Browse by author
Lookup NU author(s): Dr Jie ZhangORCiD
Full text for this publication is not currently held within this repository. Alternative links are provided below where available.
© 2017 IEEE. This paper presents reliable multi-objective on-linere-optimisation control of a fed-batch fermentation processusing bootstrap aggregated neural networks. To overcome thedifficulty in developing mechanistic models, data driven neuralnetwork models are developed from process operational data.However, a single neural network can lack robustness in thatits performance on unseen data might be unsatisfactoryespecially when the amount of training data is limited. Toovercome this problem, multiple neural networks aredeveloped from bootstrap re-sampling replications of theoriginal training data and they are combined. A furtheradvantage of bootstrap aggregated neural networks is thatmodel prediction confidence bounds can be obtained fromindividual network predictions. Model prediction confidencebounds are incorporated into a multi-objective optimisationframework in order to enhance the reliability of optimisationresults. In order to further reduce the effect of model plantmismatches and unknown disturbances on the optimisationresults, on-line re-optimisation is used in this study. Theproposed method is applied to a simulated industrial scale fed-batch fermentation process for producing baker's yeast.
Author(s): Zhang J, Fisher R
Publication type: Conference Proceedings (inc. Abstract)
Publication status: Published
Conference Name: 2017 International Symposium on Computer Science and Intelligent Controls, ISCSIC 2017
Year of Conference: 2018
Online publication date: 19/02/2018
Acceptance date: 20/10/2017
Publisher: Institute of Electrical and Electronics Engineers Inc.
URL: https://doi.org/10.1109/ISCSIC.2017.23
DOI: 10.1109/ISCSIC.2017.23
Library holdings: Search Newcastle University Library for this item
ISBN: 9781538629413
