Continuous Casting Defect Prediction Across the Process Chain
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In continuous casting plants, surface defects such as cracks and peelings in products like thick slabs and long products often go undetected until later stages, causing quality issues and waste, due to spatial and temporal separation of process steps, and existing methods lack effective means to identify and prevent causal process parameters.
Innovation Solution
A method involving recording relevant control variables across the steelworks, ladle furnace, and rolling mill, using AI-based prediction models to detect surface defects, evaluating their causes, and adjusting these variables via an expert system to prevent future defects, ensuring optimal surface quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If process optimization is performed only within a single plant section, then the complexity of the optimization system is reduced, but the ability to prevent surface defects is insufficient because defects in previous sections impact subsequent sections
Solution Approach 1:
The patent merges multiple plant sections (steelworks, ladle furnace, casting machine, rolling mill) into a unified optimization system that shares data and knowledge across sections. This allows the system to identify and address root causes of surface defects that originate in upstream sections, thereby improving defect prevention capability while maintaining manageable complexity through integrated architecture.
Solution Approach 2:
The optimization system is designed with universal functionality to handle multiple plant sections and various types of surface defects simultaneously. The system can process data from different sections, identify defects, determine root causes, and generate optimized process parameters applicable across the entire production chain, making it adaptable to different defect types and locations.
2Measurement precision
If AI and machine learning are used to evaluate process parameters across the entire production chain, then the ability to identify causal process parameters is improved, but the device complexity increases significantly
Solution Approach 1:
The patent segments the complex AI evaluation task into distinct functional modules: data collection from plant sections, surface defect detection, root cause analysis, and optimization parameter generation. Each module handles a specific aspect of the evaluation process, making the overall system more manageable while maintaining high identification accuracy through specialized processing in each segment.
Solution Approach 2:
The patent introduces an intermediary optimization system that acts as a mediator between raw process data and final defect prevention actions. This intermediary layer processes and correlates data from multiple plant sections, uses AI to identify causal relationships, and translates findings into actionable optimized parameters, thereby managing complexity while preserving analytical precision.
3Device complexity
If surface defects are detected only at the end of the process chain, then the inspection process is simplified, but product quality and waste reduction are compromised because defects cannot be prevented
Solution Approach 1:
The patent implements preliminary defect detection and prevention by analyzing process parameters and predicting potential surface defects before they occur or propagate through the production chain. The system identifies root causes in upstream sections and adjusts process parameters proactively, preventing defect formation rather than merely detecting them at the end, thereby reducing waste while maintaining inspection simplicity.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The invention relates to a method (100) "Cracks Preventer" for preventing surface defects in the production of products such as thick slabs, thin slabs, or long products, based on the following steps: - prediction (104) of surface defects using at least one prediction model (13) based on recorded relevant control variables, the positions on the manufactured product assigned to the control variables, and recorded surface defects and their position on the manufactured product, wherein the at least one prediction model (13) is based on artificial intelligence; - evaluation (105) of the prediction of the prediction model (13) using an evaluation model (14) to determine a cause of the surface defect predicted by the prediction model (13), wherein the cause of the defect comprises at least one control variable of the plant components steelworks (2), ladle furnace (3), casting machine (4), and rolling mill (5);and - Determining (106) adapted control variables using an expert system (15) to avoid the predicted surface defect, wherein the expert system (15) adapts the control variables that were determined by the evaluation model (14) as the cause of the error.;