Cascade

Control
14.1 n INTRODUCTION
Cascade control is one of the most successful methods for enhancing single-loop
control performance. It can dramatically improve the performance of control strate
gies, reducing both the maximum deviation and the integral error for disturbance
responses. Since the calculations required are simple, cascade control can be im
plemented with a wide variety of analog and digital equipment. This combination
of ease of implementation and potentially large control performance improvement
has led to the widespread application of cascade control for many decades. In
this chapter, cascade control is fully explained with special emphasis placed on
clear guidelines that, when followed, ensure that the cascade method is properly
designed and is employed only where appropriate.
As explained in the introduction to this part, single-loop enhancements take
advantage of extra information to improve on the performance of the PID feedback
control system. Cascade uses an additional measurement of a process variable to
assist in the control system. The selection of this extra measurement, which is
based on information about the most common disturbances and about the process
dynamic responses, is critical to the success of the cascade controller. Therefore,
insight into the process operation and dynamics is essential for proper cascade
control design.
The basic concepts of cascade control are presented in the next section. Sub
sequent sections provide concise explanations of the design criteria, performance
expectations, tuning methods, and implementation issues. All of the methods and

-i a

458

guidelines are presented for continuous systems but are applicable to digital con
trol systems. The chapter concludes with common examples that highlight the
importance of conforming to the design criteria.

CHAPTER 14
Cascade Control

14.2 m AN EXAMPLE OF CASCADE CONTROL

The best way to introduce cascade control is with reference to a simple process
example, which will be the stirred-tank heat exchanger shown in Figure 14.1.
The goal is to provide tight control of the exit temperature. The conventional
feedback controller, with integral mode, attempts to maintain the exit temperature
near its set point in response to all disturbances and ensures zero steady-state
offset for steplike disturbances. Suppose that one particularly frequent and large
disturbance is the heating oil pressure. When this pressure increases, the initial
response of the oil flow and the heat transferred is to increase. Ultimately, the
tank exit temperature increases, and the feedback controller reduces the control
valve opening to compensate for the increased pressure. While the effect of the
disturbance is ultimately compensated by the single-loop strategy, the response is
slow because the exit temperature must be disturbed before the feedback controller
can respond.
Cascade control design considers the likely disturbances and tailors the control
system to the disturbance(s) that strongly degrades performance. Cascade control
uses an additional, "secondary" measured process input variable that has the im
portant characteristic that it indicates the occurrence of the key disturbance. For
the stirred-tank heat exchanger, all measured variables are shown in Figure 14.1.
The secondary variable is selected to be the heating oil flow, because it responds
in a predictable way to the disturbances in the oil pressure, which is not measured
in this case. The control objective (tight control of the outlet temperature) and the
final element are unchanged.
The manner in which the additional measurement is used is shown in Figure
14.2. The control system employs two feedback controllers, both of which can
Feed

U
do

^ Product

(rc)

ob
nsm

& *

Heating oil

<hin

<>

Heating ^o
oil =

SP, from
operator
MV,

CV2

SP2=MV,

<FCK

FIGURE 14.1

Stirred-tank heat exchanger with single-loop

temperature control.

FIGURE 14.2

Stirred-tank heat exchanger with cascade control.

use the standard PID controller algorithm. The important feature in the cascade
structure is the way in which the controllers are connected. The output of the exit
temperature controller adjusts the set point of the flow controller in the cascade
structure; that is, the secondary controller set point is equal to the primary con
troller output. Thus, the secondary flow control loop is essentially the manipulated
variable for the primary temperature controller. The net feedback effect is the same
for single-loop or cascade control; in either case, the heating oil valve is adjusted
ultimately by the feedback. Therefore, the ability to control the exit temperature
has not been changed with cascade.
As described previously, the single-loop structure makes no correction for
the oil pressure disturbance until the tank exit temperature is upset. The cascade
structure makes a much faster correction, which provides better control perfor
mance. The reason for the better performance can be seen by analyzing the initial
response of the cascade system to an oil pressure increase. The valve position is
initially constant; therefore, the oil flow increases. The oil flow sensor quickly
detects the increased flow. Since the flow controller set point would be unchanged,
the controller would respond by closing the valve to return the flow to its desired
value. Because the sensor and valve constitute a very fast process, the flow con
troller can rapidly achieve its desired flow of oil. By responding quickly to the
pressure increase and compensating by closing the control valve, the secondary
controller corrects for the disturbance before the tank exit temperature is signifi
cantly affected by the disturbance. Typical dynamic responses of the single-loop
and cascade control systems are given in Figure 14.3a and b for a decrease in oil
pressure.
A few important features of the cascade structure should be emphasized. First,
the flow controller is much faster than the temperature controller. The improvement
results from the much shorter dead time in the secondary loop than in the original
single-loop system; as discussed in Chapter 13, shorter dead times improve single-

459
An Example of
Cascade Control

'- Temperature

_L_I

Time
ia)
FIGURE 14.3

Dynamic response of stirred-tank heat exchanger to a disturbance in oil pressure: (a) with single-loop control;
ib) with cascade control.

460 loop control. If the flow controller were not faster, the cascade design would have
w^$xs*w.*Km^v&M^\ no advantage. Second, the temperature controller with an integral mode remains
CHAPTER 14 in the design to ensure zero offset for all disturbance sources. The primary conCascade Control trailer is essential, because (1) the secondary variable may not totally eliminate
the effect of the disturbance, (2) other disturbances that are not affected by the
cascade will also occur, and (3) the ability to change the primary set point must
be retained. Remember that the secondary variable is selected for one (or a few)
common disturbances; in the example, a heat exchanger feed temperature distur
bance would affect the tank outlet temperature but does not influence the heating
oil measurement. Finally, the judicious selection of the secondary variable has
made the improvement possible without using a model of the effect of pressure on
exit temperature in the control calculation; the only models used were the process
models used to tune the two feedback controllers. As a result, cascade control
is not strongly sensitive to modelling errors, although large errors could lead to
oscillations or instability in one of the feedback controllers.
The two controllers in the cascade are referred to by various names. The three
pairs of names in the most commonly used terminology are presented as they
would be applied to the stirred-tank heat exchanger:

Te m p e r a t u r e F l o w
Primary Secondary
Outer
Inner
Master
Slave

At first encounter, it may seem improper to use two feedback controllers to

achieve one objective; however, the propriety of cascade control can be established
by analyzing the degrees of freedom of the system. For the heat exchanger in Figure
14.2, the material and energy balances were derived in Example 3.7 (for cooling)
and are repeated here for heating.
F0
dT
dt

Fi
(14.1)
aF^^
_
a F uh
Fh +
ZphCph
The heating flow is related to the valve position (v) according to the following
general equation:

Fh

CvvJP
V Ph

P]

(14.3)

where we assume that the pressures and the coefficient Cv are constant, although
they can be variables (see Chapter 16). The final equations are the two cascade
controllers:
FhsP = Kd \iTsp -T) + ^r~ f iTsp - T) dA + IFh (14, 4)

v = Kc2 [(Fsp - Fh) + ^- j (F,,sp - Fh) dA + Iv (14.5)

Variables: Fx, Fh, T, (F/,)sp, v DOF = 5-5 = 0

External variables: F0, To, 7/,jn, Tsp
Parameters: V, p, Cp, ph, Cph,a, b, Cv, P0, P\, Kc[, Tn,Kc2,
7/2 /fAi h

The number of degrees of freedom is equal to the number of variables minus

the number of equations; thus, the system is exactly specified when the primary
temperature controller set point has been defined. Note that the cascade secondary
controller was placed between the primary controller output and the valve, which
added one variable (F/,sp) and one equation (14.5).
14.3 m CASCADE DESIGN CRITERIA
The principles of cascade control have been introduced with respect to the example
stirred-tank heater. In Table 14.1, the design criteria are summarized in a concise
form so that they can be applied in general. Adherence to these criteria ensures that
cascade control is designed properly and used only where appropriate. The first
two items address the selection of cascade control. Naturally, only when singleloop control does not provide acceptable control performance is an enhancement
such as cascade control necessary. As described in Chapter 13, single-loop control
provides good performance when the dynamics are fast, the fraction dead time
is small, and disturbances are small and slow. Also, the second criterion requires
an acceptable measured secondary variable to be available or added at reason
able cost.
A potential secondary variable must satisfy three criteria. First, it must in
dicate the occurrence of an important disturbance; that is, the secondary variable
must respond in a predictable manner every time the disturbance occurs. Naturally,
the disturbance must be important (i.e., have a significant effect on the controlled
variable and occur frequently), or there would be no reason to attenuate its effect.
Second, the secondary variable must be influenced by the manipulated variable.
This causal relationship is required so that a secondary feedback control loop

TABLE 14.1

Cascade control design criteria

Cascade control is desired when
1. Single-loop control does not provide satisfactory control performance.
2. A measured secondary variable is available.
A secondary variable must satisfy the following criteria:
1. The secondary variable must indicate the occurrence of an important disturbance.
2. There must be a causal relationship between the manipulated and secondary variables.
3. The secondary variable dynamics must be faster than the primary variable dynamics.

461
Cascade Design
Criteria

462
CHAPTER 14
Cascade Control

functions properly. Finally, the dynamics between the final element and the sec
ondary must be much faster than the dynamics between the secondary variable and
the primary controlled variable. The secondary must be relatively quick so that it
can attenuate a disturbance before the disturbance affects the primary controlled
variable. A general guideline is that the secondary should be three times as fast
as the primary. This could be roughly interpreted as the secondary reaching its
steady state in one-third the time of the primary after an open-loop step change in
the manipulated variable. A more proper comparison is the critical frequency of
each loop; cascade is recommended when the critical frequency of the secondary
is at least three times that of the primary. Using critical frequencies accounts for
differences in the fraction dead time as well as the speed of response.
A cascade control strategy combines two feedback controllers, with the primary
controller's output serving as the secondary controller's set point. The design should
conform to the design criteria in Table 14.1, which provide a simple, step-by-step
procedure for selection.

14.4 CASCADE PERFORMANCE

In the introduction to this chapter, cascade was described as simple and effec
tive. The foregoing material has demonstrated how simple a cascade strategy is to
design. In this section, its effectiveness is shown by calculating its performance
using simulation and frequency response for a few cascade systems and comparing
with single-loop control performance on the same systems. Because the number of
parameters in a cascade systemprimary dynamics, secondary dynamics, distur
bance dynamicsmake general performance correlations intractable, this section
presents sample results for typical process dynamics. The general trends showed
by these results should be expected for most realistic processes.
The block diagram in Figure 14.4 presents the structure of a cascade control
system, which summarizes the flow of information and can be used to evaluate
important properties such as stability and frequency response. Transfer functions
can be derived from this block diagram for the relationships between the primary
controlled variable CVi is) and the secondary disturbance D2is), the primary dis
turbance D\ is), and the primary set point SPj (s), as follows:
CVi(j)
=
Gd2Gpds)
D2is) 1 + Gc2is)Gds)Gp2is)Gs2is) + Gcds)Gc2is)Gds)Gpds)Gp2is)Gsds)
(14.6)
CVifr) = Gdljs)[l + Gc2is)Gds)Gp2is)Gs2is)]
Dds) 1 + Gc2is)Gds)Gp2is)Gs2is) + Gclis)Gt2is)Gds)Gpiis)Gp2is)Gsds)
(14.7)
CVil?) = Gcds)Gc2is)Gds)Gp2is)Gpds)
SP,(j) 1 + Gc2is)Gds)Gp2is)Gs2is) + Gcds)Gc2is)Gds)Gpds)Gp2is)Gslis)
(14.8)

D,

>.

Secondary loop

Cascade Performance

Gais)
Exis) SP2(ar)
SP,(5)

E2is)

MV(i)

Celis)

Gc2(5)

Cv(*)

II-

G</,(s)

CV,(5)

CV2(j)
Gp2is) -<f>

G'pi

<i>

CVm2(s)

Gj2W

cvmlw

G,i(*)

FIGURE 14.4
Block diagram of cascade control.

As apparent from the introductory example, a key factor in cascade control

is the relative dynamic responses of the secondary and primary processes. Since
the main reason for cascade is secondary disturbances, the studies in this section
evaluate the responses to secondary disturbances: step, sine, and stochastic. For
these simulation studies, the models for the sensors GSiis) and valve Gds) were
taken to be unity, and the dynamics of the plant models and disturbance model are
given below, with all times scaled so that the process models have a common value
of the fraction dead time. The relative dynamics between the secondary and primary
are defined by a variable n, which will be allowed to vary in the following models:

Cascade System
Process
1.0e(--3/*)s

Control

Secondary: Gp2is) = 1 + (0.7/i|)*

\.0e(~03)s
Gp\ = 1 + 0.7*

Primary:

PI controller tuned accordingly

PI controller tuned accordingly

lkfrfa^*^^

Single-Loop System
Process
1 ,Oe~(0-3+03/'')5

Gp= [1 + (0.7/i7)*](1+0.7j)

463

Control
PI controller tuned accordingly

464
CHAPTER 14
Cascade Control

For All Cases

Disturbance:

1.0
Gd2is) = -1 + i0.1/r))s

Instrumentation:

Gsds) = Gs2is) = Gds) = \.0

Response to Step Disturbance in D2

In the first cascade system studied, a step disturbance was introduced in the sec
ondary loop, and no noise was added to the measurements, so that only the effect
of the cascade could be determined. Both primary and secondary controllers used
PI algorithms with conventional tuning. The control performance measure is the
integral of the absolute value of the error, IAE, of the primary controlled variable; it
is reported as a ratio of cascade to single-loop IAE to characterize the improvement
achieved through cascade. The resulting control performance is shown in Figure
14.5 as a function of the relative secondary/primary process dynamics, n. As ex
pected, the performance is very good when the secondary is fast. For example,
the integral error is reduced by 95 percent or more for cascade versus single-loop
control when the secondary is more than 20 times faster. This large ratio in primary
to secondary dynamics is typical when the secondary is a fast loop such as a flow
or pressure controller, which is often the case. However, many cascade control
systems cannot achieve such a remarkable improvement because the secondary
loop is not so fast, and some potential secondary loop dynamics are so slow as to
prohibit cascade control.
Sample dynamic responses from cascade control are shown in Figure 14.6a
and b for a step disturbance in the secondary loop, D2is) = \/s at time = 10.

10 15 20 25 30 35
Relative primary-to-secondary dynamics
TI

FIGURE 14.5
Relative performance (IAEcasc/IAEsi) of cascade and
single-loop control for a step disturbance in the secondary
loop.

40

465
Cascade Performance
Primary
variable,
CV,
Secondary
variable,
CV2
Manipulated J
variable,
MV

Primary
variable,
CV,
Secondary
variable,
CV2
Manipulated
variable,
MV

FIGURE 14.6

Performance of cascade control for a disturbance in the

secondary loop: (a) with /; = 10; ib) with ^ = 1.0. [Scales for the
plots: One tick (10%) is 0.15 for primary, 0.50 for secondary, and
2.5 for manipulated variable.]

The case with a very fast secondary demonstrates how quickly the secondary con
troller attenuates the effect of the disturbance. The case of a much slower secondary
shows much poorer performance, especially the highly oscillatory response. These
oscillations, which are more troublesome with the continuous disturbances experi
enced in industrial plants, usually prohibit the use of cascades with PID controllers
when n is less than about 3, although Figure 14.5 shows that some improvement
in performance may be possible. (See Chapter 19 for the use of predictive con
trollers in cascade control, which can increase the region of acceptable cascade
performance.)

466
CHAPTER 14
Cascade Control

Response to Stochastic Disturbance in D2

In a second study, the same process was considered with a stochastic disturbance,
which is more representative of the responses encountered in a continuously operat
ing plant. The block diagram and models were the same as for the step disturbance,
and the disturbance enters in the secondary loop. Again, the system was simulated
with single-loop PI control and cascade control tuning. The control performance
in Figure 14.7 is expressed as standard deviation from the set point

,A(_SP,-cv,y

0"sp =

N /=0

The standard deviation of the primary variable is plotted as a function of the

relative secondary/primary dynamics, n. Again, the faster the secondary, the better
the performance of the cascade. Dynamic responses for this system are given for
n = 10 in Figure 14.8a through c for open-loop, single-loop, and cascade control,
respectively. It is important to recognize that the results in Figure 14.7 are limited
to the specific process and disturbance studied; other disturbances, with different
frequency components, would give different results, although the general trend
would be unchanged.

Response to Sine Disturbance in D2

The third cascade study investigates the frequency response, which evaluates the
control performance of a cascade control system for a range of disturbance fre
quencies. As described in Chapter 13, the amplitude ratio gives the magnitude
of the variation in the controlled variable for a unit sine input; thus, the smaller
the amplitude ratio for a disturbance response, the better the control performance.

0.80

5.00

Relative primary-to-secondary dynamics

10.00

T|

FIGURE 14.7

Relative control performance of single-loop and cascade

(ocasc/osi) for a stochastic disturbance in the secondary loop.

The amplitude ratios for the cascade control system were calculated for a range of
frequencies using equation (14.6). Because of the complexity of the algebra, the
amplitude ratios were evaluated using a computer program similar to the one in
Table 13.1, and the results are plotted in Figure 14.9.
The smaller amplitude ratio for cascade clearly demonstrates the advantage
of cascade control, especially when the secondary process is much faster than the
primary (here, n = 10). The cascade system is very effective for slower disturbance
frequencies. Both systems have little deviation for very fast disturbances, because
the process attenuates these disturbances. Also, the effect of the resonant frequency,
which was discussed in Chapter 13, is attenuated but not eliminated by the cascade
system.
Finally, the performance of a cascade control strategy must be evaluated for
circumstances for which this enhancement was not specifically designedthat is,

467
Cascade Performance

CV, ^wftuftagw <m *!< **> **mV"WVfc >* + v****'"***** * **w*i <

CV,

'

'

'

i
140

Time
ic)
FIGURE 14.8

Dynamic responses for stochastic secondary disturbance with n = 10: (a) open-loop process; ib) single-loop
control; ic) cascade control.

468
CHAPTER 14
Cascade Control

10 e 1 | | llllll 1 I I llllll 1 I I llllll 1 I IIIII

1 io-' =
O
- ^ io-*
,3
-
>
U 10-3
o
2
o io-4
a
o.
a
< 10-5
1Q-6

"

10-2

'

"llll

io-i

llllll

ioo
101
Frequency (rad/time)

IIIII

102

FIGURE 14.9
Closed-loop frequency responses for single-loop (solid curve)
and cascade (dotted curve) control with ri = 10.

primary disturbances that do not directly affect the secondary variable and changes
to the primary set point. By analyzing the cascade block diagram, it is apparent
that the primary controller can respond to other types of disturbance in the cascade
design; the only difference is that it manipulates the secondary set point rather
than the valve directly. One would expect that the responses to unmeasured distur
bances and set points are not substantially changed. This is the case, with cascade
providing slightly better performance because it increases the critical frequency
of the secondary loop (Krishnaswamy et al., 1990). In conclusion,

Cascade control can substantially improve control performance for disturbances

entering the secondary loop and is recommended for use when the secondary loop
is much faster than the primary loop.

14.5 CONTROLLER ALGORITHM AND TUNING

Cascade control can use the standard feedback control PID algorithm; naturally,
the correct modes must be selected for each controller. The secondary must have
the proportional mode, but it does not require the integral mode, because the
overall control objective is to maintain the primary variable at its set point. How
ever, integral mode is often used in the secondary, for two reasons. First, since a
proportional-only controller results in offset, the secondary must have an integral
mode if it is to attenuate the effect of a disturbance completely, preventing the
disturbance from propagating to the primary. Second, the cascade is often oper
ated in a partial manner with the primary controller not in operation, for example,
when the primary sensor is not functioning or is being calibrated. A negative side
of including integral mode in the secondary controller is that it tends to induce os-

dilatory behavior in the cascade system, but the result is not significant when the
secondary is much faster than the primary. Studies have demonstrated the effec
tiveness of the integral mode in the secondary loop (Krishnaswamy et al., 1992).
The secondary may have derivative mode if required, but the fast secondary loop
almost never has a large enough fraction dead time to justify a derivative mode.
The modes of the primary controller are selected as for any feedback PID
controller. It is again emphasized that the integral mode is essential for zero offset
of the primary variable.
The cascade strategy is tuned in a sequential manner. The secondary controller
is tuned first, because the secondary affects the open-loop dynamics of the primary,
CV\is)/S?2is). During the first identification experiment (e.g., process reaction
curve), the primary controller is not in operation (i.e., the primary controller is
in manual or the cascade is "open"), which breaks the connection between the
primary and secondary controllers. The secondary is tuned in the conventional
manner as described in Chapter 9. This involves a plant experiment, initial tuning
calculation, and fine tuning based on a closed-loop dynamic response.
When the secondary has been satisfactorily tuned, the primary can be tuned.
The initial plant experiment perturbs the variable that the primary controller ad
justs; in this case, the secondary set point is perturbed in a step for the process
reaction curve. The calculation of the initial tuning constants and the fine tun
ing follow the conventional procedures. Naturally, the secondary must be tuned
satisfactorily before the primary can be tuned.

Tuning a cascade control system involves two steps; first, the secondary controller
is tuned; then, the primary controller is tuned. Conventional initial tuning guidelines
and fine-tuning heuristics apply.

1 4 . 6 I M P L E M E N TAT I O N I S S U E S

When properly displayed for the operator, cascade control is very easy to under
stand and to monitor. Since it uses standard PID control algorithms, the operator
displays do not have to be altered substantially. The secondary controller requires
one additional feature: a new status termed "cascade" in addition to automatic and
manual. When the status switch is in the cascade position (cascade closed), the
secondary set point is connected to the primary controller output; in this situation
the operator cannot adjust the secondary set point. When the status switch is in the
automatic or manual positions (cascade open), the secondary set point is provided
by the operator; in this situation the cascade is not functional.
Cascade control is shown in a very straightforward way in engineering draw
ings. Basically, each controller is drawn using the same symbols as a single-loop
controller, with the difference that the primary controller output is directed to the
secondary controller as shown in Figure 14.2. Often, the signal from the primary
controller output is annotated with "reset" or "SP" to indicate that it is adjusting
or resetting the secondary set point.
The calculations required for cascade control, basically a PID control algo
rithm, are very simple and can be executed by any commercial analog or digital

469
Implementation Issues

470

control system. Two special features contribute to the success of cascade. The
first is anti-reset windup. The potential exists for any controller in a cascade to

CHAPTER 14
Cascade Control

experience integral windup due to a limitation in the control loop. Analysis for the
secondary is the same as for a single-loop design; however, reset windup can occur
for one of several reasons in the primary controller. The primary controller output
can fail to move the valve because of limits on (1) the secondary set point, (2) the
secondary controller output, or (3) the valve (fully opened or closed). Thus, the po
tential for reaching limits and encountering reset windup, along with the need for
anti-reset windup, is much greater in cascade designs. Standard anti-reset windup
methods described in Chapter 12 provide satisfactory anti-reset windup protection.
The second feature is "bumpless" initialization. Note that changing the sec
ondary status switch to and from the cascade position could immediately change
the value of the secondary set point, which is not desired. The desired approach is
to recalculate the primary controller output to be equal to the secondary set point
on initialization. Many commercial controllers include calculations to ensure that
the secondary set point is not immediately changed (bumplessly transferred) when
the secondary mode switch is changed.
Digital control equipment can use the standard forms of the PID algorithm
presented in Chapter 11 for cascade control. In addition to the execution period of
each controller, the scheduling of the primary and secondary influences cascade
control performance. To reduce delays due to control processing, the secondary
should be scheduled to execute immediately after the primary. Naturally, it makes
no sense to execute the primary controller at a higher frequency (i.e., with a shorter

Feed

--,

Product

y_?
.*)
P

Heating oil
FIGURE 14.10
Control design with a switch to bypass
the cascade and convert to single-loop
control.

period) than the secondary, because the primary can affect the process (move the
valve) only when the secondary is executed.
The cascade control system uses more control equipmenttwo sensors and
controllersthan the equivalent single-loop system. Since the cascade requires all
of this equipment to function properly, its reliability can be expected to be lower
than the equivalent single-loop system, although the slightly lower reliability is not
usually a deterrent to the use of cascade. If feedback control must be maintained
when the secondary sensor or controller is not functioning, the flexibility to bypass
the secondary and have the primary output directly to the valve can be included
in the design. This option is shown in Figure 14.10, where the positions of both
switches are coordinated.
Since the cascade involves more equipment, it costs slightly more than the
single-loop system. The increased costs include a field sensor and transmission
to the control house (if the variable were not already available for monitoring
purposes), a controller (whose cost may be essentially zero if a digital system with
spare capacity is used), and costs for installation and documentation. These costs
are not usually significant compared to the benefits achieved through a properly
designed cascade control strategy.

Cascade control, where applicable, provides a simple method for substantial im

provements in control performance. The additional costs and slighdy lower reliabil
ity are not normally deterrents to implementing cascade control.

14.7 m FURTHER CASCADE EXAMPLES

471

The concept of cascade control is consolidated and a few new features are presented
through further examples in this section.
EXAMPLE 14.1. Packed-bed reactor.
The first example is the packed-bed reactor shown in Figure 14.11. The goal is to
tightly control the exit concentration measured by AC-1. Suppose that the singleloop controller does not provide adequate control performance and that the most
significant disturbance is the heating medium temperature, T2. The goal is to de
sign a cascade control strategy for this process using the sensors and manipulated
variables given. The reader is encouraged to design a cascade control system
before reading further.
Since we are dealing with a cascade control strategy, the key decision is
the selection of the secondary variable. Therefore, the first step is to evaluate the
potential measured variables using the design criteria in Table 14.1; the results of
this evaluation are summarized in Table 14.2, with Y(N) indicating that the item is
(is not) satisfied. Since all of the criteria must be satisfied for a variable to be used
as a secondary, only the reactor inlet temperature, T3, is a satisfactory secondary
variable.
The resulting cascade control strategy is shown in Figure 14.12. Given the
cascade design, an interesting and important question is, "How well does it re
spond to other disturbances for which it was not specifically designed?" Several
disturbances are discussed qualitatively in the following paragraphs.

Further Cascade
Examples

FIGURE 14.11
Single-loop packed-bed reactor control.

Feed temperature, Tf. A change in the feed temperature affects the outlet
concentration through its influence on the reactor inlet temperature, T3. Therefore,
the cascade controller is effective in attenuating the feed temperature disturbance.
Heating oil pressure (not measured). A change in the oil pressure influences
the oil flow and, therefore, the heat transferred. As a result the reactor inlet tem
perature, T3, is affected. Again, the cascade controller is effective in attenuating
the oil pressure disturbance.
Feed flow rate, F1. A change in the feed flow rate influences the reactor outlet
concentration in two ways: it changes the inlet temperature T3, and it changes
the residence time in the reactor. The cascade controller is effective in attenuating
the effect of the disturbance on T3 but is not effective in compensating for the
residence time change. The residence time effect must be compensated by the
primary controller, AC-1.
TABLE 14.2

Evaluation of potential secondary variables

A2

F1

F2

T1

T2

1. Single-loop control is not satisfactory Y

2 . Va r i a b l e i s m e a s u r e d Y
3. Indicates a key disturbance N
4 . I n fl u e n c e d b y M V N
5. Secondary dynamics faster
N/A

Y
Y
N
N
N/A

Y
Y
N
Y
Y

Y
Y
N
N
N/A

Y
Y
Y
N
N/A

Criterion

T3

FIGURE 14.12
Cascade packed-bed reactor control.

472
CHAPTER 14
Cascade Control

Feed composition, A2. A change in the feed composition clearly changes the
reactor outlet concentration. The cascade has no effect on the feed composition
disturbance, because the composition does not influence T3. Therefore, this dis
turbance must be totally compensated by the primary feedback controller, AC-1.

A single cascade control system can be effective in compensating for the

effects of several disturbances, and given several possible secondary vari
ables, the one that attenuates the most important disturbances is the best
choice.

Conclusions and tuning. In some cases, the attenuation is complete (at least
in the steady-state sense); in other cases, the attenuation is partial. Thus, a welldesigned cascade strategy can produce a major improvement in the control sys
tem performance. This example is completed by describing the tuning procedure.
1. Both controllers are placed in the manual mode, and a process reaction curve
experiment is performed to obtain a model for tuning the secondary controller.
The model parameters and tuning based on Figure 9.9a and b are

Model relating valve to T3 T3 controller

Kp2 = 0.57C/%
62 = 8 s
r2 = 20 S

Kc2 = 2.4%/C
TI2 = 23 s

2. The tuning constants are entered into the secondary controller, which is finetuned by placing it in automatic and entering small set point changes.
3. A process reaction curve experiment is performed to obtain the model for
tuning the primary controller. The model and parameters are
Model relating T3 set point to A} A, controller
KpX =-0.19mole/m3/C
0i = 20 s
t, = 50 s

Kc\ = -7.9C/mole/m3

(changed to -3.7 in fine tuning)

Tn = 54 s (changed to 70 in fine tuning)

4. The tuning constants are entered into the primary controller, and the controller
is fine-tuned by placing it in automatic and entering small set point changes.
Note that the primary loop was somewhat oscillatory, so that the primary con
troller gain and time constant were modified as noted in the foregoing list.
5. The response to a disturbance is observed and further fine tuning is applied
to improve the response, if necessary. A dynamic response to a disturbance

in the secondary loop is shown in Figure 14.13: D2is) = -1.8/^; Gd2is) =

1/(1+20j).
Finally, we note that the secondary variable is sensed at the outlet of the
heat exchanger. This contributes to its effectiveness, because it can sense the
influence of many inlet temperature and flow disturbances. However, the heat
exchanger dynamics make the secondary dynamic response somewhat slow. One
improvement in the design is to add another level of cascade to compensate for
the oil pressure disturbance, which can be sensed by the flow sensor F2. The
three-level cascade is shown in Figure 14.14. Industrial designs with three to four
cascade levels (and sometimes more) are not unusual and function well.

473
Further Cascade
Examples

0.05 -

o
U
-0.05
Time

There is no theoretical or practical limit to the number of cascade levels used as long
as each level conforms to the design criteria in Table 14.1.

Cascade control can be applied to a variety of processes. A few more examples

are presented briefly to demonstrate the diversity of the cascade approach. In each p
case, an analysis similar to the method shown in Table 14.2 was performed to
design
the
cascade
s t r a t e g y.
-1
EXAMPLE 14.2. Fired heater.
Another typical cascade design is given in Figure 14.15 for furnace control. A
single-loop temperature controller would adjust the fuel valve directly, making the
fuel flow subject to pressure disturbances. A cascade control strategy is possible
that satisfies all of the design criteria. In the cascade, the outlet temperature of the
fluid in the coil is controlled tightly by adjusting the fuel flow controller set point,

Time

Flue gas

Time
Feed

FIGURE 14.13
Dynamic response of the reactor cascade
controller to a disturbance in the heating
medium temperature in Example 14.1.

I-*-

Fuel oil

FIGURE 14.14

FIGURE 14.15

Three-level cascade control for

packed-bed reactor.

Cascade control design for outlet

temperature.

which adjusts the valve position. An additional advantage of the cascade becomes
apparent when considering the performance of many real control valves; the valve
does not always move exactly the amount directed by the controller, because
friction occasionally causes sticking, which degrades control performance. The
cascade design with a flow controller as its fast secondary corrects quickly for both
fuel pressure disturbances and the effects of a sticking valve and substantially
improves control performance over the single-loop strategy.

474
CHAPTER 14
Cascade Control

Remote location

"

S,t /,!

Positioner located at valve

FIGURE 14.16
Schematic of a valve
positioner.

Superheated
steam

Flue gas

;---<)
&

-T^] Treated
water

EXAMPLE 14.3. Valve positioner.

Cascade control principles are used to enhance the performance of control valves
when fast secondary variables cannot be included in the design. Because of
a difference between static and dynamic friction, valves often stick and do not
exactly achieve the percent stem position demanded by the controller output. The
result is that the valve may remain stationary and then "jump" to a value beyond
necessary to bring the controlled variable to its set point. A standard control valve
can have a dead zone of up to 3 percent (Buckley, 1970), which can lead to poor
control and cycling in many control systems. If the valve is being adjusted by a
fast control loop (and the process is not sensitive to high-frequency cycling), no
corrective action may be necessary; however, valve sticking can lead to severe
control performance degradation.
The effects of valve sticking can be reduced by including a cascade con
troller called a valve positioner, which is included as part of the valve equipment
as shown in Figure 14.16. The primary controller, which can be controlling any
measured process variable, sends a signal, which is interpreted as the desired
valve stem position (0-100%), to the valve positioner. The positioner senses the ac
tual valve stem position and adjusts the air pressure until the desired stem position
is (nearly) achieved. Since a valve positioner uses a proportional-only algorithm,
it does not give perfect compensation for sticking, but the fast dynamics allow
a very high controller gain, which reduces the dead zone to about one-tenth of
that experienced without a positioner. It is worth noting that this improvement is
achieved with minimal investment.
Other advantages are provided by valve positioners, such as faster valve
dynamics and overcoming large pressure drops. They are also used when split
range control (see Chapter 22) or changes in the valve characteristic (see Chapter
16) are required. There is no consensus concerning the application of positioners
on fast loops; some practitioners recommend them on essentially all control valves,
whereas others recommend them only on slow loops.

L_

Treated
water

Fuel

FIGURE 14.17
Cascade control design for boiler
superheated steam temperature control.

EXAMPLE 14.4. Steam superheater.

Industrial processes consume large quantities of steam for heating, and machin
ery consumers require superheated steam for power. As shown in Figure 14.17,
steam is generated by vaporizing water in a boiler where the heat transfer is by
radiation. The saturated steam temperature is then raised further by convecrive
heat exchange with the hot combustion flue gases. The final steam temperature
is controlled by injecting water in the steam. The primary temperature controller
could adjust the spray water valve directly, but the control performance would not
be good, because of the long dynamic response to disturbances in water flow

and heat transfer. The control performance is good for the cascade control with a
secondary that responds quickly to both types of disturbances.

475
Further Cascade
Examples

EXAMPLE 14.5. Heater exchanger.

A process stream can be cooled by exchanging heat with a refrigerant, which is
vaporized in the exchanger. An example is shown schematically in Figure 14.18,
which shows that the rate of heat transfer can be controlled by adjusting the heat
transfer area (i.e., the liquid refrigerant level in the exchanger). This operating
policy is implemented by the cascade control strategy in the figure, where the
secondary controller responds quickly to disturbances in liquid flow resulting from
pressure variations. An additional advantage is that the level controller maintains
the liquid level within acceptable limits; in contrast, a single-loop temperature
controller directly adjusting the valve might cause liquid refrigerant to carry over
and damage downstream equipment.

EXAMPLE 14.6. Pressure control.

Normally, pressure control involves fast process responses and does not require
cascade control. However, processes sometimes have large, integrated systems
with a single valve regulating the pressure. An example is shown in Figure 14.19,
where the most important pressure is at the initial unit, and the pressure control
valve is located far downstream. In this case, pressure disturbances near the
control valve can cause a relatively large, prolonged disturbance in the initial unit's
pressure. The cascade strategy shown in the figure rapidly senses and corrects
for downstream disturbances before they upset the integrated upstream unit.

EXAMPLE 14.7. Jacketed CSTR.

An often-noted example of cascade control is the jacketed continuous-flow stirredtank reactor shown in Figure 14.20. The dynamics related to the thermal capac
itance of the jacket fluid and metal could lead to poor control performance with

Process
outlet
Vapor
refrigerant

Process
inlet

js*_ Liquid
refrigerant

Pressure
disturbances

FIGURE 14.18

FIGURE 14.19

Cascade control design for temperature

control.

Cascade control design for pressure

control.

476
CHAPTER 14
Cascade Control

Coolant

t&r
Product

FIGURE 14.20
Cascade control for a stirred-tank reactor with
cooling jacket

University
president

Engineering
dean

Strive for better

education

14.8 CASCADE CONTROL INTERPRETED

AS DISTRIBUTED DECISION MAKING

Cover more
material

Give more
Process
control
professor

single-loop control from the reactor temperature directly to the valve that controls
the inlet temperature of the jacket fluid. A cascade controller uses a secondary
variable to sense and quickly correct a disturbance in the jacket fluid inlet tem
perature.

thought-provoking,
open-ended
assignments

Feedback

Engineering
students

FIGURE 14.21
A cascade process at a university.

Cascade control is essentially a way to delegate decision making to the lowest

level possible. In the packed bed reactor (Figure 14.12), the outlet analyzer AC-1
determines the desired value for the inlet temperature TC-3. The inlet temperature
controller is free to determine the valve position that is required to achieve the
desired value of TC-3. Since the T3 measurement provides rapid indication of
every effect on the reactor inlet temperature, it can achieve the required inlet
temperature control better than AC-1.
Cascade control concepts are not limited to engineering control systems. So
cial and business organizations also benefit from distributed decision making. A
hypothetical example of university decision making is given in Figure 14.21. The
president of a university decides to improve the education of the engineering stu
dents. Rather than tell each student how and what to learn, the president informs
the Dean of Engineering. The Dean of Engineering gives directions to the Process
Control professor, who finally gives directions to the students. The students then
implement the decision by adjusting their studying to satisfy the requirements set
by the professor. This distribution allows quick response to disturbances, such as
competing demands of other courses, and provides frequent feedback from class
discussion and course quizzes. The distributed system surely functions better than
the single-loop feedback approach in which the president would obtain feedback
every few years and then give directions to every student in the university. It also
might clarify why the secondary controller is sometimes called the "slave"!

14.9

CONCLUSIONS

In this chapter, the principles of cascade control have been presented, and the
excellent performance of cascade control for disturbance rejection has been es
tablished. Cascade control employs the principle of feedback control, since the
secondary variable is a process output that depends on the manipulated variable in
a causal manner. Cascade control can improve performance when the dynamics,
mainly the dead time, of the secondary loop is much shorter than in the primary. In
this situation, some disturbances can be measured and compensated quickly. As
shown in Figure 14.3a and b, this improved performance of the controlled vari
able is achieved without significantly increasing the variability of the manipulated
variable. Based on this performance improvement and simplicity of implementa
tion, the engineer is well advised to evaluate cascade control as the first potential
single-loop enhancement.
The first few times new control engineers evaluate cascades, they should per
form a careful study like the one in Table 14.2, but after some experience they
will be able to design cascade controls quickly by applying the design principles
without explicitly writing the criteria and table.
However, cascade control is not universally applicable; the design criteria in
Table 14.1 can be used to determine whether cascade is appropriate and if so, select
the best secondary variable. If it is not immediately possible and a significant
improvement in control performance is desired, the engineer should investigate
the possibility of adding the necessary secondary sensor. Even with improved
sensors, cascade is not always possible; for example, a causal relationship between
the manipulated variable and a measurement indicating the disturbance may not
exist. Thus, while cascade is usually the preferred choice for enhancing control
performance, further enhancement approaches are often required, and some of
these are introduced in subsequent chapters.

REFERENCES
Buckley, P., "A Control Engineer Looks at Control Valves," in Lovett, O.
(ed.), Paper 2.1, Final Control Elements, First ISA Final Control Elements
Symposium, May 14-16, 1970.
Krishnaswamy, P., and G. Rangaiah, "Role of Secondary Integral Action in
Cascade Control," Trans. IChemE, 70, 149-152 (1992).
Krishnaswamy, P., G. Rangaiah, R. Jha, and P. Despande, "When to Use
Cascade Control," IEC Res., 29, 2163-2166 (1990).
Verhaegen, S., "When to Use Cascade Control," In. Tech., 38-40 (October
1991).

ADDITIONAL RESOURCES
Cascade control has been used for many decades and seems to have been developed
by industrial practitioners, who often do not publish their results. Therefore, the
inventor(s) of cascade are not known. A few of the earlier papers are listed here.
Franks, R., and C. Worley, "Quantitative Analysis of Cascade Control," IEC,
48,1074(1956).

477
Additional Resources

4 7 8 We b b , P. , " R e d u c i n g P r o c e s s D i s t u r b a n c e s w i t h C a s c a d e C o n t r o l , " C o n t r o l
CHAPTER 14
Cascade Control

QUESTIONS
14.1. id) In your own words, discuss each of the cascade design criteria. Give a
process example in which cascade control is appropriate.
ib) Identify the elements of the cascade block diagram in Figure 14.4 that
are process, instrumentation, and control calculations.
ic) Discuss the topics in Table 13.3 that are influenced by cascade control,
explaining how cascade improves performance for each.
id) For the mixing system in Figure 13.4 and a disturbance in the feed con
centration, discuss how you would add one sensor to improve control
performance through cascade control.
14.2. Derive the transfer function in equations (14.6) to (14.8) based on the block
diagram in Figure 14.4.
14.3. Figure 14.9 presents the frequency responses for single-loop and cascade
control with a disturbance in the secondary loop.
id) Sketch the general shapes and discuss the frequency responses for
cascade and single-loop control for (1) a set point change and (2) a
disturbance in the primary loop.
ib) Calculate the frequency responses for (a), with Gd\ is) = Gp\ is) and
n = 10.
14.4. Based on the transfer function in equation (14.6), mathematically demon
strate the assertion made in this chapter that cascade control performance
for a secondary disturbance improves as the secondary becomes faster. For
this question, assume that the secondary controller is a PI. (Hint: Evaluate
the amplitude ratio as a function of frequency.)
14.5. Answer the following questions based on the transfer function in equation
(14.6).
id) Mathematically demonstrate the assertion made in this chapter that
integral mode in the secondary controller is not required for zero offset
in the primary.
ib) Demonstrate the assertion that the secondary controller must be tuned
before the primary is tuned.
ic) This question addresses why the integral mode is often included for
the secondary controller. Consider the secondary controller and its
initial response to a disturbance before any feedback from the primary.
Calculate the amount that a P-only and a PI controller attenuate the
same disturbance at the limit of low frequency (i.e., at steady state).
Based on this analysis, which controller is more effective in attenuating
a disturbance?
id) When the secondary has an integral mode, the integral error of the
primary is zero for a step disturbance. The oscillatory effect of the
integral error being zero is apparent in Figure 14.6. For the transfer
function in equation (14.6), prove this statement. You may use the

following relationship (see Appendix D, equation (D.4)):

Jo

Edt'W = Eds)\s=0

14.6. Discuss the proposed cascade control designs. In particular, apply the cas
cade control criteria to each proposed design, and estimate whether the
cascade design would provide better performance than single-loop for dis
turbance response. Consider each of the disturbances separately. To assist
in the analysis, prepare a block diagram for each process showing the ap
propriate cascade control systems.
ia) Jacketed stirred-tank reactor in Figure 14.20; disturbances are (1)
coolant pressure, (2) coolant temperature, (3) recycle pump outlet pres
sure, (4) reaction rate (e.g., feed concentration), and (5) feed flow rate.
ib) Furnace coil outlet temperature control in Figure 14.15; disturbances
are (1) fuel pressure, (2) fuel density (composition), (3) valve sticking,
and (4) feed temperature.
ic) Repeat ib) with the temperature controller cascaded to a valve posi
tioner, without the flow controller.
id) Figure Ql .9a modified for cascade control with a level to flow to valve
control structure; disturbances are (1) pump outlet pressure and (2)
second outflow valve percent open.
ie) Analyzer to reboiler flow cascade for distillation in Figure Q14.6;
disturbances are (1) heating medium temperature, (2) feed temperature,
(3) tower pressure, and (4) heating medium downstream pressure.
14.7. Assume that the dynamic behavior in Figure 14.8a through c are for a fired
heater in Example 2.1 that has the goal of maximizing temperature without
exceeding a maximum constraint of 864C; a performance correlation is
provided in the example. The primary variable is temperature, which is
plotted as the top variable in Figure 14.8a through c. Assume that 10 percent
of the scale represents 5C and that the top of the scale is 864C.
ia) Estimate the benefit for (1) single-loop and (2) cascade control due
to the reduction in the variability in the temperature using the data in
Figure 14.8a through c.
ib) Suggest changes to the operating conditions (set points) for both con
trol designs and repeat the estimation done in id).
14.8. The cascade control design shown in Figure 14.12 should have anti-reset
windup protection.
ia) Discuss the potential causes for integral windup in this strategy.
ib) Assume that the feedback algorithms are of the form that use external
feedback, as described in Chapter 12. Which signal should be used for
external feedback for each controller? Explain your answer.
(c) In Chapter 11, the use of digital PI algorithms was explained. Discuss
the performance of the incremental (velocity) algorithm as the primary
temperature controller when the valve reaches its maximum or mini
mum position. Does the velocity form of the PI controller satisfactorily
prevent reset windup?

479
Questions

FIGURE Q14.6

480
CHAPTER 14
Cascade Control

14.9. The initial design for the packed-bed reactor, Figure 14.11, included a con
troller for the outlet concentration and a single control valve. The cascade
control design added a controller but did not change the number of control
valves. This might seem to violate the degrees of freedom of the process.
id) In your own words, discuss why the cascade is possible.
ib) Perform a degrees-of-freedom analysis to demonstrate that the cascade
control is possible. You may use the transfer function model of the
reactor for this analysis.
14.10. For the following control designs determine which valves should have
positioners and explain why: id) Figure 13.19, ib) Figure 2.2, (c) Figure
14.1, and id) Figure 14.2.
14.11. A stirred-tank chemical reactor is shown in Figure Q14.11 with the follow
ing reaction.
A -+ B AH = 0
CA
l+kCx
The available sensors and control valves are shown in the figure, and no
changes to these are allowed. The goal is to control the reactor concentration
of A tightly by single-loop or cascade control, whichever is better. For each
of the following disturbances, design the best control system and explain
your design: (a) the heating medium pressure (Pi), ib) the solvent feed
temperature iTs), ic) the reactant feed pressure iP2), and id) the inhibitor
concentration iCx) which enters in the solvent. (To assist in the analysis,
prepare a block diagram for each case showing the appropriate single-loop
or cascade control system.)
rA = -k0e-ElRT

Reactant

FIGUREQ14.il

14.12. A set of cascade design criteria are presented in an article by Verhaegen

(1991). Discuss the similarities and differences between the criteria in this
chapter and in the article.

14.13. The chemical reactor with separate solvent and reactant feed flows in Fig
ure Q14.13 has the following properties: well-mixed, isothermal, constant
volume, constant density, and Fs % FA. The chemical reaction occurring
is A - B with the reaction rate, rA = kCA. The concentrations of the
reactant and the product can be measured without delay. Follow the steps in
this question to evaluate two possible cascade control designs to select the
best for controlling the concentration of product B in the effluent, measured
byA2.
id) The total feed flow (F1) and the feed concentration (A 1) are the poten
tial secondary variables for the reactor effluent primary composition
control. Construct a control scheme and sketch it on the figure that will
control these two variables (Fl, Al) simultaneously to independent
set point values. You must add two feedback controllers. If needed,
you may move the locations of the sensors and valves.
ib) Derive the dynamic model Cnis)/CAois) [or A2(.s)/A1(j)] under the
control in ia), i.e., with Fl constant. Analyze the model regarding (i)
order, (ii) stability, (iii) periodicity, and (iv) step response characteris
tics.
(c) Derive the dynamic model for C$is)/Fis) [or A2is)/F 1 is)] under the
control in ia), i.e., with Al constant. Analyze the model regarding (i)
order, (ii) stability, (iii) periodicity, and (iv) step response characteris
tics.
id) The results in ib) and (c) provide the dynamics of the primary feedback
process for the two designs. Based on these results, select which of the
secondary variables would provide the best feedback control for a set
point change in A2. Your answer should be either A2 - Al or A2 -
Fl. Sketch the feedback (cascade) structure on the figure prepared in
ia).

H&h

ir(j) Q)

'AO

u (_

- &
FIGURE Q14.13

481
Questions

482

ie) Compare with the solution in Example 13.8 and discuss whether your
design is expected to provide better or worse control performance for
A2.

CHAPTER 14
Cascade Control

14.14. In the packed-bed reactor example (14.1) an alternative approach would

be possible. In the alternative approach, the oil temperature (T2) would be
measured and the effect on the outlet analyzer due to changes in oil tem
perature could be calculated and used to adjust the valve. Which design
cascade control or the alternative described herewould you prefer? Dis
cuss why.

Sources

Sinks

^
-A-^
Liquid
supply

FIGURE Q14.16

L-C&}- He*ing
medium

14.15. Prepare a digital computer program to perform the control calculations

for the cascade control system in Figure 14.12. Include initialization, reset
windup, and other factors required for good implementation.
14.16. A vaporizer process is shown in Figure Q14.16. The gas pipe (header)
has several sources and sinks of gas, and the pressure in the pipe is to be
controlled by adjusting the amount vaporized.
id) A cascade control design has been suggested from the pressure to the
flow of vapor to the heating medium valve. Evaluate this design using
the cascade design criteria. Correct it if necessary.
ib) Discuss the response of this cascade design to a disturbance in the
heating medium pressure, upstream of the control valve.
ic) Discuss the response of this cascade design to a disturbance in one of
the source or sink flows.
id) Discuss the response of this cascade design to a disturbance in the
liquid temperature.
ie) Generalize the results from ib) through id) and develop a further cas
cade design criterion to be added to those in Table 14.1.