Single-loop control—Still the mainstay of advanced process control
By Jim Ford, PhD
Just over a decade ago, an article titled, "Advanced Control Strategies Move into the Field" (Control, October 2008), highlighted three evolving trends in the process control world that would "make our dependence on single-loop control part of history." The three trends cited were the:
- movement of APC controllers to the field control devices
- increased availability of inexpensive sensors
- increased use of fieldbus and wireless in the control network.
Ten years later, how much of this prediction has come true? Predictably, not much. And, why?
Single-loop control (i.e., a control loop with one input and one output) as implemented in various versions of the proportional, integral, derivative control algorithm in modern distributed control systems has two primary functions:
- servo-control-reacting to changes in the set point (SP) to move the process variable (PV) to its new target
- feedback control-reacting to changes in the PV to return the PV to SP
Servo-control is easy. Choose the right amount of integral action (in combination with the gain action, and with or without gain action on SP changes), and the control algorithm will adjust its output (OP) to move the PV to its new target at the chosen rate with or without overshoot, as desired.
Feedback control is much more important and more complex. It is the only mechanism in the entire process control world that can react satisfactorily to changes in unmeasured disturbance variables, such as changes in ambient temperature, rainfall, stream composition, pump load, exchanger fouling, field operator moves, board operator moves, and many others.
APC technological advances
Advanced process control (APC) got its start and made its mark in the 1960s primarily by reacting to measured disturbance variables. This "reaction" was coined feedforward control. If a disturbance variable to a unit operation can be measured, then the key control variable (CV) for that operation can be kept close to its SP by adjusting the associated manipulated variable (MV) using a simple "model."
For example, consider a catalytic reactor where the key CV is the reactor inlet temperature, which is controlled at the basic level on feedback by adjusting the fuel flow to a fired heater. If the reactor feed rate changes, then a feedforward controller can adjust the fuel flow (the MV) using a dynamic model between the feed rate and the fuel flow. The intended result is little or no change in the reactor inlet temperature (the CV)-the primary benefit of all APC-a reduction in process variance.
Initially, feedforward control was implemented using one input - one output relationships. As the technology developed in the 1970s and 1980s, more complex approaches and strategies developed, involving multiple inputs and outputs with more complex models. Eventually, multivariable, model-predictive control replaced simpler APC approaches and algorithms for both feedforward and feedback control of complex process control applications.
But, throughout the past 50 years of APC technology advances, those unmeasured disturbances have not gone away. The APC controllers can now be implemented in field control devices (trend number 1); there are cheaper sensors (trend number 2); and fieldbus and wireless (trend number 3) are now realities. But none of these technological achievements have been able to mitigate the process instabilities created by unmeasured disturbances. Their presence still confounds the most experienced control system engineers. That is why single-loop control is still the mainstay of process control, and why those trends discussed earlier (or any others) will never result in APC being implemented successfully in its absence.
If APC needs single-loop control to reject unmeasured disturbances, then how best to utilize it?
APC control objectives
The most important APC control objectives are related to production rate and product quality, because these variables are directly related to operating profitability. Production rate is limited by constraints (e.g., maximum temperature and control valve position). Product quality is normally controlled by temperature, analyzers, lab analyses, or indirectly by "soft sensors" (inferred properties). Vessel levels are integrating, inventory-related variables and are almost never included as CVs in an APC controller. The same can be said for vessel pressure, unless it is a constraint variable related to pushing the production rate. (There are some exceptions-fired heater controls often use burner pressure as a substitute for fuel gas flow to control heater temperature.)
The lone outlier is flow, which is a true "extensive" variable, independent of product quality or operating profitability. In all but truly exceptional cases, flow is always adjusted to achieve some other process control objective. It is almost always the secondary, or slave, in a one-on-one basic cascade. If the flow controller is standalone, then its SP is adjusted by the operator (or an APC controller) to achieve a higher-level control objective.
At the same time, the flow controller is a true mitigator of unmeasured disturbances. It is typically characterized as a high-frequency loop, meaning that a change in OP is followed almost immediately by a change in PV. When tuned properly, and when challenged by unmeasured disturbances, it returns the PV to its SP very quickly and with little overshoot or oscillation. As such, it is normally the variable of first choice as an MV for any higher-level control strategy, especially for an APC controller.
There are some exceptions. Although flow is a true extensive variable, there are instances where it can be used to "create" an intensive variable. Ratio variables are used quite often in process control. For example, reboiler duty on a distillation column can be calculated from flow and temperatures and then ratioed to the column charge rate. The column reboiler duty/charge ratio can then be used as an MV in an APC application. Same for treat ratios in absorbers, product yields in fractionators, and so on. The single-loop flow controller rejects unmeasured disturbances and thereby stabilizes the created intensive variable.
Status quo: Do not disturb
Today, even after 50 years, APC continues to rely on the lowly flow control loop, the most basic single-loop control, as the best rejector of unmeasured disturbances and the most stable platform for the APC/optimization control hierarchy. So, the next time somebody suggests getting rid of single-loop control in an APC application, just ask, “What about unmeasured disturbance variables?” Do not expect a righteous reply.
Single-loop control (shown) is the mainstay of process control, and APC can never be successfully implemented in its absence.