Devising Effective Control Strategies: A Case Study
Dan Hebert | May 05, 2015Creating a stable process often depends on selecting the best process variables to help regulate a critical element of a larger system, as with this lime kiln.
Engineers and designers often discover unanticipated negative side effects as a result of trying to solve a control problem. Here’s an example where this happened and was addressed, with company names omitted.
The company was a lime producer wanting to reopen an old plant situated with its own quarry. The process could not be more straightforward: There was a long rotary kiln where limestone gravel moved toward the burner. Product coming out glowed orange and was quickly cooled. Since the production facility was more than 50 years old, it reflected the technology of its time and had essentially nothing in the way of air pollution abatement equipment. The company kept the basic rotary kiln, but added a baghouse to capture particulates. The kiln had been fired with pulverized coal but could also burn natural gas.
Process engineers analyzing the equipment and designing the refurbishments calculated the flue gas off the kiln would be too hot to go to the baghouse directly. So they added a cooling vessel equipped with an atomized water spray cooling system to reduce the gas temperature and volume to a level the baghouse could handle.
A system integrator was selected to work with a spray nozzle company to design the pumping system to drive the nozzles. These nozzles would be controlled by a temperature loop to modulate the spray volume and achieve the desired setpoint, defined as the gas temperature at the entry of the baghouse.
The control system also measured gas temperature at the cooling vessel entrance, but this was simply for reference and was not used in the control strategy. The concept was simple and should have worked without any major problems. However, reality intervened and two difficulties emerged.
Incorrect Fuel Calculations
The first problem was that once the process started up, it became clear the heat recovery capabilities of the old kiln were not what the designers had expected. It took a lot more fuel than the calculations had suggested to heat the limestone to the desired temperatures. This reduced the efficiency of the process overall which was not good, but it also created a problem for the gas cooling system.
The temperature difference from the kiln exit to baghouse entry turned out to be much greater than the system had been designed to handle. The cooling system had enough excess capacity to inject enough water to cover the difference, but the cooling vessel was now effectively too small. Such cooling systems are designed for complete water evaporation in the hot gas stream. The finely atomized droplets must be turned into steam before wetting any internal surfaces, otherwise dust sticks on the wet surface and can quickly build up. This system now ran on a very delicate edge between dry and wet, good and bad.
Unexpected Dust Loading
The second problem was that the level of dust loading in the gas stream was higher than expected. It tended to deposit in some areas in the ducts and the baghouse had to be emptied more often than planned. This had a compounding effect since more feedstock had to be fed into the kiln to maintain production as product was lost to dust.
The ultimate effect was a compounding of all these problems as all the elements ratcheted up in order to reach necessary production levels. Even though these problems were not trivial and added to production costs, once operators got some experience with the refurbished plant, they proved to be manageable when things were running well.
But there was one more problem: the kiln’s burner system proved to be temperamental and had occasional flameouts. When those happened, operators had to restart the burner manually, which took at least several minutes. Meanwhile, the induced-draft fan would continue to suck air through the kiln, cooling the vessel and baghouse so that the gas temperature would begin to fall quickly. When this cooler air hit the gas cooling vessel, the amount of heat available was no longer enough to flash off the water which would drench the lower walls and bottom of the vessel, causing airborne dust to build up quickly.
The cooling system was controlled by the baghouse entry temperature, so the cool gas had to reach the temperature sensor before it would turn the spray down. This was made worse by the natural time lag of the sensor itself. It could take two to four minutes before the system would compensate by itself and reduce water flow in the cooling vessel.
Once operators got the burner going again, the tower drenching sequence was followed by an opposite situation where a slug of much hotter air would come through the system, hit the cooling vessel with the sprays turned down and reach the baghouse at an excessive temperature until the system was able to stabilize itself.
Eventually, the operators created a manual sequence they followed whenever the flameout alarm sounded. While one operator restarted the burner, a second had to count down the time for the cold air to reach the cooling vessel, turn the sprays down to a predetermined level before wetting the cooling vessel and then turn the sprays back up when the burner got going again. Over time, flameouts became less frequent, but never entirely went away.
Lessons Learned
The biggest avoidable problem was the nature of the control strategy itself. Running the cooling system based on a process variable subject to such long dead time was asking for trouble. It would have been a simple matter to monitor temperature somewhere farther upstream and incorporate it into the cooling calculation. Moreover, the most useful bit of information (“Is the burner on?”) did not enter into the calculation at all.
Based on typical airflow velocities, it would have been a simple matter to build in a time lag and then automatically turn the sprays down some number of seconds after a flameout, then turn the sprays back up when the burner restarted. The latent heat of the equipment and product moving through the kiln would provide a stabilizing effect to keep temperature changes from being large and immediate, and so provide time to react.
Solution
The solution was an automated version of the manual procedure the operators created. The status of the flame became a feed-forward indication, and the amount of water injected into the tower moved automatically to a predetermined level after a specific period of time in the event of a flameout.
The amount of time to make the adjustment was calculated based on the amount of airflow through the kiln as it determined when cooler air would reach the cooling vessel. The airflow measurement was already monitored, so no instrumentation additions to the system were necessary.
The temperature characteristics of the kiln were such that temperature changes were not immediate, so there was some period of adjustment before disaster struck in the form of a wet tower or burned bags.
Over time, the company made modifications to the kiln to improve its heat recovery characteristics, which improved process efficiency and reduced the need for gas cooling. Process performance became more stable and a deeper integration of the gas cooling system into the control strategy made life much easier for the operators.