(95h) Optimization and Control of An Industrial Packed-Bed Reactor for the Production of Ethylene Oxide

This work tackles the control problem of an existing industrial multi-tubular packed-bed reactor for the production of ethylene oxide. The production of ethylene oxide is based on the partial oxidation the ethylene with oxygen. Complete combustion of ethylene, as well as further oxidation of ethylene oxide results in the formation of carbon dioxide and water (no desirable products). This specific control problem can be re-formulated as the steady production of the ethylene oxide without compromising personnel safety and equipment integrity. However, temperature excursions at dangerous levels have been experienced due to variations on the inlet oxygen concentration in the fresh feed to the reactor and other no measurable disturbances. The stabilization of the process is achieved by manipulation of the inlet coolant and feed temperatures to the reactor and the addition of 1, 2-dichloroethene that acts as reaction moderator. Currently, the physical sensors for gas pressure and composition are located in the inlet and outlet streams of the reactor, respectively. The temperature, pressure and oxygen concentration measurements may be regarded as real-time, suitable for control purposes. However, the delayed response from the concentration analyser for ethylene oxide and other reaction by-products is suitable only for process monitoring. Since, the chemical reactions involved in the synthesis of ethylene oxide are highly exothermic, it makes the internal temperature control of this reaction unit a challenging task due to dangerous transient responses of the catalyst bed core temperature. The lack of real-time measurements for composition and temperature along the reactor handicaps the effort to implement control strategies. In practice, a model-based control strategy requires real-time measurements of the temperature and composition profile along the reactor which are not available in this reaction unit. In order to exploit the fact that some measurements are available, the problem associated with the need of real-time measurements along the reactor can be overcome with the design of an observer. This method combines a priori knowledge about the system (plant data) with a mathematical model to provide a real-time estimation of the temperature and composition profiles along the reaction vessel. The observer synthesis derives from traditional Extended Kalman Filter and more recent approaches based-on high-gain techniques. The observer or soft-sensor is capable of estimating the ethylene oxide concentration in the reactor outlet stream and the catalyst bed temperature along the reactor from the real-time measurements. The observer is also corrected when the composition analyser produces an output. The observer-based control is robust and it presents a good closed-performance despite modelling mismatch and sustained disturbances.