Soft Sensor Based on Fusion Modeling with the Balanced Spatio-temporal Feature Network

In recent years, multi－feature integrated method based on deep neutral network has been widely applied to soft sensor． However, existing methods rarely consider the independence between feature extracted so that same features can be affected or even submerged．In this paper, a stacked variational aut...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Chinese Control Conference S. 6851 - 6856
Hauptverfasser: Tang, Xiaochu, Wang, Yiting, Tao, Na, Yu, Yang
Format: Tagungsbericht
Sprache:Englisch
Veröffentlicht: Technical Committee on Control Theory, Chinese Association of Automation 28.07.2025
Schlagworte:
Accuracy
Autoencoders
BST-network
Convolution
convolution neural network
Correlation
Decoding
Feature extraction
Long short term memory
long short-term memory network
Neural networks
soft sensor
Soft sensors
stacked variational autoencoder
Time series analysis
ISSN:1934-1768
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:In recent years, multi－feature integrated method based on deep neutral network has been widely applied to soft sensor． However, existing methods rarely consider the independence between feature extracted so that same features can be affected or even submerged．In this paper, a stacked variational autoencoder based on balanced spatio－temporal feature network（BST－ SVAE）is proposed to keep spatio feature and temporal feature in a balance state．In this way, the original features can be well preserved avoiding mutual influence．Firstly, a balanced parallel network including CNN and LSTM is used to design encoders and decoders to form a balanced spatio－temporal feature network．It can consider the interaction between internal relations and mine the time series relationship．Furthermore, stacked variational autoencoder（SVAE）is construed based on a balanced spatio－ temporal feature network．Finally, the proposed method is applied to the actual industrial process．Experimental results demonstrate the superiority and effectiveness of the proposed method．
ISSN:1934-1768
DOI:10.23919/CCC64809.2025.11178652