Portnov

Control and optimisation of
hydrocyclone classification performance

with predictive models
Victor Portnov, Chief Product Officer & Co-Founder, Conundrum Industrial Limited
An essential part of mineral processing is the physical separation of coarse and fine fractions using
hydrocyclones. As it is well known, in hydrocyclones, creating a vortex cyclone facilitates the
separation between large and smaller particles, with the finer particles and most of the water exiting
through the overflow to the next flotation stage. This separation process is crucial for ensuring the
efficiency of subsequent flotation operations, as it directly impacts the particle size distribution and
the overall recovery rates of valuable metals and minerals.
While working with a hydrocyclone circuit, consider the main challenges that usually impact the
quality of any control you want to apply.
Key Challenges in Circuit Process Control:

● High Variability of Slurry Flow from SAG Mill—The SAG mill product can vary
significantly in throughput and particle size distribution, making it challenging to maintain
stable conditions.
● Small Sump Capacity - Often, the sump is relatively small, necessitating rapid response to
prevent overloading and potential overflow, which can disrupt the process.
● Insufficient Technical Water Supply—Occasionally, there is not enough technical
water available at the plant, limiting the ability to manage slurry density effectively.
● Degradation of Hydrocyclone Nozzles—Over time, the nozzles in hydrocyclones
degrade, impacting their performance and efficiency.
● Impact of Opening and Closing Hydrocyclones—Opening and closing individual
hydrocyclones causes significant disturbances in the process, leading to instability.
These challenges can lead to several issues and also lead to product losses:
● High Variability in P80 and Density of Hydrocyclone Overflow - The particle size (P80)
and density of the hydrocyclone overflow can vary widely, affecting downstream processes.
● Instability in the Flotation Circuit - Waves and instability in the flotation circuit can occur,
disrupting the separation process.
● Losses in Metal Recovery - The overall extraction of metals can suffer, leading to reduced
recovery rates and potential economic losses.
Typical control strategy for hydrocyclones:

In nearly all observed cases, the control of hydrocyclones is configured such that two separate PID
controllers operate with two independent objectives. One controller maintains constant pressure in
the hydrocyclones using the pump speed, while the other controller maintains a specific water level
in the sump through water flow regulation. This dual-controller system is designed to ensure stable
operation but often falls short when dealing with dynamic changes in the process conditions.
From our perspective, this approach is not optimal because fluctuations and changes in slurry flow
properties, such as variations in grain size or increased feed from the mill and increased
recirculation in the hydrocyclone circuit, lead to severe fluctuations in the feed density to the
hydrocyclones. This, in turn, causes significant changes in the density of the hydrocyclone overflow,
affecting the stability and efficiency of flotation operations. These fluctuations can also result in
suboptimal performance of the grinding circuit, leading to higher energy consumption and reduced
throughput.
Circuit overview:
Let’s first look at the operating circuit, which is typical for most existing concentrator plants:
The manipulated parameters (MVs) were few: adding water to the sump, hydrocyclone feed pump
speed, and the number of open hydrocyclones in the hydrocyclone battery. This model could predict
the optimal operating conditions based on real-time process data, enabling proactive adjustments to
maintain optimal performance.
The controlled parameters (CVs) were the sump level, slurry pressure and flow to the
hydrocyclones, density and flow in the hydrocyclone overflow, the percentage of the desired particle
size, and the power of the ball mills. By monitoring these parameters continuously, we could identify
trends and deviations that indicated potential issues, allowing for timely interventions.
Additional parameters (DVs) considered in the model included the undersized product
from the SAG mill, calculated recirculation from the hydrocyclones and ball mill, and the content of
valuable metals, namely copper and nickel, in the flotation feed. Incorporating these parameters into
the model provided a more comprehensive understanding of the process dynamics and improved
the accuracy of the control strategy.
Description of our predictive process control solution:

What is better to stabilise for flotation: flow rate, magnitude of flow, or stability and magnitude of
density? In any case, it seems that the sump level itself does not affect the qualitative characteristics
of the circuit’s operation as long as there is no overflow or critically low sump level. In this article, we
will demonstrate how to generally improve circuit performance with automatic predictive solutions for
density control. By shifting the focus from maintaining a fixed sump level to dynamically adjusting the
feed density, we aim to enhance the overall process efficiency and recovery rates.
The key idea, implemented and tested at two processing plants, is that while we still maintain
consistent hydrocyclone pressure using pumps, instead of fixing a single sump level, we define a
range within which this level should be maintained. We then adjust the setpoint for a specific
required level depending on the desired feed density for the hydrocyclones. This adaptive control
strategy allows for more flexible and responsive operation, accommodating variations in ore
properties and process conditions.
Regarding pressure, we decided to work with the most stable pressure possible because significant
decreases lead to instability in hydrocyclone operation, and significant increases in pressure
increase the flow to flotation, reducing flotation time and thereby decreasing recovery. Maintaining
stable pressure ensures that the separation efficiency of the hydrocyclones remains high, minimising
the risk of coarse particles being misplaced in the overflow.
Based on the data collected from sensors, we built a model of the reaction of controlled parameters
depending on the manipulated ones. This adaptive machine learning model also considered
changes in hydrocyclone efficiency, which, as it turned out, were mainly related to wear on their sand
nozzles.
Here is an example of a model matrix (step responses based on the linearisation of our predictive
model) for use in MPC (model predictive control) at some period of time. The model is not
parameterised and learns automatically using machine learning algorithms.
(valve_201 - Dart valve of head flotation machine)
To sum it up, in most cases, applied control might be represented by the following scheme:
Thus, the larger the sump, the more manoeuver space we have and the more stable the density we
can maintain. Of course, in situations where the sump is small, and density is already high, it is
difficult to significantly reduce it because the required sump level, i.e., its actual volume for slurry
dilution, can exceed our acceptable range, hitting the upper limit. This limitation underscores the
importance of designing sumps with adequate capacity to handle fluctuations in slurry density.
During the experiment, we also tested the possibility of not only stabilising density but also
maintaining the required density depending on external factors. For example, successful
experiments were conducted on density control based on the metal content in the flotation feed
(obtained using flow analysers). Thus, when metal content changed, the recommended density was
calculated, and we adjusted the sump level to proactively achieve the most optimal flotation feed
density (instead of just the most stable). This approach allows for real-time optimisation of the
flotation feed, enhancing the recovery of valuable minerals.
During actual implementation, we faced a few additional limitations, such as:
1. The discrete change in the number of open hydrocyclones – is a rather complex control
element because opening or closing a hydrocyclone creates a shock load and shock effect
on the flotation feed and the milling circuit as a whole. This transient response must be
carefully managed to avoid destabilising the entire process.
2. There were limitations on recirculating technical water, as sometimes the maximum setpoint
for water was reduced due to its shortage within the circuit. This constraint necessitated the
development of water-saving strategies to ensure continuous operation without
compromising the quality of the separation process.
Here is a screenshot of a dashboard used during the test. You may see how the system tried to
control density, but when the amount of technical water limited it, it closed one of the hydrocyclones:
Circuit control was automatic, but operators could adjust requirements for controlled parameters and
partially restrict the optimisation algorithm if they deemed it necessary for the situation. This
human-in-the-loop approach ensured that the expertise and intuition of experienced operators
complemented the automated control system, leading to better overall performance.
The main experiment lasted two months, with control-experiment switches based on plant shifts.
Given the quick response and low inertia of this circuit, one conclusion is that the duration of the
experiment could have been significantly reduced by changing control-experiment pairs every four
hours instead of every 12 hours. Shorter intervals would have allowed for more data points and
quicker identification of optimal control settings.
Adjustments were also made for preventive maintenance, significant changes in ore properties, or
unscheduled downtimes. These adjustments ensured that the experimental conditions remained
consistent and that external factors did not skew the results.
In this article, we wouldn’t conclude whether stabilising density or flow rate is better for the entire
beneficiation process or whether various density recommendations based on metal content in the
ore are better. Instead, we are trying to share the insights and thoughts we gained from our
experiments. These findings provide a foundation for further research and development in advanced
process control strategies for mineral processing.
Results:
Summarising the results, during the controlled experiment aimed at stabilising the hydrocyclone feed
density, we stabilised the density by 23%, reducing the standard deviation by 23%. We also
increased the share of ready class (-71mkm) by 0.58%. This increased the percentage of the
desired particle size in the hydrocyclone overflow, and stabilisation of the process increased rough
flotation recovery by 0.74% and total flotation by 0.5%. These improvements demonstrate the
effectiveness of the advanced control strategy in enhancing process stability and efficiency.
Example 1 - hydrocyclones overflow density stabilisation:
Example 2 - results of the test on extraction in rough flotation, total flotation and % of the main class
-71mkm:
This approach appears progressive and more optimal than the classical approach of simply fixing a
specific sump level. It is important to note that if the sump volume were just 30% larger, we could
maintain the required density in twice as many cases. This means twice as many cases where the
required density should have been lower, but we could not achieve it due to hitting the upper sump
level limit. This finding highlights the importance of considering sump design and capacity in the
overall optimisation strategy.

Portnov

Uploaded by

Copyright:

Available Formats

Portnov

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Portnov

Uploaded by

Copyright:

Available Formats

Control and optimisation of

hydrocyclone classification performance

Key Challenges in Circuit Process Control:

Typical control strategy for hydrocyclones:

Description of our predictive process control solution:

During actual implementation, we faced a few additional limitations, such as:

Example 1 - hydrocyclones overflow density stabilisation:

You might also like