Advanced Regulatory Control: Decoupling

AT A GLANCE

Feedforward control

Controller tuning

Operator effects

Economic considerations

Feedforward control systems act to isolate the controlled variable from influences that would otherwise disturb it. When a process has multiple loops and the various control loops upset each other, use of a reciprocal system that isolates the loops from each other may be required. Such a design is referred to as decoupling control.

The feedforward system developed for the reactor (shown in the January 2005 installment) isolates product temperature from changes in product flow rate and composition. Two questions frame the remaining issues:

What is required to control product composition without affecting total production rate?

What is required to control product flow rate without affecting product composition?

The answers point the way to a fully decoupled system wherein:

While adjusting the flow of ingredient B to control composition, the system should also make equal and opposite changes in the flow rate of ingredient A;

While adjusting the flow of ingredient A to control total flow, the system should also change the flow of ingredient B to keep the ratio of A/B constant.

Structure of the reactor decoupler shows how measurements of each ingredient flow alter the control signal to the other set point.

Figure 1 shows the structure of the decoupler for product flow and composition.In this setup, measurements of each ingredient flow alter the control signal to the other set point according to these principles.

Implementation

Figure 2 shows the structure of the complete system for the reactor. It presents a system for fully decoupling all of the control loops from each other. Since steam flow does not affect either product flow or composition, decoupling signals from steam flow are not necessary. Implementing and using this structure, however, require addressing several critical issues.1. External integral feedback . Integral action in controllers requires a feedback signal from the controller output. In many controllers, this signal is obtained internally and is transparent to the user. But this limits the functionality of the controller to simple loops; more capable platforms accept an external integral signal for their PID algorithms. External integral feedback can provide protection against windup in cascade systems and enables more advanced control structures such as auto-selector, feedforward, and decoupling controls.

This control setup delivers a system for fully decoupling all of the control loops from each other.

The external integral signal must be numerically equal to the controller output signal. Using the output signal directly only duplicates an internal signal and offers no advantages. To gain the benefits of external integral feedback, the signal must come from the lower level structures.For cascade loops, the measurement to the secondary controller is the signal to use. For more complicated structures, the external integral signal must often be calculated. This is the case for these decoupling controls.For this purpose, the feedforward scheme requires a separate calculation, derived from the design equation, to back-calculate the controller output from the steam flow measurement. The controller output is the feedback trim signal. Solving the design equation for the feedback trim signal yields

T bc = [F s /4(F a +F b )]+(T a F a +T b F b )/(F a +F b )

where T bc = back-calculated external feedback signal.

When a calculation is this complicated, it can be done only in a custom-designed structure, as indicated in Figure 2.

However, many PID algorithms provide connections for basic feedforward and decoupling inputs. This is the case for the composition and flow controllers in this implementation. Summation and multiplication on these controller outputs can be accomplished through standard connections, and the necessary calculation for the integral feedback is handled internally. This makes additional external calculations unnecessary and simplifies the configuration of advanced control schemes.

2. Controller tuning. In effect, the presence of underlying feedforward and decoupling structures changes the meaning of the controller output variables. For example, in the simple cascade structure for product flow control, the output of the flow controller was the set point for the flow of ingredient A, which was Fa. Adding the summation function to the output of this controller means that it manipulates the sum of ingredient flows (Fa + Fb), before the subtraction of Fb.

This, in turn, effectively changes the process gain as perceived by the controller. Compared to the basic regulatory structures, the process gain appears different, but more constant. This requires initial retuning, but thereafter, the loop stability is much better.

Controllers which function as feedback trim controllers can typically be detuned because the feedforward system provides good dynamic response to upsets and tight feedback tuning will interfere with its performance. Set point response may suffer, but because this is a quality specification variable, set point changes are rare.

Conversely, controllers that are fully decoupled, such as the composition controller, can often be tuned more tightly because they no longer have to be detuned to allow for the delayed effects of interaction. This can improve set point response if it changes often.

3. Bumpless manual-auto transfer. The implementation must also force the higher-level control functions into a tracking mode when the lower-level controllers are in manual or local set control. In this state, the output signal from the upper controllers must follow a signal equal to the external integral feedback. This allows bumpless transfer into automatic mode when permitted by the operator.

Advanced regulatory controls

To test performance of these controls, the same upsets were used here as were used for the regulatory controls in the first article (January 2005). Figure 3 shows a trend of the performance of this structure.

Based on the regulatory control upsets used in Part 1 of this series (January 2005), this graphic shows a trend in the performance of the advanced reactor control structure.

The “Performance index values” graphic compares the performance index for this set up with the values for basic regulatory control.

Taken together, these data show the significant benefits, and some limitations, of this advanced regulatory control scheme.

Applying feedforward and decoupling controls provides significant improvement in the stability of product quality during production rate changes. As the trend shows, changing the flow set point causes an immediate change in all the manipulated flows. The decoupling causes the ingredient flows to change in a ratio, which keeps the composition constant while the change in ingredient flows is immediately compensated by in increase in steam flow.

Results include:

The composition index is reduced by a factor of 35; the temperature index by a factor of 10. Overall, the total index is reduced by a factor of 20, or 95%.

Compared to the previous response, the swing in temperature is both faster and smaller, so the system returns to steady state much more quickly.

Feedforward and decoupling controls also improved performance during composition set point changes. Changing the composition set point causes a change in the flow of ingredient B, which is quickly compensated by the total flow controller and the feedforward calculation for steam flow set point.

This benefit was gained entirely through more stable temperature control. The decoupling system does not significantly improve set point response for composition control because:

The composition index is essentially unchanged, while the temperature index is reduced by a factor of 6. Overall, the quality index improved by approximately 20%;

The composition index did not improve for set point changes because the structure for this change is essentially unchanged. The decoupling from product flow does not significantly affect the dynamic response of the composition controller to a set point change; and

Control of product flow was essentially unchanged because the flow loop is so much faster than the composition loop. As the output of the composition controller slowly changes, the much faster flow controller can quickly readjust the flow of ingredient A to compensate.

In general, it is not necessary to decouple fast loops from relatively slower ones, since feedback alone can make the necessary adjustments.

Effect on the operator

The impact of advanced controls on plant operators is often overlooked. Commissioning a more complex control system changes their interaction with the control system in several significant ways.

If the application works well, the controls will be in automatic mode more of the time. Less manual control will be needed to handle transitions and upsets in the operating conditions.

Operators will need to understand the functions of the advanced control system so they can correctly interpret changes in the manipulated variables. With feedforward and decoupling controls in place, they may see changes in the manipulated variables occur for no immediately obvious reason.

New procedures may be required to put the system into operation following startup and other unusual conditions. The structures for initialization and bumpless transfer will add new flags and state indicators into their interaction with the controls. The operator will see state transitions occur, which may be new to them.

Feedforward control introduces several new tuning concepts for dynamic compensation and feedback trim; old perceptions of how to affect the system response will need to be enlarged.

Most significantly, advanced control will improve the stability of plant conditions during upset conditions. This will create opportunities to improve the plant economic performance by changing normal operating conditions. But if operators cling to the same operating points, most of the potential economic benefit will remain unrecovered.

Not all loops are equally important to the plant’s economic performance. The best operators will try to identify loops that can be the source of economic benefits through the application of advanced control. Once advanced control is in place, they will have to reconsider the normal operating points for the process, and operate closer to limits that had previously been given a wide berth.

Appropriateness

Advanced regulatory control is the least expensive and most familiar approach with which to achieve significant improvements in the stability of the operating conditions, depending on the scope of the problem. The reactor that is the target application for this series is a relatively small problem—only three controlled and manipulated variables. And yet the complete system can quickly become quite complex. This can make documenting, configuring, and operating these systems a challenge. For problems of a larger scope, this approach can become too unwieldy, and techniques that are at once more powerful and more compact can be used.

And yet, when advanced regulatory control can be applied to move the process to a point of improved economic performance, this technique can be the most cost-effective approach to improving economic performance through control. The necessary tools and functions are typically available in any DCS already installed (though rarely in PLCs), so it is usually not necessary to purchase any new hardware of software. Generally, these systems can be installed and commissioned online, without interrupting process operations.

References

Shinskey, F.G. Feedback Controllers in the Process Industries, McGraw-Hill Publishing, New York, 1994

Shinskey, F.G. Process Control Systems, 3 rd edition, McGraw-Hill Publishing, New York, 1988

Performance index values

Control technology
Change production rate

Composition ISE
Temperature ISE
Total ISE

Basic regulatory control
.53
.28
.81

Advanced regulatory control
0.015
0.028
0.04

Change product composition

Composition ISE
Temperature ISE
Total ISE

Basic regulatory control
1.79
0.55
2.34

Advanced regulatory control
1.83
0.08
1.89

Author Information

Lew Gordon is a principal application engineer at Invensys;

Do you have experience and expertise with the topics mentioned in this content? You should consider contributing to our WTWH Media editorial team and getting the recognition you and your company deserve. Click here to start this process.