.png){kind=icon}
.png){kind=icon}
Return
Enhancing interpretable soft sensing with embedded hybrid modeling: the GraphTrans approach for industrial processes
.png){kind=icon}
PRE
AI
P
B
K
李
C
DOI：10.1016/j.isatra.2026.06.036.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
• An interpretable hybrid framework with GraphTrans is proposed for parameter identification and accurate soft sensing. • GraphTrans uses graph convolutions and multi-head attention for long-range dependencies and higher-order interactions. • A graph mask matrix in multi-head attention selectively focuses on causally relevant interactions among variables. • The kernel projection module mitigates missing data effects by mapping incomplete observations to high-dimensional space. • The hybrid framework detects abnormal samples via parameter variations, enabling fault diagnosis and process monitoring.
Keywords:
Interpretable soft sensing
Embedded hybrid modeling
Graph convolution
Multi-head attention
Graph mask matrix
Kernel projection
Journal
IF:
6.5
Papers:
5.9K
Citations:
2.0W
Organization
No organization information available
Cited Papers