BOSE, JUDY ANN D.

09-15-17
BSCHE-V

NONLINEAR CONTROL OF A BATCH POLYMERIZATION REACTOR: An

Experimental Study

In this experimental study, the GLC method is implemented to control

the temperature of a batch polymerization reactor, in which solution

polymerization of methyl methyacrylate (MMA) takes place. The initiator and

solvent are azo-bis-isobutyronitrile (AIBN) and toluene, respectively. This

work is the first experimental study, in which a nonlinear model based

control method is applied to a polymerization reactor.

The control objective is to track the calculated optimal temperature

profile in the presence of disturbances by manipulating the heater power and

the cooling water flow rate. In this heating/cooling scheme, the inlet flowrate

of cooling water and the input power to the heater is the manipulated inputs.

And the controlled variable is the reactor temperature.

BOSE, JUDY ANN D. 09-15-17
BSCHE-V

T
RTD
T
N2 Gas

T
RTD

Reactor

Heater
Coolant

D/A A/D D/A A/D A/D A/D

Schematic Diagram of a Batch Reactor System

BOSE, JUDY ANN D. 09-15-17
BSCHE-V

Control Strategies

1. FEEDFORWARD CONTROL

A feedback controller responds only after it detects a deviation in the

value of the controlled output from its desired set point. On the other hand, a

feedforward controller detects the disturbance directly and takes an

appropriate control action in order to eliminate its effect on the process output.

Consider the distillation column shown in Fig (1) The control objective

is to keep the distillate concentration at a desired set point despite any

changes in the inlet feed stream.

(a) Feedback control configuration (b) Feedforward control configuration

Feedback and Feedforward control configuration of a distillation column

BOSE, JUDY ANN D. 09-15-17
BSCHE-V

Fig(a) shows the conventional feedback loop, which measures the

distillate concentration and after comparing it with the desired setpoint,

increases or decreases the reflux ratio. A feedforward control system uses a

different approach. It measures the changes in the inlet feed stream

(disturbance) and adjusts the reflux ratio appropriately. Fig (b) shows the

feedforward control configuration.

2. CASCADE CONTROL

The primary disadvantage of conventional feedback control is that the

corrective action for disturbances does not begin until after the controlled

variable deviates from the setpoint. In other words, the disturbance must be

felt by the process before the control system responds. Feedforward control

offers large improvements over feedback control for processes that have large
BOSE, JUDY ANN D. 09-15-17
BSCHE-V

time constant and/or delay. However, feedforward control requires that the

disturbances be measured explicitly and that a model be available to calculate

the controller output

Cascade Control Structure

3. RATIO CONTROL

A ratio controller is a special type of feedforward controller where

disturbances are measured and their ratio is held at a desired set point by

controlling one of the streams. The other uncontrolled stream is called wild

stream. Figure shows the schematic of a ratio controller. The ratio of flow rates

of two streams are being held at a desired ratio by controlling the flow rate of

one stream. The flow rates are measured through flow transmitters (FTs).
BOSE, JUDY ANN D. 09-15-17
BSCHE-V

The schematic of a ratio control structure

The chemical process industries have various applications for ratio controllers.

Following are a few such examples:

Reflux ratio and reboiler feed ratio in a distillation column

Maintaining the stoichiometric ratio of reactants in a reactor

Keeping air/fuel ratio in a combustion process

4. OVERRIDE CONTROL

During the operation of a process plant it is possible that a dangerous

situation may arise due to unacceptable process conditions which may

destruct the process or its personnel. In such case the normal operation should

temporarily be stopped and preventive measures should be initiated to avert

the unacceptable situation. In order to facilitate such measures, a single-

BOSE, JUDY ANN D. 09-15-17
BSCHE-V

purpose switch can be used that can take preferential instruction from one

controller over the others to manipulate the final control element in such a

way that the dangerous situation can be averted. This is called override

control. The technique can be illustrated with the following example.

Override control of a boiler

Consider a boiler shown in the Fig. It has one water inlet and one steam

outlet. The steam outlet is regulated by the valve in the discharge line that

takes the control signal from the control mechanism in Loop1 (pressure

transducer and pressure controller). In other words, the discharge of steam is

regulated on the basis of its pressure desired in the supply line elsewhere.

However, the water is boiled using a heating coil that needs to be always

submerged below the water level so that the heating coil does not burn out.

Hence, in order to ensure a certain minimum level of water inside the boiler,

the control Loop 2 is set in place that contains a level transducer and a level
BOSE, JUDY ANN D. 09-15-17
BSCHE-V

controller. Both level controller and pressure controller give the control signal

to the valve through an intermediate switch LSS (Low Selection Switch) that

takes the preferential signal from the level controller. In other words, Loop 2

remains inactive during the normal operation and the Loop 1 regulates the

process. Nevertheless, at critical situation when the water level drops below

the minimum allowable limit, the Loop 2 takes over and takes corrective

measures.

5. AUCTIONEERING CONTROL

There are conditions in process plant where multiple process

measurements are available for a particular variable that needs to be

regulated through a single control action. Thus it is evident that the said

control action should be given based on the most critical measurement

condition for the process variable. This is termed as Auctioneering Control.

The technique can be illustrated with the following example.

BOSE, JUDY ANN D. 09-15-17
BSCHE-V

Auctioneering control of a tubular reactor

Consider a tubular reactor shown in the Fig. The reaction is exothermic and

hence the temperature inside the reactor needs to be regulated. However, the

temperature varies along the length of the tube and if the corrective action,

i.e. the coolant flow rate, is taken on the basis of highest temperature

measurement, it will ensure that the other temperature zones are also

guarded against overheating.

6. SPLIT RANGE CONTROL

When there are more than one manipulated inputs available for one

single controlled output then Split Range Control scheme can be implemented.

The technique can be illustrated with the following example.

BOSE, JUDY ANN D. 09-15-17
BSCHE-V

Split Range control

Consider a reactor shown in the Fig . The pressure inside the reactor

needs to be controlled. The control valve is available in both inlet and outlet

lines. Hence the control action can be coordination between two valves. When

the pressure is low, inlet valve should be FULLY OPEN and when the pressure

is high, outlet vale should be FULLY OPEN In either care the other valve needs

to be PARTIALLY OPEN/CLOSED depending upon the requirement of control

action.

7. ADAPTIVE CONTROL

It is understood in the previous chapters that task of controller design

for a process is very much domain specific. First the model of the process is

linearized around a certain nominal point and the controller is designed on the
BOSE, JUDY ANN D. 09-15-17
BSCHE-V

basis of that linearized model and finally implemented on the process. Hence,

the controller is applicable for certain domain around the nominal operating

point around which the model has been linearized.

Adaptive control

One of the most popular adaptive control techniques is gain scheduling

technique. The overall gain of an open loop process is usually given as

K overall K c K p K m K f

It is customary to keep the overall gain constant. In case of changes in the

process (or valve characteristics or measuring element), Kc should be tuned

in such a manner that overall gain remains constant.

K overall
Kc
K p KmK f
BOSE, JUDY ANN D. 09-15-17
BSCHE-V

The above is called the gain scheduling control law.

When the process is poorly known, one cannot rely much on the correctness

of the value of K In such cases, the self-adaptive control may be helpful. A

self adaptive controller optimizes the value of certain objective function

(criterion) in order to obtain updated controller parameters. Two examples of

self adaptive controllers are Model Reference Adaptive Control (MRAC) and

Self-Tuning Regulator (STR)

8. INFERENTIAL CONTROL

Often the process plant has certain variables that cannot be measured

on-line, however, needs to be controlled on-line. In such cases, the

unmeasured variables to be controlled can be estimated by using other

measurements available from the process. Consider the following example:

BOSE, JUDY ANN D. 09-15-17
BSCHE-V

Inferential Control

The process has two outputs y1 (unmeasured) and y2 (measured). The

disturbance d affects the process adversely that needs to be nullified by

manipulating input u. The open loop model equations can be written as

y1 G y1u u G y1d d

y 2 G y 2u u G y 2 d d

In this case disturbance, which is usually not measured, can be expressed in

terms of two measurable quantities

1 G
d y 2 yzu u
Gy d Gy d
z z

And hence,

1 G
y1 G y1u u G y1d y 2 yzu u
G yz d

Gy d
z


G y1d G y1d
y 2 G y u G y u u
Gy d
1 Gy d z
z z
BOSE, JUDY ANN D. 09-15-17
BSCHE-V

In other words, the variable y1 is estimated through two measurable quantities

y2 and u. This control mechanism is termed as inferential control because here

the controlled variable y1 is never measured; rather it has been estimated

through the inference drawn from measurement of other variables ( y2 and u

in this case).