Identification of Process Networks with Recycle Streams

Abstract

Processes with material and energy recycle are frequently encountered in chemical process industries. Most of the existing data-driven model identification techniques assume that the underlying process units are without any recycle. While the ideas of traditional closed-loop system identification are useful in dealing with recycle streams, identifying each of the subsystem models embedded in the process network in the presence of unmeasured disturbances and measurement noise requires careful consideration of the underlying process topology and a systematic step-wise procedure. In this work, we utilize the existing ideas of two-stage (SISO) identification of dynamical networks in combination with process topology to efficiently identify process networks with recycle streams. We demonstrate the efficacy of the approach through simulation case studies comprising different levels of process network complexity.