Title: Neural network based iterative learning control for product qualities in batch processes

Authors: Zhihua Xiong, Yixin Xu, Jin Dong, Jie Zhang

Addresses: Department of Automation, Tsinghua University, Beijing 100084, China. ' Department of Automation, Tsinghua University, Beijing 100084, China. ' Supply Chain Management and Logistics, IBM China Research Lab, Beijing, 100094, China. ' School of Chemical Engineering and Advanced Materials, Newcastle University, Newcastle upon Tyne, NE1 7RU, UK

Abstract: A neural network based iterative learning control (NN-ILC) strategy is proposed to improve the product qualities in batch processes. Based on the repetitive nature of batch processes, iterative learning control (ILC) is used to improve product qualities gradually from batch to batch. The learning gain in the ILC is usually determined according to a linearised model. Instead of building a model for the system dynamics, a feed-forward neural network (FNN) is used directly as a non-linear learning gain in the ILC law. The tracking error profile of the previous batch is used as the input of the FNN, while the output of the network is the control change profile for the next batch run. It has been proved that if the network is trained properly based on the historical operation data, the tracking error under the proposed NN-ILC can converge to zero gradually with respect to the batch number. The neural network can also be retrained during the ILC to renew the learning gain in order to handle model uncertainties of the batch processes. The proposed control strategy is illustrated on a typical batch reactor.

Keywords: iterative learning control; ILC; neural networks; batch processing; product quality; nonlinear learning gain; tracking error; model uncertainties; control strategy; batch reactors.

DOI: 10.1504/IJMIC.2010.035285

International Journal of Modelling, Identification and Control, 2010 Vol.11 No.1/2, pp.107 - 114

Published online: 20 Sep 2010 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article