Rolling Train Geometry Control via Neural Network Feedback
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing rolling processes face challenges in maintaining uniform geometry of metal strips during width and thickness reduction, leading to deviations from target dimensions and material waste due to asymmetric material flow and 'dog bone' formations at strip ends.
Innovation Solution
A method involving higher-level process automation that measures actual geometry, calculates and adjusts working variables using a computer-aided process model to align with target geometry, iteratively refining these variables until specified tolerances are met, and applies them to control units in the rolling train.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If conventional rolling processes are used without real-time measurement and adjustment, then the rolling process is simple to operate, but geometric deviations and material waste increase
Solution Approach 1:
The system performs preliminary measurement of the actual geometry before rolling and calculates optimal rolling curve parameters in advance. The neural network determines preliminary rolling curve parameters based on the measured actual geometry, which are then used to guide the rolling process, preventing geometric deviations before they occur.
Solution Approach 2:
The system implements a closed-loop feedback mechanism where the actual geometry is measured, compared with target geometry to determine geometry deviations, and these deviations are used to adjust the rolling curve parameters in real-time. This feedback loop continuously optimizes the rolling process to maintain geometric uniformity.
2Manufacturing precision
If real-time measurement and iterative adjustment are implemented, then manufacturing precision improves, but processing time and system complexity increase
Solution Approach 1:
The system performs measurement and calculation of optimal parameters before the actual rolling operation. By determining the rolling curve parameters in advance based on measured geometry, the system avoids time-consuming adjustments during the rolling process itself, reducing overall processing time while maintaining precision.
Solution Approach 2:
The system replaces traditional mechanical trial-and-error adjustment methods with automated optical measurement and neural network-based parameter optimization. This substitution of mechanical adjustment with intelligent algorithms reduces processing time while improving dimensional accuracy.
3Loss of substance
If traditional width reduction methods are used, then the rolling process is straightforward, but material waste increases due to dogbone formations and trimming
Solution Approach 1:
The system applies different rolling curve parameters to different locations along the strip width based on measured geometry variations. The neural network determines location-specific parameters that account for local geometric deviations, preventing dogbone formations at the edges while maintaining uniformity in the center, thereby reducing material waste.
Solution Approach 2:
The system dynamically changes rolling parameters based on the actual measured geometry of each workpiece. By adjusting rolling curve parameters according to the specific geometry deviations detected, the system optimizes material flow during rolling to eliminate dogbone formations and reduce the need for trimming and skimming.
4Reliability
If geometric deviations are detected only after rolling, then measurement simplicity is maintained, but corrective action is delayed to subsequent products
Solution Approach 1:
The system performs measurement and calculates corrective parameters before the rolling operation begins. By determining the optimal rolling curve parameters in advance based on pre-rolling geometry measurement, the system ensures that corrective actions are applied during the current rolling process rather than being delayed to subsequent products.
Solution Approach 2:
The system implements real-time feedback by measuring geometry, calculating deviations, and adjusting rolling parameters during the same rolling cycle. This immediate feedback loop ensures that corrective actions are applied without delay, improving process control reliability and eliminating the need to wait for subsequent products.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The invention relates to a method for controlling a rolling mill (1) and a rolling mill in which a metallic rolled material, preferably in the form of slabs (12) and ingots, is formed in one or more rolling passes in units of the rolling mill by reducing its width and/or thickness to a finished metal strip as an end product, wherein the method comprises the control of individual units of the rolling mill by means of a higher-level process automation system and wherein the method comprises measuring and recording an actual geometry of the rolled material at at least one measuring point at least before and/or after a forming step in a unit of the rolling mill and calculating an expected geometry of the rolled material using a computer-aided process model, which models the flow behavior of the rolled material, particularly near the edges, under a given deformation effect, and deriving adapted working parameters for at least one forming process.if the deviation of the model-based determined expected geometry from the metrologically measured target geometry lies outside a given error tolerance.