Toggle Main Menu Toggle Search

Open Access padlockePrints

Reliable data-driven modelling and optimisation of a batch reactor using bootstrap aggregated deep belief networks

Lookup NU author(s): Dr Jie ZhangORCiD

Downloads

Full text for this publication is not currently held within this repository. Alternative links are provided below where available.


Abstract

To enhance the generalisation performance of DBN models, instead of building just one DBN model, several DBN models are developed from bootstrap re-sampling replication of the original modelling data and these DBN models are combined together to form a bootstrap aggregated DBN model (BAGDBN). BAGDBN is used for the modelling of a batch reactor and its generalisation performance is significantly better that that of a DBN model. Furthermore, model prediction confidence bounds can be readily obtained from the individual DBN model predictions and can be incorporated into the batch reactor optimisation framework to enhance the reliability of the resulting optimal control policy. Wide model prediction confidence bound is penalised to enhance the reliability of optimisation.


Publication metadata

Author(s): Zhu C, Zhang J

Publication type: Conference Proceedings (inc. Abstract)

Publication status: Published

Conference Name: 50th Meeting of the Italian Statistical Society (SIS 2021)

Year of Conference: 2021

Pages: 94-99

Acceptance date: 30/05/2021

Publisher: Pearson

Library holdings: Search Newcastle University Library for this item

ISBN: 9788891927361


Share