas in the range of 0.2-150ppm. The gas intake area on the monitor was
2
lined with a rubber seal, which fit fairly tightly over the top of the 15mL sample tubes used
during testing. Since the ppm reading was not an exact measurement but more of an SO
2
indicator, the following procedure was used to determine whether or not break-through SO
2
was present in a sample.
• A 15mL sample tube was filled to the 14mL mark with effluent directly from the
reactor and immediately capped.
• The sample was allowed to sit and cool for 15 minutes. Due to the partial pressure of
SO , if any was present, it would readily come out of solution and fill the 1mL space
2
in the capped tube.
51 |
Virginia Tech | • The sample was uncapped and the SO monitor was immediately fitted over it.
2
If the monitor read above 2ppm, the sample was considered to contain break-through SO ;
2
2ppm or less was considered negligible. For the purposes of this paper, “minimum residence
time” then refers to the shortest residence time allowing 2ppm or less of break-through SO .
2
This method of SO detection was quite conservative considering the sample and air volumes
2
used and the current SO allowances set by OSHA (PEL = 5ppm, TLV = 2ppm), (OSHA,
2
2005).
To determine the change in Fe(II) concentration during the regeneration process for
each experiment, a Fe(II) titration was completed on un-reacted and reacted electrolyte
samples using potassium permanganate (KMnO ) as the titrant. This titration method is
4
rather common for Fe(II) and well documented (ASM, 1994). All titrations were done using
a Metler-Toledo DL50 auto-titrator, which was regularly calibrated and standardized, before
use and between about every 12 samples. Additionally, atomic absorption spectrometry
(AAS) was used to determine total iron and Cu(II) concentrations in all samples for two
purposes. First, total iron measurement served as a method of checking and validating initial
Fe(III) concentrations (i.e., Fe(III) is the difference between the total iron and Fe(II)
concentrations). Second, both total iron and copper measurements were helpful in
determining whether evaporation during testing was significant (i.e., if total iron and Cu(II)
concentrations were elevated in the effluent, then evaporation had to be considered). AAS
analysis is a common technique for measuring many types of metal ion concentrations in
solution (ASM, 1992). It was completed using a Perkin Elmer 1100 AA, also calibrated
regularly. Finally, some sample duplicates were taken and sent to an outside lab throughout
the testing to periodically verify titration and AAS results; the other lab also used titration
and AAS to determine Fe(II) and total iron, respectively.
Results and Discussion
For each experiment, individual data sheets were created containing the following
information: test protocol values for process factors, initial and effluent electrolyte Fe(II),
total iron and Cu(II) assays, and break-through SO trends with changing flow rate (reactor
2
52 |
Virginia Tech | residence time). The effluent assays were only taken on samples from the minimum
residence time test for each experiment. A sample data sheet is shown below in Figure 3.4.
date 14-Oct-04 Exp 15
TEST CONDITIONS
carbon Fe3+
volume 253.69 mL conc. 6.18 g/L
Total Fe
SO2 conc. 0.50 g/L conc. 30.72 g/L
SO2 Fe2+
density 0.003017 g/cc conc. 24.54 g/L
outlet
pressure 2.00 psi Cu conc. 38.43 g/L
atm
pressure 14.70 psi Acid conc. 159 g/L
solution
atm temp 294.00 K temp 110 F
Parameter Settings Measured Responses (Effluent
electrolyte SO2 residence
flowrate flowrate time Fe2+ Total Fe Fe3+ Cu Acid
mL/min cc/min min SO2 g/L g/L g/L g/L g/L
0.00 0.00 0.00 0 24.54 30.72 6.18 38.43 159
142.00 23.50 1.79 1.4 25.22 30.93 5.71 38.6 159.8
169.00 27.80 1.50 2.6
183.50 30.40 1.38 3.3
209.00 34.60 1.21 8
238.00 39.50 1.07 12.6
snoitidnoc
deriuqer/nevig
elpmas
laitini
morf
derusaem
Breakthrough SO2 for Varying Residence Times
200
150
100
50
0
0.00 1.00 2.00 3.00
Residence Time (min)
)mpp(
2OS
Figure 3.4 – Sample Data Sheet for an Individual Experiment
The data sheets like the one shown in Figure 3.4 were used to assess the trend of
break-through SO with flow rate, as well as compile an overall summary. For all
2
experiments during which multiple flow rates were applied to determine the minimum
residence time, a clear exponential trend was observed in break-through SO with increasing
2
flow rate. Although too few data were collected to quantify a relationship between the
process factors and the exponential trend in break-through SO , the general trend is
2
important. It indicates that if the Fe(II) regeneration process is operated at residence times
53 |
Virginia Tech | Figure 3.7 – Design Expert Model Predicted vs. Actual Values
As seen above, the correlation between the model and experimental data is very reasonable
with the natural log transformation. The model predicts all points within 20% of the actual
values, with only three points predicted to be more than 10% from the actual value.
With respect to the actual model expression, all process factors were included;
however, only initial Fe(III) concentration and carbon particle size were included as linear
terms. Initial SO concentration and temperature were included as squared or two factor
2
interaction (2FI) terms. A 2FI was also significant between the Fe(III) concentration and
carbon size. The model expression is as follows in Equation 8, where C and C are the
Fe3+ SO
2
initial Fe(III) and SO concentrations, respectively, T is the reaction temperature, and S is the
2
mean carbon particle size.
R =−3.64+[0.47C ]−[0.54S]−[1.24T2]−[0.27C ×T]+[0.33C ×S] (8)
model Fe3+ SO2 Fe3+
The model coefficients are given in terms of coded factor values between -1 and 1, which
correspond to the highest and lowest tested values of a factor. For example, coded values for
the mean carbon particle size are -1, 0 and 1, which correspond to 499 μm, 649 μm and 843
μm, respectively.
59 |
Virginia Tech | The linear effects of both carbon particle size and initial Fe(III) concentration, and the
relatively small influence of initial SO concentration, are not surprising. As mentioned
2
previously, since the carbon catalyzes the Fe(II) regeneration via the provision of surface
area on which the reactants can adsorb and then react, a larger surface area promotes greater
adsorption and, hence, faster reaction. Separate studies or the regeneration reaction kinetics
show that the reaction is mass transfer-controlled by Fe(III) because the diffusivity of Fe(III)
is much smaller than that of SO . This mechanism is confirmed by the significant model
2(aq)
terms (Equation 8), including the 2FI between carbon particle size and Fe(III) concentration.
This term can be logically explained considering that changes in carbon size will
significantly affect the reaction rate by changing adsorption Fe(III), the rate limiting reactant.
Additionally, the highest reaction rates were observed during experiments at a
temperature of about 120°F, with moderate SO concentrations. This phenomenon is
2
captured by the squared temperature term in the model expression (Equation 8); at a
moderate temperature, the negative effect of the squared temperature term is minimized such
that the Fe(II) regeneration rate is increased.
The fact that reaction temperature was not found to have a linear influence on the
Fe(II) regeneration rate is rather surprising. Generally, increased temperature increases a
chemical reaction rate, as well as diffusivity. However, the experimental data suggests that
temperature alone does not significantly affect the reaction rate, but that the combination of
temperature and initial SO concentration do; this is confirmed by the 2FI model term that
2
pairs the two quantities. At low SO concentrations, increased reaction rate tended to
2
correspond with increased temperature, as expected; however, at higher SO concentrations,
2
reaction rate decreased with increased temperature. These unpredicted results cannot be
conclusively rationalized given the amount of data collected, but the dependence of reactant
diffusivities on temperature might provide some explanation. At high temperature the SO
2
might reach the catalyst surface so much faster than the Fe3+ that it could inhibit Fe3+
adsorption, and hence the reaction. It seems that this would only become problematic if
carbon surface area was low, but the temperature trends with SO concentration did not
2
60 |
Virginia Tech | appear to differ much between the largest and smallest carbon particle sizes tested. Another
explanation might focus on the solubility of SO , since as temperature increases, SO
2 2
solubility decreases; if air space was present at the top of the regeneration reactor, SO may
2
have collected there, reducing the concentration available for the reaction.
Given the relatively small significance of the 2FI term combining temperature and
SO concentration, indicated by its small coefficient in Equation 8, it may actually be
2
removed from the model expression without substantially affecting the model fit for the
ranges of factors tested. The other term coefficients are only changed slightly and the
reduced model expression then becomes Equation 9.
R =−3.64+[0.47C ]−[0.53S]−[1.25T2]+[0.33C ×S] (9)
model Fe3+ Fe3+
Optimization
The Design Expert software offers three methods of optimization: point prediction,
graphical and numerical. The numerical method determines the optimal combination(s) of
process factors for user defined goals (i.e., desired responses), and generates a set of 10
solutions, which are ranked by desirability. This method was chosen for optimizing the
Fe(II) regeneration since the stated goal of the problem was to maximize R , thereby
Fe2+
maximizing the regeneration rate without allowing break-through SO in the reactor effluent.
2
The most significant model expression (Equation 8) was used for optimization. The model
was constrained to the boundaries of the design space for optimization such that it could not
find solutions by extrapolating process factor values outside of their tested ranges.
At first, optimization was completed with the carbon particle size being allowed to
fluctuate as a numeric factor (i.e., the model was not required to find a solution with the
particle size being one of the discreet mean sizes tested). Although this is not practical
because activated carbon is typically sold in established, discreet mesh sizes, allowing the
particle size to fluctuate continuously showed whether a particular size was overwhelmingly
desirable. Table 3.3 gives the 10 most desirable solutions for the combinations of process
factors that maximize R .
Fe2+
61 |
Virginia Tech | Table 3.3 – Optimization Solutions with Carbon Particle Size as a Numeric Factor
Process Factors
Solution Initial SO 2 Initial Fe(III) Reaction Mean Carbon R model R Fe2+ Desirability
Concentration Concentration Temperature Particle Size
# (g/L) (g/L) (°F) (μm) (mol/L-min) (mol/L-min)
1 0.78 5.97 121 506 -3.000 0.0498 1
2 0.76 5.90 120 509 -2.996 0.0500 1
3 0.82 5.95 120 499 -3.000 0.0498 1
4 0.72 5.94 122 499 -2.999 0.0498 1
5 0.71 5.92 121 512 -3.000 0.0498 1
6 0.75 5.99 121 519 -2.995 0.0501 1
7 0.71 5.90 120 499 -2.988 0.0504 1
8 0.74 5.98 119 507 -2.985 0.0505 1
9 0.81 5.99 120 505 -2.999 0.0498 1
10 0.78 3.09 120 499 -3.176 0.0418 0.94
The solutions in Table 3.3 confirm the expectation that the smallest carbon particle
size (20x50 mesh, mean size 499 μm) is the optimal size of those tested. It provides much
more surface area on which the Fe(II) regeneration reaction can occur, and probably a
smaller size would allow an even faster reaction. The numerical optimization was completed
again with the mean carbon particle size specified as 499 μm. Table 3.4 shows the 10 most
desirable solutions.
Table 3.4 – Optimization Solutions with Carbon Particle Size Fixed at 499 μm
Process Factors
Solution Initial SO Initial Fe(III) Reaction Mean Carbon R R
2 model Fe2+ Desirability
Concentration Concentration Temperature Particle Size
# (g/L) (g/L) (°F) (μm) (mol/L-min) (mol/L-min)
1 0.81 6.00 120 499 -2.990 0.0503 1
2 0.75 5.96 122 499 -2.992 0.0502 1
3 0.78 5.99 120 499 -2.976 0.0510 1
4 0.71 5.99 120 499 -2.980 0.0508 1
5 0.75 5.81 120 499 -2.995 0.0501 1
6 0.78 5.97 118 499 -2.997 0.0499 1
7 0.72 6.00 121 499 -2.974 0.0511 1
8 0.75 5.98 121 499 -2.985 0.0506 1
9 0.77 5.81 120 499 -2.996 0.0500 1
10 0.78 2.00 120 499 -3.171 0.0420 0.94
Initial Fe(III) concentration is one of the two most influential model factors, the other
being the carbon particle size. Since the reaction rate is controlled by the transport of Fe(III)
to the carbon surface, it is reasonable that the highest Fe(III) concentration is the optimal
value for maximizing the reaction rate. Likewise, the initial SO concentration and
2
temperature optimal values, which are both in the middle of the ranges tested, seem
appropriate since the values are close to those used in experiments where the highest reaction
rates were observed.
62 |
Virginia Tech | To actually optimize the Fe(II) regeneration process for a larger scale operation,
reactor size and flow rate will require consideration. For a desired amount of Fe(II)
regeneration, given the R value predicted by the model for any combination (within the
Fe2+
design space) of the process factors, the reactor residence time can be determined (Equation
6). From the residence time, the reactor volume and flow rate can then be determined
(Equation 5). It should be noted that, while many possible combinations of volumes and
flow rates could be utilized to achieve a desired residence time, large changes to the flow rate
may invalidate the model due to changes in reaction kinetics (i.e., the reactor flow rate can
affect how easily reactants adsorb onto the carbon surface). This issue was not specifically
addressed during the test-work for this optimization problem, but it was in a separate study,
which is presented in Chapter 2.
Conclusions and Recommendations
The Fe(II) regeneration process is an important aspect of AART using FFS for copper
hydrometallurgy; however little has been done to study it specifically. SO has been chosen
2
as the most attractive reducing agent for use in the process and, because the homogenous
reaction between Fe(III) and SO in solution is very slow, activated carbon is utilized as a
2
surface catalyst. Although several factors may affect the regeneration reaction, few may be
varied due to the place of the reaction within the scheme of the entire hydrometallurgical
process. It is important to understand how those that are variable may be manipulated to
produce desirable responses, for the purposes of process design and improvement. As such,
the goals of this work were to examine and optimize a set of four variable factors: carbon
particle size (i.e., surface area), temperature, and the initial concentrations of the reactants,
Fe(III) and SO
2.
The most desirable response was defined as the maximum rate of Fe(II) regeneration
at which there is no break-through SO in the process effluent. Experiments were conducted
2
to capture the effects of each factor on the regeneration rate. The experimental data were
analyzed and modeled using Design Expert software. Finally, the model was optimized to
determine a factor combination which resulted in the most desirable response.
63 |
Virginia Tech | The following conclusions can be made with respect to the major observations during
experimentation, analysis and modeling:
• Break-through SO in the process effluent increases exponentially with increased flow
2
rate.
• The two-part experimental procedure appears to be capable of producing precise
results, although there is likely some constant error associated with each experiment
due to the inability to avoid air space at the top of the reactor.
• Data analysis and modeling using Design Expert software proved successful methods
for relating multiple factors to a response for the Fe(II) regeneration process.
• The generated model is statistically significant, and predicts all experimental data
points within 20% of their values.
• Carbon particle size and initial Fe(III) concentration are the most influential process
factors and are included as linear terms in the model.
• Temperature is also significant and included as a squared term in the model, which
results in an optimal temperature within the range of tested values.
• Initial SO and temperature are not linearly influential; however, the combination of
2
the two factors was determined to slightly affect the Fe(II) regeneration rate, but the
effect cannot be conclusively explained.
Using the software-generated model to optimize the Fe(II) regeneration process within
the tested variable ranges, it was confirmed that the smallest carbon particle size in
combination with the highest initial Fe(III) concentration generally produces the fastest
regeneration rates. Additionally, by operating the process with temperature and initial SO
2
concentration at moderate values, the regeneration rate is maximized.
Furthermore, several recommendations can be made regarding additional test-work and
scale-up operations for the Fe(II) regeneration. First, the results of the current work might be
used as the design basis for future experiments. The most influential factors may be tested
within a wider range of values, and some conclusions made with respect to the economy of
changing operating parameters to improve process results. A better understanding of the
64 |
Virginia Tech | Chapter 4: Summary, Conclusions and Recommendations
Summary
Copper is essential to modern society and demand for the commodity will continue to grow
with global populations and economies (Demler, 2005). Due to its unique properties, the
metal has an array of end uses, most importantly as an electrical conductor and a corrosion-
resistant construction material. A major method of processing mined copper ore is
hydrometallurgy, which utilizes the electrical properties of copper to finally extract the pure
metal during a process called electrowinning (EW). EW accounts for a large percentage of
the total processing costs for hydrometallurgy, as it has substantial power requirements. In
an effort to reduce the power consumption by copper EW operations, Alternate Anode
Reaction Technology (AART) is being investigated. This technology is attractive because of
its considerably lower energy requirement and added benefits, namely elimination of acid
misting in EW tank-houses and by-production of recoverable acid. A major aspect of AART
is regeneration of the anode reactant, Fe(II), which is achieved by reducing Fe(III) with SO .
2
The Fe(III)-SO reaction is catalyzed by activated carbon particles, since the homogenous
2
reaction is slow in aqueous solution (Sandoval and Dolinar, 1996).
Until now, little has been done to specifically the study the Fe(II) regeneration process,
however a better understanding may provide insight into improvement of the current process
and/or associated materials or equipment. Given its place within the scheme of the EW
circuit, and the overall hydrometallurgical process, many Fe(II) regeneration process
parameters are fixed or have limited variability. Fixed conditions, including acid and sulfate
concentrations, are relatively extreme, making comparison to other studied processes
difficult. While the regeneration process has been successfully operated at both the bench
and pilot scale, variable factors, like reactant concentrations and carbon particle size, have
not been examined to determine their effects. Additionally, the mechanism of Fe(III)-SO
2
reaction in the presence of the carbon catalyst has not been verified.
66 |
Virginia Tech | To gain a better understanding of the Fe(II) regeneration and improve process results, two
studies were conducted and the details are presented in this work. The first study surrounded
the fundamental kinetics of the Fe(II) regeneration reaction with the purpose of developing
and validating a rate model. The purpose of the second study was to determine the basic
effects several variable factors, and then optimize the regeneration process by finding a
combination of the factors which produced the most desirable results.
Conclusions
The study surrounding the fundamental kinetics of the Fe(II) regeneration resulted in
validation of the hypothesized mass transfer model shown below in Equation 1:
⎡ 1 2 ⎤
dFe2+ 6M 2D V 2D 3
= ⎢ f +0.6 t f ⎥[C ]x (1)
dt ρdV ⎢ d 1 1 ⎥ Fe3+
d 2v 6
⎣ ⎦
dFe2+
which predicts the rate of Fe(II) regeneration, (mol/L-min), under the tested
dt
conditions. M (g) is the mass of the carbon, d (cm) is the mean diameter of carbon particles,
V (mL) is the volume of the solution within the carbon bed, p (g/cm3) is the carbon density,
D (cm2/min) is the diffusion coefficient of Fe(III), V (cm/min) is the slip velocity of the
f t
solution, v (cm2/min) is the kinematic viscosity of the solution, and C (g/mL) is the initial
Fe3+
Fe(III) concentration. The observed D was 1.1x10-7 cm2/s (6.6x10-6 cm2/min). Additionally,
f
the following conclusions were made about the Fe(II) regeneration reaction:
• Because it accounts for the reaction mechanism, a mass transfer rate model predicts
the Fe(II) regeneration rate more accurately and precisely than a general model.
• The reaction rate is first-order (i.e., x = 1 in the above model equation), and is
limited by the rate at which Fe(III) diffuses onto the carbon particle surfaces.
• Initial concentration of SO does not significantly affect the reaction rate.
2(aq)
• The reaction rate is significantly increased by increased initial Fe(III) concentration
or decreased carbon particle size.
• Flow rate significantly affects the diffusivity of a species; as a result, reaction rate is
inhibited at very high flow rates.
67 |
Virginia Tech | The optimization study also has an associated model to predict Fe(II) regeneration rate,
which was generated by analyzing data from a series of experiments within a defined design
space. The regeneration rate investigated during this study was constrained; it was
considered the fastest rate at which there was no or negligible SO in the process effluent.
2
Experimental design, data analysis, modeling and optimization were completed using Design
Expert software by Stat-Ease. Under the tested conditions – varied ranges of carbon particle
size, temperature and initial Fe(III) and SO concentrations – the model response is a
2
function of all four factors. The model is given by the following (Equation 2):
(2)
R =ln(R )=−3.64+[0.47C ]−[0.54S]−[1.24T2]−[0.27C ×T]+[0.33C ×S]
model Fe2+ Fe3+ SO2 Fe3+
which actually predicts the natural logarithm of the Fe(II) regeneration rate, R . S is
Fe2+
carbon particle size, T is temperature, and C and C are the initial concentrations of
Fe3+ SO
2
Fe(III) and SO , respectively. Equation 2 is given in terms of coded factor values between -1
2
and 1, which correspond to the highest and lowest tested values of a factor. It has only been
verified within the tested ranges of each factor.
The following conclusions were made regarding the major observations, data analysis
and the generated model:
• Break-through SO in the process effluent increases exponentially with increased flow
2
rate.
• The two-part experimental procedure appears to be capable of producing precise
results, although there is likely some constant error associated with each experiment
due to the inability to avoid air space at the top of the reactor.
• Data analysis and modeling using Design Expert software proved successful methods
for relating multiple factors to a response for the Fe(II) regeneration process.
• The generated model is statistically significant, and predicts all experimental data
points within 20% of their values.
• Carbon particle size and initial Fe(III) concentration are the most influential process
factors and are included as linear terms in the model.
• Temperature is also significant and included as a squared term in the model, which
results in an optimal temperature within the range of tested values.
68 |
Virginia Tech | • Initial SO and temperature are not linearly influential; however, the combination of
2
the two factors was determined to slightly affect the Fe(II) regeneration rate, but the
effect cannot be conclusively explained.
Upon optimization of the model (Equation 2) to maximizeR , the following combination
Fe2+
of process factors was determined optimal: maximum initial Fe(III) concentration, minimum
carbon particle size, and moderate values of temperature and initial SO concentration.
2
The findings of both studies coincide well with each other. Each determined that carbon
particle size and initial Fe(III) concentration were the most influential factors for the Fe(II)
regeneration process. Also, the initial SO concentration was determined to be of least
2
significance to the reaction rate. The reaction mechanism – reactant adsorption and
subsequent reaction on the carbon surface – was confirmed by the kinetics study and
supported by the results of the optimization study. While the results of each study are only
known to be applicable under the tested conditions, they provide some insight into Fe(II)
regeneration in larger scale processes. Through a better understanding of the regeneration
process, commercial processes may be better designed or improved.
Recommendations
Following is a list of topics surrounding the Fe(II) regeneration process for AART using FFS
that may require further work:
• determination and quantification of the effects of reaction temperature on the process
responses;
• quantification of the effects of solution flow rate through the regeneration reactor on
reactant diffusivities, and, hence, process responses;
• determination of the effects of carbon type on the process responses;
• determination of factors contributing to carbon deactivation, and quantification of
carbon lifetime; and,
• validation of the Fe(II) regeneration rate models outside of the factor ranges tested in
this work.
69 |
Virginia Tech | values plotted against the observed values over the entire tested range of initial Fe(III)
concentration and flow rate; the solid line is a linear trend-line.
0.0600
0.0500
0.0400
0.0300
0.0200
0.0100
0.0000
0.0000 0.0500 0.1000 0.1500 0.2000
dFe(II)
dt
)II(eFd
nim-L/lom
,
td
Predicted , mol/L-min
Figure B.2 – Experimental Fe(II) Regeneration Rate Values vs. Mass Transfer Model Over
Entire Initial Fe(III) Concentration and Flow Rate Ranges
The three circled points in Figure B.2 clearly throw off the trend of the experimental
vs. predicted rate values, which are associated with three of the experiments utilizing the
12x40 mesh carbon. The model severely over-estimates the Fe(II) regeneration for these
points, which represent experiments with both very high initial Fe(III) to SO ratios and flow
2
rates. These conditions may explain the over-estimation of the model as follows. Due to the
relatively large Fe(III) to SO ratio used in generating these points, all of the SO may have
2 2
been consumed in a shorter time than the tested residence time for these experiments. With
respect to the high flow rates, they were applied in an effort to determine the reaction rate,
since at low flow rates the reaction would have definitely been completed faster than allowed
by the residence time. However, extreme flows may have negatively affected Fe(III)
diffusivity. It seems logical that as the flow rate is dramatically increased, turbidity and
shear speed of electrolyte flowing past the carbon will inhibit Fe(III) adsorption.
Although not to the extent of the circled points in Figure B.2, the model under-
estimates the Fe(II) regeneration rate for experiments with the lowest Fe(III) concentrations
77 |
Virginia Tech | and flow rates tested. Similar effects as proposed above might also explain the poor
correlation between the model and experimental data at the low end of the initial Fe(III)
concentration and flow rate ranges. It is possible that the reaction may have neared
completion (i.e., most Fe(III) was consumed) faster than allowed by the tested residence
times. As well, at very low flow rates, Fe(III) diffusivity may be increased and result in
higher Fe(II) regeneration rates than those predicted by the mass transfer model. While
more data would be necessary to conclusively explain this lack of fit between the model and
experimental data discussed above, opposite trends at opposite ends of the tested Fe(III)
concentrations and flow rate ranges support the offered explanations. Because the currently
operable range of initial Fe(III) concentration and residence time (corresponding to tested
flow rate) are well below those used to generate the circled points in Figure B.2, they can be
removed from the data set without sacrificing the worth of the model for use in commercial
design and operation of the Fe(II) regeneration process.
For the reduced data set, the average D value of 1.1x10-7 cm2/s makes the mass
f
transfer model fit the data well. This value is much lower than that found in the literature,
3.5x10-6 cm2/s (Han, 1990). While the solute (sulfate) concentration for the current work and
the study in the literature are similar, other conditions are quite different. Particularly the
continuously flowing nature of the process in the current work, as opposed to more batch-
type experiments used to determine diffusivity in the literature, provides some explanation
for the large spread in the calculated and published D values. The flow of electrolyte
f
(containing reactants) past the carbon particles inhibits adsorption of both species onto the
catalyst, thereby inhibiting reaction. Countering the reduced diffusivity, it is conceivable that
turbid conditions might also encourage reaction somewhat by increasing collision of
reactants, and perhaps even reactants and carbon particles near the top of the carbon bed;
however, this effect would probably not be nearly as noticeable as decreased diffusivity. For
the current Fe(II) regeneration operating parameters, including residence times (flow rates)
and Fe(III) concentrations, the model appears to predict the regeneration rate fairly
accurately; however, to model such a process under more extreme conditions, it may be
necessary to include D as a function of flow rate.
f
78 |
Virginia Tech | Part 3: Model Fit for Each Tested Carbon Particle Size
In addition to evaluating the overall fit of the mass transfer model to the experimental data, it
should also be evaluated for each tested carbon particle size. Figure B.3 shows the model
predicted Fe(II) regeneration rate values plotted against the observed values. The data is
separated by carbon particle size used to generate it.
0.0600
0.0500
0.0400
0.0300
0.0200
0.0100
0.0000
0.0000 0.0100 0.0200 0.0300 0.0400 0.0500 0.0600
dFe(II)
dt
)II(eFd
td
12x45 carbon (843um)
16x45 carbon (649um)
20x50 carbon (499um)
nim-L/lom
,
Predicted , mol/L-min
Figure B.3 – Experimental Fe(II) Regeneration Rate Values vs. Mass Transfer Model
The slope of the trend-lines for the large (12x40 mesh) and small (20x50 mesh)
carbon sizes are still relatively close to one (0.70 and 0.90, respectively), indicating good
model fit. Scatter in the 12x40 mesh data (R2 = 0.72) is most likely due to difficulties during
experimentation, including being able to ensure an airtight reactor and maintain a constant
temperature. There are only a small number of data points for the 20x50 mesh carbon (R2 =
0.91), so the model could be validated with more confidence for this carbon size with
additional tests.
As for the 16x45 mesh carbon (R2 = 0.63), the slope of the trend for the experimental
vs. model values is flat (0.35), and it should be equal to one. From the data summary, the
model under-estimates the Fe(II) regeneration rate at low initial Fe(III) concentration and
over-estimates at high initial Fe(III) concentration. Since the model predicts too low at one
79 |
Virginia Tech | end and too high at the other, inclusion of the mid-sized carbon data does not significantly
influence the overall mass transfer model (Figure 2.8); leaving out the data points only
increases the slope of the plot from 0.75 to 0.78. Additionally, there does not appear to be
significant influence by flow rate as was the case for more severe lack of fit issues seen in
Figure 2.7. Since there are very few data points for this carbon size, one possibility for poor
correlation is erroneous data, which could only be ruled out by conducting more tests.
Another explanation may be attributable to the source of the carbon, which was crushed and
screened during the current work since 16x45 mesh was not available from the carbon
supplier for the experiments. Although care was taken to size the material accurately (e.g.,
using ASTM certified screens and mechanical shaker), the trend of the experimental data
indicates that perhaps the carbon had a mean size larger than 649μm, the geometric mean
size for 16x45 mesh particles. While it seems more likely that fines may bias an experiment
with carbon sized by the user, it is possible that the size distribution was skewed toward the
larger end of the mesh class. It may prove beneficial to test more carbon sizes which are all
obtained from the same source; this will help eliminate errors caused by skewed size
distributions and provide more data to further validate the kinetic model.
80 |
Virginia Tech | Table C.2 – Comparison of Major Statistical Measurements for Central Composite and
“Historical” Design Types for Experimental Data Set
Central
Statistical "Historical"
Description* Composite
Measure Design
Design
calculated from the design points, this is the average
prediction variance as a percentage of the maximum
G-Efficiency 75.9% 64.1%
prediction variance. It is desirable to have an
efficiency of at least 50%.
this indicates the degree of multicollinearity present in
the design matrix. If the value equals one, there is no
Condition Number multicollinearity and the design is orthogonal. If the
in value is less than 100 then there is not a serious 11.4 14.8
Correlation Matrix problem. Values in the range of 100 to 1000 indicate
moderate to severe multicollinearity and over 1000
indicates a severe problem.
Aliased Model these terms cannot be separated during analysis, due
none none
Terms to too few data points
VIF measures how much the variance of a model
coefficient increases due to the lack of orthogonality in
the design. Specifically the standard error of a model
coefficient increases in proportion to the square root of
Large Variance
the VIF. If a coefficient is orthogonal to the remaining
Inflation Factors none none
model terms, its VIF is one. One or more large VIFs
(VIF)
indicate multicollinearity. VIFs exceeding ten indicate
problems due to multicollinearity. (For example, if a
coefficient has a VIF of 16, its standard error is 4 times
as large as it would be in an orthogonal design.)
* Descriptions were obtained from the Design Expert 6 User’s Guide (2002).
82 |
Virginia Tech | CLEANING AND DEWATERING FINE COAL
USING HYDROPHOBIC DISPLACEMENT
Kara E. Smith
ABSTRACT
A new processing technique, known as hydrophobic displacement, was explored as a
means of simultaneously removing both mineral matter and surface moisture from coal in a
single process. Previous thermodynamic analysis suggests that coal moisture will be
spontaneously displaced by any oil with a contact angle greater than ninety degrees in water.
Based on these results, six methods of hydrophobic displacement were evaluated: hand shaking,
screening, air classification, centrifugation, filtration, and displacement. In the first five methods
hydrophobic displacement took place during the cleaning stage. A recyclable non-polar liquid
(i.e. pentane) was used to agglomerate coal fines followed by a physical separation step to
remove the coal agglomerates from the mineral-laden slurry. Bench-scale tests were performed
to identify the conditions required to create stable agglomerates. Only the last method,
displacement, did not utilized agglomeration and performed hydrophobic displacement during
dewatering, not cleaning. A procedure was also developed for determining moisture content
from evaporation curves so that the contents of water and pentane remaining in a sample could
be accurately distinguished.
Two primary coal samples were evaluated in the test program, i.e., dry pulverized 80
mesh x 0 clean coal and 100 mesh x 0 flotation feed. These samples were further screened or
aged (oxidized) to provide additional test samples. The lowest moisture, 7.5%, was achieved
with centrifugation of the pulverized 80 mesh x 0 clean coal sample. Centrifugation provided the
most reliable separation method since it consistently produced low moisture, high combustible
recoveries, and high ash rejections. Hand shaking produced the next lowest moisture at 16.2%;
however, the low moistures were associated with a drop in combustible recovery. There was also
a great deal of error in this process due to its arbitrary nature. Factors such as oxidation, size
distribution, and contact angle hysteresis influenced the concentrate moistures, regardless of the
method utilized. |
Virginia Tech | ACKNOWLEDGMENTS
First and foremost, I would like to thank my advisor Dr. Jerry Luttrell for the guidance he
has provided. I could not have finished this project without his willingness to share his
knowledge and his patience in letting me find my own path.
I would also like to thank my other committee members, Dr. Yoon and Dr. Adel. I have
truly been blessed to have the opportunity to work with such a wonderful department, and I
appreciate the help everyone has lent during both my undergraduate and graduate degrees.
This project would not have been possible without funding from the U.S. Department of
Energy through Virginia Tech’s Center for Advanced Separation Technologies.
I am indebted to Kerem Eraydin for his help and previous experience on the project.
I would like to thank Jim Waddell. His patience and willingness to make equipment for
my project were invaluable.
I would also like to acknowledge Alpha Natural Resources’ generosity in providing
numerous coal samples. Without the assistance of the group at the Tom’s Creek processing plant,
this project would not have been possible.
Finally, thanks go out to my family and Chris for their patience and support.
iii |
Virginia Tech | 1. INTRODUCTION
1.1. Preamble
Cleaning and dewatering of fine coal is currently one of the coal mining industry’s
greatest challenges. The United States’ dependence on coal for 52% of its electricity in 2006
mandates that coal be utilized as efficiently as possible (EIA, 2007). Though 92% of coal in the
United States is used for the generation of electricity, the remaining percentage is used for a
variety of purposes including heating and coke for steel blast furnaces (EIA, 2007). As an
integral component to many industries, it is vital to the country’s economy that a steady, low-
cost supply of coal is maintained. Many industries would be hard pressed to come up with an
economic substitute for coal.
As easily accessible, high quality coal reserves are depleted, the mining industry needs a
low cost solution which will allow it to utilize more of the fines generated during processing and
to recover fine coal stored in refuse impoundments (Hazra, 1988). The high water content of
ultrafine coal often makes it uneconomical to sell due to the associated contract penalties, and
this coal contributes to the 70-90 million tons of fine wastes produced each year (Orr, 2002). In
2002, there was already more than 2.5 billion tons of fine coal wastes discarded into
impoundments.
Conventionally, fine coal undergoes separate cleaning and dewatering phases. Water-
based density separators such as spirals and/or water-only cyclones are used to treat small
particles in the 1 mm x 100 mesh size range, while froth flotation is used to clean minus 100
mesh coal. The water-rich products from these cleaning processes are dried using centrifugation,
filtration, or thermal drying. Unfortunately, these methods become increasingly expensive as the
material becomes finer.
Screen-bowl centrifuges, which are commonly used to treat 1 mm x 0 coals, are the most
popular fine coal dewatering method used in the United States. The amount of minus 325 mesh
material in the feed controls the final product moisture. For example, only 30% ultrafines results
in a surface moisture around 18% (Osborne 1988). Also, screen-bowl centrifuges cannot achieve
high coal recoveries since they lose nearly half of the ultrafines present in the feed as an effluent
stream.
1 |
Virginia Tech | Vacuum filtration, which can achieve nearly complete recoveries of coal solids, usually
results in a product with 20-35% moisture. This value rises as the amount of ultrafines (minus
325 mesh material) increases, resulting in higher water penalties. The fines also filter more
slowly and require more power. Unfortunately, the higher capacity, lower cost filtration units,
such as disc filters, often produce higher moisture products than lower capacity, higher cost
filtration processes (Wills, 1997). As a result, the popularity of vacuum filters has declined
dramatically in the United States where high moisture values cannot be tolerated due to strict
contract specification and difficulties associated with handling and freezing of damp coal.
Thermal drying can produce single digit moisture without the same size restrictions as
centrifugation and filtration. However, this process is expensive and increasingly harder to
permit (Osborne, 1988). As a result of the difficulties associated with dewatering ultrafines, the
minus 325 mesh stream is often discarded in coal plants even though it contains the most well-
liberated material. This not only wastes valuable coal, but also creates potential environmental
problems associated with the disposal of fine coal wastes.
1.2. Objectives
This project seeks to replace conventional coal cleaning and dewatering technologies
with a single solid-solid and solid-liquid separation process which results in products with less
than 10% moisture. The basis for this process is displacement of water and hydrophilic material
by a hydrophobic liquid. For this study, pentane was selected as the hydrophobic liquid since it
was affordable, met the thermodynamic requirements (i.e. a contact angle on coal greater than
ninety degrees in water), and could be easily recycled via evaporation and condensation. Though
the volatility of pentane necessitates a more complicated, closed system, it makes recovery of the
oil less energy intensive and less expensive.
Six cleaning-dewatering processes were evaluated: hand shaking, screening, air
classification, centrifugation, filtration, and displacement. Most of the tests revolved around oil
agglomeration with pentane in which hydrophobic displacement took place during the cleaning
stage; the dewatering stage then consisted of physical separation of oil-coated agglomerates and
free water droplets. Only the displacement process utilized hydrophobic displacement during the
dewatering stage in which oil sought to strip moisture from fine coal’s surface. This project
consists of completing the bench-scale and batch testing and evaluating the best method for
2 |
Virginia Tech | continuous operation. The project also includes an investigation into how to accurately measure
moisture due only to water, the effects of oil dosage, and the optimum dosage for the chosen
dewatering method.
1.3. Organization
This thesis is divided into seven major sections. The proceeding introductory section
seeks to explain the need for improvement in fine coal processing and how this project will
attempt to fill the gap in technology.
The second literary review section summarizes the current states of technology. This
section contains four subsections: conventional fine coal cleaning, conventional fine coal
dewatering, oil agglomeration, and hydrophobic displacement. The cleaning section includes
information on froth flotation, the current industry practice for fine coal recovery. The
dewatering section covers drying through centrifugation, filtration, and thermal drying. The oil
agglomeration section reviews the history, theory, and practice of coal agglomeration and also
touches on previous testing with pentane, the chosen oil for this project. The hydrophobic
displacement section reviews the precursor to this project at Virginia Tech.
The third, experimental section covers the samples, apparatus, and procedures used in
this study. In particular, this section covers water content determination, agglomerate formation,
and removal of agglomerates in both bench-scale and batch testing.
The fourth and fifth sections contain the experimental results and subsequent discussion
of them. These sections focus heavily on the feasibility of the different methods and the results
of the final batch centrifuge testing.
The sixth section contains a brief summary of the project, while section seven provides
recommendations for future testing of this process.
3 |
Virginia Tech | 2. LITERATURE REVIEW
The literature review is split into four subsections: conventional fine coal cleaning,
conventional fine coal dewatering, oil agglomeration, and hydrophobic displacement. The first
two sections are meant to provide a brief overview of the currently accepted practices. However,
the main focus of this literature review is oil agglomeration since it is the basis for much the
reported work. The final section reviews this project’s precursor at Virginia Tech. Though it did
not utilized pentane, it contained a thermodynamic analysis of the hydrophobic displacement
process.
2.1. Conventional Fine Coal Cleaning
Currently, froth flotation is the only commercially practiced method for cleaning fine
coal in the United States. This section will not cover other novel cleaning methods. It should be
noted that oil agglomeration is also an extensively studied method for cleaning coal; however,
due to its combined cleaning and dewatering capabilities and importance to this project, it will be
reviewed in a separate section of this document.
2.1.1. Froth Flotation
Froth flotation (Figure 1-1) is currently the preferred method for cleaning minus 100
mesh coal. It is based on the differential wettability of particles; this surface-based process
distinguishes between hydrophobic coal and hydrophilic clays. Air bubbles passing through a
coal slurry selectively attach to coal particles, carrying them to the surface froth phase, while
hydrophilic tailings remain in the water or pulp phase. The froth phase is then removed,
effectively separating the coal and impurities.
Flotation is controlled by chemical, operational, and design variables. Chemical variables
include coal rank, pulp chemistry, surface oxidation, and reagent dosage. Operational variables
include particle size, feed rate, pulp density, pulp level, froth height, impeller speed, aeration
rate, and conditioning time. Design variables are based on the type of cell and configuration.
4 |
Virginia Tech | Since flotation is a surface-based process,
the ability to control surface chemistry is
essential. This is done through reagents known as
collectors, depressants, activators, pH modifiers,
and frothers. Though complex ore flotation
applications may utilize all of these reagents, coal
flotation is simpler and often only requires a
collector and frother. A collector is a chemical
that adsorbs on coal rendering it more
hydrophobic in order to facilitate bubble
attachment (Wills, 1997). A frother is a surfactant Coal
which helps to stabilize a froth and prevent bubble
Feed
breakage once the loaded bubble reaches the
Reject
surface (Wills, 1997). Without a frother, bubbles Air
would break or coalesce and release coal back
Figure 1-1. Conventional flotation bank used to clean
ultrafine coal.
into the pulp phase, preventing separation.
Next to surface chemistry, particle and bubble size are two of the most important
variables. Flotation works best for fine particles about 0.1-0.25 mm in diameter. Larger particles
have a high probability of bubble-particle detachment, while smaller ones have a low probability
of bubble-particle collision. While particle size determines which particles are most likely to
float, bubble size controls the amount of particles that are able to float. The total surface area of
the bubbles determines the carrying capacity of the froth. If there is no free area on a bubble for a
coal particle to attach to, it will be misplaced to the tailings. Since surface area can be drastically
increased by decreasing the size of a bubble, modern flotation equipment are typically designed
to produce small bubbles to maximize flotation kinetics and the carrying capacity of the air
volume.
Though flotation is a selective process, not all of the material reporting to the concentrate
is coal. Three mechanisms contribute to the concentrate: attachment, entrapment, and
entrainment. Attachment refers to selective bubble-particle attachment. These particles may be
coal or a combination of unliberated coal and ash. Attachment is the only mechanism which
selectively contributes to the desired components of a concentrate. Entrapment, the most rare of
5 |
Virginia Tech | the mechanisms, refers to small particles that are trapped between agglomerated, coarser
particles or between coarser particles and a bubble. Entrainment occurs due to hydraulic
transport of ultrafine material. Water is required to form the liquid films in the froth, and free
floating, ultrafine material may exist in these channels.
There are two major types of froth flotation cells: mechanical and column. A mechanical
or conventional cell produces bubbles through mechanical agitation. The froth layer is relatively
short and is scrapped off by paddles. Due to the large water recovery and possibility for
entrainment, large banks of cells are usually setup with multiple stages of flotation (Osborne,
1988). Column cells produce bubbles through spargers and allow bubbles to rise through tall
quiescent tanks. Froths are considerably deeper in column cells, so wash water may be used to
remove entrained material from the froth (Osborne, 1988). Both types of cells are currently in
use in the United States.
2.2. Conventional Fine Coal Dewatering
Three methods of drying will be reviewed: screen-bowl centrifugation, vacuum filtration,
and thermal drying. Though screen-bowls are reviewed and able to handle some ultrafines, they
are usually reserved for coarser feeds than those studied in this project. Vacuum filtration is the
most common method for dewatering ultrafines
and is the mostly likely candidate for treating the
types of feed size distributions used in the project.
Finally, though thermal drying produces the driest
product, it is the least used of the three methods
due to problems with expense and permitting.
2.2.1. Screen-Bowl Centrifugation
Centrifuges combine centrifugal
sedimentation and filtration. High g-forces cause
solids to settle quickly into a compact cake and
force water out through the pores (Osborne,
1988). Screen-bowl centrifuges (Figure 1-2)
consist of a horizontal tube with a screw inside to
Figure 1-2. Screen-bowl centrifuge used to dewater
fine coal.
6 |
Virginia Tech | move the material. The first section of the tube is solid and removes the bulk of the water. As the
feed comes into this section, it quickly forms a cake while the majority of the liquid and about
half of the minus 325 mesh material flow over the adjustable weirs in the back of the machine.
The screw pulls the material up a small ramp to the front section of the machine which consists
of a screen for further dewatering. A screen-bowl centrifuge is a hybrid centrifuge; the solid
bowl section enables the machine to handle large volumes of water with a high solids recovery
while the screen basket section allows drainage aided by centrifugal force (Osborne, 1988).
These centrifuges are high capacity, long life machines that can provide low moistures.
The final moisture is directly related to the amount of minus 325 mesh feed material. For
example, if a feed contains 30% minus 325 mesh, the product’s moisture will be around 18%
(Osborne, 1988). It should also be noted that some of this ultrafine material is discarded with the
main effluent. Typically this effluent is not recycled, and any material in it is lost to the tailings.
Final product moisture is also dependent on the centrifugal force. A higher operating speed will
lead to lower moisture and a finer cut; however, screen-bowl centrifuges are generally not
operated above 500g due to excessive wear. Due to the strong dependence of product moisture
on feed size and limited centrifugal force, screen-
bowl centrifuges are generally used for
dewatering fine material coming off of spirals.
2.2.2. Vacuum Filtration
Vacuum filtration (Figure 1-3) is the most
effective method for dewatering fine coal
containing a large proportion of minus 325 mesh
solids. Disc filers, the most common type in the
United States, consist of vertical discs with fan-
shaped sectors covered in fine cloth or mesh. The
hollow discs are under vacuum and submerged
about half way in slurry. As the discs rotate, they
pick up solids from the slurry, the cake dries as it
is carried into the air, and then the dried cake is
Figure 1-3. Disc vacuum filter used to dewater fine
coal.
7 |
Virginia Tech | blown off before the fan segment is again dipped into the slurry (Osborne, 1988). Fine solids are
trapped in the cake against the filter cloth, and recoveries are usually greater than 97%. They
typically produce moistures in the 25-35% range, and reagents may be needed to reach the lower
moistures. Flocculants are usually added to reduce screen blinding, reduce ultrafine losses, and
aid in cake release, while cationic coagulants are occasionally used to increase the filtration rate.
Disc filters are popular in the United States due to their small footprints, high capacities,
and low cost; however, they produce higher moistures and require more maintenance compared
to some other filters. The vertical nature of disc filters also prevents cake washing (Wills, 1997).
Other continuous vacuum filters include rotary drums and horizontal belt filters. Filtration may
also be done by applying positive pressure instead of a vacuum; however, these filters are more
expensive and are used rarely in the coal industry for dewatering clean coal products.
2.2.3. Thermal Drying
Thermal drying (Figure 1-4) is not common in the United States. It is the most expensive
unit operation in coal preparation (Osborne, 1988), and it is extremely difficult to permit new
units. They are generally only used on ultrafine
coals whose large surface areas lead to high
moisture contents. Thermal dryers are the only
unit that can consistently provide single digit
moisture with ultrafine feed. This low moisture
may be worth the cost to reduce the possibility of
freezing, to reduce heat loss during combustion,
and to prepare the coal for coke making among
other reasons (Osborne, 1988). Industrial coal
dryers usually employ convection in direct heat-
exchange type dryers in which wet coal is
continuously brought into contact with hot gases
in order to evaporate surface moisture (Osborne,
1988).
Figure 1-4. Thermal dryer used to dry coal to low
moisture contents.
8 |
Virginia Tech | 2.3. Oil Agglomeration
2.3.1. History of Oil Agglomeration
Oil agglomeration was first performed on coal in the early 1920’s (Mehrotra et al., 1983);
however, it was not until the 1970’s energy crisis that the United States invested significant
amounts of time and money into the potential uses of oil agglomeration. The sharp increase in oil
prices spurred the need for an alternative source of energy to run equipment such as turbines and
diesel engines. It was discovered that the fine particulates in coal slurry were not problematic,
but the residue due to ash was unacceptable. Oil agglomeration was investigated as a method to
produce the ultraclean coal needed. Though most of the testing during the 1980’s focused on the
cleaning ability of oil agglomeration, dewatering and oil recovery were also explored. Several
pilot plants were even created to test the feasibility of continuous, larger scale processing
(Mehrotra et al., 1983). The inherently expensive process could not compete with the falling oil
prices in the late 1980’s, and oil agglomeration was largely abandoned.
Much of the prohibitive cost associated with oil agglomeration is due to the need to finely
grind the feed, sometimes as fine as a few microns, for ultracleaning (Nguyen et al., 1983). The
price and consumption of refined oil are also major disadvantages. The finer the feed, the more
surface area is created, increasing the oil consumed. Complete recovery of the oil is often
impossible or prohibitively expensive. For these reasons, commercial oil agglomeration with
coal is not currently practiced in the United States.
2.3.2. Theory of Oil Agglomeration
Oil agglomeration is based on the ability of an oil to preferentially wet hydrophobic or
oleophilic surfaces. This selectivity enables an oil to coat the hydrophobic sites of coal particles,
while rejecting hydrophilic material such as clays and pyrite (Good et al., 1991). The hydrophilic
particles remain in an aqueous suspension, while the hydrophobic coal particles combine into
agglomerates to minimize the surface in contact with water.
Despite the fact that oil agglomeration has been studied extensively, the microscopic
interactions are still not well understood. Coal is not homogenous and consists of a patchwork of
hydrophilic and hydrophobic sites (Keller et al., 1987); therefore, several conflicting theories
exist on which liquid, oil or water, acts as the bridging mechanism to form the agglomerates. The
9 |
Virginia Tech | first popular theory is that oil acts as a liquid bridge between coal particles (Keller et al., 1987).
The oil envelopes the coal and bridges over the hydrophilic sites. Though small droplets of water
may remain bound to the hydrophilic sites, oil displaces the water from the hydrophobic sites and
remains the dominant liquid in the agglomerates. As two oil coated particles collide during
mixing, the oil and capillary attraction of the oil causes the particles to stick together and
eventually form agglomerates.
The second opposing theory is that water actually acts as the bridging liquid. Many oils
simply spread on hydrophobic coal surfaces. In contrast, when surrounded by oil water sticking
to the hydrophilic sites forms water droplets with contact angles greater than 90 degrees (Good et
al., 1991). When two of these droplets meet, they form a bridge and the surface tension of the
water pulls the coal particles together. The more the particles are pulled apart, the more the
surface tension increases and forces the particles back together. In contrast, hydrophobic liquids
will break apart into two droplets when the bridge is stretched (Good et al., 1991). Oil simply
coats the particles and provides an environment for the water bridges. Finally, there is little
discussion on whether these theories are mutually exclusive or may both contribute to
agglomerate formation.
The location of agglomerate water is dependent on which theory of bridging liquids is
ascribed to. If oil is considered to be the bridging liquid, then only minor amounts of water will
be trapped in the agglomerates at the hydrophilic sites. If water is considered to be the bridging
liquid, then there is an inherent amount of water needed for an agglomerate to keep its shape. In
both cases, free water droplets may be located in the voids and pores between agglomerates. As
more oil is added, these water droplets may be completely enclosed and trapped.
The macroscopic interactions of oil agglomeration are better understood than interactions
within agglomerates, and most of the papers and patents associated with oil agglomeration are
written at this level. The appearances of the agglomerates vary greatly with changes in the
operating parameters. These operating parameters include oil type, oil dosage, mixing time,
mixing speed, coal surface oxidation, particle size, and solids content (Boni et al., 1994; Hazra et
al., 1986).
The type of oil being used makes a significant impact on the speed and selectivity of
agglomeration. Oils are commonly divided into light and heavy oils. Heavy oils are less selective
and used with lignite and other low-rank coals. Lighter low viscosity paraffinic oils are more
10 |
Virginia Tech | selective and are used with higher rank coals such as anthracite and bituminous coal (Capes,
1991). The type of oil also affects whether it will be recoverable and whether a binder may need
to be added for dewatering. Oil dosage varied widely in previous testing and has a large effect on
the appearance of the agglomerates. Lower dosages create small, compact, rounded agglomerates
while higher dosages tend to result in large, soft, irregular curds (Osborne, 1988). Enough oil is
needed to cover the coal and displace water from its surface, but too much oil may actually trap
water globules between the curds. Mixing time and speed also vary greatly in oil agglomeration.
High shear mixers are often used under the theory that more energy provides more opportunity
for the oil to come into contact with the coal (Keller et al., 1987). The high shear reworks and
cleans the agglomerates. However, lower shear mixing may also be used to prevent the
agglomerates from being ripped apart.
Coal surface oxidation affects the ability of oil to coat the coal. Oxidation makes coal
more hydrophilic and may prevent oil from binding to its surface (Drzymala et al., 1994). The
more oxidized sites, the more water will be incorporated into the agglomerate. Particle size
affects the dosage of oil needed. As particles become smaller, the surface area increases rapidly.
It is the surface of the coal that determines now much oil will be needed to coat the particles
(Osborne, 1988). Particle size also influences the amount of water in the agglomerates. Smaller
particles will be more liberated and may have less hydrophilic sites for water. Conversely, if the
coal is oxidized, the increase in surface area will lead to an increase in trapped water and perhaps
suspended hydrophilic material. There is not a single optimum set of operating conditions. The
amount of interdependent operating parameters means several sets of conditions can produce
similar results and may be customized for the particular coal being used.
2.3.3. Practice of Oil Agglomeration
There are several different aspects of oil agglomeration that have been the focus of
research including cleaning, dewatering, and oil recovery. The ability of oil agglomeration to
clean a coal has been the dominant and original focus of research in the United States. Oil
agglomeration cleans a coal by agglomerating only hydrophobic coal and leaving hydrophilic
waste in suspension. Cleaning often involves finely grinding the coal to as little as five microns
in order to liberate all of the clays, pyrite, and other hydrophilic impurities (Mehrotra et al.,
1983). A typical process uses higher dosages of oil and high shear mixing to ensure that all of
11 |
Virginia Tech | hydrophilic material is expelled (Mehrotra et al., 1983). The curds are then skimmed from the
surface or rinsed on a screen. The Otisca-T Process is one of the most well-known patents in this
particular area.
The dewatering aspect of oil agglomeration has not received as much attention in
research. Dewatering assumes that water is not the bridging liquid and may be expelled from the
interior of the agglomerates if enough oil is present. Only minor amounts of water remain at the
hydrophilic sites; the majority of the water is located in the pores between the individual
agglomerates. This water may be removed with conventional dewatering equipment including
centrifuges and vibratory screens (Mehrotra et al., 1983). Binders may be added to strengthen the
agglomerates before they are dewatered. Research by the National Research Council of Canada
resulted in some of the lowest moistures ranging from 3-12%. They focused on spherical
agglomeration to simplify dewatering. Micro-agglomerates, or flocs, were dewatered on a screen
before being pelletized with heavy, cheap oil binders (Mehrotra et al., 1983).
Oil recovery is a popular research topic for oil agglomeration. Oil agglomeration
consumes large amounts of oil, and this may be prohibitively expensive. Some processes do not
attempt to recover the oil. Instead, the oil is allowed to burn with the concentrate, contributing to
the overall heating value of the coal product. There are even multi-stage oil agglomeration
processes in which a light refined oil and a heavier binder are utilized together (Mehrotra et al.,
1983). While processes such as these make dewatering and cleaning easier, their expensive
nature prompted research into oil recovery through heating and condensation. Many papers in the
late 1980’s were based on the economics of oil agglomeration, a topic closely related to the loss
and cost of oil (Mehrotra et al., 1983).
2.3.4. Previous Testing with Pentane
Pentane is not a widely used hydrocarbon in oil agglomeration. Most labs, including the
Convertol Lab, NRCC Lab, Shell Lab, and BHP Lab, tended to focus on cheaper oils including
diesel, kerosene, fish oil, and fuel oil. Although lighter oils are more selective, heavier oils were
utilized in order to minimize loss due to evaporation (Mehrotra et al., 1983). One of the most
popular processes utilizing pentane is the Otisca-T process. This process focused on coal
beneficiation, not dewatering. The ultrafine feed was agglomerated using large amounts of
pentane or heptane in order to ensure coal particles were coated in oil and hydrophilic material
12 |
Virginia Tech | was rejected. When using finely micronized coals, the resulting curd-like agglomerates formed
by this process often contained less than 1% ash with a combustible recovery greater than 95%
(Keller et al., 1990).
2.4. Hydrophobic Displacement
Studies in hydrophobic displacement were initiated at Virginia Tech in 1995 and included
thermodynamic analysis and batch testing with butane. The thermodynamic analysis compared a
beginning state of coal (1) in water (3) and an end state of coal in a hydrophobic liquid (2).
Application of Young’s equation yielded the following condition for spontaneous dewatering
(Sohn et al., 1997):
dG/dA=γ cosθ<0 [1]
23
In other words, a hydrophobic liquid will displace water from coal when its contact angle (cid:2016) is
greater than 90 degrees (Figure 2-1.). In response the contact angles for several hydrocarbon
liquids (C4-C10) were measured or calculated. The contact angles increased as the carbon
number decreased. Liquefied butane (C4) had the greatest contact angle at 110°. Pentane had
the next highest at 106° (Sohn et al, 1997).
γ
Coal (1) γ 23
12
Hydropobic
Liquid (2)
Water (3) Water (3) Hydropobic γ θ γ 13
Liquid (2) 12
Coal (1)
Coal (1) γ
13
(a) (b)
Figure 2-1. a) A schematic representation of the removal of a coal particle from an aqueous phase to a
hydrophobic liquid phase with a change in free energy from γ to γ . b) The angle θ represents the equilibrium
13 12
hydrophobic liquid-in-water contact angle (Sohn et al., 1997).
13 |
Virginia Tech | 3. EXPERIMENTAL
3.1. Coal Samples
The majority of the tests were performed on coal from a single preparation plant
in order to maintain consistency for comparison. The minus 100 mesh coal sample was collected
from the flotation feed at Alpha Natural Resources’ Tom’s Creek Preparation Plant in Coeburn,
Virginia. The Tom’s Creek plant processes coal from two mines operating in the Lower Banner
and Dorchester seams. The feed slurry contained 6.0-7.1% solids with an average ash of 38%.
The slurry contained about 72% minus 325 mesh, and it was not deslimed at the plant site before
flotation due to its ease of cleaning. The slurry contained minor amounts of frother since it was
obtained from the cyclone overflow sump where frother was first introduced before flotation.
Most of the samples did not undergo further processing before being utilized. In order to vary the
solids content for testing, the buckets of slurry were allowed to sit until the solids settled and
water could be decanted. Chemicals such as flocculants or coagulants were not added to speed
settling because they would interfere with the oil agglomeration process. Three solid contents
were used for the extraction by shaking tests, i.e., the original solids content, 15%, and 30%. In
order to prevent oxidation, bucket samples were used within a few weeks. Samples that were not
used immediately were stored in sealed metal drums.
Two other samples were prepared from the original minus 100 mesh Tom’s Creek
flotation feed. One bucket was wet screened at 325 mesh, and the 325 mesh x 0 product was used
in order to evaluate the dewatering ability of oil agglomeration on slime. It was only used for the
extraction by shaking method. The second sample was used only for dewatering by
centrifugation. Though it remained a 100 mesh x 0 sample, it only contained about 28% minus
325 mesh material. The sample was created by screening the feed to remove all of the minus 325
mesh ultrafines. Some of the original unscreened feed was then mixed back into the new,
screened agglomeration feed. The final feed sample was about 6.1% solids.
Since previous hydrophobic displacement testing at Virginia Tech was performed on dry
coal, a dry coal sample was also prepared. The coal originated from the clean coal stockpile at
Tom’s Creek Preparation Plant in Coeburn, Virginia. The ¼” coal was ground using a pulverizer.
After going through the pulverizer once, the product was dry screened. The minus 80 mesh
material was saved and the oversize was discarded. The product was approximately 19% ash and
15 |
Virginia Tech | 34% minus 325 mesh. Water was added to create a 6.0% solids slurry, assuming the original
solids were dry. Two samples were prepared with this material. The first was wetted by shaking
the coal and water in a closed container for 1 minute and used immediately. Each 900 ml feed
was made separately. The second sample was created by letting the water and coal sit for three
days in a bucket. It was stirred occasionally with a rod and all of the lumps were broken to allow
wetting.
Two other preparation plants provided samples for testing. The first came from
Dickenson-Russell Coal Company’s Moss No. 3 Prep Plant in Clinchfield, Virginia. The minus
100 mesh flotation feed came from the same coal splits as those feeding Tom’s Creek. The slurry
was 6.6% solids. The second sample originated from Arch Coal’s Pardee Preparation Plant in
Appalachia, Virginia. It was cut from the six inch cyclone overflow in order to produce an
ultrafine, minus 325 mesh sample. The slurry was 9.3% solids. In both cases, chemicals, likely
cationic coagulants, were added at the plant in order to decant water and increase the solids
content.
Agglomeration tests were also performed on several anthracite and bituminous tailing
pond samples. Their origin was undisclosed. The semi-dry material was mixed with water to
create a 7.0% solids slurry and a coarse screen was used to remove foreign twigs and leaves. The
anthracite sample was 55% ash, the Pittsburgh seam sample was 33% ash, and the Sewell seam
sample was 15% ash.
3.2. Materials and Supplies
3.2.1. Pentane
The oil used in agglomeration was n-Pentane produced by Alfa Aesar Co. The pentane is
HPLC grade and 99% min. The clear liquid has a density of .626 g/ml and is immiscible in
water. Its boiling point is 36ºC, and explosive mixtures of air and pentane exist when pentane
composes 1.4-8.0% by volume. Pentane is considered a very fast evaporating liquid and has a
vapor pressure of 416 mm Hg at 20ºC. It is very flammable and may cause skin irritation. More
detailed information may be found in Alfa Aesar’s MSDS.
16 |
Virginia Tech | 3.3. Experimental Apparatus
3.3.1. Moisture Determination
In order to report moisture due only to water, not pentane and water, a special weighing
platform was designed. A bottom-loading Denver Instruments scale was placed on top of a small
oven within a fume hood. A metal platform was connected to the base of the scale via a hole in
the top of the oven. The scale was connected to a laptop so changes in weights could be recorded
and graphed in real time. A remote thermometer was used to monitor the temperature in the
oven, and the scale was partially surrounded by Plexiglas to protect it from the breeze in the
fumehood (Figure 3-1).
A condenser was also setup to capture the evaporating moisture. The liquid was
condensed in graduated container, so the volumes of pentane and water could be compared. A
500 ml flask, large enough to hold the centrifuge concentrate, held the original sample and sat on
a laboratory heater. Quarter inch tubing connected the flask to a condenser. The condenser was
chilled by cold tap water, and the condensed liquid fell into a graduated container chilled by ice
water. The system was closed to prevent vapor loss (Figure 3-2).
Figure 3-1. Weighing platform used for moisture Figure 3-2. Condenser apparatus.
determinations.
18 |
Virginia Tech | 3.3.2. Agitation and Mixing
3.3.2.1. Hand Shaking
Most oil agglomeration was done by simple hand shaking. Pentane is a volatile liquid and
evaporates quickly if open to the atmosphere. Hand shaking provided a simple solution to
mixing in a closed container. The original oil agglomeration tests were performed in a 500 ml
cylindrical separatory funnel. The cylindrical shape was chosen because it created a thicker bed
of agglomerates. Later centrifuge testing required larger samples and a 1000 ml separatory
funnel was utilized. The funnels were each used with a rubber stopper and stopcock.
3.3.2.2. Impeller Mixing
Impeller mixing was also explored in order to provide more consistent agitation than
hand shaking. First, a glass separatory funnel with a ground glass and o-ring sealed bearing was
used. Though the bearing held against the vapor pressure, the mixing blades hit the edges of the
container at high speeds. For safety reasons, a cylindrical plexiglass container with a top and
bottom port was created for continuous mixing. A ground bearing with vacuum grease was used
to seal the mixing shaft. However, the bearing around the mixing shaft could not withstand the
vapor pressure, and the container leaked pentane. Finally, impeller mixing open to atmosphere
was attempted with a laboratory Denver flotation cell. Air was not used during the mixing.
3.3.3. Cleaning and Water Transfer
During the centrifugation process, a pseudo-enclosed water transfer system was created
to minimize pentane loss. After mixing, the 1000 ml separatory funnel used for hand shaking
was connected between two other containers for liquid exchange. The top container held clean
wash water and the bottom flask held the exchanged clay water. Quarter inch rubber tubing
completed the circuit between the three containers to prevent a vacuum, allow liquid exchange,
and prevent pentane loss. Stopcocks above and below the funnel containing the agglomerates
allowed it to be sealed while water was added or emptied from the other two glasses. Small
amounts of pentane vapor were lost from these transfer containers whenever they were open to
atmosphere (Figures 3-3 and 3-4).
19 |
Virginia Tech | Figure 3-3. Base of water fransfer apparatus with Figure 3-4. Top of water transfer apparatus with
Fluorescein dye. Fluorescein dye.
3.3.4. Centrifugation
A sealed centrifuge was the main component of this process. The centrifuge was driven
by a ½ hp, variable speed DC motor. A one to four pulley ratio enabled the centrifuge to operate
between 0 and 4360 rpm. The centrifuge contained an inner basket lined with centrifuge bars,
and the sample was poured into this basket through a port in the middle of the lid. The inner
basket contained an insert in the middle which forced the coal to the sides of the container. It
was added to assist in even cake formation. The inner basket was enclosed in a solid outer shell.
The entire centrifuge was made of aluminum and was held by bearings on the top and bottom.
Though the centrifuge could hold about 450 ml of liquid while still; if it was spinning, it could
only hold 160 ml before the liquid would touch the outside of the inner basket. The centrifuge
contained a plug in the bottom so water could be drained after centrifugation was complete, and
a rubber stopper in the top of the lid sealed the centrifuge (Figures 3-5 and 3-6). Scaled
drawings of the centrifuge are located in the Appendix (Figures A-1 through A-6).
20 |
Virginia Tech | Figure 3-5. Centrifuge and stand. Figure 3-6. Inner centrifuge assembly.
3.3.5. Filtering and Displacement
Two types of vacuum filters were used in the testing. i) The main Denver filter was
approximately twelve inches in diameter and was used with filter paper to dewater tailings before
they were placed in the oven. ii) A small Peterson filter was used to measure concentrate
moisture due to filtering and to determine the void space of a packed coal cake. The Peterson
filter was only 6.3 cm in diameter and could hold about 500 ml of slurry prior to filtering. It was
generally used with filter paper. Both filters were connected to a four liter flask which stored the
removed water, and a flask of desiccant was located before the vacuum pump to protect the
machinery. Stiff, quarter inch tubing connected the components. The vacuum pump could pull
up to 20 in Hg.
The displacement process also utilized the Peterson filter, but it was powered by a water
aspirator. The Peterson filter sat on a four liter flask which was connected directly to the water
aspirator. Since displacement utilized large amounts of oil, all of the filtering equipment,
including the aspirator, was located within a fumehood (Figure 3-7).
21 |
Virginia Tech | 3.3. Experimental Procedures
3.3.1. Moisture Determination
Correct moisture determination played a critical role in the experiments. In a commercial
application, pentane would be recovered before the product was shipped; therefore, it would be
deceptive if the reported moisture included anything other than water. Four different methods
were explored to determine the product’s water content. A moisture balance was used to identify
the original problem. A weighing platform constituted a solution to the problem, and the
condenser and fluorescein methods were introduced to attempt to validate the weighing platform
method.
3.3.1.1. Moisture Balance
The original purpose of the Ohaus moisture balance was to confirm the variable
evaporation rates of pentane and water. Its other purpose was to determine if the produced
baseline evaporation curves could be used to predict the time needed to remove pentane. In
order to reach these goals, the balance was set at a constant 40ºC and was connected to a laptop
for live recording of data. Thirty tests were run with mixtures of 0g, 5g, 10g, or 20g of water
and/or pentane and 0g or 15g of coal. Once the desired weights were determined, the coal (a dry,
clean flotation concentrate) was placed into the aluminum weighing dish. Densities of 1 g/ml for
water and .626 g/ml for pentane were used to calculate the desired volumes of the liquids. Next,
the appropriate amount of water was measured and stirred into the coal. Pentane was added next,
and after a quick mixing, it was placed into the balance before too much pentane could
evaporate. Weights were recorded until either the sample weight remained constant or 60
minutes passed. Once the testing was completed, the resulting change in weights and
evaporation rates were plotted against time to identify trends and attempt to predict needed
evaporation times.
23 |
Virginia Tech | 3.3.1.2. Weighing Platform
The weighing platform was created to estimate the water-only moisture content in large
concentrate samples. Before beginning, the oven was set at 40ºC. This temperature is close to
the boiling point of pentane (36ºC); a higher temperature was not used due to the risks associated
with the flammable liquid. A remote digital thermometer was used to monitor the temperature,
and once the oven reached the desired temperature, the scale was connected to a laptop.
Concentrate samples were placed in deep, 500ml glass petri dishes. After the concentrate was
placed on the platform, the oven temperature was manually adjusted to maintain the temperature
within 1ºC of 40ºC. As pentane evaporated, it chilled the container and surrounding air. To
compensate for this immediate drop in temperature and the heat loss due to opening the oven
door, the oven was usually set a few degrees above the desired temperature and changed back to
the desired set point after the concentrate was in the oven.
Excel was used to plot the changing weight against time and calculate the evaporation
rates (Figure 3-1). Unlike water, pentane is considered a very fast evaporating liquid. The
32
30
28
26
24
22
20
18
16
14
12
0 10 20 30 40 50 60
Time (Minutes)
Figure 3-8. Example of an evaporation curve obtained from the weighing platform.
24
)mg(
ssaM
80
70
60
50
40
30
20
10
0
)s/mg(
etaR
noitaropavE
15 gm Coal
10 gm Water Mass (gm)
5 gm Pentane
Rate (gm/s)
Moisture
Mass (gm) |
Virginia Tech | was exchanged for clean water, the clean water also contained dye. The sample underwent
centrifugation, and the resulting cake was examined in a dark room with an ultraviolet light.
Several agglomerates were also removed from the cake and examined under a microscope with
ultraviolet light.
Finally, fluorescein-rich water was mixed with dry centrifuge concentrate to create a 20%
moisture mixture. This represented the high moistures coming out of the centrifuge. The sample
was stirred until the coal was thoroughly wetted, and the sample was examined in a dark room to
see if fluorescein was visible when spread over such a large surface area.
3.3.2. Agitation and Mixing
3.3.2.1. Hand Shaking
In this project, the standard method for agglomeration agitation was hand shaking. It was
difficult to create a small sealed mixing container for pentane, so a simple separatory funnel and
rubber stopper were used. Since hand shaking is not consistent and to ensure complete
agglomerate formation, the slurry sample was shaken for about five minutes. It was shaken the
full time regardless of whether or not the agglomerates formed in a shorter amount of time.
3.3.2.2. Impeller Mixing
Two impeller mixing methods were attempted for oil agglomeration. First, a plexiglass
container described in the apparatus section was used to agglomerate with pentane. After the two
ports were sealed, slurry and pentane were added, and a variable speed motor provided mixing.
Second, agglomeration open to atmosphere was attempted. Since pentane evaporates quickly,
heptane was used instead. Slurry was placed in a 1.5 liter Denver flotation cell. The flotation cell
was used only for agitation; air was not used.
3.3.3. Cleaning and Water Transfer
Water transfer consists of three stages: clay water removal, cleaning, and water
replacement. Before transfer began, the stopper in the shaking separatory funnel was replaced
with a new stopper that had a tube running through it. The tube was designed to allow clean
water to enter while clay water drained through the stop cock at the base of the glass. The first
26 |
Virginia Tech | step in cleaning was to remove the bulk of the clay water. The upper chamber remained empty
during this stage. Both stop cocks were opened, and the agglomerates were allowed to gravity
drain until no water remained. Then the stop cocks were closed again. To begin the cleaning
stage, about 600 ml of clean water were placed in the upper chamber while the bottom was
emptied of the clay water. The top stop cock was opened, allowing water to enter the
agglomerate chamber, and the agglomerates were shaken slightly to free the clay trapped
between them. The bottom stop cock was opened once about 400 ml of clean water had been
transferred. The remaining clean water continued to wash the agglomerates as water drained
through the base. Once all of the water had been drained from the agglomerates, the stop cocks
were closed again to isolate the chamber. To begin the water replacement stage, about 100 ml of
clean water were added to the top chamber. The top stop cock was opened, and the clean water
moved into the agglomerate chamber. The stopper was quickly exchanged for the original, solid
stopper, and the separatory funnel was disconnected from the bottom chamber. The water
transfer was complete.
3.3.4. Agglomerate Extraction and Dewatering
3.3.4.1. Shaking
This method was used in conjunction with agitation by hand mixing. No cleaning stage
was utilized; instead clay water was simply drained from the bottom stop cock. The
agglomerates were gently shaken up and down a few times to release the remaining trapped
water. Once the bulk of the water was removed, the cylindrical separatory funnel was held
above a metal pan at about 20 degrees from horizontal. The glass was shaken in an irregular
motion. The agglomerates were thrown up the glass and out the opening while the bulk of the
water remained in a pool at the other, lower end. The shaking time was arbitrary; since the goal
was dewatering, shaking stopped when too many water droplets exited the glass. The material
shaken out was considered the concentrate, and it was scraped into a glass dish with care being
taken to also remove the water smears from the metal pan. The concentrate’s moisture was then
measured with the weighing platform. Any solids still in the original container were considered
middlings. Though they were clean, they had a high moisture content due to the pooled water.
Unfortunately, the middling’s moisture could not be determined because wash water was used to
remove the solids from the glass, raising the water content.
27 |
Virginia Tech | It should be noted that small variations in this process were needed based on the dosage
of oil used. As the oil dosage increased and the balls became soft, additional shakes were not
used to remove extra water after the main draining because shaking collapsed the drainage
channels. As dosage further increased, the consistency of the material was not conducive to
shaking. In these cases, all of the material was simply poured into the concentrate dish,
essentially eliminating the middlings.
3.3.4.2. Screening
This screening method is different from the classical method; instead of water falling
through the screen, water rich solids remained on top of the screen while dry material fell
through. This method also used the finest screens first and proceeded coarser, the opposite of
traditional screening used for sizing. Four screens were utilized: 28, 65, 100, and 270 Tyler
mesh. Fitted solid metal pans were placed under each screen to prevent material loss. The
powdery flocs constituting the feed were produced by oil agglomeration with a low dosage of oil.
Some of the flocs were poured over the smallest screen, and the screen was gently shaken back
and forth to sieve the dry powder through. More feed was slowly added until the entire batch had
been screened. Shaking caused the water droplets to coalesce and roll on the surface of the
screen. Screening stopped once no more material passed through, the screen became blinded, or
the screen began to wet. The undersize was weighed for moisture; the weighing platform was
not utilized because initial weighing with it indicated no pentane was present after screening.
Meanwhile, the oversize was placed on the next larger screen, and the process was repeated for
each screen. Care was taken when transferring the oversize in order to prevent the water droplets
from wetting the metal. Since water could not be used to rinse the oversize onto the next screen,
any solids or water remaining on the screens were considered lost.
3.3.4.3. Air Classification
In this process, agglomerates were removed from the water phase with an air hose. The
hose was inserted into the separatory funnel and used to blow the top agglomerates out. The aim
was to remove the agglomerates sitting above the liquid level. As agglomerates were removed
from the top of the bed, more filled the gaps due to buoyancy.
28 |
Virginia Tech | 3.3.4.4. Centrifugation
Centrifugation of spherical agglomerates was performed only after a sample had been
cleaned with the water transfer procedure. A funnel was placed in the top of the sealed
centrifuge, and the agglomerates were poured out of the separatory funnel. Due to the limited
volume in the centrifuge, only 50 ml of wash water were used in addition to the original 100 ml
of water with the agglomerates. The rubber stopper was replaced once the agglomerates were
inside the centrifuge, and the time required to pour the material was recorded in case a sample
took an excessive amount of time to pour, allowing pentane to evaporate. The variable speed
centrifuge was then turned on for a specified amount of time. It should be noted that the
centrifuge was generally off during the pouring stage in order to create a more even cake.
3.3.4.5. Filtering
Filtering was performed with clean compact spherical agglomerates floating in clear
water. Once the slurry was placed in the Peterson filter, the vacuum pump was turned on, and
the valve was opened. The initial filtering time was recorded, and the agglomerates were
generally allowed to dry for an additional one minute after the entire cake’s surface was exposed
to air. The pump operated at maximum vacuum in all of the tests, and the vacuum pressures
were recorded for the filtering and drying phases. The dewatered sample was collected from the
filter and subjected to a standard moisture determination. After initial testing with the weighing
platform indicated pentane had evaporated during filtering, the initial weight for moisture
determination was simply taken at time zero.
3.3.4.6. Displacement
Displacement consisted of filtering coal slurry with large amounts of pentane. Dry
pulverized clean coal from Tom’s Creek was utilized for all of the tests. Feed samples were
prepared by shaking coal and water together for one minute; the resulting 400 ml slurry
contained 6.0% solids. After the slurry was poured into the filter, about 80 ml or one inch of
pentane was slowly poured on top. Care was taken to pour the oil slowly to prevent the
formation of coal-covered water droplets in the pentane phase. The valve was then opened to
begin filtering. Several times corresponding to important events were recorded during the course
29 |
Virginia Tech | of each experiment. These included the approximate time when water finished filtering, the total
filtering time, and the drying time, usually 30 or 60 seconds. Since a water aspirator was used
and multiple liquids were being filtered, the pressure varied throughout the test. Only the
starting pressure could be controlled by water flow. Pressures were recorded at the beginning
and end of the water and pentane phases.
3.3.5. Conventional Cleaning and Dewatering
3.3.5.1. Flotation
Two sets of flotation testing were performed: release analysis and single-stage cleaning.
A release analysis was performed on the minus 100 mesh flotation feed from Tom’s Creek and
on each of the tailing pond samples. In each case, the collector was diesel, and the frother was
MIBC. Though recorded, the amount of collector was not regulated during release analysis since
the goal was complete cleaning. Each sample underwent five cleaning stages; the concentrate
was cleaned five times with one cumulative tailings. During the fifth stage, the concentrate was
removed in eight batches of approximately equal weights. These samples were weighed and
ashed, and the resulting recoveries were plotted against the ash content.
Single stage cleaning was also performed to represent conventional cleaning. The
product then underwent filtration, and the final moisture could be used to compare conventional
and pentane dewatering. The 1.5 liter flotation cell was agitated at 1100 rpm, and all of the tests
used 3µl of MIBC frother. Several Nalco reagents and diesel were screened as collectors at 200
and 400 g/ton or at 20 and 40 µl. The cells were run to exhaustion.
3.3.5.2. Filtering
Three conventional filtering series were performed with the use of a Peterson filter. The
first filter series utilized flotation concentrate from single-stage laboratory flotation. The dry
concentrate was mixed with water for a solids content of 10%. Each feed sample contained
250ml of slurry. Baseline filtering was performed without any dewatering aids. Tests were also
performed with the RV and RW dewatering aids developed at Virginia Tech. The samples were
conditioned with the chemicals for two minutes prior to filtering. All filtration was performed at
maximum vacuum pressure with a drying time of two minutes.
30 |
Virginia Tech | 4. RESULTS
4.1. Samples
Testing revolved around three main samples from Alpha Natural Resources’ Tom’s
Creek Plant. A Microtrac size analyzer was used to determine the size distributions for each
sample. The surface areas of the concentrates were then estimated from the size distribution data
by assuming spherical particles.
The first sample consisted of unaltered flotation feed from the Tom’s Creek plant. This
minus 100 mesh sample was used to evaluate all of the dewatering methods explored in this
study. The sample had an average ash of 38% and was approximately 72% minus 325 mesh.
Figure 4-1 shows the relative size distributions of a centrifuge feed, concentrate, and tailings
originating from this sample. It has an estimated surface area of 14,300 m2. This sample also
underwent BET analysis, which used gas adsorption to determine the total surface area,
including pore area, of the sample. Unfortunately, the data from the BET analysis was not
available at the time this report was prepared.
The second sample also originated from Tom’s Creek’s minus 100 mesh flotation feed;
however, a portion of the ultrafines were removed by sieving. The deslimed sample contained
19% ash, 28% minus 325 mesh, and had an estimated surface area of 320 m2. Figure 4-2 shows
the relative size distributions of a centrifuge feed, concentrate, and tailings originating from this
sample.
The third sample was obtained by dry pulverizing a sample of clean coal from the Tom’s
Creek plant. The resulting minus 80 mesh sample contained 19% ash, 34% minus 325 mesh, and
had a surface area of 380 m2. The centrifuge size distributions are provided in Figure 4-3.
Figure B-1 in the Appendix provides further information on the Microtrac’s sizing
methods and accuracy, and Figures B-2 through B-4 contain further size distribution information
for the three samples.
32 |
Virginia Tech | 4.2. Moisture Determination
The preferred method of moisture determination utilized a weighing platform. Before
this method was implemented, the initial weight for moisture determination was found by
placing a sample in an oven for 20 minutes at 40ºC to evaporate pentane. No water was assumed
to evaporate during this initial period. Any loss in weight after this period was attributed to
moisture remaining in the sample. Unfortunately, this approach tended to misrepresent moisture.
Moistures were underestimated for concentrates produced with small oil dosages and were
overestimated for concentrates with large oil dosages. To overcome this problem, moisture
determinations for future samples were conducted using the weighing platform. This device
provides an evaporation curve showing the sample weight loss as a function of time.
Evaporation rates were also calculated from the recorded weights and times (Figure 4-4).
The example evaporation curve in Figure 4-4 is from a concentrate with a large oil
dosage. It displays a sharp elbow, and the two legs are due to two distinct evaporation schemes.
The first leg has high evaporation rates which fall from 2.9 g/min to .4 g/min. The initial rise in
240 3.0
238
236
2.5
234
232
230 2.0
228
226
1.5
224
222
220 1.0
218
216
0.5
214
212
210 0.0
0 5 10 15 20 25 30 35 40 45
Figure 4-4. Example how the initial weight is chosen for moisture determination from a weighing platform evaporation
curve.
34
)g(
thgieW
)nim/g(
etaR
noitaropavE
initial weight
weight evaporation rate
Time (min) |
Virginia Tech | small amounts of pentane were used. It was assumed the pentane evaporated during handling.
The original moistures displayed errors of 10-55%. Therefore, all of the moisture contents
reported in this document were determined using the weighing platform. Tables B-1 and B-2 in
the Appendix contain detailed data on the original incorrect moistures.
4.3. Agitation and Mixing
Agglomeration using pentane was performed with a wide range of dosages. The dosages
were roughly based on pentane-to-solid mass ratios of 1:9, 1:7, 1:5, 1:4, 1:3, and 1:1. Based on a
feed of 400 ml at 15% solids and an ash of 39%, these ratios correspond to pentane-to-coal mass
ratios (M :M ) of 0.18, 0.23, 0.32, 0.40, 0.53, and 1.60.
p c
Figure 4-5 roughly shows the trend in agglomerate size as dosage changes. The lowest
dosage produces floc-like agglomerates less than 0.5 mm in diameter. The agglomerates appear
powdery and float on the surface of the water phase; however, large amounts of water are
trapped within the powder. The next three dosages produce spherical agglomerates that increase
in size from 0.5 to 2 mm in diameter as the oil level increases. The agglomerates float in the
40
35
30
25
20
15
10
5
0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8
Figure 4-5. Estimated change in agglomerate size and appearance with pentane dosage.
36
)mm(
retemaiD
etaremolggA
Mass Ratio (Pentane:Coal)
scolf
setaremolggalacirehps
sdruc |
Virginia Tech | water near the water-air interface. As the oil level increases, they become shinier and clump
slightly. There are distinct drainage channels between the agglomerates. The next dosage
produces black, cottage-cheese-like curds. These semi-solid curds are extremely soft and tend to
trap large amounts of water. Finally, the highest dosage of pentane tends to forms a semi-solid
layer of coal, water, and pentane near the top of the water phase. There are no distinct
agglomerates, and a clear layer of pentane lies on top of the coal.
4.4. Pellet Extraction and Dewatering
4.4.1. Shaking
Hand shaking tests were performed on four different samples, which all originated from
the flotation feed at the Tom’s Creek plant. The first two were 100 mesh x 0 samples of flotation
feed containing solids contents of about 6% and 15% by weight. The third was identical to the
first two except that the samples had been aged (oxidized) for four months prior to testing. The
fourth was a much finer sized sample (i.e., 325 mesh x 0). In order to evaluate any
reproducibility errors in moisture determination, each series of tests was performed multiple
times. The first two series of tests were repeated three times; however, the tests performed with
oxidized and 325 mesh x 0 samples were only performed once due to lack of sample.
Tables 4-2a and 4-2b shows the results of the first series of tests which were performed
on fresh 100 mesh x 0 feed without using the weighing platform to determine moisture. The tests
were repeated using the weighing platform and the results are summarized in Tables 4-3a and 4-
3b. A comparison of the data from Tables 4-2 and 4-3 show that the true moistures obtained
using the balance are significantly higher than those obtained without the balance. The tests
conducted at lower percent solids provided slightly lower moistures. The lowest, at an average
moisture of 16.2%, occurred at a M :M ratio of 0.32, which corresponded to a dosage of about
p c
1:5 pentane to solids by weight.
37 |
Virginia Tech | The data reported in Tables 4-4a and 4-4b utilized the same 100 mesh x 0 flotation
sample after it had been aged four months. This particular set of experiments was conducted at
14.9% solids, which is very close to the solids content of 14.8% solids used to obtain the lower
set of data for the fresh sample reported in Table 4-3. As shown, the moistures obtained using the
oxidized samples increased up to 137% when compared to the moistures obtained using the fresh
samples. The difference was particularly notable for the experiments conducted at the lower
range of pentane dosages. On the other hand, the moisture values obtained at the higher dosage
levels were relatively unaffected by the aging and potential oxidation of the sample.
The last series of agglomeration tests were conducted using a 325 mesh x 0 split of the
Toms Creek flotation feed. The results obtained for the sized sample are shown in Tables 4-5a
and 4-5b. As shown, the moisture values obtained using the minus 325 mesh split were
Table 44a. Measured values for hand shaking with oxidized 100 mesh x 0 flotation feed.
Mass Ratio Ash Content (%)
(Pentane:Coal) Tailings Middlings Concentrate
14.9% solids by weight
0.12 90.76 15.45 8.71
0.16 81.50 9.52 7.44
0.22 83.68 9.62 6.43
0.28 82.02 10.31 5.46
0.37 84.90 27.09 6.67
1.11 76.18 48.32 21.22
Table 44b. Calculated values for hand shaking with oxidized 100 mesh x 0 flotation feed.
Mass Ratio Yield (%) Recovery (%) Ash Rejection (%) Moisture
(Pentane:Coal) M+C C M+C C M+C C (%)
14.9% solids by weight
0.12 77.7 70.7 97.1 89.2 73.3 77.7 46.2
0.16 73.5 52.3 93.2 67.0 78.1 85.9 46.6
0.22 73.2 62.7 93.9 81.0 81.2 85.4 35.4
0.28 71.7 63.2 93.0 82.5 83.9 87.5 25.2
0.37 73.5 72.5 94.5 93.5 81.4 82.5 42.0
1.11 88.7 88.1 96.3 95.9 31.1 32.4 82.9
*Note: M+C indicates the yield, combustible recovery, or ash rejection’s concentrate consists of the
shaking middlings and concentrate. C indicates only the shaking concentrate is used in the
calculations.
40 |
Virginia Tech | Table 45a. Summary of measured values for hand shaking with 325 mesh x 0 flotation feed.
Mass Ratio Ash Content (%)
(Pentane:Coal) Tailings Middlings Concentrate
7.3% solids by weight
0.22 91.39 6.68 7.59
0.28 91.46 6.22 6.63
0.40 90.52 6.26 6.22
0.50 89.78 7.53 6.63
0.66 84.52 14.38 9.31
1.99 81.24 14.82 14.83
Table 45b. Summary of calculated values for hand shaking with 325 mesh x 0 flotation feed.
Mass Ratio Yield (%) Recovery (%) Ash Rejection (%) Moisture
(Pentane:Coal) M+C C M+C C M+C C (%)
7.3% solids by weight
0.22 47.4 30.7 90.7 58.6 93.4 95.5 34.0
0.28 47.0 24.6 90.7 47.3 94.1 96.9 28.7
0.40 46.3 23.4 89.5 45.2 94.4 97.2 27.7
0.50 46.3 28.5 88.7 54.9 93.7 96.3 35.1
0.66 44.1 41.2 82.2 77.0 91.7 92.6 47.6
1.99 44.8 42.9 78.6 75.3 87.0 87.7 72.0
*Note: M+C indicates the yield, combustible recovery, or ash rejection’s concentrate consists of the
shaking middlings and concentrate. C indicates only the shaking concentrate is used in the
calculations.
considerably higher than those obtained using the coarser 100 mesh x 0 sample. This finding
shows that agglomeration procedure is sensitive to particle size. The lowest moisture for the
ultrafine sample occurred at a M :M ratio of 0.4, which provided a moisture of 27.7%. The
p c
27.7% moisture was 70% higher than that obtained for the 100 mesh x 0 sample. This large
increase cannot be attributed to only the slight increase in solids content of the sample.
Figure 4-6 shows the recovery-rejection curve for the hand shaking tests conducted using
the fresh 100 mesh x 0 sample at different pentane dosage levels. Only the data from this sample
was included in the summary plot since it provided the lowest overall moisture values. As
shown, most of the data points fall along a single recovery-rejection curve, which suggests that
increasing the oil dosage does not shift the separation curve. On the other hand, higher dosage
values did tend to move each grouping of data points to a higher recovery level on the curve.
41 |
Virginia Tech | 4.4.2. Screening
Another method used to recover pentane agglomerated coal was screening. In this
method, floc-like agglomerates were placed into individual screens with aperture openings
ranging from 30 to 270 mesh. Dry solids passed through the sieve and were collected in pans,
while coal-covered water droplets were retained atop the sieve (i.e., shaking caused the water
droplets to coalesce and roll over the surface of the sieve). The screening tests were performed
using a 100 mesh x 0 sample of flotation feed from the Toms Creek facility.
The test results from the screening tests are summarized in Table 4-6. It should be noted
that many of the size classes contained less than one gram of sample; therefore, there are high
errors associated with all of this data. Nevertheless, this procedure generated products with
moistures contents ranging from a high of 50% for the coarsest sieve (30 mesh) down to a low of
6.1% for the finest sieve (270 mesh). Unfortunately, the procedure provided an overall
combustible recovery of only 67.8%. Moreover, the bulk of the recovered solids were obtained
from size fractions that corresponded to relatively high moisture values. The recoveries
associated with products having 6.5% moisture or lower accounted for just 29.3% of the
cumulative recovery (i.e., 23.1 + 6.2 = 29.3).
Table 46. Example of screening with 100 mesh x 0 flotation feed.
Particle Size Class Measured Calculated
Pass Retain Mean Ash Dry Wt Moisture Yield Recovery
(mesh) (mesh) (mm) (%) (g) (%) (%) (%)
30 0.600 6.44 0.30 50.0 3.7 5.9
30 70 0.406 5.91 0.74 27.5 9.2 14.6
70 100 0.181 5.10 0.91 12.5 11.3 18.1
100 270 0.102 4.76 1.16 6.5 14.4 23.1
270 0.053 4.60 0.31 6.1 3.8 6.2
Tailings 87.49 4.65 97.4 57.6 12.2
Total 42.4 67.8
*Note: The feed consisted of 200 ml of 41% ash 5.5% solids slurry. Approximately 8.5% of the
material was lost to the screens and was not included in the yields and combustible recoveries. All of
the screens are US mesh.
44 |
Virginia Tech | 4.4.3. Centrifugation
One of the most promising methods used to recover agglomerates was centrifugation.
Three samples were dewatered using this technique: (i) the original 100 mesh x 0 sample
containing 49% minus 325 mesh solids, (ii) a sample of 80 mesh x 0 pulverized coal containing
about 32% minus 325 mesh solids, and (iii) deslimed 100 mesh x 0 sample containing 19%
minus 325 mesh solids. The original and deslimed samples were conditioned with pentane using
an average M :M ratio of 0.32, while a slightly lower average M :M ratio of 0.21 was utilized
p c p c
for the coarser 80 mesh x 0 sample. Pentane was occasionally added after the cleaning stage to
replace any pentane that may have been lost to evaporation. Although the added pentane helped
to reduce the moisture, it also made the agglomerates softer and harder to handle. When an
excessive dosage of pentane was added, the agglomerates tended to clump together and become
stuck when feeding the centrifuge.
The data given in Table 4-7 shows that the moistures obtained by centrifugation using the
original 100 mesh x 0 sample were in the 18-19% range. These moisture values are significantly
Table 47. Reduction in moisture based on centrifuge sample.
Spin Time (sec) Moisture (%) Moisture Reduction (%)
original 100 mesh x 0 with 49% ‐325 mesh
15 20.3 ‐‐
30 18.5 ‐‐
60 18.8 ‐‐
120 19.4 ‐‐
deslimed 100 mesh x 0 with 19% ‐325 mesh
5 12.6 38.2
15 13.3 28.2
30 13.2 29.5
60 14.0 27.7
pulverized 80 mesh x 0 with 32% ‐325 mesh; soaked 1 min
5 9.9 51.4
15 8.4 54.5
30 7.8 58.3
60 7.5 61.1
pulverized 80 mesh x 0 with 32% ‐325 mesh; soaked 3 days
5 8.6 57.5
15 8.2 55.8
30 8.6 54.4
60 8.7 55.2
*Note: Moisture reduction was based on the difference from the original 100 mesh x 0. The minus
325 mesh values refer to the ultrafines in the concentrate.
45 |
Virginia Tech | Table 48a. Measured centrifugation results for 100 mesh x 0 flotation feed.
Mass Ratio Pentane Centrifuge Drying Ash Content (%)
(Pentane:Coal) (ml) Time (sec) Speed (rpm) Tailings Concentrate
0.32 0 30 3280 79.70 4.11
0.32 1 30 3280 81.20 4.11
0.32 2 30 3280 77.69 4.00
0.32 0 60 3280 81.41 4.08
0.32 1 60 3280 79.12 3.93
0.32 2 60 3280 76.93 4.07
0.32 3 60 3280 75.73 3.74
0.32 2 60 3280 73.78 3.82
0.33 2 15 880 75.65 3.95
0.33 2 30 880 83.41 4.15
0.33 2 60 880 81.98 4.05
0.33 2 120 880 84.94 4.15
0.32 2 15 3280 82.05 4.00
0.32 2 30 3280 80.17 3.87
0.32 2 60 3280 81.04 3.89
0.32 2 120 3280 81.06 3.92
0.31 2 15 2040 79.00 0.90
0.31 2 30 2040 80.76 3.92
0.31 2 60 2040 85.15 4.04
0.31 2 120 2040 82.44 4.08
*Note: All of these tests were conducted with 900 ml at 6.0% solids. The variations in mass ratio were
due to changes in feed ash; 17.2 ml of pentane were used with each test. The pentane listed above
was added after the agglomerates were cleaned. The centrifuge was still during feeding except for
the eighth test which was fed at 880 rpm.
47 |
Virginia Tech | Table 49a. Measured centrifugation results for screened 100 mesh x 0 flotation feed.
Centrifuge Drying Ash Content (%)
Mass Ratio
(Pentane:Coal)
Time (sec) Speed (rpm) Tailings Concentrate
0.34 5 3280 89.09 3.90
0.34 15 3280 88.53 3.91
0.34 30 3280 88.88 3.94
0.34 60 3280 88.99 3.91
*Note: All of these tests were conducted with 600 ml at 6.1% solids. No extra pentane was added
after the agglomerates were cleaned.
Table 49b. Calculated centrifugation results for screened 100 mesh x 0 flotation feed.
Mass Ratio Yield (%) Recovery (%) Rejection (%) Moisture (%)
(Pentane:Coal) Cleaning Centrifuge Cleaning Centrifuge
0.34 81.7 100 97.5 100 83.6 12.6
0.34 81.6 100 97.4 100 83.6 13.3
0.34 81.7 100 97.5 100 83.5 13.2
0.34 81.7 100 97.5 100 83.6 14.0
*Note: The yields were calculated based on ash content. The centrifuge yield and combustible
recovery are 100% because there were no measureable tailings. The ash rejection refers to the
cleaning stage since no tailings were produced in the centrifuge. The concentrate for the cleaning
stage calculations include the centrifuge concentrate and the material lost during centrifugation.
Table 410a. Measured centrifugation results for pulverized 80 mesh x 0 clean coal.
Centrifuge Drying Ash Content (%)
Mass Ratio
(Pentane:Coal)
Time (sec) Speed (rpm) Tailings Concentrate
coal soaked 1 minute
0.21 5 3280 76.36 8.61
0.21 15 3280 79.30 8.02
0.21 30 3280 60.67 7.78
0.21 60 3280 40.45 7.71
coal soaked 3 days
0.21 5 3280 77.08 6.59
0.21 15 3280 79.73 7.30
0.21 30 3280 84.69 7.21
0.21 60 3280 79.21 7.33
*Note: All of these tests were conducted with 900 ml at 6.0% solids. No extra pentane was added
after the agglomerates were cleaned.
49 |
Virginia Tech | Table 410b. Calculated centrifugation results for pulverized 80 mesh x 0 clean coal.
Mass Ratio Yield (%) Recovery (%) Rejection (%) Moisture (%)
(Pentane:Coal) Cleaning Centrifuge Cleaning Centrifuge
coal soaked 1 minute
0.21 82.6 100 94.8 100 65.2 9.9
0.21 82.6 100 95.5 100 67.6 8.4
0.21 76.1 100 88.2 100 71.0 7.8
0.21 61.1 100 70.9 100 76.9 7.5
coal soaked 3 days
0.21 83.2 100 95.3 100 70.3 8.6
0.21 84.6 100 96.2 100 66.5 8.2
0.21 85.5 100 97.3 100 66.6 8.6
0.21 84.5 100 96.1 100 66.4 8.7
*Note: The yields were calculated based on ash content. The centrifuge yield and combustible
recovery are 100% because there were no measureable tailings. The ash rejection refers to the
cleaning stage since no tailings were produced in the centrifuge. The concentrate for the cleaning
stage calculations include the centrifuge concentrate and the material lost during centrifugation.
Figure 4-9 shows the effect of centrifuge speed and spin time on moisture. Three
different rotation speeds (880, 2040 and 3280 rpm) and four different spin times (15, 30, 60 and
120 seconds) were examined. Since no replicate tests were performed, each point in the plot
represents the data for only one test run. As a result, the impact of spin time on product moisture
is not apparent due to the large degree of scatter in the experimental data. Nonetheless, logic
would suggest that a longer spin time should result in lower overall moisture values (i.e., there is
no reason to believe that a sample could regain moisture by extending the spin time). On the
other hand, there appears to be a visible decrease in moisture content as the speed was increased
from the lowest value of 880 rpm to the highest value of 3280 rpm. The lower moistures can be
attributed to an increase in centrifugal g-force with increasing rotation speed. The ash and fines
(minus 325 mesh) contents for these samples are summarized in Table 4-11.
50 |
Virginia Tech | Finally, the data plotted in Figure 4-10 summaries the effects of particle size and spin
time on cake moisture for a fixed rotation speed of 880 rpm. The data clearly shows that
moisture decreases sharply with increasing particle size. For the 100 mesh x 0 sample, a
reduction in the ultrafines content reduced the moisture from about 18-20% down to 12-14%. A
further reduction down to single-digit values in the 7-9% moisture range was obtained using the
80 mesh x 0 coal. It is interesting to note that this sample actually contained a higher percentage
of minus 325 mesh solids than the deslimed 100 mesh x 0 sample (Figure 4-11). The data also
tends to suggest that both the spin time and soaking time (for the case of the 80 mesh x 0 sample)
has a comparatively low impact on the final moisture content.
25
20
15
10
5
0
0 20 40 60 80 100 120
Figure 4-10. Effect of centrifuge sample on moisture.
*Note: Tests conducted with pentane‐to‐coal ratios of 0.32‐0.34 for the
100 mesh x 0 samples and 0.21 for the pulverized 80 mesh x 0 samples.
52
)%(
erutsioM
Spin Time (sec)
original 100 mesh x 0 screened 100 mesh x 0
pulverized 80 mesh x 0 ‐soaked 1 min pulverized 80 mesh x 0 ‐soaked 3 days |
Virginia Tech | It is interesting to note that though moisture does not appear to be directly related to ultrafines
and surface area (Figure 4-11), which is unusual for a fine dewatering method, moisture does
appear to have a semi-linear relationship to void space (Figure 4-12). In other words, the smaller
the void space and available volume for water, the lower the moisture. However, this apparent
dependence needs further investigation to ensure that void space is calculated correctly. It is
currently based on the packing in a filter cake. It should also be noted that the pentane-to-voids
volume ratio is higher for the 80 mesh (.67) than the 100 mesh (.52) due to different packing.
The 80 mesh sample has the least available volume for water to fill. This may be a controlling
factor instead of just void space.
4.4.4. Filtering
The use of vacuum filtration was also examined as a possible method for dewatering and
recovering coal-pentane agglomerates. The filtration tests were performed using all three of the
samples examined in the centrifugation tests (i.e., original 100 mesh x 0 sample, deslimed 100
mesh x 0 sample, and pulverized 80 mesh x 0 sample). As shown in Table 4-12, the filtration of
the spherical coal-pentane agglomerates resulted in moistures varying from 20.1 to 32.0%
moisture. Lower moisture values were obtained for samples with less ultrafines and more coarse
solids. Also, no material was lost to tailings with this dewatering technique since pre-cleaned
agglomerates were utilized. Unfortunately, the moisture values obtained from the filtration tests
were considerably larger than those obtained in the centrifugation experiments. As such, this
method is not as effective for agglomerate recovery and separation as centrifugation. The higher
Table 412. Summary of agglomerate filtration results.
Run Sample Thickness (mm) Dry (g) Filtration Time (sec) Moisture (%)
c1 original 100 mesh x 0 15 15.71 2 32.0
c2 original 100 mesh x 0 12 14.66 2 27.2
c4 original 100 mesh x 0 13 15.88 3 27.3
c5 pulverized 80 mesh x 0 17 20.43 3 20.1
c6 pulverized 80 mesh x 0 20 21.11 3 21.7
c7 deslimed 100 mesh x 0 20 23.32 2 24.4
*Note: All tests were conducted with 500 ml and 60 seconds of drying time.
54 |
Virginia Tech | moistures are attributed to the presence of coal-covered water droplets that were visually
observed within the cake after vacuum filtration. Tests conducted using the centrifuge did not
appear to suffer from the same problem.
4.4.5. Displacement
The displacement method of dewatering consisted of filtering coal slurry with large
amounts of pentane. Unlike the previous techniques, oil agglomeration was not involved in
displacement since pentane was added on top of a homogenous coal slurry phase being filtered.
In this particular set of tests, only the slurry prepared from the dry pulverized 80 mesh x 0 Tom’s
Creek clean coal was used. The results obtained for this method of dewatering are provided in
Table 4-13. Most of the tests conducted using pentane resulted in moistures values in the 22-28%
range, while those conducted without pentane were substantially higher in the 31-34% range.
There was no measureable material loss, so the recovery for this dewatering technique is 100%.
The lowest moisture of 19.7% obtained using pentane was achieved by increasing the vacuum
pressure and drying time to their largest values (i.e., 24 mmHg and 1 minute, respectively);
however, the centrifuge method provided lower moistures in the single-digit range when used to
treat the same sample.
Table 413. Summary of displacement results for pulverized 80 mesh x 0 clean coal.
Pentane Thickness Pressure (mmHg) Filtering time (min) Drying time (%)
(mm) A B C D Water Total (min) Liquid Moisture
yes 6 24 ‐‐ ‐‐ 6 2.75 5.67 0.5 28.43 28.43
yes 6 24 21 8 8 2.58 6.63 0.5 25.75 22.21
yes 7 24 20 10 8 1.25 5.42 1.0 22.59 19.68
yes 8 20 15 8 7 2.08 8.50 0.5 27.66 27.66
yes 6 20 19 8 8 2.25 6.28 1.0 26.93 25.17
yes 7 12 10 8 6 4.33 9.83 0.5 27.87 26.01
yes 6 12 10 10 10 3.50 9.50 1.0 27.26 26.27
no 6 24 19 ‐‐ 14 ‐‐ 3.17 1.0 34.58 34.58
no 5 20 21 ‐‐ 9 ‐‐ 2.67 1.0 31.06 31.06
no 6 12 12 ‐‐ 5 ‐‐ 4.13 1.0 33.62 33.62
*Note: The first result displays moisture if all the water beads are not removed prior to the drying time; it
was redone in the next test. The pressures correspond to different liquid phases: A‐beginning of the test
(water phase), B‐end of the water phase, C‐beginning of the pentane phase, and D‐beginning of the air
phase.
55 |
Virginia Tech | 4.5. Comparison with Conventional Cleaning and Dewatering
The original Tom’s Creek 100 mesh x 0 flotation feed sample was subjected to
conventional cleaning and dewatering tests in order to obtain data that could be used as a
baseline to compare with data obtained from the various pentane dewatering tests. A flotation
test conducted using a Denver Model D-12 laboratory flotation cell was used to evaluate the
cleanability of the sample. Flotation tests were performed using the release analysis procedure to
obtain the best possible separation results for flotation. Dewatering of flotation concentrate was
assessed by performing laboratory vacuum filter tests. The dewatering tests were performed
using two different dewatering aids (reagents RV and RW) in an attempt to achieve the lowest
possible moistures for comparison to the pentane displacement technology.
The results of the laboratory flotation release analysis tests are plotted in Figure 4-13. As
shown, the release analysis curve suggests that ash contents below 5% are attainable in a
“perfect” flotation process. Very efficient real-world processes, such as column flotation, would
be expected to provide ash contents that were slightly above this level. Fortunately, the pentane
agglomeration tests also reduced ash contents to values that matched or exceeded this level of
separation performance (see Table 4-8a). In fact, many of the pentane tests provided concentrates
containing less than 4% ash. This finding indicates that this approach would be a very selective
100
90
80
70
60
50
40
30
20
10
0
0 5 10 15 20 25 30 35
Figure 4-13. Release analysis for 100 mesh x 0 flotation feed.
*Note: Data from three different flotation trials are displayed.
56
)%(
yrevoceR
Cumulative Ash (%) |
Virginia Tech | Table 414. Results for conventional vacuum filtration using flotation concentrate.
Dewatering Time (min) Thickness (mm) Dry (g) Moisture (%)
Aid Conditioning Formation Drying
‐‐ 0 5.8 2 7 18.44 29.2
RW 2 2.5 2 8 18.27 24.8
RV 2 3.7 2 10 20.16 30.5
*Note: Tests were conducted with 250 ml samples using a dewatering aid active‐to‐diesel ratio of 0.5.
The ratio corresponds to 3 lb/ton or 39 µl of prepared solution.
process for separating unwanted mineral matter from valuable coal.
A more significant advantage of the pentane agglomeration process is the ability to
provide cleaned products that are relatively low in moisture. The test results summarized in
Table 4-14 show that conventional vacuum filtration was not capable of providing moisture
contents lower than about 30% by weight. The addition of a dewatering aid (reagent RW) made
it possible to further reduce the moisture down to about 25%, while another dewatering aid
(reagent RV) did not significantly impact the resultant moisture. These values are significantly
worse than the 18-20% moisture achieved using pentane agglomeration combined with
centrifugation (see Table 4-7). The combined ability of the agglomeration process to reduce both
ash and moisture make it a very attractive alternative for coal upgrading.
The vacuum filter technique was also utilized to perform a void space analysis on the
concentrates from hand shaking and centrifuge agglomeration tests. Each sample was cleaned by
agglomeration and then re-homogenized into slurry form. The compact cake produced by
filtration was used to estimate the packing of coal particles within the agglomerates.
Table 4-15 shows the void spacing data obtained for the agglomerated samples produced
by the centrifuge technique. As shown, the original 100 mesh x 0 flotation feed sample gave the
largest void space (59.9% average), while the pulverized clean coal gave the smallest void
spacing (38.0% average). The deslimed 100 mesh x 0 sample gave a void spacing of 42.3%,
which is between these two values. It is interesting to note that the void spacing values fall in the
same sequential order as the moisture values for these particular samples. This correlation
suggests that the observed differences in moisture content for these three samples may be due to
the variations in void spacing, particularly since no such correlation could be established
57 |
Virginia Tech | Table 415. Void space analysis for centrifuge concentrates.
Sample Coal Volume Thickness Cake Volume Void Volume Void Space
(cm3) (mm) (cm3) (cm3) (%)
original 100 mesh x 0 12.3 9 28.1 15.7 56.0
original 100 mesh x 0 11.1 11 34.3 23.2 67.8
original 100 mesh x 0 12.3 9 28.1 15.7 56.1
average 11.9 9.7 30.1 18.2 59.9
pulverized 80 mesh x 0 18.2 8 24.9 6.8 27.2
pulverized 80 mesh x 0 15.9 10 31.2 15.2 48.9
average 17.0 9.0 28.1 11.0 38.0
deslimed 100 mesh x 0 18.0 10 31.2 13.2 42.3
*Note: The volumes were calculated with dry coal weights, a density of 1.31 g/cm3, thicknesses, and a
diameter of 6.3 cm. The void space is a volume percentage.
between moisture and the percentage of ultrafine (minus 325 mesh) solids in each of the three
samples. Additional experimentation is recommended to further investigate this possibility.
For reference, Table 4-16 provides the void space data obtained for the sample of
agglomerates recovered by hand shaking. An average void spacing of about 55% by volume was
obtained for the 100 mesh x 0 sample used in these experiments. This void spacing value is very
similar to that obtained using the centrifugation technique.
Table 416. Void space analysis for hand shaking concentrate from 100 mesh x 0 flotation feed.
Coal Volume Thickness Cake Volume Void Volume Void Space
(cm3) (mm) (cm3) (cm3) (%)
20.6 4 44.5 23.9 53.8
21.7 5 55.6 33.9 61.0
21.8 4 44.5 22.7 51.1
21.3 4.3 48.2 26.9 55.3
*Note: The last line contains the average values for the three tests. The volumes were calculated
with dry coal weights, a density of 1.31 g/cm3, thicknesses, and a diameter of 11.9 cm. The void space
is a volume percentage.
58 |
Virginia Tech | 5. DISCUSSION
Six methods of hydrophobic displacement were evaluated for their cleaning and
dewatering capabilities: hand shaking, screening, air classification, centrifugation, filtration, and
displacement. The first five methods all utilize oil agglomeration to perform displacement of
water during the cleaning stage. The following extraction methods then removed the oil-coated
agglomerates from the remaining bulk water. Only the last method, displacement, performed
hydrophobic displacement during the dewatering stage.
5.1. Moisture Determination
Correct moisture determination is a critical step in this process. Since dewatering is one
of the main goals, inaccurately portraying moisture content defeats the purpose of this project.
The difficulty arises in reporting moisture due only to water, not pentane and water. Fortunately,
the two clear liquids have very different evaporation rates. Though they have other widely
distinguishable properties such as their refractive indices, evaporation rates are easy to record in
real time and do not require a closed system.
This problem was identified early in the project. Originally, moistures were reported after
twenty minutes at 40ºC. However, this procedure had a major flaw: the time for pentane to
evaporate varies widely with dosage and surface area. For example, with low dosages pentane
will evaporate before the concentrate is even placed in the oven. Low dosage tests will therefore
report a lower moisture than realistic because water will be evaporated for twenty minutes.
Conversely, with large dosages twenty minutes is not enough time to remove all of the pentane.
The reported moisture will be much higher than the real water-only moisture. This skews the
moistures in favor of low oil dosages. Only the moisture due to water is important because in a
commercial application, pentane will be evaporated and recovered.
The solution to the problem was to implement a weighing platform. The change-in-
weights were plotted in real-time, and the moisture was chosen where the curve began to level
off as seen in Figure 4-4. It should be noted that this procedure does remove some water.
Unfortunately, though water evaporates more slowly than pentane, there is no way to prevent it
from evaporating while the pentane does; however, in a commercial application it is still
preferable to remove this negligible amount of water in favor of recovering all of the pentane due
to the high cost of the oil.
59 |
Virginia Tech | Even after all pentane and moisture had been removed, the agglomerates maintained their shape
until shaken.
The other main variable tested was the solids content of the feed slurry. Higher solid
contents tended to speed the creation of agglomerates; however, similar size agglomerates
formed for 6%, 15%, and 30% solids. It did not have a major impact on agglomerate appearance.
Finally, it is important that the feed be relatively clean of chemicals, especially cationic
coagulants. Chemicals used to clarify water interfered with agglomeration and prevented
spherical agglomerates from forming; therefore, most of the pond samples experienced
difficulties due to the lingering presence of ions.
5.2.2. Impeller Mixing
Impeller mixing was tested in order to provide a more consistent method of
agglomeration. Unfortunately, it was difficult to create such a small sealed mixing chamber.
Heptane, which evaporates more slowly than pentane, was agitated using a Denver cell to
demonstrate that similar spherical agglomerates formed with impeller mixing; however, once this
was proved, it was decided to delay impeller mixing until a continuous, sealed process was
designed.
5.3. Pellet Extraction and Dewatering
5.3.1. Shaking
Pellet extraction by hand shaking dominated the original dewatering methods tested. The
lowest average moisture, 16.2%, occurred at an M :M of .32. This moisture was obtained with
p c
the original Tom’s Creek 100 mesh x 0 flotation feed with a solids content of 6.3%. Increasing
the solids content to 14.8% raised the moisture by 14%. Oxidation and size distribution had an
even greater impact. After the slurry sat for four months, the moistures increased up to 137%,
depending on the dosage used. Using a screened 325 mesh x 0 sample increased the moisture by
70%. Most of the moisture in these samples was due to free water droplets being shaken out
with the concentrate. The agglomerates also had to bounce up the damp glass of the container
which increased the opportunity for agglomerates to pick up small water droplets. Tiny pin-
pricks of water were occasionally seen on the agglomerates.
61 |
Virginia Tech | The hand shaking method has a high inherent error, hence the reason repetitive testing
was utilized. Since shaking was not consistent, the decision on when to stop was arbitrary. This
led to large variations in combustible recovery as material was left inside the container. The coal
inside the container was considered middlings based on its higher water content even though it
had a similar ash content. Since yield varied greatly depending on the amount of water shaken
out, concentrate weights also varied greatly. Small concentrate weights could produce large
errors as a single droplet would have a large impact on the moisture. Finally, the samples were
shaken into a metal pan to prevent agglomerate loss, and the agglomerates were later transferred
into a small glass dish for platform weighing. Though care was taken to transfer all of the water
to the dish, loss was impossible to prevent. Droplets could easily smear on the metal and
evaporate.
An attempt was made to mechanize this method, but the irregular movement was difficult
to duplicate; eventually the method was abandoned in favor of the higher g-forces produced by a
centrifuge.
5.3.2. Screening
This unique screening method was first used by Kerem Eraydin at Virginia Tech. The
goal of the method is to screen out dry coal while coalescing water droplets on the surface of the
screen. Coatings of coal protected the coalesced water from wetting the screen. This method has
the ability to produce single digit moisture; however, the associated recoveries are also single
digit. The method was not deemed practical due to the low recoveries and process difficulties.
The process suffers from blinding due to fine dry particles, and there is always a risk of wetting
the screen which would further encourage blinding. It would be difficult to prevent screen
wetting in a plant environment.
5.3.3. Air Classification
Air classification was briefly examined as a method of pellet extraction. No quantitative
testing was completed. This method utilized spherical agglomerates, and air was used to remove
the top agglomerates sitting above the water. Unfortunately, the majority of the agglomerates sat
in the water phase, so it was difficult and slow to blow out individual agglomerates. Also, there
62 |
Virginia Tech | was always a risk of blowing out water droplets. This method only removed the agglomerates
from the main water phase. It did not exert any force to remove the tiny water droplets attached
to the agglomerates.
5.3.4. Centrifugation
Dewatering by centrifugation was one of the main methods examined. Centrifugation
provided a means to strip the agglomerates of their tiny water droplets by applying g-forces as
high as 447 g’s. The spherical agglomerates were surprisingly durable under the high pressure
and maintained their shape. This was critical in order to maintain the pores for water to drain. A
dosage higher than a M :M of 0.21 to 0.34 would have resulted in agglomerates too soft; they
p c
would have deformed, effectively trapping the moisture. Centrifugation provided moistures as
low as 7.5% with a combustible recovery of 70.9%, ash rejection of 76.9%, and concentrate ash
of 7.71% (Table 4-11). These results were based on the pulverized 80 mesh x 0 Tom’s Creek
clean coal. The screened 100 mesh x 0 sample provided a higher moisture at 13.2% but a more
favorable recovery of 97.5%, rejection of 83.5%, and ash of 3.94%. These recoveries represent
the entire process, including cleaning. The centrifuge recoveries are about 100% since the
centrifuge tailings did not contain measureable solids.
Oxidation, size distribution, and g-force all played major roles in centrifugation. The dry
pulverized 80 mesh x 0 clean coal resulted in the lowest moistures (Figure 4-10). It was dry
ground, unlike the flotation feed, and had been stored in a freezer to slow oxidation. Though it
contained more fines than the screened 100 mesh x 0 sample, it still had a lower moisture by
5.7% for a 43% difference. Both of the samples had similar estimated surface areas. The original
100 mesh x 0 flotation feed produced the worst moistures, 18.5% being the lowest. This sample
contained 72% minus 325 mesh versus the 34% and 28% in the previous samples. This resulted
in a two orders of magnitude increase in surface area based on an estimation of spherical
particles. Though moisture was not directly related to ultrafine content, it appears to be strongly
related to void space and pentane-to-void volume ratio. The 80 mesh sample had the lowest
moisture and void space and the highest pentane-to-void volume ratio (Figure 4-12). Finally,
Figure 4-9 shows that an increase in speed and g-force tend to produce dryer cake. As speed
increased, the cake became thinner and taller (Figure B-6). At slow speeds, the agglomerates
simply filled the trough around the insert.
63 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.