Process Control Progresses at Dow Chemical

Jan. 27, 2003
A mere columnist dares to comment on loop tuning

It is sometimes said that columnists rush in where angels fear to tread. The arcane subject of control loop tuning is a thicket wherein unwary novices are quickly ensnared. Loop tuning is also perhaps the most fundamental cerebral of our profession. In the vigorous mental conflict between powerful intellects, innocent bystanders (this writer included) often sustain the worst injuries.

The eighth Pharmaceuticals Automation Roundtable (hosted by John Radler, [email protected]) was held in Midland, Mich., in mid-September. I had the opportunity to talk with some of Dow Chemical's process control luminaries as well as tour several facilities in the Midland complex. I also walked to the Herbert H. Dow Historical Museum on the Tittabawassee River. I sprinted the last quarter mile (not breaking any Olympic records) as the clock was fast approaching the 5 p.m. closing. On arriving I found that the museum, which features a period cable-tool percussion drilling derrick, is closed on Monday but a friendly curator gave me access to some descriptive literature.

Herbert Dow had successfully produced bromine, by electrolysis, from subterranean Michigan brine in 1891. Bromine, of course, was a raw material for early medicinal preparations (bromides), among other uses. After a series of ventures and partnerships, The Dow Chemical Co. was incorporated in 1897 and later absorbed the Midland Chemical Co., which had inherited Dow's bromine technology.

World War I was a period of great expansion by the then-nascent U.S. chemical industry as imports from Germany, a major supplier, were abruptly ended. The founder oversaw expansion during the 1920s as Dow became a leading supplier of phenol, caustic soda, magnesium, and other chemical products. The demand for bromine (ethylene dibromide) in the production of tetraethyl lead for anti-knock gasoline outstripped the Michigan source and led to the establishment of a seawater source at Freeport, Texas. In one of his last official functions, Herbert Dow was awarded the Perkins Medal by his chemical industry colleagues at The Chemists' Club in New York in January 1930.

Since the 1950s, Dow Chemical has been a leader in developing advanced process control technology. Recently, with the acquisition of Union Carbide, another process control technology arm with an illustrious history has been added to the Dow portfolio. While in Midland, I met with Margaret Walker, Dow's director of process automation, and Eric Cosman, senior automation architect. After several decades of internal manufacturing of DCS systems and software, Dow has opted to start deploying commercial control systems as economies of scale and third-party software compatibility have become key factors.

The bedrock of feedback controller tuning was set down by J. G. Ziegler and N. B. Nichols of Taylor Instrument in the Transactions of ASME in November 1942. In January 2002 in Chemical Engineering Progress, Lanny Robbins ([email protected]) published his tuning methodology, which has been employed extensively within Dow achieving up to a 40% reduction in loop variability as compared to the Ziegler-Nichols tuning rules.

Dr. Robbins (Ph.D. in chemical engineering from Iowa State University) joined Dow's R&D Dept. in Midland in 1966. He began to specialize in separations-purification operations about 1970. In the context of this work, in 1991, he took on the challenge of developing an improved controller tuning method for distillation operations. The resulting Robbins Method has been vetted on thousands of simulated loops and hundreds of actual production plant loops over the intervening decade. He observed that a control loop gave minimum variability to a step change in setpoint when the height between the first peak-to-valley was 16% of the height of the step change.

In a response to a reader's critique of the CEP article, Robbins makes two rather interesting observations:

* The Robbins Method is based on the controller's proportional band (gain) acting on the error signal. Using a proportional band that acts on the process variable appears to be rather uncommon in CPI applications.

* Applications of PI (proportional and integral) outnumber applications of PID (proportional, integral and derivative) for improved process response by at least 3,000:1.

In line with my first remarks, I could be walking into a buzz saw, but this technique appears to be an innovation that process control professionals should investigate.

Terrence K. McMahon

McMahon Technology Associates

135 Fort Lee Road

Leonia, NJ 07605

Tel: 201-585-2050

Fax: 201-585-1968

[email protected]