Introduction To The Process Control Problem
Introduction To The Process Control Problem
Introduction To The Process Control Problem
PROCESS CONTROL
PROBLEM
Contents
2.1: Introductory Example
2.3.1: Controller
2.3.2: Control Valve
2.7: Summary
2. INTRODUCTION:
In this chapter, an introductory process control problem will be discussed. Two broad
approaches for control namely the feedback and feed forward control will be considered for
tackling the control problem
Uncontrolled disturbances: There are uncontrolled disturbances due to the variations in the
values of the following variables:
q : Inflow rate q, l/min
TI : Inlet temperature, o C
Thi : Inlet temperature of the heating liquid
These are called as the load variables. The heat transfer coefficient associated with the
transfer of heat from the jacket to the heating vessel can also change due to fouling. This can
also be treated as a load variable. Since the effect of fouling is a very slow process, this will
not be considered as a significant load variable in our present analysis of the control problem.
Manipulated variable: The variable amenable for regulation by the operator is the flow rate
qh of the hot liquid. This is called as the manipulated variable.
Statement of the control problem: Maintain the controlled variable T at the set point
Tset in the presence of load variables q, TI, Thi .
There are two broad control philosophies used for solution of the control problem namely the
feedback control and feedforward control philosophies.
n o o f v a r 4 x y 1 y 2 y 3 y 4
0
2 0 0 . 2 3 . 8 1 3 5 3 8 1 2 . 9 6 7 5 7 0 . 3 0 5 7 8 0 . 0 9 5 2 9 4
1 0 0 0 . 4 6 . 3 8 2 8 0 7 1 9 . 7 0 1 5 2 1 . 0 0 6 6 6 1 0 . 6 4 0 1 4 1
kc 2 0 . 2 0 . 6 6 . 3 1 9 0 7 1 1 6 . 7 7 7 2 3 1 . 7 1 6 5 8 4 1 . 3 6 4 3 1 8
t i 1 0 . 8 3 . 6 4 3 7 4 2 5 . 8 9 2 9 2 9 2 . 0 4 7 7 6 1 . 8 8 7 1 0 1
y s e t 1 1 - 0 . 2 1 3 9 5 - 7 . 0 0 1 9 5 1 . 8 1 9 2 1 9 1 . 9 3 0 7 5 8
d i s t 0 1 . 2 - 3 . 1 8 7 2 5 - 1 4 . 9 0 4 8 1 . 1 5 3 7 6 8 1 . 4 7 4 1 2 6
1 . 4 - 3 . 6 8 8 7 9 - 1 3 . 5 3 6 2 0 . 4 1 0 7 8 5 0 . 7 6 5 3 1 4
1 . 6 - 1 . 4 6 7 8 - 3 . 6 4 3 8 8 - 0 . 0 0 8 6 9 0 . 1 8 7 8 7 6
1 . 8 2 . 2 5 6 2 8 2 9 . 4 1 2 5 7 6 0 . 1 2 0 9 9 8 0 . 0 5 3 5 0 8
F i g . 2 . 9 : D y n a n m i c2 r e 5 s. 4p 5o 5 n3 8s 8e o f1 8 c. 5l 6o 9s 0 e9 d 0 . 7 2 8 2 7 1 0 . 4 3 3 7 4 6
y 1 , y 2 , y 3 , y 4
2 . 2 6 . 3 9 0 6 3 2 1 8 . 8 8 1 1 6 1 . 4 8 3 4 8 7 1 . 1 2 1 7 1 5
3 0 2 . 4 4 . 5 5 1 9 9 1 0 . 1 9 5 0 8 1 . 9 7 7 5 9 3 1 . 7 4 4 4 8 3
2 0 2 . 6 0 . 9 3 1 9 0 7 - 2 . 7 7 6 3 2 1 . 9 4 3 4 7 1 1 . 9 6 5 0 2 5
2 . 8 - 2 . 5 1 4 0 8 - 1 3 . 0 1 0 1 1 . 4 0 0 1 7 1 1 . 6 6 4 4 3 9
1 0 3 - 3 . 9 2 6 7 2 - 1 4 . 9 7 5 2 0 . 6 4 2 1 4 3 1 . 0 0 5 9 0 7
3 . 2 - 2 . 5 4 8 6 3 - 7 . 6 2 1 4 7 0 . 0 7 9 5 4 5 0 . 3 4 5 9 8 5
0 3 . 4 0 . 8 6 7 6 4 5 . 0 5 9 5 1 0 . 0 1 6 2 1 9 0 . 0 4 1 4 4 8
0 5 1 0 1 5 2 0 25
- 1 0 3 . 6 4 . 4 6 8 9 2 9 1 6 . 2 0 1 1 1 0 . 4 8 5 5 2 4 0 . 2 5 6 3 3
3 . 8 6 . 3 0 5 3 4 1 1 9 . 7 8 0 0 7 1 . 2 3 2 7 1 3 0 . 8 7 3 4 9 8
- 2 0 4 5 . 3 8 4 9 2 9 1 3 . 8 7 2 8 2 1 . 8 5 3 1 6 5 1 . 5 5 8 4 4 8
t i m e 4 . 2 2 . 2 0 8 6 1 1 . 6 8 9 3 4 9 2 . 0 1 1 4 5 3 1 . 9 4 0 5 5 8
4 . 4 - 1 . 5 0 3 1 6 - 1 0 . 1 7 0 9 1 . 6 2 2 6 5 2 1 . 8 1 3 5 9 3
4 . 6 - 3 . 7 4 2 8 6 - 1 5 . 2 9 3 9 0 . 8 9 7 8 6 8 1 . 2 4 6 9 8 1
T I to P
converte
Process liquid q,Ti r
Process liquid q,T p
It is possible to get a better understanding of the operational principles of the control system
through the representation of the control system in terms of block diagram representation
Load variables
TI q Thi
Control
Tset valve
Iset e Controll I I to P qh Heater T
T to I Final
+ er converter Control PROCESS
- Element
Feedback path
Measuring System &
Transmitter
IT
has one or more input signals and has an output. In the block the input signal is transformed
to an output signal. For example process block that is the heater has 4 input signals namely
three load variables (q, TI, Thi) and one manipulated variable (qh ). The output of the process
block is T, the controlled variable.
Assume that the process is initially at steady state with T=Tset . Let the initial value of Ti be
Ti0 . Error e =0. Consider a sudden change in Ti from Ti0 to Ti1 . The temperature T tends to
increase. This variation in temperature causes a negative error signal. The controller acts to
generate a corrective action in a direction which stops the increase of temperature away from
the set point and move the temperature towards the set point. The corrective action will cause
a reduction in the value of qh from its initial value. From the block diagram it can be seen that
there is a feedback path which continuously informs the controller about the effect of its
action on the process output. This is the reason that this control structure is called as negative
feedback system( because of the presence of a negative feedback path) or closed loop control
system (presence of a closed loop).
In the above control circuit, the function of all the elements can be understood very clearly as
there is a well defined function for each element. It is only in the case of the controller the no
quantitative relation has been given so far.
2.3.1: CONTROLLER:
We will consider a number of basic control options
Though this is a logical form of control action, this has a major limitation. Assume that the
process is at steady state at Ti = Ti0 , Thi = Thi0 and q =q0 . The controller is set such that it
generates a signal I0 for which the value of qh = qh0 . At this value of qh0 , the temperature T
will at T0 = Tset and the error signal will be zero. Assume that the temperature Ti increases to
Ti1 .This causes an increase in temperature driving the error signal to be negative. This causes
I to be less than Io and qh less than qho . The temperature rise will be arrested by this action.
Ultimately the system will reach a steady state with qh = qh1 < qh0 . The new value of qh1 will
not bring T to Tset , since a finite error signal is required to change qh from its original value
qh0 . For qh = qh1 , the steady state value of T = T1 > Tset . The deviation of the steady state
value of T from Tset is called as offset. Offset for a change in Ti from to Ti0 to Ti1 is T1 –
Tset . Hence proportional control has the drawback that it gives no offset only for one set of
values of the load variables. Offset occurs for any other set of load variables. In many
applications, steady state error (offset) may not be tolerated. Therefore one needs to look for
a control action that does not give offset irrespective of the values of the load variables.
(b) INTEGRAL ACTION: In this action, the control action uses the history of the error
signal. The integral control action is governed by the equation
t
Kc
I − I0 =
TI ∫ edt
t =0
(2.2)
TI = Integral time
An integral controller will be at steady state only if e = 0. If e > 0, then I continues to
increase. On the other hand if e < 0, then I decreases continuously. In a closed loop in which
various elements are connected, steady state exists if all the elements of the closed loop are at
steady state. Since an integral controller is an element of the closed loop, it should also be at
steady state. This happens only when e = 0 i.e. when the temperature T is at the set point.
Hence an integral controller ensures that there will be no steady state offset in the presence of
variation in the load variables.
Though integral action eliminates offset, there is some problem in basing the controller
totally on integral control action. This can be understood by analysing the relative speeds at
which corrective actions are generated by proportional and integral controllers.
Consider a proportional controller with Kc = 1 and an integral controller with TI = 1 min.
Assume that the error signal e = 0 initially. The error signal is suddenly changed to 1. The
responses of the two controllers are shown in Figure 2.3 below
Output of Integral
∆I
Controller
Output of Proportional
Controller
1 2
Time, min
It can be seen that the integral controller will take one minute to develop a corrective action
to a sudden change in e while the proportional controller develops the same change
immediately after the change in e. Hence it may not be advisable to use integral action alone
as the benefit of the immediate corrective action on the appearance of an error signal will not
be available. This can be overcome by combining proportional and integral action to generate
proportional + integral control action
Kc t
TI ∫0
I − I 0 = Kc e + edt
(2.3)
With Kc = 1 and TI = 1min, for a sudden change in error signal, the controller output will
vary as shown in Fig. 2.4
1 2
Sudden change in
output due to
Time, min
Assume that the heater process is initially at steady state with Ti = Ti0 = 300 C with T = Tset =
60 o C. The error signal e = 0. Assume that Ti changes from 30o C to 35 0 C at time t = 0. At
time t = 0+ , e = 0 since T has not yet changed from the its initial value of 600 C. As a result of
this, neither the proportional controller nor the Integral controller can generate a corrective at
t = 0+. This is not desirable since after the onset of a disturbance in Ti, one would like to
initiate a corrective action as quickly as possible. One can take advantage of the fact that at t=
0+, though T remains at T = 600 C and error signal e =0, dT/dt is not equal to zero. Since
dT/dt is > 0 at t = 0+, then de/dt < 0 (Fig 2.5). Hence one can generate a corrective action
based on the derivative of the error signal.
dT/dt
T
Time t
However derivative action cannot be used alone since at steady state, derivative of the error
signal de/dt =0. Hence the corrective action generated by the derivative action at the
beginning of the disturbance will disappear as steady state is reached. Hence derivative action
can be used with either a Proportional controller or a proportional integral controller.
PROPORTIONAL + INTEGRAL + DERIVATIVE CONTROLLER (PID)
The control law is given by
Kc t de
I − I0 = K c e + ∫
TI 0
edt + KcTD
dt
(2.4)
2.3.2: CONTROL VALVE:
The control valve is an air operated diaphragm control valve (Fig 2.6). Pneumatic signal from
the I to P converter goes to the topworks of the air operated diaphragm control valve.
The air pressure exerts a force on the diaphragm. This is opposed by a spring. The net
displacement of the diaphragm is communicated to the valve trim (valve plug) through the
valve stem movement. The valve stem moves through the packing. As the valve stem moves,
the area of opening of the valve port changes causing a change in the flow rate.
Air pressure
Topwork
Diaphrag
Spring
Valve
Packing
Valve trim
With the process placed in the closed loop with a controller, it is desirable to find the effect of
changes in disturbance on the steady state value of the controlled variable T. To perform
steady state analysis of the closed loop, it is necessary to develop steady state models for each
element in the control loop.
For the average jacket temperature one can use the arithmetic mean
(T + T )
Thav = hi hout
2
(2.7)
Degrees of freedom: One can start the analysis of steady state with the determination of the
degrees of freedom of the heater.
Degrees of freedom = Number of variables – Number of independent equations
Variables: q, qh , ρ, ρh , Cp , Cph , T,Ti , Thi, Thav, Thout , U, A
Number of variables = 13
Number of equations: 3
Degrees of freedom = 13 –3 = 10
Variables assumed known = ρ ,ρh , Cp , Cph , U, A
This consumes 6 degrees of freedom
Final degrees of freedom = 10 – 6 = 4
Four variables need to be specified to completely describe the steady state of the heater.
While any set of 4 variables can be chosen, the most logical choice is Ti, Thi, q, qh .
Since q, Ti , Thi are the load variables, their values vary in an uncontrolled manner. For each
instance of the values of these variables, the specification of qh will completely specify the
system and one can calculate T from the set of equations described above for the heater.
By eliminating Thav and Tout using the equations 2.5 to 2.7 one can get an explicit
expression for T.
1 2UAqh ρhC ph
T = Ti + (Thi − T )
qρC p 2qh ρ hC ph + UA
(2.8)
Temperature Sensor-Transmitter: As the temperature varies from 0 to 1000 C, the output IT
varies from 4 mA to 20 mA.
The gain KT of the transmitter is given by
20 − 4
KT = = 0.16
100 − 0
The transmitter equation is
I T − I T 0 = KT (T − T0 )
(2.9)
Error signal, e = Iset - IT
Controller: Proportional controller, I – I0 = Kce (2.10)
I to P converter: As the controller output goes from 4 to 20 mA, the output p of the I to P
converter varies from 0.2 to 1 kg/cm2 . The gain of the I to P converter KI is
1 .0 − 0 .2
KI = = 0.05
20 − 4
p − p0 = K I ( p − p0 )
(2.11)
Control valve: The control valve is assumed to have linear characteristics. When the pressure
varies from 0.2 to 1 kg/cm2 , the flow rate qh varies from 0 to 30 kg/cm2 . The gain of the
valve, Kv
30 − 0
KV = = 37.5
1 − 0 .2
To perform steady state analysis, we can start with nominal steady state conditions:
Thi0 = 1200 C; Ti0 = 300 C; T = 600 C; q=10 l/min; UA=15 kcal/(0 C.min); qh = l/min; Tset = 600
C; Kc= 1 mA/1mA
For a change in Ti , Thi and q from their nominal values to new value one can solve the
steady state model equations and get the new steady state value T. For illustrative purposes,
we will consider the cases of changes in values of Ti while the remaining two load variables
have their nominal values. Two other values of Ti are considered namely Ti1 = 350 C and Ti2 =
250 C. For this simplified case, one can use a graphical approach as described below.
The heater equations can be solved for T for variation of qh over its range 0 to 30 l/min for
three values of Ti namely Ti0 , Ti1 and Ti2 .
The steady state characteristics (static characteristics) of the heater are shown in Fig. 2.7 and
Table 2.1 presented in Excel worksheet
One can combine the gains of the temperature measurement and transmitter, controller, I to P
converter and the control valve to give the equation
q h − qh0 = Kc K I K v KT (Tset − T )
(2.13)
For Kc=1, this equation can be written as
various elements in the control loop. Because of the presence of dynamics, there is
a possibility of the closed loop system becoming unstable at a high value of Kc >
Kcmax.
This will be illustrated later in section 2.5
Because of these problems, it is not possible to depend on high values of Kc to restrict the
magnitude of offset.
Note : It is not necessary to extend the steady state analysis for a PI controller or a PID
controller. The control characteristics of a controller with integral action will be the
horizontal line corresponding to T = Tset . Thus the point of intersection of this characteristics
with the heater characteristics will always be at T = Tset for all values of Ti.
Since derivative action has no role to play in the steady state characteristics of a controller, it
is not necessary to extend this analysis for controllers with derivative action.
Graphical analysis outlined in the previous section becomes complicated when one has to
consider variations in more than one load variable. The equation solving approach becomes
complicated because of the nonlinearity of the heater characteristics defined in equation 2.8.
It is possible to employ an approximate analysis which can permit analytical solution to the
steady state analysis.
This approach is outlined below:
Assume that the process is initially at steady state with the following nominal conditions:
It is necessary to find the effect of changes in Ti, Thi, q and qh on T. To facilitate this analysis
one can apply deviation variables as shown below
2 qh0 ρh C phUA
q 0 ρ C p T 0 − q0 ρ C p T i 0 = (Thi0 − T0 )
2q h0 ρ h C ph + UA
Qout0 − Q i0 = Q ht 0
(2.16)
Where Qi0 = Sensible heat input, Qi at the nominal conditions
Qout0= Sensible heat output, Qout at the nominal conditions.
Qht0 = heat transfer term, Qht
With perturbations in TI, Thi, q and qh , one can write the heat balance equations as
Qi 0 + ∆Qi − Qout0 − ∆Qout = Qht0 + ∆Qht
(2.17)
To evaluate the deviation terms one can assume a simplification.
For example Qi is a function of q and Ti. To get the deviation of Qi from its base value, one
can apply Taylor series expansion of Qi as shown below:
∂Qi ∂Q i ∂ 2Q i ∂ 2Q i ∂ 2Q i
∆Qi = ∆( qρCpTi ) = ∆q + ∆ Ti + ( ∆ q ) 2
+ ( ∆ T 2
) + 2 ( ∆T )( ∆q) + ........ i
∂q ∂Ti ∂q 2 ∂Ti 2 ∂Ti ∂q
i
(2.18)
For small signals the higher order terms can be neglected so that equation 2.18 can be
approximated as
∂Qi ∂Q
∆Qi ≈ ∆( qρCpTi ) = ∆q + i ∆Ti = ρC pTi 0∆q + q0 ρC p ∆Ti
∂q ∂Ti
(2.19)
In the above equation the partial derivatives are evaluated at the nominal condition. Hence
equation 2.19 can be expressed as
600
∆Q ht = ∆q h + 10∆Thi − 10∆T
45
(2.22)
Substituting these values in the incremental heat balance
600
60∆q + 20∆T − 30 ∆q − 20∆Ti = ∆qh + 10 ∆Thi − 10∆T
45
(2.23)
Rearranging the above equation
∆Thi
∆q Kq
Kh
∆Ti Ki
+ + +
∆p ∆qh
∆Tse e ∆I ∆T
KT Kc KI Kv Kp
t +
+
-
∆IT
KT
Fig 2.8: BLOCK DIAGRAM OF THE HEATER CONTROL LOOP WITH STEADY
STATE GAINS
It is possible to write now the effect of a change in the values of load variables on ∆T.
Ki∆Ti + K q∆q + K h∆Thi KT K c K I Kv K p
∆T = + ∆Tset
1 + K c K I K v K p KT 1 + Kc Kv K I K p KT
(2.27)
The first term on RHS of the above equation indicates the ability of the closed loop control
system to reject disturbances in the form of load variable changes. This is called as the
regulatory control operation (∆Tset = 0) .The second term on the RHS corresponds to the
ability of the closed loop to respond to change in the set point. This is called as the servo
operation.
The numerator terms associated with the regulator operation corresponds to the effect of each
of the load variable on ∆T in the absence of control. The denominator term corresponds to the
effect of control action on both disturbance rejection and set point following.
For the case treated in section , ∆TI = 5 0 C while ∆q , ∆Thi and ∆Tset = 0. For Kc=1, the
steady state value of ∆T is
2
∆T = 3 5 = 2.94
20
1 + 1* 0.05 * 37.5 * * 0.16
45
(2.28)
When Kc is increased to 5, ∆T for the same change in TI gets reduced to
2
∆T = 3 5 = 1.99
20
1 + 5 * 0.05 * 37.5 * * 0.16
45
(2.29)
70
Temperature T
Series1
60
Series2
50 Series3
Series4
40
Series5
30
0 10 20 30 40
qh, l/min
As stated earlier, study of steady state performance alone will not guarantee the applicability
of a feedback control scheme for solution of a control problem. It is necessary to take into
account the dynamic behaviour of the closed loop in the presence of disturbances. Elements
in the closed loop interact with each other continuously as the signals pass from one element
to the next. This interactions can cause the signals passing through the loop to either subside
so that a steady state is finally reached or the signals start getting amplified leading to larger
and larger values as time progresses. This leads to an unstable condition, a situation not at all
desirable. It is only through consideration of the dynamics that one can find out the dynamic
behavior of the closed loop.
To illustrate this point, dynamic behaviour of individual elements in the closed loop is
incorporated and the resultant dynamic model equations of the closed loop are solved. Excel
worksheet is provided to study the dynamic behaviour of the closed loop. The controller
tuning parameters are given in the following cells.
Cell (5,2)= Kc, Proportional Sensitivity
Cell (6,2)= TI, Integral time.
Cell (7,2)= Change in set point
Cell (8,2) = Change in disturbance
The output variables in the worksheet are:
Integral action output = y1
Output of the control valve = y2
Output of the process heater = y3
Output of the measuring system = y4
You can see that for the settings Kc > 20.2 and TI= 1 min, the system will be unstable for a
sudden change in the set point or disturbance. You will see that as the control is tightened
with either increasing value of Kc or decreasing value of TI, the closed loop response will
become more oscillatory and will become unstable eventually. An unstable system or a
system that is too oscillatory is not acceptable. Thus it becomes imperative to undertake a
systematic study of the dynamic behaviour of the closed loop.
no of var 4 x y1 y2 y3 y4
0
20 0.2 3.813538 12.96757 0.30578 0.095294
100 0.4 6.382807 19.70152 1.006661 0.640141
kc 20.2 0.6 6.319071 16.77723 1.716584 1.364318
ti 1 0.8 3.643742 5.892929 2.04776 1.887101
yset 1 1 -0.21395 -7.00195 1.819219 1.930758
dist 0 1.2 -3.18725 -14.9048 1.153768 1.474126
1.4 -3.68879 -13.5362 0.410785 0.765314
1.6 -1.4678 -3.64388 -0.00869 0.187876
1.8 2.256282 9.412576 0.120998 0.053508
Fig. 2.9: Dynanmic response of closed
2 5.455388 18.56909 0.728271 0.433746
2.2 6.390632 18.88116 1.483487 1.121715
30 2.4 4.55199 10.19508 1.977593 1.744483
2.6 0.931907 -2.77632 1.943471 1.965025
20
y1,y2,y3,y4
2.8 -2.51408 -13.0101 1.400171 1.664439
10 3 -3.92672 -14.9752 0.642143 1.005907
3.2 -2.54863 -7.62147 0.079545 0.345985
0
3.4 0.86764 5.05951 0.016219 0.041448
0 5 10 3.6154.468929 2016.20111 25
0.485524 0.25633
-10
3.8 6.305341 19.78007 1.232713 0.873498
-20 4 5.384929 13.87282 1.853165 1.558448
time4.2 2.20861 1.689349 2.011453 1.940558
4.4 -1.50316 -10.1709 1.622652 1.813593
4.6 -3.74286 -15.2939 0.897868 1.246981
Feedback control structure has the disadvantage that after the onset of disturbance, the control
action can only be initiated after the controlled variable has started varying and not
immediately after the disturbance has set in. Feedforward is based on alternate philosophy
that requires monitoring of disturbances. Assume that TI and q are the load variables. Thus
for a complete implementation of feedforward control action, load variables TI and q are
monitored. For perfect steady state compensation of the effect of changes in load variables on
the controlled variable, steady state model of the heater is assumed to be known accurately.
The steady state model equation of the heater can be solved to calculate the value of qh which
is required to maintain T at Tset in the presence of variation in TI and q. The value of qh is
obtained as
1 1
qh =
2 ρ hC ph Thi − Tset 1
−
qρC p (Thi − Tset ) UA
(2.30)
In modern computer controlled system it is possible to perform such computations in real
time to provide the required value of qh to bring the process to steady state at the desired
value of T = Tset .
Thus with the availability of perfect model for the heater and the measurements of the load
variables, a perfect steady state compensation is possible. The feedforward control structure
is shown in Figure 2.10
TI q
PROCESS
pf Control qh HEATER
Valve
(a) A perfect model is seldom available so that uncertainty in the model parameters (like
density, specific heat etc) will result in an error in the feedforward control law. Hence
the steady state value of T will not be at the set point Tset .
(b) Additional measurements are required to implement feedforward control. This calls
for additional investment.
(c) In the structure, there is no explicit measurement of the controlled variable. This is
disadvantageous since the key variable of interest is not measured and could deviate
from the set point without this inaccuracy in control being noticed.
(d) The control law does not consider the dynamics and hence after a disturbance the
corrective action may not be correctly synchronised with the disturbance effect on the
output. This dynamic mismatch will lead to control inaccuracies in the steady part of
the response.
There are some simplifications made to utilise some of the positive aspects of feedforward
control philosophy.
(a) Apply feedforward to tackle the effect of fast disturbances for which feedback control
may be wanting. In the present case q is the fast disturbance while Ti variations are
much slower. Thus only q is measured while Ti is unmeasured. Thus feedforward
control law is calculated using the measured value of q and a nominal value of Ti
namely Ti.
(b) Simplify the feedforward control law calculation using a linearised model instead of
the rigorous model.
For the heater a linearised model relating T, q and qh will be of the form
∆T = K q∆q + K p ∆qh
(2.31)
For ∆T to be zero in the presence of changes in ∆q,
Kq
∆qh = −
Kp
(2.32)
∆qh required is calculated from the above equation as
If the valve gain is Kv , then the feedforward signal to the control valve will be
(2.33)
Kq
∆p f = −
KpKq
(c) Add feedback control to account for uncertainty in the model resulting in inexact
feedforward control so that accuracy in control is maintained (Fig. 2.11).
Feedforward control provides fast but approximate corrective action while the feedback
controller will provide the slower but surer correction to bring the controlled variable to the
set point.
(d) For incorporating the effect of dynamics in the feedforward control law, it will be
necessary to study the dynamics of the process and the control valve.
T q
I
Feedforward
control law
pf
pb + PROCESS
Iset Control T
Controll IP HEATER
+ er Valve
+
-
IT
Measuring System &
Transmitter
2.6: SUMMARY
• In the feedback structure there is a continuous feedback of information about the process
behaviour enabling the controller to take appropriate corrective action
• The controller forms an important component of the control loop. The most widely used
form of the controller is the PID controller. Proportional controller alone introduces an
offset when there is a change in the load variable and the set point. Integral action
eliminates offset but is slower than Proportional action. Derivative action acts earlier than
either Proportional or Integral action after the onset of a disturbance but cannot exist on
its own.
• Steady state can determine the steady state the closed loop will attain after the onset of a
disturbance. However it does not address the problem of instability of the closed loop
under intense control action. Nor does it provide a clue as to how the manner the
controlled varies with time after the process is disturbed. Dynamic analysis of the closed
loop will be necessary to answer these problems.
• Feedforward control provides faster corrective action than feedback control especially
fast disturbances but requires exact process knowledge for perfect control. Approximate
versions of feedforward can be configured to give fast corrective action. Long-term
accuracy of control can be ensured by adding feedback to feedforward control.
Consideration of dynamics is again important for design of feedforward controllers.
This will be addressed in a later chapter.