期刊
ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE
卷 128, 期 -, 页码 -出版社
PERGAMON-ELSEVIER SCIENCE LTD
DOI: 10.1016/j.engappai.2023.107517
关键词
Particle size distribution; Blast furnace; Semantic segmentation; Kinetic modeling
In this paper, a method for estimating the overall particle size distribution (PSD) based on mechanistic modeling and local-global fused prediction networks is proposed. The method can accurately predict the PSD in harsh production environments with limited detection conditions.
Estimating the overall particle size distribution (PSD) on the conveyor is one of the most important means of monitoring blast furnace production. Due to the harsh production environment and limited detection conditions, existing online detection methods can only estimate part of the PSD, but it is difficult to predict the overall PSD. Unlike segmentation methods that can only obtain surface PSD, we propose an overall PSD estimation method based on mechanistic modeling and local-global fused prediction networks. First, we constructed a kinetic mechanism model of particle accumulation, from which global field distribution features such as coordination number, moment of inertia, and spatial distribution were extracted. Then, a particle segmentation model is trained using surface images collected by detection means such as cameras to obtain surface PSDs as local surface features. Next, to predict the overall PSD, we propose a Local surface and Global field distribution feature Fusion-based Transformer network (LGFT). Among them, in order to better aggregate local and global features, we introduce orthogonal fusion and point cloud extractors in the encoder. Finally, we verify the accuracy and efficiency of our method through extensive ablation experiments.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据