The University of Southampton
University of Southampton Institutional Repository

Efficient adaptive deep gradient RBF network For multi-output nonlinear and nonstationary industrial processes

Efficient adaptive deep gradient RBF network For multi-output nonlinear and nonstationary industrial processes
Efficient adaptive deep gradient RBF network For multi-output nonlinear and nonstationary industrial processes
Due to the complexity of process operation, industrial process data are often nonlinear and nonstationary, high dimensional, and multivariate with complex interactions between multiple outputs. To address all these issues, this paper proposes a novel industrial predictive model that integrates deep feature extraction and fast online adaptation, and can effectively deal with multiple process outputs. Specifically, a multi-output gradient radial basis function network (MGRBF) with excellent predictive capacity of nonstationary data is first used to provide preliminary prediction of target outputs. This prior quality information is combined with the original process input for deep feature learning and dimensional reduction. Through layer-wise feature extraction by the stacked autoencoder (SAE), deep quality-enhanced features can be obtained, which is further fed into a MGRBF tracker for online prediction. In order to timely capture the fast-changing process characteristics, the first two modules, namely, preliminary MGRBF predictor and SAE feature extractor are frozen after training, while the structure and parameters of the MGRBF tracker are updated online in an efficient manner. Two industrial case studies demonstrate that the proposed adaptive deep MGRBF network outperforms existing state-of-the-art online modeling approaches as well as deep learning models, in terms of both multi-output modeling accuracy and online computational complexity.
Multi-output gradient radial basis function network, Multivariate nonlinear and nonstationary industrial process, Online adaptive tracking, Quality-enhanced feature extraction, Stacked autoencoder
0959-1524
1-11
Liu, Tong
17f1a70b-449d-4078-af64-957a5b374698
Chen, Sheng
9310a111-f79a-48b8-98c7-383ca93cbb80
Yang, Po
5d8c6606-f845-4c52-8116-8d4a8cd22463
Zhu, Yunpeng
363c26ba-f671-48f6-a771-26bf881d0d1a
Harris, Chris J.
daa59e88-2e26-42df-bf2e-f8e6792ecb18
Liu, Tong
17f1a70b-449d-4078-af64-957a5b374698
Chen, Sheng
9310a111-f79a-48b8-98c7-383ca93cbb80
Yang, Po
5d8c6606-f845-4c52-8116-8d4a8cd22463
Zhu, Yunpeng
363c26ba-f671-48f6-a771-26bf881d0d1a
Harris, Chris J.
daa59e88-2e26-42df-bf2e-f8e6792ecb18

Liu, Tong, Chen, Sheng, Yang, Po, Zhu, Yunpeng and Harris, Chris J. (2023) Efficient adaptive deep gradient RBF network For multi-output nonlinear and nonstationary industrial processes. Journal of Process Control, 126, 1-11. (doi:10.1016/j.jprocont.2023.04.002).

Record type: Article

Abstract

Due to the complexity of process operation, industrial process data are often nonlinear and nonstationary, high dimensional, and multivariate with complex interactions between multiple outputs. To address all these issues, this paper proposes a novel industrial predictive model that integrates deep feature extraction and fast online adaptation, and can effectively deal with multiple process outputs. Specifically, a multi-output gradient radial basis function network (MGRBF) with excellent predictive capacity of nonstationary data is first used to provide preliminary prediction of target outputs. This prior quality information is combined with the original process input for deep feature learning and dimensional reduction. Through layer-wise feature extraction by the stacked autoencoder (SAE), deep quality-enhanced features can be obtained, which is further fed into a MGRBF tracker for online prediction. In order to timely capture the fast-changing process characteristics, the first two modules, namely, preliminary MGRBF predictor and SAE feature extractor are frozen after training, while the structure and parameters of the MGRBF tracker are updated online in an efficient manner. Two industrial case studies demonstrate that the proposed adaptive deep MGRBF network outperforms existing state-of-the-art online modeling approaches as well as deep learning models, in terms of both multi-output modeling accuracy and online computational complexity.

Text
DMGRBF_JPC-re1 - Accepted Manuscript
Available under License Creative Commons Attribution.
Download (855kB)
Text
1-s2.0-S0959152423000604-main - Version of Record
Restricted to Repository staff only
Request a copy

More information

Accepted/In Press date: 10 April 2023
e-pub ahead of print date: 2 May 2023
Published date: 1 June 2023
Additional Information: Funding Information: This work was supported by the Innovate UK under Project 107462 . All authors approved the version of the manuscript to be published.
Keywords: Multi-output gradient radial basis function network, Multivariate nonlinear and nonstationary industrial process, Online adaptive tracking, Quality-enhanced feature extraction, Stacked autoencoder

Identifiers

Local EPrints ID: 476691
URI: http://eprints.soton.ac.uk/id/eprint/476691
ISSN: 0959-1524
PURE UUID: b8414989-90e8-4c0f-88bc-95c8e3c862c2

Catalogue record

Date deposited: 11 May 2023 16:44
Last modified: 17 Mar 2024 01:43

Export record

Altmetrics

Contributors

Author: Tong Liu
Author: Sheng Chen
Author: Po Yang
Author: Yunpeng Zhu
Author: Chris J. Harris

Download statistics

Downloads from ePrints over the past year. Other digital versions may also be available to download e.g. from the publisher's website.

View more statistics

Atom RSS 1.0 RSS 2.0

Contact ePrints Soton: eprints@soton.ac.uk

ePrints Soton supports OAI 2.0 with a base URL of http://eprints.soton.ac.uk/cgi/oai2

This repository has been built using EPrints software, developed at the University of Southampton, but available to everyone to use.

We use cookies to ensure that we give you the best experience on our website. If you continue without changing your settings, we will assume that you are happy to receive cookies on the University of Southampton website.

×