About this Abstract |
| Meeting |
2026 TMS Annual Meeting & Exhibition
|
| Symposium
|
Measurement & Control of High Temperature Processes
|
| Presentation Title |
Process-Aware Multiscale Temporal Modeling for Working Condition Identification in the Traveling Grate Pelletizing Processes |
| Author(s) |
XuLing Chen, LuanFeng Li, ZhenXiang Feng, XiaoHui Fan, XiaoXian Huang |
| On-Site Speaker (Planned) |
|
| Abstract Scope |
Accurate identification of working conditions in traveling grate pelletizing processes is critical for ensuring pellet quality and enhancing economic efficiency. However, feed fluctuations cause variable pellet residence times, disrupting the actual reaction atmosphere experienced. Traditional alignment methods link process data to product quality timestamps but overlook transport delays and internal dynamics, leading to inaccurate working condition identification. To address this issue, this paper proposes a process-aware multiscale temporal modeling approach that integrates dynamic time-window alignment with a multiscale temporal perception network. Parallel temporal convolutional networks are employed to extract dynamic features across multiple time scales, while a spatio-temporal attention mechanism adaptively focuses on critical time segments and important variables, thereby enhancing the identification of complex working conditions. Validation on industrial data from a large-scale traveling grate pelletizing process demonstrates that the proposed method shows robust and accurate performance, underscoring its strong potential for practical application. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Iron and Steel, Modeling and Simulation, Machine Learning |