In the 35 years since they introduced distributed control systems, automation system suppliers have provided their process industry customers with far more than the ubiquitous PID control function block in their suite of available function blocks.
Many of these additional DCS function blocks originated back in the days before multivariable or model predictive control (MPC) was commercially available.
We now refer to putting together functions such as lead-lag, dead time, or selectors to formulate advanced control as "advanced regulatory control" or "ARC." ARC and MPC are both types of advanced process control (APC).
While distributed control systems provide advanced control capabilities in the many useful function blocks available, some process industry end users moved away from function blocks in favour of developing MPC models for advanced process control.
Before MPC, if the process required advanced control techniques, the process control practitioners had to assemble the correct control functions to achieve the required results. These evolved into what we now call advanced regulatory control.
Modern process control assumes that the regulatory control provided by the process control system is solid, with well-tuned loops and the controllers operating in the correct mode. Without good basic regulatory control, advanced process control will not perform well.
But what about advanced regulatory control? Now that MPC is so widely accepted and used, are control engineers and other practitioners still taking advantage of the ARC capabilities built into most DCSs?
An entire culture has evolved around the use of MPC. MPC has earned widespread respect and acceptance due to its often-spectacular payback. Despite the considerable implementation cost often involved, users have cited MPC projects that have delivered return on investment in 18 months or less.
As a result, the "culture of MPC," might unduly influence some companies to implement MPC, when in some cases, advanced regulatory control implemented right in the DCS control blocks might actually provide the best solution.
Not highlighted as much is the fact that MPC requires a continued investment. The models are built based on the set of process conditions, feedstocks, ambient conditions, variable interactions, and business objectives that exist at that point in time. Over time, however, any or all of these may change, requiring the model to be modified or rebuilt for the MPC to continuing driving value.
So here's a good question: If all the process needs to perform better is feed-forward, do you really need to build (and maintain) a model to accomplish this?
In many instances, advanced control could be accomplished by configuring function blocks and tuning each to remove loop interaction, provide feedforward action, and other advanced regulatory control techniques. If so, could this possibly be a better approach than MPC? More to the point, how's a plant to decide?
There is no substitute for knowing your process and the skills of your control personnel. The process control culture in any plant or company has evolved over many years with deep roots in "how we've always done it".
All plants must continuously maintain basic regulatory control to provide the stable, well-tuned base layer. However, above this foundation, plants have options for performing APC functions.
It's important to note here that the choice of which technology to use for advanced process control – MPC or ARC – is not an "either/or" decision, but rather a matter of understanding which tool, or combination of tools, is best for each application. But how do you know if you have the right balance?
The decision to implement an advanced control application using model predictive control, advanced regulatory control or some combination is more than a technology decision. Both tool sets can perform well for a variety of applications.
Users must take support for the implemented application into consideration. If the staff at the site is not trained in maintaining MPC models, then ARC could be a better choice. If uniformity across the enterprise is important, it is important that the company considers how it will support remote locations, regardless of whether it chooses MPC or ARC.
In the end, both MPC and ARC may be the right decision. Some applications will lend themselves to using DCS function blocks, while others are best solved using model predictive controllers.
If you question whether you're using the right combination, benchmarking can help you at least determine if you are alone or with the majority. End users in process plants may find it worthwhile to participate in ARC Advisory Group's Benchmarking Consortium, which addresses MPC performance and other key issues.
[Dick Hill is Vice President, ARC Advisory Group.]