Batch-to-Batch Iterative Learning Control Based on Kernel Independent Component Regression Model

摘要

A model-based batch-to-batch iterative learning control (ILC) strategy for batch processes is proposed in this paper. The data-driven model of batch process is developed using kernel independent component regression (KICR) method when the operating data of batch process have a non-Gaussian distribution. The ILC algorithm is derived based on the linearization of the KICR model around the control profile. Applications to a simulated nonlinear batch reactor demonstrate that the proposed ILC strategy can improve process performance from batch to batch when the operating data of batch process follow non-Gaussian distribution. Comparisons between KICR model based and support vector regression (SVR) model based ILC strategies are also made in the simulation. The results show the KICR model based ILC has better performance.