Priori Knowledge-Based Online Batch-to-Batch Identification in a Closed Loop and an Application to Injection Molding

Abstract

Online identification of a plant based on the input–output data obtained under closed-loop operation conditions is a fundamental problem in many industrial applications, because it paves the way for process monitoring, controller calibration, controller redesign, etc. This paper proposes a recursive batch-to-batch identification method for modeling a batch process regulated by a within-batch controller by exploiting its intrinsic repeatable pattern. To overcome severe variations of parameter estimates, three kinds of priori plant knowledge are included, namely, time constant, static gain, and stability. The constraints formulated from priori plant knowledge are efficiently handled by a technique–sequential projection. Additionally, the approach’s robustness against interbatch dynamics drift is analyzed mathematically. Finally, the performance of the proposed method is demonstrated on a numerical benchmark and a real industrial application–injection molding.