Adding a collector-distributor allowed direct measurement of vapor flow, which was critical to achieving stable control.
A variety of factors complicated the control problem:
* An uncontrolled hot-oil system with large variations in hot-oil temperature was causing large heat-input disturbances;
* The column was grossly oversized;
* Internals had been selected to allow for future vacuum service, even though current operation was at atmospheric pressure;
* The differential pressure between the column and the vapor side-draw condenser was very low;
* The temperature profile in the column was extremely sharp;
* Temperature control points were located incorrectly for inferential control;
* The liquid inventory in the column bottoms was too low for the recirculating heater control;
* The feed preheat was unstable;
* No control valve existed on the vapor side-draw, so vapor draw rate was instead controlled by operating pressure;
* The overhead condenser had an insufficient liquid-level range (6 in.); and
* The side-draw condenser had an insufficient liquid-level range (6 in.).
The problem here was that we had to control a vapor draw that was a large fraction of the total column feed using highly unstable heat input. Not only was the heat input unstable, but constraints on the hot-oil system made measuring it essentially impossible except by back-calculating from the tower performance.
To arrive at a fix, we did not restrict ourselves to the information we already had. Instead, we asked ourselves what information the process had to generate for a controllable tower. We needed a measure of vapor traffic up the column , so we could make sure that the vapor side-draw rate was sufficiently less than the column vapor traffic to allow for proper distillation. Vapor traffic in a distillation column directly depends upon heat input into the column. Most units can measure heat input; we could not.
The solution was to directly measure vapor rate (Figure). A collector-distributor was installed in the column that was specially designed to impose a significant back-pressure. Essentially, we added an orifice plate inside the column. The differential pressure across the collector enabled direct measurement of vapor flow. The collector was sized to allow both for current atmospheric-pressure service and future vacuum operation without flooding.
Due to massive oversizing of the column, the packing pressure drop was nearly constant at atmospheric pressure operation. Without the added restriction at the collector, pressure drop did not provide enough signal gain to allow for control.
Yes, this system is highly interactive, but it works for this case. The results: capacity increased to about 70% of design rates and product purity increased to more than 98%. Capacity is now restricted by too small a reboiler and hot-oil loop problems (more on these another time).
A partial list of supporting changes included a complete redesign of the column internals; relocation of temperature control points; switching to a dynamic temperature threshold for composition control; use of an on/off level control for bottoms; installing a control valve in the vapor draw line; and adding level controls to the overhead condenser and the side-draw condenser.
We will discuss this troubled column several more times, because there are many other things we can learn from its host of problems.
