.png){kind=icon}
.png){kind=icon}
Return
A multimodal data fusion machine learning model for recognising top gas flow distribution patterns in blast furnaces
.png){kind=icon}
PRE
AI
Y
C
S
X
F
D
S
DOI：10.1007/s42243-026-01901-5.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
A stable and reasonable gas flow distribution is critical to ensure the stable and efficient operation of blast furnaces, which directly impacts gas utilisation rate and overall smelting performance. However, owing to the harsh internal environment of blast furnaces and the limitations of current sensing technologies, real-time recognition of gas flow distribution patterns is still a major challenge. To address this issue, a central gas flow distribution pattern recognition model was established based on machine learning, which employed multimodal data fusion to integrate infrared images with cross temperature measurements. Specifically, three key feature parameters of the central gas flow, namely, area, temperature, and offset degree, were extracted using entropy and neighbourhood valley-enhanced Otsu method, heat transfer principles, and dual-point tilt correction method, respectively. The convolutional neural network–long short-term memory model was used to forecast the temporal dynamics of these parameters. Through the feature parameters, the proposed Euclidean-weighted fuzzy C-means algorithm was applied to recognise the central gas flow distribution patterns. The results show that the model achieves high accuracy over 95% in parameters prediction and exceeds 90% in pattern recognition, highlighting its potential for intelligent monitoring and control of central gas flow behaviour in blast furnaces.
Keywords:
Furnace infrared image
Cross temperature measurement
Gas flow distribution
Feature extraction
Multimodal data fusion
Journal
IF:
3.6
Papers:
3.6K
Citations:
6.1K
Organization
Cited Papers