Model Predictive Control for Ethylene Oxide Explosion Limits Under Plant Upset Conditions

In this paper, a model predictive controller is designed to for an ethylene oxide plant using industry-standard distributed control system and dynamic process simulation in tandem. The controller is tested for ability to remain within explosion limits and ability to maintain optimal operation parameters under normal operation and plant upset conditions, including sudden failure in the lean cycle water stripping column and cooling duty availability in the reactor cooling systems.