Plant-Wide Control Using an Inferential Approach

This paper focuses on an industrial test problem, the Tennessee Eastman problem. A decentralized PID control system for the Tennessee Eastman plant was developed in an earlier paper. An inferential feedforward/feedback control structure added on the base control is propoed here to improve the control of product properties during random feed disturbances. PLS models are developed based on steady state data and used to implicitly predict the unmeasurable disturbances. Dynamic simulation results show that control of product flow rate and compositions is significantly improved by using this control structure.