Toggle Main Menu Toggle Search

Open Access padlockePrints

Nonlinear Data-Driven Process Modelling Using Slow Feature Analysis and Neural Networks

Lookup NU author(s): Jeremiah Corrigan, Dr Jie Zhang

Downloads

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


Abstract

Slow feature analysis is a technique that extracts slowly varying latent variables from a dataset. These latentvariables, known as slow features, can capture underlying dynamics when applied to process data, leading toimproved generalisation when a data-driven model is built with these slow features. A method utilising slowfeature analysis with neural networks is proposed in this paper for improving generalisation in nonlineardynamic process modelling. Additionally, a method for selecting the number of dominant slow features usingchanges in slowness is proposed. The proposed method is applied to creating a soft sensor for estimatingpolymer melt index in an industrial polymerisation process to validate the method’s performance. Theproposed method is compared with principal component analysis-neural network and a neural networkwithout any latent variable method. The results from this industrial application demonstrate the effectivenessof the proposed method for improving model generalisation capability and reducing dimensionality.


Publication metadata

Author(s): Corrigan J, Zhang J

Editor(s): Oleg Gusikhin, Kurosh Madani, Janan Zaytoon

Publication type: Conference Proceedings (inc. Abstract)

Publication status: Published

Conference Name: 16th International Conference on Informatics in Control, Automation and Robotics (ICINCO 2019)

Year of Conference: 2019

Number of Volumes: 2

Pages: 439-446

Online publication date: 29/07/2019

Acceptance date: 23/05/2019

Publisher: SciTePress

URL: http://insticc.org/node/TechnicalProgram/icinco/presentationDetails/79589

Library holdings: Search Newcastle University Library for this item

ISBN: 9789897583803


Actions

Link to this publication


Share