Rolling Load Prediction Using Temperature-Aware Neural Networks

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing rolling load prediction models struggle to accurately predict rolling loads due to the difficulty in measuring steel temperature and limited operation record data, leading to suboptimal rolling quality and efficiency.

Innovation Solution

Incorporating temperature-related factors, such as steel, slab, and skid rail temperatures, into the input variables for a neural network-based rolling load prediction model, allowing for more accurate predictions using machine learning methods.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If conventional rolling load prediction models are used without temperature data, then the model complexity remains low, but the prediction accuracy deteriorates

Engineering Contradiction:
Improveprediction accuracyVSAvoidmodel complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The system performs preliminary measurement and recording of steel temperature, slab temperature, and skid rail temperature during the rolling process. These temperature data are collected and stored in advance as training data for the neural network model, enabling the model to learn temperature-related patterns before actual prediction is needed.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

A neural network model is introduced as an intermediary between the input parameters (including temperature data) and the rolling load prediction. The neural network processes the complex relationships between multiple parameters and temperature effects, providing accurate predictions without requiring direct complex physical modeling of thermal effects.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Measurement precision

If temperature measurement and data collection systems are added, then prediction accuracy improves, but the ease of operation deteriorates

Engineering Contradiction:
Improveprediction accuracyVSAvoidease of operation
Core Design Contradiction:
Measurement precisionVSEase of operation

Solution Approach 1:

The neural network model is designed to handle multiple input parameters simultaneously, including steel temperature, slab temperature, skid rail temperature, and other rolling parameters. This multi-functional approach consolidates various measurement requirements into a single unified prediction system, reducing the operational burden despite the increased data collection needs.

Inventive Principle:
Principle #6Universality (Multi-functionality)

3Measurement precision

If more operation record data including temperature are collected, then the prediction model accuracy improves, but the loss of time for data collection increases

Engineering Contradiction:
Improveprediction accuracyVSAvoiddata collection time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The system continuously collects temperature data and operation record data during the rolling process without interrupting the production flow. Sensors continuously monitor steel temperature, slab temperature, and skid rail temperature, and this continuous data stream is used for both real-time predictions and ongoing model training, eliminating the need for separate data collection phases.

Inventive Principle:
Principle #20Continuity of useful action

Data Source

PatentEP3928885B1Rolling load predicting method, rolling load predicting device, and rolling control method
Publication Date: 2025.01.15 JFE STEEL CORP
  • EP3928885B1 patent drawingFigure 1~2
  • EP3928885B1 patent drawingFigure 3
  • EP3928885B1 patent drawingFigure 4(a)~4(c)

AI summary

A rolling load prediction device according to the present invention is a rolling load prediction device for predicting a rolling load of a rolling mill for rolling steel. The rolling load prediction device includes a unit predicting the rolling load of the rolling mill in a case where the steel is rolled under an operating condition for prediction, by inputting the operating condition for prediction into a rolling load prediction model that has been trained with operation record data including at least a factor related to a temperature of the steel as an input variable and an actual value of the rolling load of the rolling mill as an output variable.