“QuickStudy significantly reduced variation in the acetic acid reactor temperature and increased throughput by more than 5 percent. And our operators love it.” – Trevor Arnold, Sterling Chemicals, Senior Process Control Engineer
Sterling Chemicals operates an acetic acid production unit at its Texas City, TX site. Carbon monoxide and methanol are reacted in the presence of a catalyst to produce acetic acid, which is purified in the associated distillation train. Greater production efficiency was achieved through the use of an improved catalyst but with the penalty of decreased stability of reactor temperature control. The PID based control strategy was adequate most of the time but could not prevent temperature excursions during upset conditions. The selection of a satisfactory advanced control product was complicated by the fact that for safety reasons, the reactor could not be bump tested because the reaction is exothermic.
The Control Objectives
There are several control objectives used for a carbonylation type acetic acid reactor, such as maintaining constant compositions and heat/mass balance. The main control variable is the reactor temperature.
Maintaining reactor temperature is complicated by the interaction of feed, effluent and recycle flows with the exothermic carbonylation reaction. The primary method of heat removal from the reactor is via the effluent flow. Acetic acid is extracted from this flow in a flash vessel and distillation train before the remaining reactants and catalyst solution are returned to the reactor at a lower enthalpy. The effluent flow cannot be manipulated directly to control temperature because it also affects the mass balance, therefore reactor fluid is circulated through a boiler which is used to remove the extra heat. PID control was initially used to control the boiler heat removal, but it suffered from instability at high unit throughput. This resulted in having to reduce the production rate, stabilize the unit and then increase rates again.
The QuickStudy™ Solution
Prior to installing a QuickStudy adaptive process controller on the acetic acid reactor, it was used to automatically create process models off-line from historical data. The QuickStudy adaptive controller was then then placed in simulate mode to determine how well the models could predict the reactor temperature using additional production data. Visually the correlation between actual and predicted temperature was very good. Also the steady state gains generated by QuickStudy were similar to those in a calibrated unit model.
The next step was to replace the reactor PID temperature controller with a QuickStudy predictive controller set-up (PCS) block which manipulated the existing boiler pressure controller. Dynamic models of the relationships between the temperature and reactor disturbance variables, which had been developed off-line, were loaded. After a brief period of on-line learning the QuickStudy block was placed in control. A significant improvement in temperature stability was seen within hours as QuickStudy further refined its internal models on-line. Disturbances that would normally have caused a temperature upset, such as an effluent flow change, were attenuated. The whole process was carried out without the need for bump tests in 2 days.
An “off the shelf” PC, running Windows NT operating system was used as the platform to implement the QuickStudy software. This was connected to a Fisher Provox DCS via Highway Data Link interface hardware and a 32 bit DDE driver software package. Operators can access operating features of QuickStudy on the Standard DCS graphics and faceplates, while control engineers access the configuration environment via the PC screen. Data and communications checks are carried out within the DCS to prevent plant upset due to hardware failures or data corruption.
Using QuickStudy to solve the reactor temperature control problem produced these results:
- Reduced temperature standard deviation from 3.6 to 0.8 • Eliminated high temperature trips
- Stabilized purification columns
- Increased sustained throughput by over 5%
- Increased unit availability QuickStudy has performed reliably in this application for over 18 months. It was fast and easy to install, without bump tests, and its adaptive capability has made model maintenance a simple task
PID V/S QuickStudy process temperature control