The Route to Model Predictive Control Success

McMillan and Weiner Ask Dennis Cima How He Approaches Making MPC Applications as Successful as Possible

By Greg McMillan, Stan Weiner

Share Print Related RSS
Page 1 of 2 « Prev 1 | 2 View on one page

Stan: Navigating the ocean of advanced control applications in the continuous process industries requires commitment and expertise. Many companies have reached their targets and some have been extremely successful. However, others have returned to port empty-handed, and some even have been lost at sea with no understanding of what really happened.

Greg: Mark Darby who was the feature interviewee for the three-part series "MPC Past, Present and Future" (Feb., Mar. and Apr. of this year), recommended we talk with Dennis Cima to get the user perspective on the best route to success achieved at a large oil and petrochemical company. Dennis is the manager of Process Control Network and Control Systems for Chevron Downstream & Chemicals. Dennis, how do you approach making MPC applications as successful as possible?

Dennis: Creating a large MPC application to run a production unit is like configuring a cruise control for a battleship. You want the best technology and the best people. Cheaper is often not better. The view that the automation specialist is a commodity is counter-productive. Chevron has built considerable in-house expertise and is self-sufficient in creating, implementing, maintaining and improving MPC. We cover all the essential aspects; development, documentation, training, maintenance and continuous improvement.

Stan: What keeps you on the best route?

Dennis: Continual measurement and monitoring MPC metrics. For MPC benefits, the ISA Transactions paper "Estimating Benefits from Advanced Control" by Latour, Sharpe and Delany (http://tinyurl.com/bqj4eg5) is the heart of our guidance system, The statistical methods for quantifying the financial gains from MPC in this paper are tried and true and have stood the test of time. The methods are applicable for any level of control. We have put together an internal course on achieving and documenting benefits.

Greg: I noticed that besides the safety margin opportunity, the paper notes the value of gaining the results of the best operator consistently by MPC. While the benefits may be less than the safety margin, the results having been achieved are more realizable. We did this in an analogous way at Monsanto and Solutia by the use of an opportunity sizing, where Glenn Mertz would analyze the cost sheets and find the best period of operation, as noted in the June 2012 column "The Human Factor." The gap between "normal and best-demonstrated" was the basis of the opportunity assessment that lead to process control improvements (PCI) that achieved on the average a 4% reduction in the cost of goods. In our case, further adventures in PCI stopped when we left the plant. What we implemented stayed online, but innovation ceased for the most part and benefits become less recognizable. How do you make MPC benefits self-sustaining?

Dennis: You need infrastructure to make the benefits consistent. Site A and B are on the same basis. We have corporate standards and systems to historize and report MPC key performance indicators (KPIs).

Stan: Do you have online metrics?

Dennis: Yes, however the online metrics need to be screened for outliers and bad inputs. KPIs need to be reviewed before being reported to management.

Greg: Why is MPC so advantageous in refining, petrochemical, chemical and the special nonlinear MPC in polymer plants?

Dennis: These are all high-volume, continuous processes. An increase of just a few percent in production rate is a huge amount of money. Some refining companies have fallen flat in using and sustaining MPC. You cannot just dump it over the fence. If the developer moves on, and maintenance is not able to deal with changes in the process or objectives, the MPC gets turned off. The same is true for at-line analyzers, which create considerable potential. If an analyzer specialist is not involved in the calibration and troubleshooting, the analyzer will fall into disuse.

Greg: I suspect the negative stories I have seen on MPC are indicative of this problem. I also suspect the difference in opinion between Hunter Vegas (a project manager) and I (a technologist) on the value of analyzers is the result of whether in-house expertise is consistently involved on startup and on an on-going basis. Hunter wanted to write a tip recommending the avoidance of at-line analyzers. We compromised by me adding exceptions and watch-outs to my "Tip #63 – Use Field Analyzers to Measure Key Component Concentrations." One of the problems with near-infrared (NIR) analyzers is that they rely on statistical models that are too mathematically complex for the technician to understand as noted in the December 2011 column with Jim Tatera "Process Analyzers. Analyze This!" The concern was echoed by Michael Chaney in the January 2012 column "Gas Chromatographs Rule." For NIR analyzers the supplier is often needed to decipher and rebuild the model. When do you need to change MPC models?

Page 1 of 2 « Prev 1 | 2 View on one page
Share Print Reprints Permissions

What are your comments?

You cannot post comments until you have logged in. Login Here.

Comments

No one has commented on this page yet.

RSS feed for comments on this page | RSS feed for all comments