ExperTune Library

Loop Optimization:
Before You Tune

To Reap the Greatest Benefits, Define Your Objectives and Understand the Limitations of Your Equipment.
By Michel Ruel, P.E.
Reprinted with permission from CONTROL Magazine, March 1999

Plant efficiency and consistent product quality depend on proper loop performance, but tuning the controller is only the last step. This is the first of a three-part series on loop optimization. In April, Part II will describe how to optimize loop characteristics. And finally, in May, Part III will cover PID tuning.

There is much to be gained by optimizing control loops. It has been estimated that 80% of process control loops are causing more variability running in automatic mode than in manual. The often-quoted EnTech study showed that some 30% of all loops oscillate due to nonlinearities such as hysteresis, stiction, deadband, and nonlinear process gain. Another 30% oscillate because of poor controller tuning.

With a poorly optimized loop, an upset in the direction towards inefficiency results in giving away product. Alternatively, a load may cause off-spec product. When a control loop is running optimally, variability is minimized. Better tuning keeps the process on spec and reduces giveaway of often-expensive ingredients.

But tuning objectives vary for different types of processes. For example, in a steam header, the pressure has to be maintained at the maximum allowable without large errors so the safety valves will not open. The PID controller must be tuned tightly to ensure the valve that controls the flow from the main header will move quickly to eliminate effects of disturbances.

On the other hand, the PID controller of a robot arm that manipulates nitroglycerin vessels has a different objective. The control loop must be optimized to change the setpoint without overshoot or cycling.

Performance Objectives

Most engineers and technicians tune process control loops using trial and error, observing the response to setpoint changes. To achieve good setpoint response takes a skilled intuitive understanding of the shape and speed of response. Only experienced people are able to achieve good setpoint response this way.

Unfortunately, once a loop is tuned for good setpoint response, the response to upset is usually very sluggish. Good setpoint tuning does not automatically result in good recovery from upsets. Unfortunately, it is upsets that usually are the source of off-spec product and poor variability.

Using modern tools to analyze a loop will give the engineer or senior technician helpful hints about the process: numbers and graphics will inform the user about design, equipment performance, and interactions with other loops. Modern tools also let the engineer or the technician select appropriate tuning parameters for the control objective. And since the algorithms used in PID controllers are different from one manufacturer to another, in many cases the algorithm is user selectable.

WHAT DO YOU WANT?

The same loop can be tuned for robustness
(green), neutral response (blue), or speed (red)
depending on the objectives.

The characteristics of good control (Table I) are difficult to obtain. When tuning a loop, one must make compromises between robustness and speed of response. Robustness is the ability of the control loop to remain stable when the process (mainly dead time or process gain) changes. Usually, to obtain robustness:

The trends in Figure 1 show the same flow loop tuned for different objectives.

A control loop consists of the process, measurement, controller, usually a current to pneumatic (I/P) transducer, and valve. Optimal process control depends on all of these components working properly. Hence, before tuning a loop, one must verify if each component is operating properly and if the design is appropriate.

Choosing the optimal PID tuning should be done after making sure all of the other components are working properly. The optimal tuning parameters ensure your equipment is used at maximum efficiency.

Questions to Be Answered

The following steps outline a procedure for approaching and optimizing a process control loop. Optimization requires observation in manual and automatic modes, and at various operating conditions. We need to answer the following questions:

1. Process gain: Is the control valve sized properly? Often, valves are oversized. If so, the controller output will be at one end of the range when the loop is in automatic. Also, oversizing the valve will amplify nonlinearities such as hysteresis, stiction, different response to small and large changes, and operating near the seat.

The process gain should be between 0.3 and 3. The ideal process gain is 1. A process gain too high will not permit the controller to work at its full potential: the controller will have to be tuned with a small proportional gain.

2. Hysteresis/stiction: Does the control valve have harmful hysteresis and/or stiction? Hysteresis is a difficulty but stiction is really the main problem. Stiction occurs when friction is present.

Hysteresis should be less than 3%, significantly less if the loop is to be tuned tightly. Stiction should be less than 1% and often 1% is too much.

3. Sensor/transmitter: Is the measurement sensor working properly? From your experience, do the numbers make sense? For example, is the dead time small enough? If a transmitter is not properly installed, the dead time can be too long; if a filter is added in the transmitter, the equivalent dead time could be longer.

4. Noise band: Is there an excessive amount of noise in the loop? When disturbances occur too fast to be removed by the PID controller, they are called noise. Filtering may help. The filter should be small enough to not increase the equivalent dead time and large enough to reduce the noise.

Selecting the filter time constant is a tradeoff between increasing the equivalent dead time and reducing the amount of noise. When the noise is reduced, the controller output is smoother.

5. Nonlinearities: How nonlinear is the loop? A loop is nonlinear when the process gain varies. All loops are somewhat nonlinear. It is the degree of nonlinearity that we are interested in. If the loop gain varies by more than a factor of two or three, then linearization will help optimize the loop.

6. Asymmetry: Does the loop respond differently in one direction than in the other? Often, a valve responds more quickly in one direction than the other. Also, in temperature processes using one fluid to add heat and another to remove heat, the two fluids are different and the characteristics of the process are different.

If the equivalent dead time or the equivalent time constant are different depending on the direction, use the worst case to tune the loop or use a special algorithm.

7. Tuning: Is the loop optimally tuned? If the loop is tuned aggressively to minimize error, the robustness is small; if the loop is tuned sluggishly to reduce variability, the recovery time after a disturbance is long.

Tuning parameters are selected to make a compromise between robustness and performance. The loops upstream could interact--selecting the appropriate tuning parameters will allow decoupling. At the opposite, if loops need to be synchronized, selecting the appropriate tuning parameters will ensure they work in accordance.

Next: Diagnosis

Each of these problems has a characteristic signature, which can be found by performing a series of tests and
analyzing the results. The tests, which will be covered in detail in the next installment of this series, start with collecting process variable and controller output data with the controller in automatic at normal operating conditions, then introducing a setpoint change. Data is also collected with the loop in manual mode.

You will be able to see how the operating range for the valve and its performance can tell you if the valve is sized correctly; whether loop cycling is being caused by hysteresis, nonlinearities, or poor tuning; and the other critical aspects of loop performance that must be understood before tuning
the controller.

Michel Ruel, P.E., at TOP Control Inc., St. Romuald, Quebec, an engineering company specializing in optimization of continuous and batch process control. Ruel has 22 years of plant experience at companies including Monsanto, Domtar Paper, Dow Corning, and Shell Oil. Author of several publications on instrumentation and control and frequent
university lecturer, Ruel is experienced in solving unusual process control problems and a pioneer in implementing fuzzy logic in process control. His e-mail address is mruel AT topcontrol.com.