XMC Multivariable control in a nutshell

Because automated multivariable control arrived cloaked in computers and complex software (ca. 1980s), it has taken a long time to realize that it’s really the same function – manual multivariable control – that has always been a central part of operating team activities.

Historically, operators adjust the single-loop controller setpoints and outputs (aka “the available handles”) to control a superset of related constraint and optimization variables (aka “controlled variables”), based on their experience, knowledge of the process, ongoing operating conditions, and input from the greater operating team, which includes supervision, process engineers, production planning, etc.

 Conventional model-predictive multivariable control (MPC) endeavors to automate this task using detailed models and embedded optimizers, but this approach has met with unexpected consequences in cost, maintenance and performance (mainly due to the long hard lesson that models change frequently – read The Big Story Behind Auto-Tuning on ControlEng.com). XMC® automates the task more simply based on historical manual multivariable control methods, which notably do not require detailed models or optimizers.

 The primary limitation in XMC® is the lack of model-predictive feedforward control action, which has always been considered a cornerstone of MPC. But feedforward is the single-loop equivalent of model-predictive control, and its lengthy history tells a different story.

The potential power of feedforward (to reject disturbances proactively) has always been well known. Feedforward function blocks have been available since industry’s first DCSs (and in PLCs, analog and pneumatic systems before that). Yet, historically, feedforward has been used only on a very limited basis, even at the much more manageable and selective single-loop level, due to the complexity, risk and maintenance it adds to any loop. In practice, feedforward has a high bar and is generally warranted only where its benefits are high and a highly reliable model is possible.

 Using the installed base of control systems throughout industry as a benchmark suggests that perhaps one in ten loops warrant the use of feedforward, and that the rest will perform satisfactorily, if not more reliably, based on feedback control alone. This sheds a different light on the MPC practice of “wholesale” feedforward – literally hundreds of mass-produced feedforward models – and suggests it might be as much a cause as a solution to MPC’s maintenance and performance record. The XMC® strategy is traditional selective use of feedforward. The top priority of advanced process control – as with single-loop control – is (or should be) to reliably close the loops, and not necessarily to use feedforward in doing so.