1.

Advanced Control Strategies

Some activities ago, it was learned how to tune a PID controller in a single feedback loop like the
one shown in Figure 1.

Figure 1: Single Feedback Loop

A good tuning is, without a doubt, a key element for the good performance of a control system.
But, in some cases a good tuning alone is not enough to achieve it and arises the necessity of
implement more sophisticate control strategies than a simple singe feedback loop. Two of such
strategies are the Cascade Control and the Feedforward control.

Cascade Control
The cascade control is used when the process can be split in 2 individual processes in series like
the one shown in Figure 2.

C1(s)

Figure 1: Single Feedback loop with two processes in series

In a single loop, the disturbances that affect the process, caused a change in the system response
(c). The controller detects this change and send the signal to correct the deviation.

Then, the disturbances affecting the process G1 are corrected until the controller detects them,
which happens when they caused a deviation in C. If the intermediate variable C1 could be
measured, a controller can be placed which can send a corrective action right away it detects a
change in C1 without having to wait until C gets affected (see figure 3).

Figure 2: Cascade feedback loop.

The controller that performs this function is called inner loop controller or slave controller (Gce),
because its function is to keep the intermediate output C1 in the R* value (inner loop reference)
that is ordered by the other controller (Gcm). The Gmc controller is known as master controller
or outer loop controller, because it is in charge of keeping C at R (outer loop reference) and
order the inner loop reference to make C reach the outer loop reference.

Cascade Control Advantages

The advantages are evident. The effects of the disturbance D1 that affect the process G1 are
corrected right away they are detected. Hence, its effects under the final process variable C are
considerable reduced.

It is worth mention that the magnitude of the improvement in the behavior of the loop is
dependent of the differences in speed between the inner and outer loop. If the outer loop is
much faster than the outer loop, the improvement will be considerable big since the correction
of D1 will be faster compared with the speed of the outer loop. If the speed of the response
between both loops are equal, the improvement will be minimum.

An empiric rule recommends to use the cascade control when the inner loop time constant (τe) of
G1 process is at least 5 time less than the outer loop time constant (τm) of process G2. If we
consider FTPDT processes the rule can be expressed as:

It is worth to say that the cascade control not only corrects the effects of the disturbances that
are introduced in the initial process but also corrects the possible non linearities that the process
can have.

Cascade Control tuning

If we observe the cascade scheme shown in Figure 2, we can notice that the process that the
inner controller sees is always G1. Hence, that shall be tuned first. In the other hand, the process
that the outer controller sees is the combination of the inner loop (inner controller+G1) and G2.
Any change in the inner controller will cause a change in the entire process that the outer loop
sees.

For this reason, the inner controller shall be tuned before the outer one. If, for any reason, the
inner controller is retuned, tuning the outer controller will be necessary

Let’s see the next example. Consider the level single loop shown in Figure 1 -3. In this process,
the controller LIC receives the level value from the transmitter LT and it orders to open or close
the valve to reach the desired level.

A “valve” process (transfer function G1) that receives a percentage for opening the valve will
result in a change in the flow coming through the valve and a tank process (transfer function G2)
that receives the flow from G1, will output the level of the tank.

Let’s analyze the feasibility and convenience for using the cascade control scheme here
considering the next points:
 Figure 1-3: Control Tank Level

1. The complete process that the single controller sees (from the input of the entire
process, the percentage of opening of the valve, until the output, level of the tank) can
be splitted in two process in series.

Figura 1-4: Splitted process

It is important to notice that the valve process (again, G1) is totally independent from
what happens to the tank process (G2), therefore we can make the splitting.
2. The valve process is much faster than the tank process. Usually, the flow dynamics have
small time constants (1 or 2 sec) while the level dynamics have much larger constants
(100 to 200secs, in big tanks).
3. The cascade control scheme will look like the following block diagram:

Figure 1-5: Cascade Control Strategy for a system level.

4. The expected advantages are the following:

a. To reduce the disturbances that are affecting to the valve over the level (i.e. inlet
pipe pressure drops)
b. Additionally, it is very likely that the flow loop will stabilize faster than the valve
alone (without the inner controller), so the outer controller will have less lag. Hence,
the outer controller may have a more aggressive tuning parameters and the response
in general will be faster.

Also observe that the disturbances affecting the tank process (like a change in the outer flow)
will not be reduced since these ones will not be observed by the inner controller.

It is very important to notice that the output of the outer controller will be the setpoint of the
inner loop.

In order to implement the cascade control scheme, all the conditions stablished in table 1-1 shall
be met.
 Table 1-1: Cascade Control criteria

Cascade Control Scheme is required when:

A single loop does not provide a satisfactory performance.

The is a secondary variable (manipulated variable) that can me measured.

A secondary variable shall met the following criteria:

The secondary flow shall indicate the presence of an important disturbance.

A causal relationship shall exist between the manipulated variable and the secondary variable.

The dynamic of the secondary variable shall be faster than the primary variable.

Inner controller tuning

In general, the controllers of an inner loop have a P or PI structure. Also, for best results, the
“Synthesis Method” is used:

5
Kc=
t0
(
K 1+ 0. 25
τ )
t0
τ i =τ +
2
t0
≤0 .2
This method requires τ

External Agents that impact the performance of a controller

Sometimes, it is almost impossible to make a controller to give a satisfactory response, even
applying different tuning techniques. Other times, it can be achieved a satisfactory response but
some hours later the behavior is no longer as good as in the beginning. In some other time, the
response is satisfactory in certain regions but not in all of them. So, what is the cause for these
behaviors to happens. The answer is clear: the nature of the process.

A conventional PID is very robust, that means, accept a diversity of hostiles environments.
However, the behavior will not always be satisfactory.

The external agents that impact the behavior of the controller process are the following:

 Nonlinearities  Coupling
 Noise  Big dead times
 Disturbances

In order to solve any of the mentioned issues, it is necessary to make a detailed analysis to
detect the issue and how it affects the process. Once, the information is obtained we can
implement different solutions.
In general, we can distinguish three types of actions:

a) Eliminate the source of the problem

b) Mitigate its effects
c) Compensate its effects

Disturbances
The disturbances are external agents that cause undesirable changes to the process response and
that cannot be controlled.

There is a difference between noise and disturbance. The noise only affects the measurement of
the sensor but not the process response. Hence it can be “ignored” by filtering its presence in the
process variable measurement. In the other hand, a disturbance does affect the process variable.
Hence, we cannot neither ignored it, nor filtering its presence in the process variable
measurement.

Feedforward Control
The feedforward control is one of the most valuable tools we have within the different control
schemes. It is used when there is a disturbance that is impacting the behavior of the process
variable. It is required to measure the disturbance in order to implement this strategy.

The basic concept of the feedforward control is the following:

To measure the disturbance before it is propagated into the process and take corrective
actions to minimize its effects under the process variable. It compensates the effect of the
disturbance over the process variable.

Figure 1-7: FeedForward block diagram

Static Feedforward control

Figure 1-8: Feedforward with static gain only

 Figure 1-8: Outer Flow feedforwatd

Feedforward control

Figure 1-9: Feedforward with LeadLag gain

Table 1-2: Feedforward criteria design

Feedforward control is adequate when:

When a single loop does not provide a satisfactory behavior.

The disturbance can me measured.