CVAE ‐ SFA : A practical dual‐channel and multi‐index comprehensive operating performance assessment method for dynamic nonstationary industrial processes
The process operating performance assessment (POPA) of process industry plays a key role in the safe, stable, and efficient operation of production process. However, the actual industrial production process often exhibits nonstationary characteristics, leading to frequent anomalies that complicate P...
Uloženo v:
| Vydáno v: | Canadian journal of chemical engineering |
|---|---|
| Hlavní autoři: | , , , , , |
| Médium: | Journal Article |
| Jazyk: | angličtina |
| Vydáno: |
15.07.2025
|
| ISSN: | 0008-4034, 1939-019X |
| On-line přístup: | Získat plný text |
| Tagy: |
Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
|
| Shrnutí: | The process operating performance assessment (POPA) of process industry plays a key role in the safe, stable, and efficient operation of production process. However, the actual industrial production process often exhibits nonstationary characteristics, leading to frequent anomalies that complicate POPA. Traditional POPA methods are mostly based on limited stable working conditions and single control limits, so it is difficult to fully evaluate the anomaly degree and lack comprehensive assessment indexes. To address these issues, this paper proposes a CVAE‐SFA dual‐channel multi‐index comprehensive POPA method for nonstationary conditions. It introduces four control limits of ‘optimal, good, general, and poor’ based on the optimal reconstruction error from a conditional variational autoencoder (CVAE) and kernel density estimation (KDE). This creates a CVAE optimal grade classification channel for simultaneous qualitative and quantitative evaluations of current operating conditions. Additionally, to monitor dynamic characteristics under nonstationary conditions, a slow feature analysis (SFA) statistic anomaly monitoring channel is established. By identifying slow‐changing essential features within the system, dynamic anomaly monitoring is achieved. Finally, a comprehensive scoring index (CSI) integrates dual‐channel monitoring results to evaluate quality, yield, energy consumption, and stability. The effectiveness of this method is validated through experiments in the cement clinker production process (CCPP). |
|---|---|
| ISSN: | 0008-4034 1939-019X |
| DOI: | 10.1002/cjce.70028 |