Tandem Rolling Mill Control for Roll Force Balance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing tandem rolling mill control methods face challenges in achieving accurate roll force prediction and maintaining roll force balance among rolling stands, leading to instability and inefficiency due to improper separation of error components and computational complexity.
Innovation Solution
A control device comprising a roll force prediction error calculation section, a roll force balance consistency value calculation section, and a roll force compensation value calculation section, which estimate roll forces, calculate balance consistency values, and adjust screw-down positions and roll speeds to maintain balance among rolling stands with simple processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If roll force prediction model separates error into material inherent component and time variance component, then prediction accuracy is improved, but error separation becomes impossible when strip thickness and width changes are small
Solution Approach 1:
The patent segments the roll force prediction error into two distinct components: a material-inherent error component (Zpk) that remains consistent for the same material type, and a time-variance error component (Zpm) that changes over time due to rolling mill conditions. This segmentation allows separate learning of each component, improving prediction accuracy while making error analysis manageable through dedicated learning coefficients for each error source.
2Measurement precision
If learning coefficients are determined separately for material inherent error and time variance error, then prediction accuracy is improved, but computational complexity increases
Solution Approach 1:
The patent divides the learning process into two separate computational paths: one for learning the material-inherent coefficient Zpk and another for learning the time-variance coefficient Zpm. This segmentation of the computational process, while increasing the number of coefficients to learn, organizes the complexity into manageable, independent learning tasks that can be executed systematically.
Solution Approach 2:
The patent implements feedback mechanisms where the learned coefficients Zpk and Zpm are continuously updated based on the difference between predicted and actual roll forces. This feedback loop allows the system to automatically adjust and improve prediction accuracy over time without requiring complex manual intervention, managing computational complexity through iterative self-correction.
3Measurement precision
If roll force prediction errors are calculated for each rolling stand independently, then prediction accuracy is improved, but roll force balance among rolling stands deteriorates
Solution Approach 1:
The patent merges the independently calculated roll force prediction errors from multiple rolling stands into a unified error model. By combining these errors and applying consistent learning coefficients across all stands, the system maintains both high prediction accuracy for each stand and overall roll force balance, preventing the deterioration that would result from purely independent calculations.
4Measurement precision
If complex error separation and learning models are applied to each rolling stand, then prediction accuracy is improved, but device complexity and computational load increase significantly
Solution Approach 1:
The patent creates a universal error learning model that can be applied across all rolling stands in the tandem mill. The learned coefficients Zpk and Zpm serve multiple functions: they correct prediction errors for different material types, adapt to time-variance in rolling conditions, and maintain balance across all stands. This multi-functionality reduces overall system complexity by using a single adaptable framework rather than separate complex models for each stand.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A roll force prediction error calculation section estimates roll forces in rolling stands utilizing rolling actual values acquired at the time when a strip is rolled, and calculates roll force prediction errors from the estimated roll forces and the rolling actual values. A roll force balance consistency value calculation section (15) calculates roll force balance consistency values that indicate the degree of differences between the roll force prediction errors in the rolling stands and roll force prediction errors in neighboring rolling-stands. A roll force compensation value calculation section (16) calculates roll force compensation values from the roll force prediction errors and the roll force balance consistency values. A control reference setup section estimates a roll force of a strip to be next rolled, compensats the estimated roll forces by the roll force compensation values which are calculated in the roll force compensation value calculation section, and calculates screw-down positions to be set in the respective rolling stands, utilizing the compensated roll forces.