Q&A: Advances in PID Tuning for Process Control

By Theresa Houck, Executive Editor

Did you know in Ancient Greece, a simple form of PID was used to move huge doors?

I learned this and a lot more during my discussion with Control Station's President & CEO Dennis Nash for an episode of our “Automation Chat” podcast, “Advances in PID Tuning & Modeling for Process Control.” We talked about advantages and drawbacks of using historical data or live data, how using "live historical" data joins the best of both worlds, and more. In this Q&A, I share some highlights of our conversation.

Theresa: What role does the PID play in process manufacturing, and how have advances in automation affected the tuning of PID controllers?

Dennis: The PID controller essentially allows process manufacturers to regulate the means of production. And the PID, or forms of it, has been around for ages. Some will point to Ancient Greece as the earliest example, where a very simple form of PID was used to move enormous doors. The most widely cited example of PID being used in an industrial application was in the late 1700s — a centrifugal fly-ball governor that allowed for the regulation of engine steam.

Back in those times, everything was analog and the process of tuning a PID controller was little more than guesswork.

Skip to the digital age. With the use of computers, tuning software came on the scene, allowing a more consistent and repeatable process by which practitioners could access data and formulate appropriate tuning parameters.

Early tuning software packages lacked the ability to handle real-world applications; real-world process dynamics are highly dynamic, noisy and oscillatory. And it wasn't until 2008 when, in fact, Control Station introduced the nonsteady state modeling innovation, that allowed for dynamic and accurate modeling of complex process data.

Theresa: What is the relationship between modeling and tuning?

Dennis: Tuning is essentially the ability to refine a controller’s responsiveness. Once you have a good model, it’s easy to tune that refinement.

Think of an industrial application with dynamic behavior, such as a surge tank. Its purpose is to absorb the impact of an upstream disturbance, so the process remains calm downstream. A controller would be tuned very aggressively so it could absorb that dynamic and smooth it out.

In contrast, a controller could be tuned conservatively if, for instance, its purpose was to maintain setpoint tracking. So, we’re just trying to make sure that within a modest band, the controller is tracking that setpoint. That can be easily done through refinement of tuning, but it’s the ability of a model to accurately describe a process’ dynamic behavior that’s the real magic.

The model can assess accurately how far, how fast and with what delay a particular controller will respond to those normal everyday disturbances. Without that good model, the tuning parameters themselves are of little value.

And fortunately, with increasing amounts of data, higher-resolution data from the process, software is now equipped with better tools with which to formulate the model and deliver those more refined tunings.

Theresa: Data is key to everything. What are the most common ways to access data?

Dennis: We're fortunate that fewer and fewer users are applying manual methods where they simply input values and cross their fingers. They’re now able to use varying forms of software to provide a more reasoned and repeatable approach to their tuning process while seeking to improve control.

To that end, however, it's historical data that remains a tried-and-true resource for many users. Most data is historized and includes those fundamental elements required when tuning out. That's information related to a set point, to a controller output, a process variable, and of course some timestamp by which you can evaluate changes.

Finding the bump test that is stored in the historian though, can present a challenge. Unless you know when the setpoint change was made — when the manual output adjustment was performed — you're essentially looking for a needle in a haystack. And that can be a challenge for those that use historical data.

Theresa: So, it seems like using historical data is a logical choice for accessing data and tuning a plant’s PID control loops. But what are some of the drawbacks to that approach?

Dennis: There are legitimate concerns over live data, so many users choose to use historical data instead, because it’s not interacting with the control network. But while on the business network, historical data does indeed have some setbacks.

Typically, it’s within the IT department's domain to assign the compression and exception guidelines. That means they’re taking all this vivid data that was once live, and extracting key components as they compress it, so they can store it.

What’s often lost within the historicized data is key attributes that can help to formulate or calculate that process model. And from the viewpoint of a software manufacturer focused on PID control and tuning, that's a major drawback.

Theresa: What roadblocks do users run into with using live data?

Dennis: In the live data access realm, users have immediate access to a facilities control network. They're installing software, something that IT departments don’t like, whether it's tuning software or any software.

And it's hard to blame IT staff. Some research shows that a cybersecurity firm called Sophos reported that in the year 2021, 55% of manufacturers were hit with ransomware. And that was up about 50% from the prior year. That’s an alarming trend.

So, on the one hand you have historical data, which can present difficulties because it lacks the resolution that can be so helpful when tuning PIDs for optimal control. And on the other hand, you have live data that has that robust, unadulterated data, but it can open up process manufacturers to cybersecurity threats.

Theresa: It sounds like you're saying users are darned if they do and darned if they don't either. What options are they left with?

Dennis: A third option recently came onto the automation scene. We refer to it as “live historical data.” It’s the ability to access live data immediately before it becomes historized.

More specifically, software now can access live data from Rockwell Automation FactoryTalk^® Historian SE on a pre-exception, precompression basis or post-exception basis. And our LOOP-PRO suite of tuning software is just one example of a tool that can do this. It can capture data via FactoryTalk links, by way of the FactoryTalk live data interface.

And since the point of access is within the confines of a production facilitiy’s business network, it evades that concern of being on the process network. So, by getting the live data just before it's going into the Historian, it’s still a robust source of data with lots of elements that verify the most accurate model, and helps assure the most effective PID controller tuning parameters.

And this is just one way Rockwall Automation and its technology partners like Control Station are innovating and working together to address changes process manufacturers are facing.  

Based in Manchester, Connecticut, Control Station offers the Loop-Pro Tuner software that simplifies tuning of PID controllers, and PlantESP™ for monitoring plant-wide performance of control loops.

Like this article? Sign up for the digital magazine (4X/year) of The Journal From Rockwell Automation and Our PartnerNetwork and get articles like this delivered right to your inbox.

^{The Journal From Rockwell Automation and Our PartnerNetwork™ is published by Endeavor Business Media.}