2030 XXXI International Mineral Processing Congress 2024 Proceedings/Washington, DC/Sep 29–Oct 3
to gravity concentrators, the model can be rearranged to
an analytical formulae which generates a general partition
surface, i.e., a 3-dimensional determination of the partition
number as a function of size and density (Rao, 2004). Any
point on the surface represents a partition number, which
corresponds to the fraction of feed of given size and den-
sity reporting to the underflow. Equations (1) and (2) relate
the partition number as a function of particle size, density
and efficiencies of separation. Equation (1) represents the
partition surface based on size and density attributes, while
Equation (2) represents Ecart probable Ep, which reflects
the separatioan inefficiency at the given particle size.
exp Y Y 100 1 1
1
p p
pd q
t
t =--e e-clnc mmc m oo (1)
ln^1
E
Y 2
.2877
p
p
pt 1.3863 0
q
d q q t
=
-
-td
`pd h-1j
(2)
where:
Y =Corrected Partition number
Yp =Pivot partition number, representing fraction
of by-pass in gravity concentrators
Ep =Ecart probable, the separation inefficiency at
the given particle size
ρp =Separation density (kg/m3)
d =Particle size (mm)
ρ =Flow parameter representing effect of viscosity
q =Flow parameter representing prevailing condi-
tions in the separator
Washability data was used to construct partition curves
which predicted DMS performance. Using this data as the
input parameters, the predicted yield, and gangue rejection
at several density cut-points were calculated. The effect of
the density cut point on yield and P2O5 grade was deter-
mined, along with the predicted Efficiency of separation
(Ep) for each cut point. The effect of particle size on the
performance of a DMS was also considered.
Pilot Dense Medium Separation (DMS) Testwork
The medium used for the pilot testing was a mixture of FeSi
(270D) and water. The work was conducted according to
the methodology described in detail in Singh et al. (2022).
To ensure steady state conditions, the circulating feed,
and the cyclone overflow and underflow were monitored
throughout the test. The feed pressure was also monitored
to ensure the cyclone feed was constant.
In this paper, the performance of the initial stage sepa-
ration is highlighted, which involved targeting a low den-
sity cut point to produce an enriched phosphate product
(float product at an SG cut point of 2.5). Figure 2 is a rep-
resentation of a simplified version of the DMS circuit for
a single stage DMS operation. Representative sub-samples
were removed from the bulk sinks and final float product,
prepared and submitted for chemical analysis.
Computational Fluid Dynamics (CFD) Modelling
The DMC, as the main driver of the separation process was
modelled using computational fluid dynamics, an tech-
nique commonly utilized to simulate hydrocyclones. A
multiphase CFD approach was used for the simulations in
this paper using the commercial package Fluent (ANSYS,
2016). This package permits for the implementation of
user-defined functions with submodels to simulate air/free
surface, mixture models to simulate distributions of par-
ticles and their associated drag effects, and Reynolds stress
model/large eddy simulation (RSM/LES) models to model
Figure 1. Outline of HLS procedure
to gravity concentrators, the model can be rearranged to
an analytical formulae which generates a general partition
surface, i.e., a 3-dimensional determination of the partition
number as a function of size and density (Rao, 2004). Any
point on the surface represents a partition number, which
corresponds to the fraction of feed of given size and den-
sity reporting to the underflow. Equations (1) and (2) relate
the partition number as a function of particle size, density
and efficiencies of separation. Equation (1) represents the
partition surface based on size and density attributes, while
Equation (2) represents Ecart probable Ep, which reflects
the separatioan inefficiency at the given particle size.
exp Y Y 100 1 1
1
p p
pd q
t
t =--e e-clnc mmc m oo (1)
ln^1
E
Y 2
.2877
p
p
pt 1.3863 0
q
d q q t
=
-
-td
`pd h-1j
(2)
where:
Y =Corrected Partition number
Yp =Pivot partition number, representing fraction
of by-pass in gravity concentrators
Ep =Ecart probable, the separation inefficiency at
the given particle size
ρp =Separation density (kg/m3)
d =Particle size (mm)
ρ =Flow parameter representing effect of viscosity
q =Flow parameter representing prevailing condi-
tions in the separator
Washability data was used to construct partition curves
which predicted DMS performance. Using this data as the
input parameters, the predicted yield, and gangue rejection
at several density cut-points were calculated. The effect of
the density cut point on yield and P2O5 grade was deter-
mined, along with the predicted Efficiency of separation
(Ep) for each cut point. The effect of particle size on the
performance of a DMS was also considered.
Pilot Dense Medium Separation (DMS) Testwork
The medium used for the pilot testing was a mixture of FeSi
(270D) and water. The work was conducted according to
the methodology described in detail in Singh et al. (2022).
To ensure steady state conditions, the circulating feed,
and the cyclone overflow and underflow were monitored
throughout the test. The feed pressure was also monitored
to ensure the cyclone feed was constant.
In this paper, the performance of the initial stage sepa-
ration is highlighted, which involved targeting a low den-
sity cut point to produce an enriched phosphate product
(float product at an SG cut point of 2.5). Figure 2 is a rep-
resentation of a simplified version of the DMS circuit for
a single stage DMS operation. Representative sub-samples
were removed from the bulk sinks and final float product,
prepared and submitted for chemical analysis.
Computational Fluid Dynamics (CFD) Modelling
The DMC, as the main driver of the separation process was
modelled using computational fluid dynamics, an tech-
nique commonly utilized to simulate hydrocyclones. A
multiphase CFD approach was used for the simulations in
this paper using the commercial package Fluent (ANSYS,
2016). This package permits for the implementation of
user-defined functions with submodels to simulate air/free
surface, mixture models to simulate distributions of par-
ticles and their associated drag effects, and Reynolds stress
model/large eddy simulation (RSM/LES) models to model
Figure 1. Outline of HLS procedure