1
25-095
Unlocking Optimal Flotation Outcomes: Leveraging AI to
Enhance Phosphate Recovery Through Predictive Modeling
Norah AlMuhaisen
Ma’aden, Riyadh, KSA
Aisha AlHarbi
Ma’aden, Riyadh, KSA
Ahmed Alghamdi
Ma’aden, Riyadh, KSA
Owain Morton
Ma’aden, Riyadh, KSA
Salem AlGharbi
Ma’aden, Riyadh, KSA
ABSTRACT
This paper addresses the challenges in optimizing the phos-
phate flotation process, specifically focusing on the variabil-
ity in P2O5 recovery. An artificial intelligence (AI) driven
Flotation Prediction Model was developed to accurately
forecast P2O5 grades. Leveraging real-time data from vari-
ous operational sources, the model provides insights into
how operators can opstimize key flotation parameters to
enhance product quality and recovery rates. The results
demonstrate the effectiveness of AI predictive modeling in
addressing operational variability, marking a crucial step
toward intelligent operations in the phosphate flotation
process.
Keywords: Flotation, AI, Phosphate Recovery, Predictive
Model, Optimization, Parameter Management, Machine
Learning.
INTRODUCTION
Phosphate beneficiation in mining is a complex, multi-stage
process that requires precise control of numerous dynamic
factors, particularly in the flotation circuit. Phosphate ben-
eficiation plants face several key challenges in the flotation
process. There are upstream semi-controllable parameters
that generally relate to Phosphorus pentoxide (P2O5) head
grade variability and particle size distributions, which affect
slurry densities. Additionally, controllable parameters can
become unmanageable due to issues such as sensor avail-
ability and instrumentation calibration, leading to a lack
of real-time measurements for critical variables. While the
AI solution is designed to tackle the primary challenge of
head grade variability, its development is also impacted by
the challenge of ensuring accurate and reliable data, which
depends on resolving instrumentation and measurement
issues.
The goal of this paper is to introduce an artificial intel-
ligence (AI) based Flotation Prediction Model that predicts
the quality of P2O5 by integrating real-time operational
data with periodic laboratory data. This comprehensive
approach allows for accurate and timely predictions of
P2O5 grades, enabling operators to plan effectively and
identify the optimal operational parameters necessary for
stabilizing the P2O5 grade. By leveraging data from both
immediate operational conditions (real-time) and detailed
lab measurements, the model provides a holistic under-
standing of the flotation process, thereby aiding operators
in making informed decisions that enhance the recovery
and quality of phosphate concentrate. This work aligns
with the broader industry movement toward smart mining,
where AI plays a central role in optimizing complex mineral
processing tasks.
LITERATURE REVIEW
The application of artificial intelligence (AI), particularly
machine learning (ML), in mineral processing has revolu-
tionized how flotation processes are understood and opti-
mized. McCoy and Auret provided an exhaustive review of
ML techniques, demonstrating their efficacy in predictive
25-095
Unlocking Optimal Flotation Outcomes: Leveraging AI to
Enhance Phosphate Recovery Through Predictive Modeling
Norah AlMuhaisen
Ma’aden, Riyadh, KSA
Aisha AlHarbi
Ma’aden, Riyadh, KSA
Ahmed Alghamdi
Ma’aden, Riyadh, KSA
Owain Morton
Ma’aden, Riyadh, KSA
Salem AlGharbi
Ma’aden, Riyadh, KSA
ABSTRACT
This paper addresses the challenges in optimizing the phos-
phate flotation process, specifically focusing on the variabil-
ity in P2O5 recovery. An artificial intelligence (AI) driven
Flotation Prediction Model was developed to accurately
forecast P2O5 grades. Leveraging real-time data from vari-
ous operational sources, the model provides insights into
how operators can opstimize key flotation parameters to
enhance product quality and recovery rates. The results
demonstrate the effectiveness of AI predictive modeling in
addressing operational variability, marking a crucial step
toward intelligent operations in the phosphate flotation
process.
Keywords: Flotation, AI, Phosphate Recovery, Predictive
Model, Optimization, Parameter Management, Machine
Learning.
INTRODUCTION
Phosphate beneficiation in mining is a complex, multi-stage
process that requires precise control of numerous dynamic
factors, particularly in the flotation circuit. Phosphate ben-
eficiation plants face several key challenges in the flotation
process. There are upstream semi-controllable parameters
that generally relate to Phosphorus pentoxide (P2O5) head
grade variability and particle size distributions, which affect
slurry densities. Additionally, controllable parameters can
become unmanageable due to issues such as sensor avail-
ability and instrumentation calibration, leading to a lack
of real-time measurements for critical variables. While the
AI solution is designed to tackle the primary challenge of
head grade variability, its development is also impacted by
the challenge of ensuring accurate and reliable data, which
depends on resolving instrumentation and measurement
issues.
The goal of this paper is to introduce an artificial intel-
ligence (AI) based Flotation Prediction Model that predicts
the quality of P2O5 by integrating real-time operational
data with periodic laboratory data. This comprehensive
approach allows for accurate and timely predictions of
P2O5 grades, enabling operators to plan effectively and
identify the optimal operational parameters necessary for
stabilizing the P2O5 grade. By leveraging data from both
immediate operational conditions (real-time) and detailed
lab measurements, the model provides a holistic under-
standing of the flotation process, thereby aiding operators
in making informed decisions that enhance the recovery
and quality of phosphate concentrate. This work aligns
with the broader industry movement toward smart mining,
where AI plays a central role in optimizing complex mineral
processing tasks.
LITERATURE REVIEW
The application of artificial intelligence (AI), particularly
machine learning (ML), in mineral processing has revolu-
tionized how flotation processes are understood and opti-
mized. McCoy and Auret provided an exhaustive review of
ML techniques, demonstrating their efficacy in predictive