Industrial IoT–enabled real-time prediction of strip cross-section shape for hot-rolling steel

Because strip cross sections cannot be obtained during hot rolling in advance, traditional automatic shape control systems can only rely on the measured shape at the exit of the final mill for feedback control, which causes a significant lag and poor adjustment effect. To accurately predict the cros...

Full description

Saved in:
Bibliographic Details
Published in:International journal of advanced manufacturing technology 2024, Vol.130 (1-2), p.961-972
Main Authors: Sun, Youzhao, Li, Jingdong, Li, Hongfan, Sun, Yamin, Wang, Xiaochen, Yang, Quan
Format: Article
Language:English
Subjects:
Accuracy
Advanced manufacturing technologies
Algorithms
Automatic control
Automation
CAE) and Design
Computer-Aided Engineering (CAD
Cross-sections
Dynamical systems
Engineering
Feedback control
Hot rolling
Hot-rolled steel
Industrial and Production Engineering
Industrial applications
Internet of Things
Manufacturing
Mechanical Engineering
Media Management
Methods
Neural networks
Nonlinear dynamics
Original Article
Prediction models
Process controls
Production lines
Real time
Shape control
Strip
Citations: Items that this one cites
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Because strip cross sections cannot be obtained during hot rolling in advance, traditional automatic shape control systems can only rely on the measured shape at the exit of the final mill for feedback control, which causes a significant lag and poor adjustment effect. To accurately predict the cross-sectional shape, an industrial Internet of things platform for steel plants is developed to collect real-time production data. A novel real-time prediction model that can determine the cross sections of strips is proposed to address the drawbacks of traditional data-driven methods that perform offline predictions. This model is established by adopting a dynamic mode decomposition algorithm (DMD) to optimize the sparse identification of nonlinear dynamics (SINDy). A practical dataset of 81 variables from a 2250-mm hot-rolling production line is utilized to validate the proposed method, and this method is compared with SINDy, EMD-optimized SINDy, and VMD-optimized SINDy. The experimental results show that the proposed method can achieve higher prediction accuracy and more minor errors.
ISSN:0268-3768
1433-3015
DOI:10.1007/s00170-023-12745-8