.png){kind=icon}
.png){kind=icon}
Return
Enhancing knowledge-data operating optimization for process industries with varying conditions
.png){kind=icon}
PRE
AI
DOI：10.1016/j.conengprac.2026.106844.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
The process industries production are time-varying and spatially coupled, with dynamic operating conditions. Relying on a single process operation makes it difficult to consistently ensure stable product quality under various changing conditions. In this paper, an operation optimization strategy under condition variation and critical parameter identification is proposed to improve capacity and stability of quality. Firstly, a conditional spatio-temporal graph attention network is designed for operating condition recognition, leveraging integrated spatio-temporal operational data to dynamically capture relationships among equipment. Secondly, a random forest algorithm is employed for optimizing the key process parameters that affect product quality, using out-of-bag permutation importance for quantitative assessment. Finally, an adaptive neuro-fuzzy inference system is constructed for improving the product qualification rate, incorporating expert knowledge to formulate dynamic adjustment rules. The effectiveness of the proposed method is validated through alcohol purification and oil refining processes.
Keywords:
condition variation
spatio-temporal graph attention network
random forest
adaptive neuro-fuzzy inference system
process optimization
Journal
IF:
4.6
Papers:
5.7K
Citations:
1.1W
Organization
No organization information available