Schneider, P.A. and Vigh, S.N.

Proc. Aust. Soc. Sugar Cane Technol., Vol. 26, 2004

________________________________________________________________________________________________

BATCH PAN CONTROL BY

MICROWAVE BRIX

By

P.A. SCHNEIDER and S.N. VIGH

School of Engineering, James Cook University, Townsville, Qld

[email protected], [email protected]

KEYWORDS: Batch Vacuum Pan, Process Control, Brix Control, Microwave Brix Sensor.

Abstract

FOR MANY decades, the massecuite brix in vacuum pans has been controlled using
simple and cost effective conductivity probes. Despite their widespread use, they suffer
from a variety of problems, including temperature sensitivity and changes in impurity
loads. Arguably, conductivity probes are not highly process relevant, since they deliver
at best only an indirect measure of fundamentally important process variables. This
paper presents results from a series of industrial trials using a microwave-based
instrument. It presents an overview of the installation and use of this class of device for
effective batch vacuum pan control. Experimental results from a series of trials run on a
100 tonne batch vacuum pan are presented.

Introduction

For many decades the Australian raw sugar industry has employed conductivity-based
sensors to elucidate the condition of boiling massecuites, in order to control batch and continuous
raw sugar vacuum pans. Despite their advantages of low capital cost, ease of installation, and low
maintenance requirements, conductivity probes are influenced by many factors that often confound
their effective use.

The practical goal of a conductivity sensor is to deliver an estimate of the massecuite

‘weight’ or ‘heaviness’ in a pan, so that pan boilers can relegate pan control to a process control
system. This gives them the ability to monitor the entire pan stage, without having to attend each
pan individually.

Massecuite ‘weight’ and ‘heaviness’ are in fact qualitative terms describing the crystal
content of the material in the vacuum pan. It is the massecuite crystal content that determines the
total surface area available for sucrose deposition. Sufficient crystal surface area (i.e. ‘heaviness’)
leads to a situation whereby the molasses oversaturation cannot feasibly reach levels necessary for
secondary nucleation. This buffering effect allows the operator to set the process controller to a
reasonably productive level of operation, without fear of filling in the massecuite with secondary
nuclei.
Schneider, P.A. and Vigh, S.N.
Proc. Aust. Soc. Sugar Cane Technol., Vol. 26, 2004
________________________________________________________________________________________________

Other sensors have been evaluated as possible alternatives to conductivity control (Miller
and Skippen, 1989), but the main conclusion was that they were ill suited to the task or
prohibitively expensive.

While conductivity-based process sensors are sensitive to changes in massecuite crystal

content, they are also sensitive to a variety of other confounding process variables (Wright, 1984),
making the level of control suffer. It is therefore desirable to employ a sensor that is more closely
related to massecuite crystal content.

The relationship between massecuite crystal content and Brix can be summarised in the
following equation.
100 (Brix massecuite − Brix molasses )
CC =
(100 − Brix molasses )
where
CC = crystal content, % on massecuite
Brix massecuite = massecuite brix
Brix molasses = molasses brix

If the molasses Brix lies within a reasonably small range, then a strong correlation exists
between massecuite Brix and crystal content. If the molasses Brix were monitored, using a process
refractometer, a highly accurate determination of crystal content should be possible. As such,
controlling massecuite Brix is an excellent surrogate for crystal content control.
Factory trials

The following sections detail experiments carried out in a 100 tonne batch vacuum pivot
pan, which executes a second seed strike, cuts out 50 tonnes of massecuite to another pan on the
stage and then runs up a 100 tonne A strike. This pan is located at CSR Mills Group’s Macknade
Sugar Mill.

Proposed microwave Brix-based vacuum pan control scheme

An alternative vacuum pan control scheme is proposed, in which the massecuite Brix is
controlled by manipulating the fresh liquor feed flow rate to the pan, pictured below in Figure 1.
This control loop replaces that of the traditional conductivity controller and has been previously
reported (Schneider, 2003). Steam flow control is identical to that employed by the standard
conductivity-controlled case.

Monitoring massecuite Brix

Measuring massecuite Brix was accomplished using a pro/M/tec1 microwave density unit.
This system is composed of an evaluation unit that transmits microwave energy between two rod-
shaped antennae of 100 mm length, separated by a 100 mm gap, to create a microwave field. Dipolar
water molecules have a high dielectric loss at microwave frequencies, leading to signal attenuation
and phase shift in the field. In the present system, the attenuation and the phase shift are inversely
proportional to the Brix of the media, and the unit can be calibrated across arbitrary Brix ranges.

1
pro/M/tec Theisen GmbH, Pforzheimer Str. 162, D – 76275 Ettlingen, Germany. (http://www.pro-m-tec.de)
Schneider, P.A. and Vigh, S.N.
Proc. Aust. Soc. Sugar Cane Technol., Vol. 26, 2004
________________________________________________________________________________________________

Fig. 1—Schematic of factory batch vacuum pan control system.

The recommended installation of the pro/M/tec is on the wall of the pan, just under the
calandria (where a level transmitter would typically be installed). While this location was initially
employed, it was soon abandoned due to a lack of dynamic response and non-representative
sensing.

The microwave device was instead flange-mounted onto an instrument tower. The concept of
the instrument tower resulted from the requirement to locate pan sensors in a more representative
location, so that they are exposed to circulating material. In this application the instrument tower was
located within the centre-well of the pan, flange mounted to the central cone of the pan floor, along
the axis of the pan. Figure 2 shows a schematic of the instrument tower, including one of a number of
flange mounting pads, upon which a variety of sensors can be placed. The microwave device can be
seen mounted onto the topmost pad. Note that the microwave sensor, but not the antennae, is located
within the instrument tower, which is actually located outside of the pan volume.

Access to sensors installed within the instrument tower is a key limitation of this concept,
since the microwave Brix device is not reachable, without first shutting down the pan and gaining
access to the interior. However, the benefit of this scheme is significant, since the process sensor is
exposed to fresh, circulating massecuite. It is worth noting that despite the harsh environmental
conditions, the pro/M/tec device ran in this location, without incident, for three crushing seasons.

Fig. 2—Schematic diagram of the instrument tower employed in the factory studies.
Schneider, P.A. and Vigh, S.N.
Proc. Aust. Soc. Sugar Cane Technol., Vol. 26, 2004
________________________________________________________________________________________________

Typical calibration data for the pro/M/tec device is shown in Figure 3. A linear relationship
shows a reasonable fit to the data.

93
92
91
Laboratory Brix

90
89
88
87 y = 0.7591x + 20.358
R2 = 0.9461
86
85
85 86 87 88 89 90 91 92 93 94 95
pro/M/tec Brix value

Fig. 3—Calibration data for pro/M/tec microwave Brix device.

Controller development and implementation

Feedback control of the Brix signal was prototyped using UNAC (now known as
ProcessACT2). A portion of the UNAC controller schematic is presented in Figure 4, indicating its
graphical nature.

Fig. 4—Brix controller prototype UNAC schematic.

UNAC was interfaced with the existing distributed control system (DCS) in such a way that
candidate controllers could be implemented without risk to the process. Bumpless transfer between
the two systems was achieved by forcing the UNAC advanced PID controller (APID) to track the
feed control valve signal (N90 Feed CO) while it was in manual mode. In order to further ‘bullet

2
Matrikon, PO Box 516, Mayfield NSW 2304 Australia. (www.matrikon.com)
Schneider, P.A. and Vigh, S.N.
Proc. Aust. Soc. Sugar Cane Technol., Vol. 26, 2004
________________________________________________________________________________________________

proof’ the system, a UNAC-generated ‘heartbeat’ (actually a square wave with period of 10
seconds) was sent to the DCS. If this signal did not change – indicating that UNAC has frozen or
‘died’ - the DCS would react by ignoring the UNAC control signals.

Since UNAC has a variety of ready-made blocks, a number of controller prototypes can be
quickly developed and analysed. Furthermore, UNAC can easily communicate with a variety of
commercially distributed and PLC-based control systems, making prototyping relatively
straightforward.

Brix setpoint selection

As mentioned before, the massecuite Brix is expected to change as the batch proceeds.
Therefore, a setpoint trajectory for massecuite Brix is required to effectively control the pan. The
vacuum pan under study was a pivot pan that boiled a second seed strike followed by an A strike.
As such, two setpoint trajectories were necessary. The massecuite Brix setpoint trajectories were
scheduled against the tonnage of massecuite in the pan according to Figure 5. Note that a constant
Brix profile is required for the second seed strike, while the A strike requires a constant Brix
setpoint followed by a ramped setpoint. This ramp is used to account for the increase in massecuite
Brix due partly to the boilback of A-molasses, but more so due to the increase in crystal content of
the massecuite which is necessary in order to bring the pan to a point where heavy-up can occur
without undue nucleation.

92
A strike trajectory
91 2nd seed trajectory

90
Brix setpoint

89

88

87

86
45 55 65 75 85 95
Masecuite (tonnes)

Fig. 5—Brix setpoint trajectories for factory experiments.

The Brix trajectories chosen were based on observed Brix profiles under conductivity
control. It should be noted that the setpoint trajectories reported are only of relative significance,
since the Brix device was calibrated only approximately. As such, the reported setpoint trajectories
should not be considered meaningful in any absolute sense. However, this does not preclude the use
of a massecuite Brix device within a closed loop controller.
Schneider, P.A. and Vigh, S.N.
Proc. Aust. Soc. Sugar Cane Technol., Vol. 26, 2004
________________________________________________________________________________________________

Controller selection, tuning and performance

A general rule of thumb when designing feedback controllers is to keep things as simple as
possible. As such, the first controller tested in the factory trials was a proportional-only controller.
When trials commenced it became apparent that this controller could be tuned with very high
proportional gains. This result was not surprising, since very high gain terms also worked in a
related simulation study (Schneider, 2003). In fact, the gains were high enough such that minimal
offset occurred, without recourse to integral control action. The final value chosen for the
proportional gain of the feedback controller balanced responsive closed-loop dynamic response
against excessive controller action, which would otherwise burden the feed control valve.

Typical closed loop performance of the alternative control system is detailed in Figure 6,
which shows a complete cycle for the second seed and A-strike of the pan. The setpoint and
measured massecuite Brix are plotted versus the sample time in the batch. The second seed occurs
between zero and about the 350th sample time. Cutting of massecuite out of the pan occurs from the
350th to the 500th sample time. The A-strike takes place from the 500th to about 950th sample time.
The significant departures of the measured massecuite Brix during start-up and cutting are due to
severe temperature departures, due to loss of process vacuum.

During the second seed and A-strike, the control of massecuite Brix is more than adequate,
especially considering that a simple proportional controller is being used. Note that the massecuite
Brix setpoint is fixed at 86.5 Brix for the first part of the pan cycle, but increases during the second
part of the cycle, due to the setpoint being scheduled against the increasing pan tonnage (see Figure
5). One point of interest is that the ramped setpoint trajectory displays a degree of random variation,
which is attributed to noise in the measured massecuite tonnage within the pan.

92 Brix
Setpoint
91
2nd Seed A-strike
90
Massecuite Brix

89
88
87

86
85
0 200 400 600 800
Sample Time

Fig. 6—Control of massecuite Brix in 100 tonne pivot pan cycle.

Figure 7 shows that the feed flow to the pan is similar to that of a conductivity controlled
pan strike. It is clear that the control system can maintain offset free control, using a very simple
feedback controller.
Schneider, P.A. and Vigh, S.N.
Proc. Aust. Soc. Sugar Cane Technol., Vol. 26, 2004
________________________________________________________________________________________________

Compared with conductivity-based control schemes, it is clear that the microwave-based

controller is worth consideration in a factory setting. One reason for this is that the massecuite Brix
signal is less sensitive to those process variables that influence electrical conductivity. As such, the
selection of a setpoint trajectory for high grade massecuite Brix should, in principle, be more
reliably than with a conductivity probe. Of course, the temperature sensitivity of the microwave
Brix sensor is still an issue, but this could be easily accommodated through a temperature
compensation circuit (or a function block in the factory’s DCS).

Another important aspect of this scheme was that the pan boilers embraced it quite
enthusiastically. In fact, during an episode of high salt levels entering the factory (due to drought
conditions in the 2002 season) in make-up water, high salt levels were noted in the feed liquor,
making conductivity control impossible (the conductivity readings were off scale). At this stage, the
pan boiler requested that the massecuite Brix be brought online, so that at least one pan on the stage
did not have to be closely monitored.

100
90
80
Feed Flow (m /h)

70
3

60
50
40
30
20
10
0
0 200 400 600 800
Sample Time

Fig. 7—Feed flow in 100 tonne pivot pan cycle.

Conclusions and outlook

A control system based on massecuite Brix is proposed, in which the conductivity controller
is replaced with a massecuite Brix controller. It was shown that this system can be used to
inferentially control the crystal content in the massecuite, without having to measure it. A 100 tonne
batch vacuum pan was used for a series of plant trials. This pan had an instrument tower within the
centre-well of the pan, which enabled the installation of the microwave Brix device (and others) in a
location that afforded highly representative sensing. A simple proportional feedback controller was
employed to maintain the massecuite Brix along predefined setpoint trajectories, scheduled against
the tonnage in the vacuum pan. Operator acceptance of the massecuite Brix control system was
encouraging.

Further work needs to be done in order to optimise the control of the pan by massecuite
Brix. The setpoint trajectories used were based on the existing conductivity control scheme. Work is
currently underway to determine optimal trajectories by computer simulation. Further industrial
experiments are recommended.
Schneider, P.A. and Vigh, S.N.
Proc. Aust. Soc. Sugar Cane Technol., Vol. 26, 2004
________________________________________________________________________________________________

Acknowledgements

The first author would like to thank the Sugar Research Development Corporation for funds
used to run this project. Both authors would like to thank everyone at Macknade Mill for their keen
interest and support of this project. Special thanks are extended to Chris Bytheway who designed
and commissioned the instrument tower, Danny Ferraris for his assistance with the UNAC/DCS
interfacing and all of the pan boilers for supporting this project.

REFERENCES

Miller, K.F. and Skippen, N.G. (1989). An evaluation of several sensors for feed control of
vacuum pans. Proc. Aust. Soc. Sugar Cane Technol., 11: 175–183.

Schneider, P.A. (2003). Evaluation of an alternative vacuum pan control scheme. Proc. Aust. Soc.
Sugar Cane Technol., 25: (CD-ROM).

Wright, P.G. (1984). Characteristics of conductivity transducers for pan control. Proc. Aust. Soc.
Sugar Cane Technol., 6: 301–307.