PC01K Control Concept

Cascade Control

© 2014 All rights reserved. This document is restricted. Neither the whole nor any part of this publication may be
reproduced, stored in any retrieval system or transmitted in any form or by any means (electronic, mechanical,
reprographic, recording or otherwise) without prior written consent of the copyright owner.

All information or advice provided as part of this course is intended to be general in nature. You should not rely on it
in connection with the making of any decision and the inclusion of information in this course does not guarantee its
accuracy. No warranty is offered regarding this course and no liability is accepted for any action you may take as a
result of relying on such information or advice or for any loss or damage suffered by you as a result of you taking this
action.
PC01K Control Concept: Cascade Control

Revision History

Date Issue Reason for Change Author Approved By

May 28, 2014 1.0 First Draft C.T. Seppala n/a

July 7, 2014 1.1 Add Level to Valve Example C.T. Seppala n/a
Fix incorrect statement in
Sept. 4, 2014 1.2 Secondary on Header Section C.T. Seppala
from D.O.W. feedback
Sept. 23, Incorporate reviewer feedback
1.3 C.T. Seppala
2014 and add Summary section

2 Restricted © Shell International BV 2014

PC01K Control Concept: Cascade Control

TABLE OF CONTENTS

1. Introduction 4
2. Examples of Cascade Control Usage 5
2.1 Secondary on a Header 5
2.2 Gain Linerisation 6
2.3 Example 8
2.4 Fast Primary Process 9
3. Configuration Details 9
3.1 PV Tracking and Initialisation 9
3.2 Outerloop Dynamic Treatment 10
3.3 Cascade Bypass 11
4. Summary 11

3 Restricted © Shell International BV 2014

PC01K Control Concept: Cascade Control

1. Introduction
This module briefly reviews the cascade control concepts introduced in PC-01-A, with additional material
on configuration and assessing scenarios where cascade control might be beneficial.

Recall from Module 5 of PC-01-A that cascade control structures consist of two connected PID loops. The
output of the primary loop sets the setpoint of the inner loop, and the output of the inner loop writes to a
final control element. A control block diagram of a cascade control loop is shown below in Figure 1. The
two loops are connected by the physical system and by control connections.

Primary or Outer Loop

Secondary or Inner Loop

Primary Secondary
SP SP
Primary Secondary Secondary Primary
- Controller - Controller Process Process
Final
Control
Primary Secondary Element
PV PV Sensor- Sensor-
Transmitter Transmitter

Figure 1: Block diagram of a cascade control loop

The secondary loop is assumed to be controlling a physical process that responds much faster than the
primary physical process. Often, the ‘fast’ inner loop is a flow or pressure controller capable of reaching
steady state within approximately one minute in open loop mode. Cascade loops add value when they
can be used to exploit this difference in dynamics, and thereby reduce the effect of disturbances and
nonlinear behaviour in the secondary loop before they have an impact on the primary loop.

The cascade control example from PC-01-A was a process heater with a primary loop for the heater
outlet temperature control and a secondary loop for the fuel gas flow control. A diagram of the heater
and the TC to FC cascade loop is shown below in Figure 1.

OIL IN OIL OUT

TC Primary or
Outer Loop

Secondary or
AIR Inner Loop
FC

FUEL GAS
FV01

Figure 2: Outlet temperature to fuel gas flow cascade on a process heater

4 Restricted © Shell International BV 2014

PC01K Control Concept: Cascade Control

The open loop time constant of the temperature response was assumed to be three minutes, while the
open loop time constant of the fuel gas response was assumed to be thirty seconds or less. With this
difference in dynamics, it was argued that the temperature controller performance would benefit from a
cascade arrangement when disturbances that directly affected the fuel gas flow occurred (supply pressure
changes, number of burners in operation, valve non-linearity). It was also noted that the cascade structure
did not offer any benefit for disturbances that did not affect the fuel gas flow (such as feed flow or feed
temperature disturbances).

In the remainder of this module, some common examples where cascade control should and should not
be used will be provided and some relevant configuration topics will be covered.

2. Examples of Cascade Control Usage

2.1 Secondary on a Header

One common potential application for cascade control occurs when the secondary controller is on a line
or header that is prone to disturbances including those which may arise due to the action of other
feedback controllers. A simple example of such a scenario is depicted in Figure 3. This process features a
vessel with two outflows: one that is routed to storage and one that is routed to a downstream unit, ‘Unit
B’.

LI LC
01 01

FI FC
01 01

STORAGE

FI FC
02 02

UNIT B

Figure 3: Example of a cascade loop with the secondary on a header

The secondary controller in the level to flow cascade is on the line to storage, while a second, stand-alone
flow controller is on the line that feeds Unit B. The Unit B charge is not on level control because it is
desired to maintain a constant charge rate to the downstream unit.

5 Restricted © Shell International BV 2014

PC01K Control Concept: Cascade Control

Cascade control is beneficial in this type of scenario because whenever a rate change is made on FC02,
the upstream pressure will change causing a disturbance in FI01. With a level to flow cascade loop in
place, the valve position of FV01 will adjust quickly to compensate for the change in upstream pressure to
keep FC01.PV close to its set value. The level controller ultimately needs to set a new setpoint for FC01
that compensates for the change in FC02, but the control performance will benefit from the fact that
LC01.OP and FI01.PV are kept close together by the action of the secondary. If a change in FC02 occurs
with no flow secondary in place, there is a temporary mismatch between the flow implied by the level
controller output and the actual flow FI01. This mismatch results in an additional level disturbance that
needs to be corrected by feedback which slows down the overall speed of response.

The concept illustrated here can be extended to other situations where multiple additional controllers are
competing with the cascade secondary.

2.2 Gain Linerisation

Cascade control can be used to linearise the gain between a final control element and a primary control
objective. What does this mean and how does it work? To illustrate this concept, imagine two identical
level control applications: One that is configured as level to valve (left image in Figure 4) and one that is
configured as a level to flow cascade (right image in Figure 4).

LI LC LI LC
01 01 01 01

FI FI FC
01 01 01

OUTFLOW OUTFLOW

Level Controller Direct to Valve Level to Flow Cascade

Figure 4: A level control application configured as level direct to valve (left) and as a level to flow cascade (right).

6 Restricted © Shell International BV 2014

PC01K Control Concept: Cascade Control

Now, suppose that the steady state gain curve was known for the valve position versus flow relationship.
This relationship could be estimated by testing or taken from valve manufacturer data. Imagine two
separate scenarios: one where the valve position versus flow relationship is linear, and one where the
relationship is non-linear, as depicted by the blue and green curves, respectively, in Figure 5.

FC.PV or FI.PV, [EU]

0 10 20 30 40 50 60 70 80 90 100

Valve Position, [%]

Figure 5: Linear (blue) and non-linear (green) flow versus valve position relationships

The important observation to make in Figure 5 is that in the linear case, a given change in valve position
will give roughly the same change in flow regardless of the absolute valve position. In contrast, the
change in flow that results from a given change in valve position for the nonlinear case is highly
dependent on the absolute value of the valve position.

If the valve position versus flow relationship is non-linear and a cascade control scheme were to be
applied, then the non-linearity would not go away but it would be handled by the much faster flow
controller. In the process of testing and tuning the flow controller, the non-linearity in the valve position
versus flow relationship would be evident to the engineer. The engineer would handle these non-linear
effects using any number of approaches, but in the end, any mismatch between the target flow and the
actual flow would be handled through the much faster flow feedback mechanism. Once the flow controller
is commissioned, the engineer would move on to test and tune the level controller by making changes to
the flow controller setpoint or level controller output. Now, in terms of the response time of the level
process, the non-linearity is barely evident as the flow controller quickly moves the valve position the
amount required to deliver the flow requested by the level controller. The relationship between the level
process variable and the final control element has been linearised by the addition of the secondary flow
controller.

Finally, if the flow versus valve position relationship is linear, then there is no case for adding cascade
control on the grounds of correcting for non-linearity between the final control element and the primary
control objective (note that there may be other reasons for implementing a level to flow cascade).
Response testing on a level to valve configuration or a level to flow cascade configuration would produce
consistent, linear process gains at all operating points along the curve. Adding cascade in such a
situation would just add unnecessary lag to the level controller response.

7 Restricted © Shell International BV 2014

PC01K Control Concept: Cascade Control

2.3 Example
In this section, the results of converting a level to valve control strategy to a level to flow cascade control
strategy will be presented. Initially, the level control application was configured as level direct to valve,
like the scheme illustrated on the left hand side of Figure 4. It was eventually converted to a level to flow
cascade scheme, like the one illustrated on the right hand side of Figure 4.

Initially, the application worked well with the level direct to valve configuration. However, over time, the
level began to cycle persistently. The cause of the problem was found to be control valve sticking. Sticking
is a type of non-linearity in the feedback control loop that manifests itself as step-like patterns in the
controller output and as a saw tooth pattern in the process variable. These effects can be seen in the
LC.PV, LC.OP and FI.PV signals on the left hand side of Figure 6. Note that this data was retrieved from a
process historian using a one minute sample interval. It is not always necessary to have fast data
collection to detect a sticking valve.
60
LC.PV, LC.SP, [PCT]

55

50

45
0 100 200 300 400 500 600 700

100
LC.OP, [PCT]

95

90

85
0 100 200 300 400 500 600 700
FI.PV, FC.PV,.SP [mlb/hr]

12

11.5

11

10.5

10
0 100 200 300 400 500 600 700

58
FC.OP, [PCT]

56

54

52
0 100 200 300 400 500 600 700
Time, [min]

Figure 6: Results of converting a level to valve control scheme to a level to flow cascade control scheme

8 Restricted © Shell International BV 2014

PC01K Control Concept: Cascade Control

At the midway point of the data set shown in Figure 6, the conversion to a cascade control scheme was
made. Prior to the conversion, the level controller output, LC.OP, was setting the control valve position
directly. After the conversion, LC.OP was setting the setpoint of the flow controller. Two new signals
appear in the right hand side of the data set after the conversion: The setpoint for the new flow controller,
FC.SP, and the output of the new flow controller, FC.OP.

Clearly, the conversion to cascade resulted in a substantial improvement to the performance of the loop.
The cycling in the level is eliminated and the variation in the flow is substantially reduced. Because the
flow controller in this example was the steam input into a column, many other key variables that are
related to the steam flow also showed improvement.

The ultimate fix for a sticking control valve that is impacting production or quality is to apply a mechanical
fix. The conversion to cascade does not get rid of the sticking problem, but as we have learned in this
module, the faster execution rate of the secondary flow controller reduced the impact of this nonlinearity
on the physical system.

2.4 Fast Primary Process

The main benefit of cascade control comes from exploiting the response time margin between a final
control element and its primary and secondary processes. In the process heater example, the final control
element was the fuel gas valve. The response between the fuel gas valve and the fuel gas flow is fast,
while the response between the fuel gas valve and the heater outlet temperature is many times slower. We
have seen how cascade control can be advantageous in that case.

There are some physical systems where the dynamic margin between the final control element and the
primary and secondary processes is not significant. It is not usually beneficial to implement cascade in
such cases, even if the secondary process measurement is available. Examples would be short residence
time vessels such as knock out drums on platforms or water boots in fractionator overhead vessels.

3. Configuration Details

3.1 PV Tracking and Initialisation

There are very basic but important configuration rules to be followed for smooth operation of cascade
control loops. The rules regarding PV Tracking and Initialisation are particularly important. These cascade
configuration rules are outlined in the DCS Basic Application Standard, DEP 32.30.20.15-Gen. They are
considered mandatory for cascade loops:

PV Tracking: The secondary controller must be configured so that its SP tracks its PV when it is in manual
or when it is being initialised by the final control element. This ensures that the primary controller’s output
is tracking the correct value via back initialisation, thus allowing for bumpless transfer when the
secondary is switched back to cascade mode.

9 Restricted © Shell International BV 2014

PC01K Control Concept: Cascade Control

Initialisation: When the secondary controller is not in cascade mode, the primary controller output is not
used and the primary control loop is essentially disabled. If secondary initialisation is enabled, then the
primary will continuously calculate its output value to match the setpoint of the secondary. This ensures a
smooth transition back to cascade mode. If secondary initialisation is not enabled, then the primary output
can be set to an arbitrary value, and returning the secondary to cascade mode will bump the secondary
setpoint to the value specified by the primary controller output.

Note: Refer to the topic on initialisation for more details on the cascade initialisation calculation.

3.2 Outerloop Dynamic Treatment

In a few cascade control applications, filtering may need to be applied to the PV of the outer loop for
dynamic compensation purposes. This type of filtering is applied not for noise reduction, but to ensure that
the dynamics of the outer loop are well-conditioned for use in cascade loops. In most cases, the objective
is to slow down the response of the outer loop so that the Epsilon value for the outer loop can be three to
five times slower than the Epsilon of the inner loop. Here are some examples:

■ Duty controllers: Some calculated exchanger duties are a direct function of the secondary
controller process value, for example a duty to steam flow cascade. Filtering the calculated duty
result gives the secondary flow controller time to track the changes set by the duty controller.

■ Separation Index (SI) controllers: The separation index is inferred as a direct function of the reflux
(R) and distillate (D) flows, given by the quantity SI = R/(R+D). In applications where an SI
controller is setting the reflux flow, digital filters are incorporated into the SI calculation so that it
does not respond instantaneously to changes in reflux. The separation index controller described
in DEP 32.30.20.16-Gen explicitly includes digital filtering of the reflux input into the SI
calculation to dynamically separate the response of reflux flow and SI to changes in the reflux
valve position.

■ Calculated total flow controllers using downstream setpoints: Total flow controllers that are
calculated by adding the setpoints of downstream secondary flow controllers have no open loop
time constant. A digital filter should be applied to give the total flow calculation a nominal,
artificial open loop time constant. Then, normal Epsilon tuning rules can be applied to the total
flow controller to yield a response time on the outer loop that is three to five times slower than the
individual flow secondary controllers.

■ Calculated total flow controllers using downstream process values: If a total flow controller
calculation is based on the process values of secondary flow controllers, then the open loop time
constant of the total flow controller is equal to the closed-loop time constant of the secondary flow
controller(s). Normal Epsilon tuning rules can be applied to the cascade.

Refer to the topic on Digital Filtering for more information on filtering for dynamic compensation.

Finally, in rare occasions process requirements or control objectives do not allow for slowing down of the
primary loop in a cascade pair (for example, a fractionator overhead pressure to compressor capacity
control cascade). In these situations, the three to five times Epsilon response margin rule may be
abandoned, but external reset windup protection will need to be added to the outer loop to ensure stable
operation of the closely coupled cascade pair.

10 Restricted © Shell International BV 2014

PC01K Control Concept: Cascade Control

3.3 Cascade Bypass

In certain applications, there is a performance benefit to be realised by using cascade control but the
secondary control loop measurement is not reliable. An example of such a system would be a distillation
column bottoms level where the outlet flow is prone to fouling due to high density, viscosity, coking or
catalyst particles. If no mitigation plan is in place, loss of the secondary PV will render the primary loop
inoperable as well, requiring manual operation or a temporary re-configuration and download to directly
connect the primary to the final control element.

Well-instrumented systems with problematic secondary loops might have redundant measurements
available for the secondary PV. In this case, the control strategy should be configured to be able to switch
the secondary PV to a backup measurement when one measurement is down for maintenance.

Cascade bypass is another mitigation option for cascade loops with failure-prone secondary
measurements. Cascade bypass involves automatically bypassing the control algorithm of the secondary
and passing the scaled primary output directly to the final control element. The advantage of cascade
bypass is that it is cheaper than dual secondary measurements and it can be implemented without a
download. Some DCSs have built-in cascade bypass that can be invoked for any cascade loop. If
cascade bypass is not available in the DCS, the function can be built by using a switch upstream of the
final control element that either selects the output of the primary or the output of the secondary. Retuning
is not normally required when invoking cascade bypass, but as we have learned in this module, all the
potential benefits of cascade control are unavailable until the secondary loop is restored to normal
operation.

4. Summary
■ In a simple cascade control configuration consisting of a primary and a secondary controller there
are two controlled variables and one manipulated variable (final control element).

■ The purpose of a cascade control configuration is to use the secondary controller to remove, or
minimise, disturbances on the primary by exploiting the difference in response dynamics between
a ‘fast’ secondary process and a ‘slow’ primary process.

■ There are some potential disadvantages of a cascade control configuration. One of them is that a
secondary controller adds dynamics to the primary control loop that may impact the overall speed
of response of the primary control loop.

■ There is (almost) no limit to the number of possible layers of control within a cascade
configuration. For example, suppose an advanced control scheme sets the setpoint of a (primary)
temperature controller, which in turn adjusts the setpoint of a (secondary) flow controller. Here,
there are three levels in the cascade. Suppose the valve is equipped with a ‘valve positioner’. Then
there would be four levels. If the APC application was steered by a plant optimiser there would be
five levels of cascade in total.

■ When tuning a cascade control scheme, start with the inner-most loop and work outwards
towards the outer most loop. Remember that the closed-loop speed of response for an inner loop
becomes part of the open loop response for the outer loop.

■ In some publications cascade control configurations are referred to as ‘Master and Slave’ control.

11 Restricted © Shell International BV 2014

PC01K Control Concept: Cascade Control

■ A cascade bypass configuration may be required to mitigate the failure of the secondary
controller due to loss of its process measurement. Without cascade bypass, loss of the secondary
control loop will render the entire cascade loop inoperable.

12 Restricted © Shell International BV 2014

PC01K Control Concept: Cascade Control

13 Restricted © Shell International BV 2014