PLANT-WIDE ADVANCED PROCESS CONTROLLER AND

OPTIMIZER FOR HINDUSTAN ZINC LTD.

Ravi Gopinath
Head – Manufacturing Practice, Tata Consultancy Services, Mumbai, India

ABSTRACT

Tata Consultancy Services has deployed a state-of-the-art online plant-wide advanced controller at the 4500 tpd
Rampura-Agucha ore beneficiation plant of Hindustan Zinc. The plant is completely automated with a
centralized DCS.

Beneficiation is a two step process - pulverization and flotation. Large and frequent variations in the ore
characteristics adversely impact the final product. Manual control is not possible due to the large number of
variables and constraints.

The plant-wide advanced process controller consists of individual controllers for grinding and flotation circuits.
Additionally there is an optimizing controller that writes target set points for the individual controllers. The
grinding controller is a combination of model-based multivariable predictive control and expert systems. It also
has a ‘softsensor’, which predicts key process variables in the absence of field sensors. The flotation controller
is based on a combination of process knowledge and heuristic knowledge of expert operators. It has been
developed using APC Toolkit, TCS’ proprietary tool for rapid development of Advanced Process Control
applications.

Benefits accruing to HZL include sustained improvement of 3.2% in zinc recovery, reduced variability in
product quality, reduced energy consumption during grinding, increased uptime and smoother running of the
plant, resulting in savings in maintenance and breakdown costs.

1. BACKGROUND

Tata Research Development and Design Center (TRDDC), a division of Tata Consultancy Services (TCS), in
collaboration with Regional Research Laboratory, Bhopal, and Hindustan Zinc Limited (HZL), has deployed a
state-of-the-art online plant-wide advanced controller for HZL’s ore beneficiation plant at Rampura-Agucha,
Rajasthan, India. The Department of Mines (DOM) and Department of Science and Technologies, both
government of India agencies, jointly sponsored this project meant to be a technology demonstrator of plant-
wide advanced process control for the first time in the Indian mineral processing sector.

2. THE PLANT

Ore beneficiation at HZL’s Rampura-Agucha plant is a two step process. Ore is pulverized to finely ground
particles in wet grinding mills in order to liberate the valuables, namely lead and zinc, from the associated
gangue. The ground particles are then selectively floated in flotation cells for individual recovery of lead and
zinc. The plant currently processes around 4500 tpd (225 tph) of lead-zinc ore with a grade ranging from 9 to
14% zinc and 1.2 to 2% lead. The plant consists of:
§ Three grinding and classification lines, each consisting of:
ù One rod mill (open circuit)
ù One ball mill (open circuit)
ù A two stage classification circuit
§ Lead flotation circuit with:
ù Conventional rougher and scavenger flotation cells
ù Conventional cleaner flotation cells
§ Zinc flotation circuit with:
ù Conventional rougher and scavenger flotation cells
ù Conventional cleaner flotation cells

Page 1
The ore beneficiation plant is completely automated, with a significant amount of instrumentation and control
hardware. All field instruments and controllers are monitored and controlled by a centralized Distributed
Control System (DCS; Honeywell TDC3000 series).

3. THE PROBLEM

Large and frequent variations in the characteristics of the ore feed to the grinding units adversely impact the
optimum grind size of the ground product. These variations in the ground product size (which forms the feed to
the flotation circuit) are detrimental to both recovery and grade of lead and zinc. Considering the wide variations
in feed quality and the number of variables, constraints and performance targets to be monitored and maintained
simultaneously, optimal performance can never be ensured through manual control. An optimizing control
system will make it possible to continuously achieve plant targets (throughput, grade and recovery)
notwithstanding the variations in feed characteristics.

4. OBJECTIVE

The objective of the project is to design and implement supervisory plant-wide control systems for the ore
beneficiation plant. This is expected to achieve 1% improvement in lead and zinc metal recoveries
(individually). Additional benefits expected are:
§ Savings in reagent consumption
§ Savings in energy consumption during grinding
§ Lower variability in plant operation
§ Smoother running of the plant resulting, in savings in maintenance and breakdown costs

5. THE SOLUTION

The plant-wide controller is a state-of-the-art software package for optimization of operations at the plant level.
The software is built using a combination of C++ and Honeywell PLANTSCAPE™ SCADA features (operator
graphical user interface (GUI), drivers for communication with the DCS and application co-ordination) and runs
on a Windows  NT platform. The combined system, consisting of individual advanced process controllers for
the grinding and flotation circuits, has advantages for the end user primarily in terms of optimized use of
resources and automatic tracking of disturbances with immediate corrective action, which help in improving
overall plant performance.

5.1. Grinding Mill Advanced Process Controller

The Grinding Advanced Process Controller (GAPC) is a state-of-the-art optimizing controller specifically
developed for grinding circuits in mineral processing plants. The GAPC has two major components as follows.

5.1.1. Circuit Supervisory Controller

The supervisory control application is a combination of a model-based multivariable predictive control (MPC)
algorithm and an expert system for process condition inference. The core technology components are embedded
into an application in the PLANTSCAPE™ SCADA environment. It computes the optimal values for the
manipulated variables (MVs) at each sample time and downloads them to the DCS.

The MPC algorithm uses models embodying dynamic relationships among all MVs and controlled variables
(CVs). The models are built using plant data collected during plant campaigns. The MPC algorithm performs
explicit optimization to arrive at a set of MVs within the operator-specified bounds, minimizing the error
between the setpoint specified by the operator and the actual values of the CVs for achieving the desired plant
performance. The above algorithm makes good use of ‘softsensors’ wherever physical sensors are not available.

5.1.2. Circuit Softsensor

The Circuit Softsensor consists of a phenomenological dynamic model for wet grinding circuits. This model,
which has been custom-built and configured for the Rampura-Agucha grinding circuits, works in tandem with
the grinding supervisory controller as a real time simulation task in the PLANTSCAPE™ SCADA environment.
The softsensor takes live data from the plant (via DCS / PLANTSCAPE™ SCADA) and predicts all the key
process variables, including size distribution of solids. Thus, this model acts as a softsensor in the absence of
field (hard) sensors to provide information on key plant process variables that are not directly measurable.
Simulation outputs are presented as report files (steady state simulation), on–screen size distribution data, and
real time trends (dynamic simulation).

Page 2
5.2. Flotation Circuit Advanced Process Controller

Flotation control is achieved by a combination of process knowledge and heuristic knowledge of expert
operators. The heuristic level of flotation control utilizes the knowledge of approximate relationships between
the process inputs (reagent dosages, pulp level, aeration rate) and concentrate grades and recoveries. These
relationships are evaluated through data collected from extensive plant campaigns using a statistical technique
called design of experiments. Once these relationships have been established, a set of knowledge-based
operational rules is coded in the fuzzy rule base. This rule base also contains fuzzy rules derived from the semi-
empirical model of the process. A fuzzy inferencing mechanism operates on these rules to infer proper action to
be taken in order to maintain acceptable grade and recovery levels.

The flotation control application is developed using APC Toolkit, an indigenous product developed by TCS for
rapid development of Advanced Process Control based applications. Though the application runs outside the
PLANTSCAPE™ SCADA, it continuously exchanges data with the SCADA system and executes on the same
level as the grinding controller.

5.3. Overall Optimizer

In the hierarchical structure of the controllers, at the topmost level is an optimizing controller overseeing and
coordinating the aforesaid controllers at the next level of hierarchy. The overall optimizer will be responsible for
minimizing a cost function based on economics and logistics of the process to arrive at targets to be provided as
set points for the individual advanced controllers at the lower level.

6. SYSTEM ARCHITECTURE

A view of the system architecture is shown in figure 6.1 below. The controller application runs on top of a
PLANTSCAPE™ SCADA/ Windows  NT environment. Data exchange with the DCS (thus eventually with the
plant) is done through interfaces provided by the PLANTSCAPE™ SCADA protocols.

PlantScape™ SCADA Supervisory

User Interface

Controller

PlantScape™
User Database
Softsensor

Setpoints
Process Data
UI DCS/PLC

User
Plant Instrumentation

Process
§ Grinding
§ Flotation

Figure 6.1

Page 3
7. PERFORMANCE, RESULTS AND DISCUSSION

The grinding circuit advanced process controller has been commissioned for all three grinding circuits at the
plant site and has been running successfully since November 2000. The deployment of the grinding controller
up-front helped to quantify the impact of optimized control in the upstream grinding circuit on the downstream
flotation circuit, and thus led to better understanding of the control requirements in the flotation circuit.

The flotation advanced process controller has been commissioned and its performance is currently being
monitored to estimate benefits.

7.1. Salient Features of the Grinding Advanced Process Controller (GAPC) Performance

7.1.1. Increased Throughput

Initial results indicate a sustained increase in current throughput levels, besides consistently meeting grinding
circuit quality requirements vis-a-vis product size distribution and percent-solids ratio. This itself will have a
strong positive influence in appreciably increasing the overall productivity of the plant, including individual
recoveries of lead and zinc. Figure 7.1 below shows that the GAPC can handle a throughput of 74 tph (upper
graph) by keeping all other key process variables steady (lower graph), as against the standard practice of much
lower throughput.

Figure 7.1

Page 4
7.1.2. Smoother Process Operation

The operating regime can be smoothly transferred from one operating point to the other without affecting key
process variables. Figure 7.2 shows how the process under control of GAPC behaves against a step change in
fresh feed (upper graph). The controller smoothly executes the change by keeping all other key process variables
steady, such as throughput, percent solids, and size fractions at grinding discharge (lower graph).

Figure 7.2

7.1.3. Improvement in Overall Metal Recovery

One of the key performance criteria in determining the benefits of the supervisory controller is its impact on the
final product. As can be seen from figure 7.3, there has been a remarkable improvement in zinc recovery during
2001 calendar year as compared to 2000. The average recovery has gone up from 87.32% to 90.11%, which is
the highest level seen by the plant. During this whole period, the supervisory control system was online with an
average uptime of about 93%. It is important to note that this increase in recovery has been achieved without
affecting the product grade, which was at 54.28% in 2000 as compared to 54.49% in 2001 (figure 7.4).

Zinc Recovery (Years 2000 - 2001)

100.0

95.0
Recovery % (Daily Averages)

90.0

85.0

80.0

75.0

70.0

65.0

60.0

55.0
Manual Control Supervisory Control Specification (88%)
50.0
1 30 59 88 117 146 175 204 233 262 291 320 349
Days

Figure 7.3
Page 5
 Zinc Grade (Years 2000 and 2001)

60.00

55.00
Grade % (Daily Averages)

50.00

45.00

Manual Control Supervisory Control Target (54%)

40.00
1 30 59 88 117 146 175 204 233 262 291 320 349
Days

Figure 7.4

7.1.4. Reduction in Recirculation Load

The GAPC successfully reduced the recirculation load, which is an indirect measure of energy consumption in
the grinding circuit. Figure 7.5 shows the trend of reduction in energy consumption for one (out of three) such
grinding circuits.

450
GAPC Installed, Nov-00

400
% Recirculation

350

300

250

200
Oct-00

Mar-01
Jul-00

Jul-01
Jan-01

May-01
Dec-00
Aug-00

Apr-01

Aug-01
Feb-01

Jun-01
Sep-00

Nov-00

Month-Year

Figure 7.5

Page 6
7.1.5. Steady Sump Levels

Variations in one of the key process variables can be observed by looking at the behavior of sump levels with
GAPC ON and OFF. Figures 7.6 and 7.7 show how sump levels are completely steady with GAPC online.

Ore Feed Sump1 Level Sump2 Level

Sump Levels With Controller OFF
100

80
% Levels

60

40

20
8:24 9:36 10:48 12:00 13:12 14:24 15:36
11-Jan-01, Time (Actual)

Figure 7.6

Ore Feed Sump1 Level Sump2 Level

Sump Levels With Controller ON
100

80
% Levels

60

40

20
8:00 9:12 10:24 11:36 12:48 14:00 15:12 16:24
17-Feb-01, Time (Actual)

Figure 7.7

Page 7
7.1.6. Control of Product Size Fractions

GAPC keeps all size fractions within specified control limits and supplies a uniform product size distribution for
the downstream flotation process. Figure 7.8 below shows that percent retained over the 53µ size is within
control limits specified by the plant (data collected from plant with controller online over almost one year).

Plant UCL LCL

Percent Retained Over 53µ Size
63
58
53
% Retained

48
43
38
33
28
23
18
1 9 17 25 33 41 49 57 65 73 81 89 97
Sample Number

Figure 7.8

7.1.7. Improved Recovery in Flotation Stage

The uniformity achieved in the grinding circuit product helped the flotation operation downstream to show a
noticeable upward trend in zinc recovery (figure 7.9) while keeping the zinc grade stable (figure 7.10).

100
GAPC Installed, Nov-00

96
% Recovery

92

88

84

80
Aug-00

Aug-01
Apr-00

Apr-01
May-00

May-01
Oct-00
Feb-00

Mar-00

Jun-00

Jul-00

Mar-01

Jun-01

Jul-01
Sep-00

Dec-00
Nov-00

Jan-01

Month-Year

Figure 7.9
Page 8
 60
GAPC Installed, Nov-00

58
% Grade

56

54

52

50
Apr-00

Apr-01
Aug-00

Oct-00
Mar-00

Jul-00

Mar-01

Jul-01
May-00

May-01
Jan-01
Feb-00

Jun-00

Feb-01

Jun-01
Sep-00

Dec-00
Month-year Nov-00

Figure 7.10

7.1.8. Other Benefits

Other than the salient points stated above, there are many qualitative benefits, some of which are:

§ Almost total elimination of alarms due to sumps overflowing or running dry. Earlier, without the controller,
these alarms were very frequent and contributed heavily to the decibel levels in the control room.

§ Ability to perform at conditions where manual control is very difficult, e.g., maintaining low percent solids
with high throughput or high percent solids with low throughput.

7.2. Salient Features of the Flotation Advanced Process Controller (FAPC) Performance

Currently, performance of the Flotation Advanced Process Controller (FAPC) is being monitored and data is
being collected to estimate the quantitative benefits. Preliminary results show smoother operations, significant
reduction in product variability, and improvements in recovery levels of both zinc and lead.

8. CONCLUSION

Deployment of the plant-wide advanced controller has resulted in realization of several qualitative and
quantitative benefits, both in the grinding circuit and in the flotation circuit. This has demonstrated the necessity
and viability of plant-wide advanced process control in the Indian mineral processing sector.

Page 9