Once upon a time I accompanied a home office technical team from the headquarters of a major oil company to investigate some problems and opportunities with an old pipeline running across a southwest desert. This installation had been doing a stellar job for decades with very little investment. It was well-designed and operated in accordance with its original purpose by people who had worked there most of their lives. The home office was interested because of some custody transfer measurement issues that were hard to resolve and some energy consumption data that suggested incredible pumping efficiency.

The first plant we visited was equipped with Foxboro pneumatic chart recorder-controllers and some pressure-actuated switches that controlled valve positions. Everything seemed to be operating in a stable and acceptable manner, and we were assured this was “normal”. One of the guys on the team was a very experienced instrument technician who was intimately familiar with the Foxboro 43-series circular chart recorder-controller. The one he was looking at obviously bothered him, and he noticed the control mechanism was not moving even though the linkage from the sensor was.

He saw something (subsequently described as rust) obstructing the linkage, and he reached in with the tip of a pen and flicked it off. The mechanism quickly adjusted to the match the sensor position, probably for the first time in years. Operators came running out of control room to try to determine why they had just lost flow into about a quarter of the system this equipment supplied. When it was explained that the instrument was mechanically hung up, so it showed a pressure that wasn’t really there, the operators were confused – mostly because that instrument always reported that same pressure.

As it turned out, there were a few similar instrument problems that pointed toward easy resolution of most of the known technical issues, and also suggested some improvements in “observability”. With some investment in maintenance and some new equipment this facility continued to provide the dependable service it always had, but the numbers it produced were different – more believable.

I was a young 20-something instrument and automation consultant at the time and from my time in college I could directly link what we could analyze and control to what we could observe, and as the recorder-controller worked so hard to point out to us, all that depended on the accuracy with which we observed it. That takes us directly into what we need to observe.

• Measurement location – is it pertinent to the operation of the process? Early in my career I worked at resolving some custody transfer measurement accuracy problems. A common issue was temperature in a meter run. The carefully designed systems, of course, included an appropriate instrument to measure the temperature of the fluid in the meter run. Sometimes, though, the instrument would be exposed to direct sunlight where, for the most part, the fluid and the meter weren’t.  Sometimes the temperature was derived from conditions in other parts of the process. Sometimes it was merely assumed.

Clearly, there must be a measurement, and it must be relevant to the subject process condition. It must be close enough to the point-of-interest so that delays in the measurement process do not introduce first order lags in the flow of control-essential data. Another common issue was the location of pressure or flow instruments in the process associated with a control scheme. When a valve moved, for example, the pressure or flow instrument that observed the change might not see the process change for a considerable amount of time while the process completed re-adjustment. This delay introduced a lag that the control equipment of the day interpreted as “no effect” from the last control input – so the equipment would automatically make another correction, and another, and another. Such control schemes didn’t produce stability, they produced continuous, cyclic adjustment of a period related to the inherent process delay between action and observation.

• Process delays between observing a “cause” and the effect we’re seeing.
• Measurement accuracy and related parameters such as repeatability and linearity – where accuracy comes from.
• Measurement bandwidth – Fourier Bandwidth shows you what you can and can’t see and whether some of what you see is real or not.
• Measurement and transmission noise.

All this says that to understand something or control something you really have to be able to see it, and usually measure it, in a way that provides enough resolution to enable the underlying picture to be seen – a view of Mona Lisa, not just a blurry sea of blue and cream pixels. Knowing what you need to do starts with knowing where you are. Some indication of how you got there can be useful, and the rapid observability of the effect of changes you make to gain order and optimization can go a long way for providing control and stability.