GETTING A MEASURE OF PLANT LOOP SUCCESS?

Regulatory controls and process measurements, although maybe unglamorous, are fundamental for large-scale controls and real-time optimisers. Yet this is rarely a strong point among practicing advanced control project engineers. Today, large-scale multi-variable predictive control projects in the refining industries include at least one major processing unit. Large-scale controller design requires a detailed examination of the […]

Regulatory controls and process measurements, although maybe unglamorous, are fundamental for large-scale controls and real-time optimisers. Yet this is rarely a strong point among practicing advanced control project engineers.

Today, large-scale multi-variable predictive control projects in the refining industries include at least one major processing unit. Large-scale controller design requires a detailed examination of the unit from basic process measurement to overall economic drivers. Successfully executing projects requires: an experienced and committed project team, clearly defined control objectives, robust and mature control software, a reliable DCS and data collection system, a well-designed regulatory control structure and accurate and, of course, reliable process measurement.

I want to focus on the last two – regulatory controls and process measurement. Although arguably the least glamorous, they act as the foundation for any large-scale controller and real time optimiser. Without this, the fastest and most elegant constrained optimal control algorithm will be of little practical value. So it’s ironic that this level of competency is normally weakest amongst practicing advanced control project engineers – and probably with good reason.

To gain a respectable level of knowledge in these two areas requires years of hands-on plant experience – working within a wide range of disciplines, often resolving problems on a trial-and-error basis. My intention here is to review related case studies I’ve encountered over several years of advanced control design and troubleshooting within the refining industry.

The real world

For the first case, we’ll look at a newly commissioned Dehexaniser 150 pound steam reboiler which failed to provide stable heat input into the column – resulting in poor temperature control. As shown here, steam flow was controlled with a throttling valve, and condensate removed from a condensate pot on level control. The condensate pot was tied to the tube-side of the exchanger with a 1in balancing line so that both condensate levels were kept the same.

Examination of the system showed the steam valve to be less than 5% open, the condensate valve and bypass both wide open, and the condensate pot level indication at full scale. As I had little experience in this area, I consulted several more senior engineers who provided me with two recommendations on how to improve the quality of temperature control.

The first was to reduce the steam throttling valve trim on the basis that is was over-sized. The second was to resize the condensate valve on the basis that it was sized for too large a pressure drop. Neither of these proposed solutions would have worked. The problem was actually solved by a senior operator, named Lou – experimenting on night-shift just three weeks before his retirement.

Lou’s idea was to connect a steam hose to the condensate pot, thereby providing the pressure required to `push’ the condensate out of the pot. I, of course, explained to Lou that this would not work as the steam would only flow back into the exchanger through the 1in balancing line, forcing the temperature controller to cut back on steam flow even further. Luckily, Lou was not deterred and proceeded to connect a steam hose to the condensate pot on night shift. When I arrived the following morning, Lou, with a smug grin, showed me a steam throttling valve and condensate valve operating close to mid range, and greatly improved temperature control.

I humbly accepted defeat and proceeded to analyse why Lou’s simple solution had worked; Q = UA T provided the answer. Given a fixed area, the steam flow is controlled by throttling steam pressure drop across the valve – hence steam temperature or exchanger T. The area term was so large that the T had to be kept small to achieve the desired Q. With such a low steam pressure, the condensate backed up into the tubes, reducing the available heat transfer area, until the steam pressure increased sufficiently to get the condensate away.

Any fluctuation in the throttling valve caused a change in T (a fast process) and condensate level (a slow process), resulting in poor temperature control. Lou’s solution worked because the steam flow back through the 1in balancing line caused sufficient pressure drop between the pot and exchanger to flood enough reboiler tubes at a controllable condensate pot level. The exchanger was over-designed.

And the moral? `Understand the process fundamentals’!

Lying instruments?

In case Two, a sponge absorber in a gas plant, designed to dissolve C4 and C5 hydrocarbons from fuel gas using a light cycle oil, was exhibiting symptoms of downcomer flooding. Flooding in a trayed column occurs when liquid flow is restricted through a downcomer due to excessive frictional losses or high tray pressure drop. Liquid then backs up and restricts flow through the downcomer above, causing it also to flood. Liquid begins to `stack’ in the column, resulting in a gradual loss of bottoms product and corresponding increase in column differential pressure.

Several times a day, the absorber’s bottoms flow rate would decrease and the differential pressure increase in a manner entirely consistent with downcomer flooding. But decreasing the liquid rate down the column, normally a quick remedy for flooding, had no effect. The column seemed to unload on its own, following which the differential pressure and bottoms flow quickly returned to normal.

After several days of this apparent flooding, the possibility of a faulty instrument was considered. A sticking base level indicator was the only scenario which seemed to fit the observed data. A quick check of the local sight glass during the next `flood’ revealed a bottoms liquid level greater than 100% – while the DCS indication remained at mid range.

The level float had been sticking with the measured value below the setpoint, causing the level controller to cut back on the bottoms flow. The base level increased until it covered the high pressure tap of the differential pressure cell, causing the reading to increase as the bottoms flow rate fell.

Moral: `Consider the lying measurement scenario’!