Professional Documents
Culture Documents
An Analytical-Numerical Method For Solving A Heap Leaching Problem of One or More Solid Reactants From Porous Pellets
An Analytical-Numerical Method For Solving A Heap Leaching Problem of One or More Solid Reactants From Porous Pellets
com
Abstract
In this paper we present an alternative method based on analytical and numerical solutions for solving the differential equations which describe
heap leaching of one or more solid reactants from porous pellets. We propose to use analytical solutions for the differential equations which describe
rate dissolutions along the pores and the surface of the particles under suitable regularity conditions. Moreover, we propose to use continuous and
discontinuous solutions for the continuity partial differential equations describing balances. We comment on the efficient numerical solving of the
remaining partial differential equations within the proposed numerical scheme. All of this, allows to obtain a numerical algorithm which is fast and
accurate for the heap leaching problem. Also, we include particle size distributions on the proposed numerical methodology. This method applies
to the case where the rate-controlling reagent is a component of the lixiviation solution only and not of the gas phase. The model includes the
effects of particle scales, kinetic factors, heap scales and several operation variables. Finally, numerical experiments are presented.
2008 Elsevier Ltd. All rights reserved.
Keywords: Heap leaching; Differential equations; Discontinuous solutions; Analytical solutions; Numerical methods
1. Introduction
Heap leaching is a hydrometallurgical process which has been
used since a long time which was originally designed for oxides
ores but today have several applications including sulphide ores
and caliche minerals. Heap leaching has been a matter of wide
research and several mathematical models which allows to simulate, analyze, design and optimize these operations. Indeed,
the extractive metallurgy by chemical dissolution is one of the
metal recuperation techniques most used in modern industry.
Nowadays, the main interest to use hydrometallurgical processes
arises from avoiding ambient contamination produced by conventional pyro-metallurgical processes and the treatment of low
grade minerals. Also, heap leaching allows low operational cost
from the energetic and production point of view. The heap leaching can use a pre-treatment of the ore. This operation generally
is based upon the trituration of the ore up to a finer level and
2396
M.E. Mellado, L.A. Cisternas / Computers and Chemical Engineering 32 (2008) 23952402
Nomenclature
bi
CA0
CE0
Cp
Cs
DAe
E
kp
ks
R
t
us
x
z
Z
Greek letters
Croft, McBride, and Gebhardt (2006), a framework was developed for multi-phase flows, gasliquidsolid, in reactive porous
media, capturing several key chemico-physicalbacterial phenomena and their interactions. In Lizama, Harlamovs, McKay,
and Dai (2005), it is shown that for sphalerite and pyrite heap
bio-leaching kinetics were proportional to the irrigation rate
3ks Css R
,
bi DAe
0 bi CA0
,
0 (1 0 )CE0
3(1 h )DAe Z
,
us R 2
p =
0 (1 0 )kp Cpp R2
bi DAe
=
b
and
0 (1 h )
M.E. Mellado, L.A. Cisternas / Computers and Chemical Engineering 32 (2008) 23952402
(1)
p (x, 0) = 1,
where t is the time, x the space coordinate, p the Damkohler II
number for solid reactant within the particle, a dimensionless
stoichiometric ratio which indicates reagent strength relative to
the grade of solid reactant, the fraction of solid reactant residing on the particle surfaces initially and is the dimensionless
reagent concentration.
The ordinary differential equation to compute s , the dimensionless grade of solid reactant on the particle surface, is given
by
ds
s s
=
b
dt
s
particle and the other one describes the interaction of all particles
through the heap. Of course, the particle time scale is faster than
the time scale which involves all particles. Therefore, the time
t in all the particular model equations must be replaced with a
dimensionless bulk flow time to solve the equations with a
common time.
Now, we present the partial differential equation to compute
the dimensionless concentration of dissolved species i , this is
3 i
2 i
i
2 i
+
=
=
+ p p
t
x2
x x
i (x, 0) = 0
and the boundary conditions:
i
(0, t) = 0.
x
We note that
us t
.
=
b Z
In the bulk lixiviant of the heap, the partial differential
equation to compute the dimensionless bulk concentration of
dissolved species ib is given by
ib
ib
i
=
(5)
z
x (x=1)
together with the initial condition:
(2)
ib (z, 0) = 0
s (0) = 1,
ib (0, ) = 0.
where s is the Damkohler II number for solid reactant on the particle surface and b stands for the dimensionless concentration
of the reagent in the bulk solution external to the particle. The
differential Eqs. (1) and (2) are complemented with four partial
differential equations to have a model for heap leaching. In fact,
the continuity equation for , the dimensionless concentration
of reagent A, is given by
2 2
= 2 +
p p
t
x
x x
(6)
z
x (x=1)
together with the initial condition:
b (z, 0) = 0
(3)
(0, t) = 0.
x
It is important to remark that the differential Eqs. (1)(3)
describe the leaching of one particle. It will be clear in the following that our analysis is also applicable to this situation (Dixon
& Hendrix, 1993a).
There exist two time scales involved in this heap leaching
model. One time scale describes the chemical reactions within a
(1, t) = b ,
(4)
i (1, t) = b ,
2397
2398
M.E. Mellado, L.A. Cisternas / Computers and Chemical Engineering 32 (2008) 23952402
s (x, ) = exp
s
3
.
(9)
and v = s1s ,
respectively.
Now we describe our approach for the partial differential Eqs.
(3) and (4). These equations are actually the radial part of the
Helmholtz equation written in spherical coordinates. Although
its is possible to solve (3) and (4), under several assumptions, in
an analytical manner, we note that this is quite expensive from the
theoretical and numerical point of view. The analytical solution
of this equation needs to consider a finite number of terms in an
infinite sum of terms and for each term, it is necessary to solve
a nonlinear equation. This procedure is, at least, so expensive
from the computational point of view than using, for instance,
finite differences. Moreover, it is easier to deal with a numerical
method than sum truncation in what concerns error studies of
the proposed methodology.
There exist a number of ways to solve numerically these equations. We note that the number 2 in Eqs. (3) and (4) corresponds
to spherical diffusion and corresponds to the radial part of the
Helmholtz equation in spherical coordinates (see Bouffard &
Dixon, 2001, Eq. (5)).
We can mention the finite difference method (Hackbusch,
1992), the finite element method (Brenner & Scott, 1994; Ciarlet,
1978; Hackbusch, 1992; Zienkiewicz & Taylor, 2000), finite
volume methods, among others. We recall that we use uniform
discretizations for solving Eqs. (3) and (4). In this sense, and
by considering standard versions, the use of the finite element
method or finite differences is equivalent. Also, as the finite
differences are easier to handle as compared with finite elements, we solve these equations with a CrankNicholson fully
implicit finite difference scheme (Hackbusch, 1992). Both equations, after the discretization procedure, require the solving of
a linear tridiagonal system of equations whose entries depends
entirely of the discretization in time and space coordinates. It
is very important to note that the generated matrices are sparse,
non symmetric and stable. Therefore, we have to deal with a
linear system of equations that must be solved in each step of
the iterations either in space or time. The linear systems can be
solved with a number of methods (Stoer & Bulirsh, 1996). We
use here a preconditioned LU solver and we recall other solvers
can be used, specially taking into account the sparse structure of
the matrices. For programming purposes, we note that it is only
necessary to assemble the matrices just once, but update their
respective right-hand sides each iteration. Therefore, there is no
computational effort in assembling the stiffness matrices.
Now we present our approach for the partial differential Eqs.
(5) and (6). First, we note that both equations have exactly the
same structure and they differs just on the initial and boundary conditions. Again, we need to mention that these equations
can be solved numerically by using, for example, the tracking of
fluid elements, finite volume methods or other similar numerical
methods. In despite of that, we solve these equations in an analytical manner. We recall that in the bulk lixiviant of the heap, the
mass balance for dissolved species is given by Eq. (5). To obtain
analytical solutions, we consider the transformation (Kesavan,
1989):
=z
and
= z + .
2 x (x=1)
Now, by using the initial and boundary value conditions, we
have
ib (z, ) = ,
<z
(10)
> z,
(11)
and
ib (z, ) = z,
where
i
=
.
2 x (x=1)
M.E. Mellado, L.A. Cisternas / Computers and Chemical Engineering 32 (2008) 23952402
<z
where
=
2 x
> z,
(12)
Now, we present the algorithm by considering heap leaching with homogeneous particle sizes. The algorithm reads as
follows, for a Z space step, solve:
(13)
and
b (z, ) = 1 z,
2399
.
(x=1)
2 2
= 2 +
p p , (x, 0) = 0,
t
x
x x
(1, t) = b ,
(0, t) = 0.
x
With obtain p through
p (x, ) = exp
p
3(1 )
.
s
3
.
Obtain i by solving
2 i
3 i
2 i
+
=
+ kp p ,
x2
x x
i
i (1, t) = b ,
(0, t) = 0.
x
i (x, 0) = 0,
<z
and
i
ib (x, t) =
z,
2 x (x=1)
> z.
Also
b (z, ) =
,
2 x (x=1)
<z
and
b (x, t) = 1
z,
2 x (x=1)
> z.
ib (1, t) dt.
2400
M.E. Mellado, L.A. Cisternas / Computers and Chemical Engineering 32 (2008) 23952402
R
,
b (z, ) = 1
= 2 ,
w = 2 w
2 2
= 2 +
p p , (x, 0) = 0,
t
x
x x
(1, t) = b ,
(0, t) = 0.
x
p
3(1 )
s
3
1
m m3 d z,
.
.
Obtain i by solving
3 i
2 i
2 i
+
=
+ p p ,
2
x
x x
i
i (x, 0) = 0, i (1, t) = b ,
(0, t) = 0.
x
ib (1, t) dt.
1
0
m m3 d ,
<z
m m3 d z,
> z.
(x=1)
and
ib (z, ) =
1
0
(x=1)
Also
b (z, ) =
0
1
(x=1)
m m3 d ,
> z.
(x=1)
4. Numerical experiments
and
<z
Fig. 1. Behaviour of i against days.
M.E. Mellado, L.A. Cisternas / Computers and Chemical Engineering 32 (2008) 23952402
2401
0 := 0.03, p := 0.000157, grade of solid reactant within particle Cp := 0.000157, particle radius r := 2.5, stoichiometric
number bi := 1.0, effective pore diffusivity of reagent DAe :=
106 , reference reagent concentration CA0 := 0.00333, initial
extractable grade of solid reactant CE0 := 0.000157, bulk solution volume fraction b := 0.03, heap void fraction h := 0.03,
superficial bulk flow velocity us := 0.000333, heap depth Z :=
1000.0.
Figs. 3 and 4 show a qualitative agreement with the standard
copper heap leaching process which validates, in some way, our
approach and allows to conclude this simple modification of
the numerical algorithm for solving (1)(6) can produce good
results at a low computational effort.
5. Conclusions
Fig. 2. Behaviour of E against days.
2402
M.E. Mellado, L.A. Cisternas / Computers and Chemical Engineering 32 (2008) 23952402