Industry
Energy, Resources & Utilities
Rebalance the energy equation
Highlights
- AI enables closed-loop hot-rolling control in steel mills by integrating geometric product models with AI-augmented automatic gauge control (AGC) and machine vision.
- Drift caused by wear and material variability can be estimated using predictive models. The integration of sensor technologies can enable temperature monitoring and the prediction of mechanical properties.
- With AI, enhanced steel surface quality, optimised product dimensions, and reduced rejections and cobbles can be achieved.
On this page
On this pageinpage
What are the complexities involved in the hot rolling of steel?
The hot rolling process is important in steel manufacturing, as it enables the shaping of metal at elevated temperatures into desired forms. It is complex, requiring precise control over steel thickness, width, length, camber, and surface quality. Factors such as variability in raw materials, equipment wear, and manual interventions may lead to defects, rejections, and cobbles. AI-led automation of quality assessment and process control can lead to smarter, more efficient operations.
How can AI mitigate the risk of defects?
A geometrical model that includes dimensions such as width, length, thickness, and camber is measured and monitored in real time throughout the hot-rolling process. AI algorithms can predict deviations and optimise control strategies when trained on historical and real-time data. The integration of AI with mill control systems enables adaptive adjustments. This ensures consistent compliance with product specifications and minimises the risk of defects.
Model structure and AI integration
Final production dimensions, such as width, length, thickness, and camber, vary with hot-rolling process variables, including roll gap settings, mill speed, and material properties. The geometric model captures this relationship.
AI-driven modelling identifies patterns and anticipates anomalies by harnessing regression, classification, and neural networks. By incorporating real-time feedback from sensors and image analytics, one can improve the model’s accuracy and responsiveness.
How can AI improve thickness control in hot rolling?
Maintaining consistent product thickness during hot rolling is possible through automatic gauge control (AGC). While traditional AGC systems depend on feedback from thickness sensors and manual interventions, AI leverages predictive analytics and adaptive control logic. Machine learning algorithms forecast deviations and recommend corrective actions by analysing historical process data and sensor readings. Implementation begins with data acquisition and model training, proceeds to integration with control systems, and concludes with continuous validation.
AI-first control strategies
An AI-first approach utilises reinforcement learning and predictive analytics to optimise roll pressure, speed, and temperature profiles. By reducing reliance on manual adjustments, the AI-led approach enables faster responses to disturbances during hot rolling and improves thickness accuracy. Stepwise implementation involves:
- Collection of high-resolution sensor data
- Development of AI models
- Integration with mill programmable logic controllers (PLCs)
- Performance validation through trial runs and feedback loops
How can AI improve surface quality inspection in hot rolling?
Advanced video and image analytics can assess surface quality and geometric parameters in hot rolling mills. High-resolution cameras and machine vision systems capture real-time images of the steel strip. AI algorithms can analyse these images to detect defects and measure camber, width, length, and thickness. Techniques include edge detection, segmentation, and deep learning-based classification. These analytics provide actionable insights, supporting real-time intervention and quality control.
Surface quality and dimension measurement
Surface quality assessment involves identifying cracks, scales, scratches, and other defects using convolutional neural networks (CNNs) and pattern-recognition techniques. Computer vision techniques such as contour extraction and pixel-based analysis can measure camber, width, length, and thickness. AI models can be trained to categorise surface quality and quantify geometric deviations, supporting automated decision-making and reducing manual inspection.
What are the advantages of AI and sensor integration?
Sensor technologies are instrumental in capturing process variables and supporting AI-driven predictions. Common sensors for rolling and finishing temperature include:
- Pyrometers: Non-contact sensors to measure high-temperature surfaces.
- Thermocouples: Embedded sensors to accurately measure temperature readings at specific locations.
- Infrared cameras: For thermal imaging to map temperatures across the strip.
Temperature data is essential for predicting mechanical properties such as yield strength, ductility, and hardness. AI models enable real-time adjustments and quality assurance by correlating temperature profiles with material properties.
How can steel CXOs achieve the vision of an autonomous rolling mill?
The following is a step-by-step approach to achieving a future-ready autonomous rolling mill:
Data acquisition: Set up sensors and vision systems to collect comprehensive process data (temperature, thickness, width, surface images).
Model development: Build geometrical and process models using AI trained on historical and real-time data.
System integration: Integrate AI models with mill control systems, such as programmable logic controller (PLCs) and supervisory control and data acquisition (SCADA) systems, for automated feedback and control.
Validation and optimisation: Validate model predictions and control strategies through pilot runs, refining algorithms based on performance feedback.
Autonomous operations: Establish closed-loop control for the mill to adjust process parameters autonomously in response to real-time data and predicted outcomes.
How does AI improve quality and yield in hot rolling?
AI and automation enable proactive process control and rapid quality assessment. This significantly minimises defects, rejections, and cobble. Predictive analytics identify potential issues before they escalate. Real-time video and sensor data facilitate immediate corrective actions.
Automated surface inspection and dimensional control can ensure that products meet specifications, reduce waste and improve yield. Over time, continuous learning and adaptation can further enhance system performance
Can mills achieve greater autonomy with AI?
Leveraging AI in hot-rolling mills can elevate steel manufacturing by offering robust solutions for quality control, process optimisation, and autonomous operation. Mills can achieve higher productivity, reduced rejections, and improved mechanical properties by integrating geometrical models, AGC, video and image analytics, and sensor technologies. Using this integrated approach, industry professionals can harness the full potential of AI and drive operational excellence with eventual autonomy.