Model predictive control—Building a bridge between theory and practice

https://doi.org/10.1016/j.compchemeng.2006.05.044Get rights and content

Abstract

This paper describes the development of an MPC product at AspenTech. A brief historical view sets the stage, describing collaboration with university researchers that formed much of the technical foundation. Gaps between theory and practice were addressed during product development. Major effort focused on eliminating the need for practitioners to understand control theory, while ensuring that controllers satisfy theoretical requirements. This allows control strategies to be designed and implemented via configuration and tuning in a domain familiar to a process operations engineer. Terms like eigenvalue, covariance, detectability, controllability, etc. are never exposed. Robust numerical algorithms and software suitable for real time execution must augment the control theory foundation. Best practices and expertise are embedded in the software realization, thus reducing the skill level required for success while still allowing experts to push the envelope. The resulting technical scalability facilitates a wide range of process control practice in a single framework.

Introduction

Model predictive control, MPC, has influenced process control practice significantly during the past 20 years. Early MPC technology (circa 1976) evolved principally in industrial settings, followed by copious numbers of academic papers analyzing and extending the underlying theoretical basis. Some MPC applications were one-of-a-kind, implemented for specific applications. Others were based on more generally applicable commercial products. Two survey papers by Qin and Badgwell, 1997, Qin and Badgwell, 2003 describe MPC products available during the past 10 years. These surveys summarize product features that relate to important issues in process control practice.

Theoretical insight into MPC increased rapidly, but actual practice evolved more slowly. Practitioners, working with tools that were firmly rooted in the past, faced increasingly challenging applications that forced them to develop clever ad hoc solutions to address real-world issues. Unfortunate side effects of evolution, today's MPC technology and products often contain “relics of the past” similar to unused genes found in living creatures. Eventually, AspenTech began development of a new product with the philosophy to start fresh, using the best available theory without being encumbered by such relics. Past practice was retained when appropriate, however. The resulting product is formally known as the APC State-Space Controller, a module of the aspenONE™ Advanced Process Control Solution. The term SSC is used in this paper for brevity.

AspenTech and predecessor companies had supported various university research consortia. Much of the theoretical basis for SSC can be traced to research from the Texas-Wisconsin Modeling and Control Consortium, TWMCC, starting with its inception in 1993. This long-term relationship resulted in a research thread that built incrementally on prior research. Incorporating new results was much easier than embracing new, potentially outstanding, results that did not quite fit into this continuum. A hallmark paper by Muske and Rawlings (1993) defines the starting point. Much of the product development involved filling theoretical and practical gaps from this initial point. Gap closure came from both academic research and internal AspenTech efforts.

The paper proceeds as follows. A short historical review of MPC provides an overall perspective. Then, the motivation for a new product is described. This includes specifying the range and types of control applications to be addressed. In a sense, this defines a certain view of process control practice and tools needed to achieve successful applications. Gaps between theory and practice are described in a separate section. An additional section is devoted to a few simple examples illustrating important technical advantages compared with previous generation products. No CPC paper would be complete without predictions about future directions. These are given in a forward-looking section, which outlines areas for future research. Finally, concluding remarks are made including answering the question: why did it take so long?

Section snippets

Historical perspective

Early papers (Cutler & Ramaker, 1980; Richalet, Rault, Testud, & Papon, 1978) document the generally accepted beginning of industrial MPC. The initial work was probably being done at about the time of CPC 1 in 1976. A long sequence of MPC-oriented papers have appeared at every CPC since then, including one on “Model Algorithmic Control” at CPC 2 (Mehra, Rouhani, Eterno, Richalet, & Rault, 1981).

Five years later, at CPC 3, a session was devoted to MPC with focus on industrial applications.

A new MPC product—why?

Why would a company invest in a new linear MPC product? And what does this have to do with process control practice? A few short answers follow.

Despite being mature, existing MPC products are not suitable for all applications or they do not capture all the benefits possible when they do apply. Compromises are inherent with finite models and/or finite prediction and control horizons. The typical approach of reconciling model predictions with measurements by simply biasing predictions is clearly

Gaps between theory and practice

Theory and “toy problem” simulations clearly indicated that a new generation of MPC could frequently outperform available products. Some important questions were:

  • Will the performance exist for practical applications?

  • What surprises will emerge using infinite state-space versus finite response models?

  • Is calculation of a “no compromise” infinite horizon, constrained move plan really feasible?

  • What special initialization steps are required?

  • How do issues like asynchronous, missing or invalid data

Examples

Performance benefits from state estimation and infinite horizon control are illustrated in this section. One example demonstrates improved disturbance rejection. It is well known that PID controllers can be tuned to reject disturbances more effectively than MPC that uses bias updates (Shinskey, 1994). Addition of state estimation eliminates this performance difference and furthermore applies to multivariable applications.

The second example demonstrates calculation of a complex “no compromise”

Future developments

This MPC environment marks a new beginning with control theory and software to support a wide range of users and applications. The framework will facilitate incorporation of recent and future control theory results for even better performance and increased ease of use. It is also suitable for extension to other classes of problems. For example, academic and practical results exist today that extend into the world of MPC using nonlinear models. The software framework has been architectured to

Conclusions

The title contains the phrase “building a bridge between theory and practice”. This describes much of the product development activity. A major objective was to develop an MPC environment, not just a controller, to enable engineers without formal control theory training to efficiently build, deploy and maintain control applications. This included the objective that the environment would be suited to a wide range of control problems, especially in industries where the benefits of MPC have not

Acknowledgments

The work described in this paper represents a large team effort, especially in the final year of product development. The “true believers” that contributed during the early phases deserve special mention. These include my current and former control engineering colleagues at AspenTech: Tom Badgwell, John Campbell, Dean Kassmann and Mike Keenan. Also, Dave Hein deserves special mention for his untiring contribution to the initial and final online real time environment architecture and software.

References (19)

There are more references available in the full text version of this article.

Cited by (69)

  • Process control practice and education: Past, present and future

    2019, Computers and Chemical Engineering
    Citation Excerpt :

    CPCVII was held in Banff, Canada in January 2006, co-chaired by Mike Henson and Tom Badgwell (2006), with the theme “Assessment of Core Technologies and Emerging Applications,” with 70 academic (including 15 students), and 43 industrial participants. Froisy (2006) provided a perspective overview of the development of a new state space-based MPC product at AspenTech; the user interface was particularly important, to shield the user from critical underlying checks and balances that assure detectability, etc. The initial prototype was developed in 1999, based on the approach presented by Muske and Rawlings (1993); several years later the fourth phase of the prototype was applied to a chemical process with 8 inputs, 20 outputs and 109 states.

  • Human factors in dairy industry process control for energy reduction

    2017, Journal of Cleaner Production
    Citation Excerpt :

    The system gathers data from across the process and controls elements (e.g. flows) in order to achieve relevant set points (e.g. temperatures). Process control systems are designed to be flexible with regard to the metrics a site wishes to control for (Froisy, 2006). However, within a standard package, it can be difficult to deal with conflicting control metrics of quality vs energy use.

  • Advanced Process and Control Strategies for Bioreactors

    2017, Current Developments in Biotechnology and Bioengineering: Bioprocesses, Bioreactors and Controls
View all citing articles on Scopus
View full text