I first met Mark Darby, principal at CMiD Solutions, at an ISA
Conference years ago. He impressed me as incredibly knowledgeable and an exceptionally nice and communicative guy. Kind of a rare combination in our profession, as our egos and intense focus get the best of us. I got his business card and made a mental note that a control talk column
with him would be fun and productive. It took me three years, but the timing was right because he had just done an excellent model predictive control
(MPC) presentation at ISA Automation Week 2012
and published a comprehensive paper in Control Engineering Practice titled "MPC: Current Practice and Challenges." Since my MPC application experience has been sporadic, and my current emphasis has returned to advanced regulatory control, I thought I, as well as the readers, could benefit from Mark's bringing us up to date. To see what I know about MPC, check out my ISA 2004 pocket guide Models Unleashed, the ISA 2001 Houston Conference paper "Constrained Multivariable Predictive Control of Plastic Sheets," the Control May 2003 article "Has Your Control Valve Responded Lately" and the Control July 2008 article "Unlocking the Secret Profiles of Batch Reactors
Stan: In this multipart part series Mark will share his thoughts on the scope of MPC applications, proper use of the regulatory level, inferential measurements, model development, economic objectives, support and maintenance. Since these Control Talk columns are designed to promote a creative capture of expertise, we use open-ended conversations rather than a script of questions so we are free to roam. You may have heard of free range chickens and how that makes for happier and tastier chickens. You can consider us free range engineers and while we may be happier than cooped up engineers, we hope not to end up at the meat counter even though upper management may just see us as bodies chewing up money better consumed as bonuses.
Greg: Where do we have MPC installations?
Mark: About 60% of MPC applications are in the refineries and petrochemicals (e.g., bulk chemical). Here the use of MPC is nearly an order of magnitude greater than advanced regulatory control (ARC). For specialty chemicals, MPC applications are growing and are approaching the number of ARC applications. MPC is moving into food processing and has even penetrated discrete manufacturing, most notably the automotive industry. If a chemical company has significant PID expertise, then ARC is often deployed. Large companies are more likely to consider MPC.
Stan: What is the success rate for new applications?
Mark: Success is nearly 100% in some plants, but is uneven despite significant advances in MPC software providing tools that previously required special programming and user skills. However, implementation naivety and disappointment has led some to reconsider the role of MPC. One of the problems is how do you go to get management support for a new APC application to get the right resources. Different silos make this difficult. A successful MPC application requires management to provide the operations training, regulatory control improvement and MPC expertise on a continuing basis. Even the best MPC software requires knowledgeable people to monitor and maintain the MPC application. Expertise must be readily accessible to insure the MPC performs well despite changes in process conditions and objectives and changes in instrumentation and equipment performance. Even if the MPC is not the cause of a problem, the MPC can be blamed. Also, operators may inadvertently disable the functionality of the MPC.
Stan: In the October Control 2012 Talk column "Bringing Advanced Process Control Home" we saw for industrial gas plants how important the support of a core company group was to success. We also had our eyes opened to how once a trustful relationship is established with operators, the MPC expert can be virtually rather than physically in the control room by remote PC applications and possibly cloud computing. What about batch operations?
Mark: While nearly all of the current applications are on continuous processes, we are seeing MPC applications in batch operations based on nonlinear hybrid models that incorporate both fundamental and empirical modeling techniques.
Greg: There are some highly successful linear MPC temperature control applications on batch distillation columns operating in a semi-continuous mode. For fed-batch reactor and bioreactor profile control, I have found that controlling the slope of the profile makes the response self-regulating and more linear. More importantly for product concentration control, using the slope of the batch profile as the controlled variable enables negative as well as positive changes that are essential for feedback control. The slope target is changed as the batch progresses.