Model-free multivariable control (XMC) aims to solve the multivariable control problem without requiring a detailed model or embedded optimizer.
Model-based multivariate control has been around for so long that many engineers are unaware that multivariate control can now be achieved without the need to work on models.
Model-free multivariate control (XMC) is not a "revolution" of multivariate control methods; rather, it represents a return to fundamental principles. XMC solves multivariate control problems in a simpler and more reliable way without requiring detailed models or online optimizers.
In traditional model-based multivariate control, model-related activities account for 90% or more of engineering and maintenance costs, but almost all reliability and performance problems are due to poor model quality, and there is little reason to expect this to change.
Currently, most advanced process control resources are used to support existing applications rather than new ones. This leaves many advanced process control needs unmet, opportunities untapped, and problems unresolved. Model-free multivariate control offers a way out of this unsustainable predicament.
The principle of model-free multivariable control
Previously, operations teams were constrained by process limitations and needed to manually optimize operational activities. Model-free multivariate control (XMC) automates these tasks without requiring detailed models or online optimizers. Important process constraints, control of key handles, driver optimization, and safe move sizes are no longer secrets gleaned from model details. They have become fundamental knowledge for operations team members (Figure 1).
Model-free multivariable control (XMC) automates multivariable constraint control and optimization methods that previously required manual execution for operations teams. Notably, this approach does not require detailed models or embedded optimizers. Model-free multivariable control internally uses patented rate predictive control (RPC). Image courtesy of ACP Performance.
• The important constraints, relevant handles, optimization objectives, and gain directions (not detailed models) captured in other conventional multivariable control matrices are sufficient to determine the correct direction of movement for each control handle in model-free multivariable control.
• Model-free multivariable control implements a pre-designed rate of movement for each handle based on established safety operating specifications (also known as “process speed limits”).
• Model-free multivariable control monitors the actual process response as it is running, rather than the response of yesterday or last year, and uses Rate Predictive Control (RPC) (a state-of-the-art, patented single-loop control algorithm) for further movement rate adjustments to predictively reduce movement.
Historically, the quality of manual multivariable control has depended on time, initiative, and the expertise of different operators. Timeliness and consistency, on the other hand, are the cornerstones of automation, and these are advantages of model-free multivariable control.
Using rate prediction control
Rate predictive control is based on automated manual control methods, combined with important process control principles and intelligent concepts. It automates traditional, mature manual control methods and integrates them with key process control principles and intelligent technologies.
Rate predictive control produces a single-loop solution, while model-free multivariable control is a multivariable control solution that internally uses rate predictive control. Therefore, all the advantages of rate predictive control also apply to model-free multivariable control.
Compared to traditional proportional-integral-derivative (PID) control, rate predictive control offers many significant advantages. It requires less PID substitution but has a high demand for multivariable control substitution. Its greatest advantage may lie in its ability to provide benefits in models-free multivariable control environments, including:
• Highly responsive and stable performance;
• When a variable gradually approaches the target value, perform forecast tapering to avoid unnecessary overshoot or fluctuations;
• Pre-selected move speed (or “process speed limit”) can be particularly important in multivariable control, where multiple key handles may be moved during the process;
• Robust feedback control characteristics, so it will take appropriate action regardless of the source of disturbance or “model matching”.
What does "model-free" mean?
Model-free control implies feedback control. All process control is essentially feedback or feedforward. Feedforward control requires models to predict and take feedforward actions. Feedback, on the other hand, responds to ongoing deviations and must have a reasonable method for responding, such as PID or rate predictive control.
Therefore, model-based and feedforward are synonymous. Model-free is synonymous with feedback. Most control engineers appreciate the potential of feedforward (seamlessly eliminating disturbances), but historical experience and modern multivariable experience suggest that it is better to use feedforward selectively, like model-free multivariable control, rather than extensively, like model predictive control. Model-free also implies adaptability, meaning that the control algorithm automatically compensates for changes in the process response (which typically change on both short-term and long-term perspectives). Adaptive control has always been a primary goal of process control. In contemporary times, industry has been attempting to achieve adaptive loop control through automatic adjustment, but with largely little success, and is trying to do the same on more complex multivariable foundations, such as adaptive modeling. Notably, rate predictive control is the only patent among countless process control patents that possesses inherent adaptability.
Furthermore, model-free control means that control is not based on a model. Therefore, by using model-free control, model-related activities can be reduced or completely eliminated, including plant testing, model identification, and model maintenance.
More advanced process control
The feasibility of model-free multivariable control offers a new multivariable control paradigm for the process industry, which is more economical, agile, scalable, reliable, and better suited to the increasingly flexible daily operational needs of modern process industries. It is based on qualitative models, lacks embedded optimizers, and combines traditional feedback and manual control principles. Table 1 compares current model-based multivariable control paradigms with the emerging model-free paradigm.
Table 1: Comparison of Model-Based Multivariable Control Mode and Model-Free Mode
In operational facilities, multivariate control applications often vary in size, ranging from a few variables to dozens of variables. Therefore, a smaller footprint solution can benefit in two ways. For traditional applications with larger footprints, it can provide scalable tools. In many industries, it can also offer alternative design strategies for legacy applications with heavy maintenance burdens.
While model-free multivariable control has not yet been fully adopted, the process industry can consider it as a backup plan and benefit from it. It should be viewed as a valuable asset and resource that promises to overcome existing models and help manufacturing companies develop agile, sustainable, and advanced process control solutions.
Key concepts:
■ Model-free multivariable control is designed to solve advanced control problems without the need for detailed models or online optimizers.
■ Rate predictive control and model-free multivariable control can be designed together and have the same benefits.
■ Model-free multivariate control can benefit modern process industries.
Think about it:
How can model-free multivariable control help your industry or field facility?
Disclaimer: This article is a reprint. If it involves copyright issues, please contact us promptly for deletion (QQ: 2737591964). We apologize for any inconvenience.
