Advanced Rougher Flotation Control
Advanced Rougher Flotation Control
Control
ABSTRACT
Many consider rougher flotation control to be more of an art than a science. Some have
even thought it impossible to make process control decisions and changes without human
interaction. Controlling flotation bank pulp levels to maintain consistent pulling-rates in
the face of changing feed slurry flow rates, varying mineral types, and inconsistent
reagent additions can be daunting. However, an advanced supervisory control strategy
was installed on the plant-floor to control an industrial rougher flotation circuit with little
or no operator interaction. Plant staff monitored the flotation circuit under supervisory
control and reported benefits of up to 4% improvement in copper recovery above that
achieved with an older supervisory control strategy. To accomplish this work, a novel use
of instrumentation to solve a flotation bank pulp level control problem was employed, a
new implementation approach was used to correct some flaws in the older approach, and
careful attention was paid to a few key process control principles. This paper discusses
important aspects of these activities.
INTRODUCTION
Think of how easy process control problems seem when they are described, but in
practice are really quite difficult to achieve. To illustrate this point, consider the control
problem that is depicted in Figure 1 and described below.
This problem is referred to as the Autonomous Flying Machine problem or the flying
machine problem for short. Figure 1 represents a large grass field about the size of a
regulation football or soccer field. The objective is to build a flying machine that
completes the mission without any human interaction. The mission is:
*
KnowledgeScape Division, Baker Process, Salt Lake City, Utah, USA
+
Minerals Division, Baker Process, Salt Lake City, Utah, USA
1. The flying machine must start from the starting box, represented in the lower
left-hand corner of Figure 1, and once the machine leaves the start box it must
fly completely autonomously.
2. The flying machine has to search for and find the closest target. Filled black
circles represent targets in Figure 1. The line labeled, “Step 1” represents one
possible flight path from the start box to the closest target.
3. The flying machine must pick-up one of the disks on the target. The white
circles in the target closest to the starting box represent disks.
4. The flying machine must search and find the furthest target. The line labeled,
“Step 2” represents one possible fight path from the closest black circle to the
furthest black circle. While searching for the furthest target, the flying
machine must clear a three-foot tall fence that separates the two black targets.
The fence is represented in Figure 1 by the thick line between the two targets.
5. The flying machine must place the disk it is carrying on the furthest target.
Step 2
Step 1
Start Here
From the description, this problem sounds relatively easy. One might think that in today’s
world of advanced science, computer technology, and global positioning instrumentation
that this control mission could easily be accomplished. After all, man first walked on the
moon 30 years ago. However, to date, in a yearly competition including competitors from
around the globe, the mission has not been successfully completed outdoors. A few have
almost completed the mission and, in the near future, will likely be successful. But once it
is successfully completed, the complexity of the mission will be extended to transport
every disk from the first target to the second target. And after that happens, the
complexity will be increased to accomplish more complex tasks.
The flying machine problem is an excellent metaphor for the control system development
discussed in this paper, specifically an operator-less, plant floor, rougher flotation control
system. The objective of the flying machine problem is to successfully complete the
mission. The objective of the rougher flotation control system is to achieve high
concentrate recoveries while maintaining adequate rougher flotation concentrate grades.
Achieving the flying machine goals is difficult because environment variables such as
wind velocity and direction are constantly changing. Achieving the rougher flotation
control goals is also influenced by process disturbances in flotation feed grade, feed
mineralogical composition and feed slurry flow rates. Once the flying machine takes off,
it must fly autonomously. When the rougher flotation control strategy is enabled, it
should control the flotation circuit without human intervention. A comprehensive suite of
instrumentation must be employed by the flying machine to complete its mission. The
same is true for rougher flotation control. After the flying machine mission is
successfully completed, the complexity of the mission will increase. The same will be
true for rougher flotation control.
Just as the flying machine problem appears to be simple, the task of implementing an
advanced supervisory control strategy to autonomously control an industrial rougher
flotation circuit sounds simple. After all, man first walked on the moon 30 years ago.
However, it is possible that the recent work, culminating in this paper, to implement an
advanced supervisory control strategy for autonomous rougher flotation control is the
first successful accomplishment of the mission. To accomplish this work, a novel use of
instrumentation to solve a flotation bank pulp level control problem was employed, a new
implementation approach was used to correct some flaws in the older approach, and
careful attention was paid to a few key process control principles. This paper discusses
important aspects of the activities accomplished while working in conjunction with the
operations and control staff of a large United States copper producing concentrator.
A NOVEL INSTRUMENTATION APPROACH
Traditional pulp level control in rougher flotation banks is the stabilizing-control-
workhorse of industrial process control for rougher flotation. A variety of level sensing
and control approaches have been investigated over the years in many plants throughout
the world. However, flotation bank pulp level control does not provide the granularity of
control necessary to optimize the performance of a rougher flotation circuit using an
advanced supervisory control system. To illustrate this point, consider the typical
flotation bank pulp level control depicted in Figure 2. Figure 2 provides a cutout of a
Process and Instrumentation Diagram (P&ID) for a flotation bank of a rougher flotation
circuit.
LIC
I/P
LIT LY
LE LV
Two flotation cells and a discharge box constitute the flotation bank shown in Figure 2.
Rougher flotation circuits are often configured with multiple stages of flotation banks
containing various flotation cells to provide the required retention time for flotation. Each
bank will have instrumentation and controls similar to that shown in Figure 2. The item
labeled “LE” represents the level element or sensor that measures the pulp level. The
item labeled “LIT” is a field-mounted level-indicating transmitter responsible for sending
the data from the level sensor to the process control system. The item labeled “LIC”
represents the level controller configured in the process control system. The input to the
level controller is the measurement of the level from the level sensor. The item labeled
“LY” represents the device that converts the level controller output signal to a pneumatic
signal that controls the position of the discharge box dart valves. The item labeled “LV”
identifies the discharge box dart valves.
The most common type of instruments used today for sensing the pulp level in a flotation
bank is an ultrasonic level sensor. Ultrasonic level sensors are quite reliable and
economical for this service. However, simple level sensing and level control does not
always guarantee that the flotation bank contributes to the overall recovery achieved by
the rougher flotation circuit. Consider the situation where the flotation bank pulp level is
6 inches below the overflow lip of the flotation bank while the froth is 5 inches deep. In
this case, the top of the froth will be one inch below the lip of the flotation bank and no
froth will be extracted from the flotation bank. When this occurs in a flotation circuit, the
flotation bank with this situation contributes to the over-all retention time, acting as a
mixing tank, but does not contribute as an extraction unit, and thereby reduces overall
flotation circuit performance.
The addition of another inexpensive level sensor can provide information for more
intelligent level control and is the foundation for adaptive, optimizing supervisory
control. Compare the following Figure 3 with Figure 2.
The additional level sensor measures the height of the froth above the overflow lip of the
flotation bank. This measurement provides direct feedback to the control system about
the extraction rate and the performance of the flotation bank. Various control loop
configurations, although their description is not the subject of this paper, can be
configured and used to intelligently control the pulp level of every flotation bank in the
flotation circuit. With the froth height level sensor, an advanced, expert control strategy
can determine which flotation banks are not extracting concentrate and can take
corrective actions on a timely basis.
LI LIC
I/P
LIT LIT LY
LE LE LV
FIGURE 3 Flotation bank froth height instrumentation and pulp level instrumentation and controls
Recall the statement made earlier that flotation bank pulp level control does not provide
the granularity of control necessary to optimize the performance of a rougher flotation
circuit using an advanced supervisory control system. With froth height measurements on
each flotation bank, the advanced supervisory control system has new flotation bank
performance data. With this data for example, the control strategy can increase the pulp
level of one particular flotation bank, while decreasing the pulp level in every other
flotation bank in the flotation circuit. The result is a more finely tuned control strategy
that can optimize the flotation circuit performance and respond to a wider range of
process disturbances. To illustrate this point, consider Figure 4 that depicts the water
additions to a Semi-Autogenous Grinding (SAG) mill. In Figure 4, only one automatic
control valve is installed in the water header piping to control the feed-end water addition
and discharge-end jet water addition.
FIC
FIT FY
FV
FE
FIGURE 4 Total water flow instrumentation and controls for a semi-autogenous grinding mill
Performance of the SAG mill is quite sensitive to the feed-end water flow rate and the
discharge-end jet water flow rate. Using only one flow meter and control valve results in
automatic control of the total flow rate of water, but does not provide fine-tuned control
of the feed-end or discharge-end water additions. Compare the Figure 4 with Figure 5.
FIC FIC
FIT FY FY FIT
FV FV
FE FE
FIGURE 5 Individual water flow instrumentation and controls for a semi-autogenous grinding mill
Figure 5 shows instrumentation and controls for each of the water flows to the SAG mill.
It takes little more than intuition to realize that this second process control solution will
result in better operation and performance of the SAG mill when managed by a
knowledgeable operator or a good supervisory control strategy.
The analogy of the SAG mill water control illustrates the same instrumentation and
control concepts for flotation bank pulp level control in rougher flotation circuits and
necessitates the addition of an additional ultrasonic level sensor to measure froth height
in each flotation bank. The additional information gleaned from the froth height data can
be used by a knowledgeable operator or advanced supervisory control strategy to improve
operation and performance of each flotation bank in the rougher flotation circuit.
In order to prepare for the implementation of a new advanced supervisory control system
in the copper producing concentrator, each of the rougher flotation banks in the rougher
flotation circuit was instrumented with froth height sensors. The sensors were connected
to the Distributed Control System (DCS) and were available to the advanced supervisory
control system through a communications channel with the DCS.
A NEW APPROACH FOR IMPLEMENTING THE ADVANCED SUPERVISORY
CONTROL SYSTEM
Advanced supervisory control systems for rougher flotation circuits, and for flotation
circuits in general, have achieved well-documented benefits that typically include higher-
grade concentrate, higher recoveries and reagent savings. The fundamental approach
taken by implementers of many supervisory control strategies for flotation is to rely on
plant-floor operators to walk through the flotation circuit at regular intervals and visually
analyze its performance and select limits that define control ranges for the control
variables in the flotation circuit. In this way, plant-floor operators provide integral input
and define the behavior of the supervisory control strategy depending on the limits that
they enter. This can be illustrated by examining Figure 6. It shows a portion of the user
interface of one supervisory control system programmed into a DCS for rougher flotation
circuit control using the G2 expert system shell.
FIGURE 6 Partial view of the user interface for a rougher flotation supervisory control strategy
The display contains three tables, including; 1) the flotation bank pulp level limits table,
2) the collector addition limits table, and 3) the frother addition limits table. The table
entries containing numeric data are input fields that can be changed by plant operators.
For example, a plant-floor operator may observe that the pulp level of flotation bank 1 is
running somewhat low and decides that the strategy should “pull” the flotation bank
harder. Therefore, he changes the high limit from 3.5 inches to 2.0 inches. Then, to keep
the strategy controlling the flotation bank pulp level setpoint within a 3.5-inch range, he
decides to change the low limit from 7.0 inches to 5.5 inches.
Before froth height measurement, this implementation approach was almost the only
viable approach available because the control system did not have reliable information
that could identify a single misbehaving flotation bank in the rougher circuit.
Additionally, the benefits achieved by and documented for these strategies justify their
installation and use. However, the key weaknesses of this implementation approach are:
All of these situations result in lost opportunity and benefits that might otherwise be
realized.
1. Remove the need for constant supervision of the supervisory control strategy
by plant-floor operators.
2. Remove as many control strategy limits that do not represent process or
equipment constraints as possible.
3. Remove any operator-entered targets and objectives that could easily be
calculated by the supervisory control strategy itself.
100
75
Copper Recovery, %
50
25
0
2.5 7.5 12.5 17.5 22.5
Concentrate Copper Grade, %Cu
FIGURE 7 Rougher flotation response curve of three different ore-types and particle size
distributions
Each trace in Figure 7 represents the highest copper recovery that can be achieved at the
stated concentrate copper grade for a given ore-type and particle size distribution. For
example, at a concentrate grade of 12.5 %Cu, the highest copper recoveries that can be
achieved are 65%, 75% and 85% for the three ore-types depicted in Figure 7. Rarely do
plants respond with the definitive consistency that is implied in Figure 7. These types of
relationships are the result of gathering and analyzing a lot of data that often has a great
deal of variability and scatter. It is this variability that requires advanced process control
to optimize the performance of the rougher flotation circuit. The point being illustrated by
the data in Figure 7 is that for a given ore-type, particle size distribution, and operating
conditions, the lowest acceptable concentrate grade will generally produce the highest
recovery and thereby meet the objective.
These are just a few of the important control principles that were identified and employed
during design and implementation of the advanced supervisory control strategy installed
in the rougher flotation circuit of the target copper concentrator. Frequently, the task of
implementing an advanced supervisory control system is greatly simplified and the
control strategy is more robust when it is tailored to the governing control principles such
as these.
The software used to implement the supervisory control strategy is the KnowledgeScape
software. KnowledgeScape is an adaptive, optimizing, process control software that
provides fuzzy logic expert system facilities in conjunction with self-learning, neural
network models and genetic algorithm optimizers. This unique set of Artificial
Intelligence tools allows KnowledgeScape to control, model and optimize a wide variety
of processes, including the rougher flotation circuit of the target copper concentrator. The
KnowledgeScape implementation replaced an older implementation using the G2
software.
Although the specifics of the data are not generally published, the benefit measured and
reported by plant staff averaged between 3% and 4% improvement in rougher flotation
circuit recovery. One means used to assess these benefits was to allow plant-floor
flotation operators to take control of the flotation circuit from the supervisory control
strategy and operate the plant as was usual before the advanced supervisory control
system was commissioned. During periods of operator control, plant staff consistently
measured a decrease in the rougher flotation circuit recovery. When the supervisory
control strategy was enabled again, an increase in the flotation circuit recovery was
observed and measured.
CONCLUSIONS
Many consider rougher flotation control to be more of an art than a science. However,
novel instrumentation, new implementation approaches and sound process control
principles make it possible to implement a supervisory control strategy that makes
process control decisions and changes without human interaction. A supervisory control
strategy was installed on the plant-floor to control an industrial rougher flotation circuit
that requires no operator interaction. Plant staff monitored the flotation circuit under
supervisory control and reported benefits of up to 4% improvement in copper recovery
above that achieved with another supervisory control system. It is possible that the work
reported in this paper, to implement an advanced supervisory control strategy for
autonomous rougher flotation control, is the first successful accomplishment of this
mission.