Up to this point, the material discussed has only focused on single-loop control without considering the impact of other associated loops and their performance or tuning. However, it is common to have one controller’s output affect the set point of another controller in what is known as cascade control. An example of cascade control can be seen in Figure 6.1.
Figure 6.1 [An example of cascade control system. The output of a primary controller adjusts the setpoint of the secondary controller.]
If you haven’t encountered a cascade control system before, it can be initially perplexing. In the provided example, the primary controller is responsible for the reactor temperature, while the secondary controller regulates the brine temperature. These controllers are often referred to as “master” and “slave,” and alternatively as “outer loop” and “inner loop.” These labels originate from the framework illustrated in Figure 6.2, which depicts a signal-flow diagram of a cascade control system.
Figure 6.2 [The signal-flow diagram of a cascade control system shows the inner and outer loops.]
- A cascade control system has the advantage of detecting certain load changes earlier and correcting them faster, resulting in improved control. For instance, in the example given, if the pressure in the brine header changes due to other users of the brine, the brine temperature controller in the cascade system will detect it sooner and take corrective action. In the absence of the cascade system, the disturbance due to the brine header pressure would remain undetected until it affects the reactor temperature, which could take much longer. As a result, the disturbance would persist for a longer duration.
- Cascade control effectively linearizes the secondary variable to a change in setpoint from the primary controller, which is generally desirable for optimum performance. In the example, a 1% change in the primary controller output will result in a defined change in jacket temperature because the brine temperature controller will take corrective action. Without cascade control, a 1% change in the primary controller output going directly to the brine valve will result in an undefined change in jacket temperature, depending on the valve characteristic and relative brine pressures. If a valve positioner is used in a simple single loop, it becomes a cascade system, as the positioner responds to its setpoint from the controller, overcoming the nonlinearity of dead band due to packing gland friction, to some extent.
- Cascade control can often decrease the natural period of the primary loop. If the secondary loop has a lag that would exist in the primary loop without cascade, then the cascade system can be faster than non-cascade control. In the example given, without cascade control, the reactor temperature controller output simply moves the brine valve, and the brine temperature in the jacket follows with some lag, which can be relatively long. On the other hand, with cascade control, the secondary controller works to achieve the temperature requested by the reactor temperature controller output. As a result, the jacket temperature changes more rapidly than it would with a simple valve position change. This reduces the lag in the primary controller loop and shortens its natural period. Figure 2.2 illustrates that a tightly tuned controller results in a lower overall lag to a setpoint change compared to a poorly tuned controller.
- There are some instrumentation techniques that can be used to optimize cascade control, such as setting limits on the secondary set point.
When considering the use of cascade control, it’s important to ensure that the natural period of the secondary loop is significantly shorter than the natural period of the primary loop without cascade. While there’s no fixed rule on the desired difference, a factor of ten is generally acceptable. Always tune the secondary loop first and ensure that it’s tightly controlled. A deliberately sluggish secondary loop would negate many of the benefits of cascade control. The amount of lag seen by the primary loop is dependent on the tuning of the secondary loop. Therefore, the tuning of the secondary loop interacts with the tuning of the primary loop, but not vice versa. It’s also worth considering not using integral action in a secondary controller to avoid complicating the tuning and to prevent problems with integral windup. With digital controllers now having the option to be either PI or P only, the pros and cons of using integral action for the secondary controller should be reviewed.