PID Control Loops Have Specific Purposes and Unique Attributes. ‘One Size Fits All’ Approaches to Tuning May Be a Recipe for Failure.
In 1909 Henry Ford famously offered customers any color of his Model T automobile as long as their choice was black. Ford’s “one size fits all” philosophy worked remarkably well at the turn of the 20th Century. For sure, the general population’s improved financial means, its appetite for mobility, and of course a dearth of alternative low-cost options played a role. In contrast, a “one size fits all” approach to PID controller tuning doesn’t work so well in today’s complex, multi-process production environments.
The Zeigler-Nichols method is widely recognized as the first formalized approach for tuning PID controllers. Introduced in 1942 the method enables process engineers to minimize the impact of disturbances by both limiting overshoot and steadily, systematically settling a process. Whether applied in open- or closed-loop the objective “quarter decay” response is synonymous with the Zeigler-Nichols methodology. But it raises an important question: What if quarter decay isn’t the goal?
Given the diversity of production processes and their associated control objectives, it may be worth considering the following:
- Stereotypes Need Not Apply
Level loops possess different dynamics than other types of control loops such as flow, temperature, pressure, etc. Similarly, the objective of one control loop is rarely the same as the objective of another. Even level control within a brewery’s surge tanks differs vastly from that of its filler systems. As effective as “quarter decay” may be for some level loops, a method that lacks flexibility to cater to a given loop’s unique characteristics and its specific purpose is less than ideal. Simply put, disturbance rejection is not the objective for every loop.
- You Want Me to Do What?
Control loops are generally designed to correct for disturbances and to maintain stable operation. Paradoxically, when performed in closed-loop the Ziegler-Nichols method requires practitioners to introduce sustained oscillations and bring a process to the brink of instability. Oscillations and instability are generally non-starters for most industrial production processes, they unnecessarily affect downstream production and can place everything from people and production to plant and profit at risk. For these reasons the closed-loop approach has few advocates.
Numerous alternative methods for control loop tuning have been introduced since the 40s. While some frequency-based approaches are better suited to electrical engineering applications than process control, others are fully in synch with the needs of today’s complex multi-process production facilities. One such approach is the Internal Model Control (IMC) or Lambda tuning method which functions equally well in open- and in closed-loop, and it requires only modest adjustments to a process’ Controller Output during dynamic testing. What’s more, one software tool based on the IMC/Lambda approach eliminates the need for a steady-state at both the start and end of testing, providing a faster and more efficient means for optimizing control.
