Hot Rolling Strip Chew Prediction Using Adaptive Rolling Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The occurrence of strip chew in hot rolling mills is difficult to predict accurately, leading to reduced productivity and roll intensity due to the reliance on operator experience and intuition, and existing control methods are ineffective in preventing this phenomenon.
Innovation Solution
A prediction system using adaptive models constructed through machine learning or statistical methods that collect and analyze data from preceding rolling paths to predict the occurrence and location of strip chew, allowing for timely adjustments to the entrance side guides to prevent the issue.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If operator experience and intuition are used to predict strip chew occurrence, then the system is simple to operate, but the prediction accuracy is insufficient
Solution Approach 1:
The patent replaces the mechanical system of operator experience and intuition with an information processing system comprising a data collection unit, model construction unit, and prediction unit. The prediction unit uses adaptive models (neural networks, support vector machines, or random forests) to process rolling condition data and automatically predict strip chew occurrence, substituting human judgment with computational analysis to achieve higher accuracy.
Solution Approach 2:
The patent introduces adaptive models as intermediaries between raw rolling condition data and prediction results. These models (neural networks, support vector machines, or random forests) serve as mediators that learn patterns from historical data and translate complex rolling parameters into accurate predictions of strip chew occurrence, enabling precise prediction without requiring direct operator intervention.
2Productivity
If strip chew is not predicted in advance, then the system operates continuously without interruption, but productivity decreases due to stoppages for inspection and roll extraction
Solution Approach 1:
The patent implements preliminary action by predicting strip chew occurrence before it actually happens. The prediction unit forecasts potential strip chew events based on current rolling conditions and historical patterns, allowing operators to take preventive measures (such as adjusting rolling parameters or preparing for quick response) before the strip chew occurs, thereby avoiding stoppages and maintaining productivity.
Solution Approach 2:
The patent establishes a feedback mechanism where prediction results are provided to operators in real-time or near-real-time. The system continuously monitors rolling conditions, updates predictions, and feeds back warning information to operators, enabling them to adjust operations proactively. This closed-loop feedback system ensures operational stability by maintaining awareness of potential issues while allowing continuous production.
3Reliability
If conventional control methods are used to prevent strip chew, then the control system remains simple, but the methods are ineffective in preventing the phenomenon
Solution Approach 1:
The patent applies preliminary action by predicting strip chew occurrence before it happens, enabling preventive control. The prediction unit identifies rolling conditions that lead to strip chew, and operators can adjust parameters (such as rolling speed, temperature, or side guide positioning) in advance to prevent the phenomenon, making control efforts effective rather than reactive.
Solution Approach 2:
The patent replaces conventional mechanical control methods with an information-based control system. Instead of relying on simple mechanical adjustments or empirical rules, the system uses data collection, adaptive modeling, and computational prediction to identify and prevent strip chew conditions, achieving effective prevention through intelligent control rather than mechanical means.
Data Source
AI summary
The prediction system of strip chew collects and stores first data and second data as adaptive model construction data. The first data indicates the occurrence or non-occurrence of the strip chew in an object rolling path and the occurrence point of the strip chew. The second data includes information on a preceding rolling path and attribute information on an object strip. The system constructs an adaptive model using the stored adaptive model construction data, and stores the constructed adaptive model as an adapted model. The system collects prediction data similar to the second data. Then, the system inputs the prediction data to the adapted model to predict the occurrence or non-occurrence of the strip chew in the object rolling path and all or some of the occurrence points of the strip chew before the prediction object strip reaches the object rolling path.