Development of a novel parallel framework upon deep dual‐enhanced autoencoder and its applications for industrial soft sensing
In recent years, data‐driven soft sensing technology has provided a cost‐effective support for industrial process monitoring, in which autoencoder plays an important role in extracting features for soft sensing technology. However, existing autoencoder models take a long time for modelling, and the...
Saved in:
| Published in: | Canadian journal of chemical engineering |
|---|---|
| Main Authors: | , , |
| Format: | Journal Article |
| Language: | English |
| Published: |
22.07.2025
|
| ISSN: | 0008-4034, 1939-019X |
| Online Access: | Get full text |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | In recent years, data‐driven soft sensing technology has provided a cost‐effective support for industrial process monitoring, in which autoencoder plays an important role in extracting features for soft sensing technology. However, existing autoencoder models take a long time for modelling, and the extracted features have difficulty considering both key variable and process variables. To address these issues, this paper proposes a parallel gated deep input‐enhanced supervised autoencoder (PGDISAE) model. Different from the standard deep autoencoder model, the proposed model improves the feature extraction performance of deep autoencoder by embedding input variables into hidden layers through supervised learning in the pre‐training stage while embedding output variable into decoder layer. The gated strategy makes full use of the abstract representation of each hidden layer. Additionally, this model adopts a data parallel training strategy which can drastically reduce the training time. The experimental results on the sulphur recovery unit and primary reformer process verify the effectiveness and feasibility of the proposed model in senses of accuracy and time‐efficiency. |
|---|---|
| ISSN: | 0008-4034 1939-019X |
| DOI: | 10.1002/cjce.70051 |