| M |
y first assignment as a newly hired control systems representative at IBM in 1961 was to evaluate the work being done at DuPont’s Repauno Works in southern New Jersey. Being an IBM test site I was sent there to see if the results of our work there could help us sell our recently announced IBM 1710 Industrial Computer Control System. The process being studied was a fixed-bed reactor for acrylonitrile synthesis. A principal control problem arose from the catalyst activity which degraded so steeply that parallel reactors were needed, one in synthesis mode while the other was being regenerated. Deciding when to switch so as to maximize productivity was an important objective.
A really talented team from IBM Research (Jack Bertram, Dick Stillman, Dick Koepke plus others) had modeled the synthesis process in a set of coupled, partial differential equations. Solving these equations numerically, however, required several minutes on an IBM 7090 to simulate a few seconds of real-time. All went back to the drawing board and the team reduced the theoretical model to a set of easily solvable empirical relationships covering the operating space of interest. Meanwhile, back at DuPont, researchers were also hard at work trying to identify an improved catalyst that didn’t degrade significantly.
“Thank you very much,” said Dupont, “it was a great project. We learned a lot about our process and how to control it. On further review we really don’t need the computer control system as we had initially thought.” Remember it was 1961, the IBM Selectric typewriter had just been introduced and airline reservations were hand-written lists handed to the gate agent. The IBM 1401 punched-card data processor dominated most accounting applications and IBM’s sales force was simply too busy selling it to pay much attention to esoteric industrial uses like process control--even for the $1,000 cash bonus (more than a month’s salary for most) a control system sale would generate.
“CLIFFTENTa mathematical method for optimizing setpoints and measuring the financial value of reduced variance for any controlled variable or key performance indicator.”
I learned right from the start that process optimization was a really tough way to sell control systems. We had recently discovered that paper-making was a very complex process that control computers could facilitate by making key performance indicators available to the machine tender in real-time. Without recourse to higher mathematics, the control system could significantly improve overall machine productivity during paper-grade changes, upset recovery (e.g., sheet-break) and other transient conditions. Process optimization became a future intangible that could close the deal if necessary.During ISA/2004 in Houston, I had a long discussion over lunch with Pierre R. Latour,a recognized authority in process automation technology and successful entrepreneur in several process control ventures. Latour began his career in the early 1960s with DuPont and Shell Oil after receiving a PhD in Chemical Engineering at Purdue. He worked on the first Shell computer control project (FCCDeer Park Refinery 1966). A two-year tour as a captain in the U.S. Army followed at NASA’s Manned Space Flight Center managing the Apollo Docking Simulator development. After mustering out, Latour co-founded Biles & Associates (later acquired by Invensys) and Setpoint (later acquired by Aspen Technology). Latour served in a business development capacity as Vice President at Aspen prior to launching his current consultancyCLIFFTENT, Inc. ([email protected]).Latour is nothing if not passionate about the value of the CLIFFTENT methodology www.groups.msn.com/CLIFFTENT. His rapid-fire discourse barely allows for dialog and I’m certain that I have only a superficial understanding of this method of selecting the optimal set-points for maximum economic yield from a process or operation.His quarrel with process control as currently practiced boils down to two points:1. No consistent, objective method for assessing the economic value of improved dynamic performance.2. No incentive for control systems implementers to deliver sustained performance improvements.Latour’s answer to the first is CLIFFTENTa mathematical method for optimizing setpoints and measuring the financial value of reduced variance for any controlled variable or key performance indicator. His answer to the second is SR2 (Shared-Risk, Shared-Reward)a method for licensing technology solutions based on their performance as measured by CLIFFTENT.This is a guy you ignore at your peril. He has demonstrated a deep technical knowledge of chemical process control, as well as the ability to survive and prosper as a business leader. If I were a chemical process manager, a control systems technologist or a business entrepreneur in a related discipline, I’d listen to Pierre Latour quite closely. Terrence K. McMahon, McMahon Technology Associates, can be reached at [email protected]
