Sunday, March 08, 2009


Many years ago, in my very first assignment with the Great Corporate Salt Mine, I helped debottleneck a plastics plant.

Debottleneck? Is dat de neck of de bottle?

Well, yes, Mr. Dialect Comedian, but at the GCSM, we used the term as both noun and verb.

To debottleneck a manufacturing process is to remove (you guessed it) bottlenecks. Narrow spots in the line. It is a way of expanding capacity by making a few, relatively inexpensive changes to an existing operation, rather than by simply throwing money at the problem and constructing a second production train. Debottlenecking makes the operation more efficient. And a “debottleneck” is a project that expands a plant’s capacity by (you guessed it again) debottlenecking it.

Got it? Good.

The GCSM had a plant, back then, that produced a certain amount of polypropylene plastic. We installed a bunch of new equipment and were able to increase capacity dramatically... by over 50%, if my recollection serves. My job (in case you’re curious) was figuring out just what to install, how much it would cost, and how much capacity improvement we would get from it... in an age of slide rules. No personal computers, no electronic spreadsheets.

Once the new equipment had been installed, it was time to start up the newly-expanded plant and let it flex its new muscles. To make sure it actually worked and that those millions of dollars we spent actually accomplished something. And that meant spending a lot of time at the plant, both in the control room and running around on the unit. Catching samples, measuring temperatures, that sort of thing.

If you have never been in the control room of a chemical plant, it’s an imposing sort of place. These days it’s a lot like being on the bridge of the USS Enterprise, with a lot of computer terminals... but 35 years ago, computer-controlled processes were still in their earliest, most primitive stages. Back then, the myriad operating controls were all manual, with many of the parameters recorded graphically on continuous plotters. Every few hours, the operators would record key settings and process parameters on a huge “horse blanket” spreadsheet the size of a tabletop... using pen and ink.

Running the process meant knowing the right settings for hundreds of temperature controllers and valves. And the “butterfly effect” - where small changes sometimes have large, unexpected results - was in full force.

There was one part of the process - a fractionation tower - that was misbehaving one morning. And so, as the resident Contact Engineer, I made a minor adjustment to a critical flow rate. It seemed trivial at the time, but it had an effect that was... not desired.

Correcting that effect created its own cascading series of changes... all of which needed their own corrections. And compounding everything was the fact that any change to a given setting took a certain amount of time - anywhere from minutes to hours - to work its effects. You could see the impact on the chart recordings, which would oscillate like a struck gong when a tweak was made, gradually settling back down to a new steady state. “Lining out,” we called it.

Getting that part of the process back under control was like wrestling a bear. In a vat of Jell-O. Dangerous, messy, and unrewarding. Eventually, I managed... but only after developing a serious respect for the sensitivities of Complex Processes.

As we watch our legislators and our new administration struggle to bring the economy under control, keep in mind that they are also trying to operate a complex system, one with mysterious lag times, uncertain cause-and-effect pathways, and that is subject to the vagaries of human behavior. The tiniest of tweaks - not to mention wholesale changes - will have unpredictable effects, effects that will manifest themselves on unpredictable timetables. And add to that the overall brokenness of the system... and the fact that it is being run not by Economic Engineers, but by Political Bumblefucks.

It’s enough to keep me awake at night, it is.

No comments: