Hunan Iron & Steel Group Boosts Steelmaking Efficiency and Intelligence with Steel Industry Model

A shift from manual calculation to large model-based intelligent computing enables new users to perform smart driving with ease.

In the 30-meter-high D building of the XISC steelmaking plant, there are three basic oxygen furnaces, three refining stations, five ladle furnaces, and five continuous casters. Rolling waves of heat pour in to the plant. Five bridge cranes with working ranges of up to 500 meters operate at great heights. A 150-ton steel ladle operating at temperatures above 1000°C is slowly transported through the air by cranes.

Bridge crane operations for steelworks are a challenging undertaking that requires continual problem solving under hazardous high-temperature conditions. Traditional bridge crane operations in steel mills are heavily dependent on manual labor. However, this reliance on dispatcher experience inevitably results in errors and inefficiencies, due to the complexity of the steelmaking process.

At XISC, this problem was resolved by switching from manual calculation to intelligent computing based on the large model. The intelligent bridge crane scheduling system deployed using the large model integrates a large amount of data, such as steelmaking production plans, bridge crane maintenance information, real-time locations of steel ladles, and service rules. The system uses algorithms to intelligently generate bridge crane scheduling plans.

Even more impressive is that if there are any last-minute production changes, the system can "think" and generate a scheduling plan for the next 30 minutes within one minute, ensuring instructions are issued on time. The large model has significantly improved the turnover rate of steel ladles, resulting in cost savings of CNY1.2 per ton of steel.

The large model safeguards the last mile of production.

In an ironworks, belts are needed to transport materials in bulk. These belts, which operate 24/7 at a speed of 2 meters per second, require regular inspection by skilled workers. If a belt breaks, the blast furnace has to shut down, resulting in losses of up to one ton of molten iron per minute.

Huawei Oil, Gas & Mining BU experts stationed at the XISC ironworks discovered that the belts would deviate during transport, setting off alarms. These belts are made of a rubber and steel wire composite material, and friction causes bulges in the belts over time.

An intelligent belt monitoring solution based on 5G and computer vision has been developed to address these issues. According to samples collected, the system can intelligently monitor belt operation around the clock, predict bulge changes, and issue warnings on time. After one month of learning, the large model's intelligent monitoring precision was improved to 98%, which allowed the plant to reduce the frequency and intensity of manual inspection.

The steel mill is transformed into a "central kitchen" for rapidly cooking up new applications.

In the steel rolling workshop of XISC's high-speed wire plant, a coil of wire can be produced every minute. In the past, workers needed to conduct sampling inspection with flashlights, a method that had low detection efficiency and carried significant safety risks due to the high temperatures of the production process. Now, eight industrial cameras are controlled by the cloud platform to detect the quality of each coil in real time.

This inspection system was developed by XISC's technical team in just 10 days, an achievement made possible by low-code development tools that translate service knowledge into technical language. These tools enable innovative AI applications to be modularly developed like "recipes" from a "central kitchen".

Data from different devices can be collected and transmitted efficiently using the high bandwidth and low latency of 5G. Combined with large models, this increase in efficiency enables data to empower production and AI to make smart decisions.

By the end of April 2024, the Pangu model for the steel Industry had been deployed at XISC to cover nine major fields, including coking, sintering, ironmaking, steelmaking, steel rolling, and quality inspection. A total of 23 intelligent applications in the categories of vision and prediction have been implemented to optimize production and control quality. With the support of the large model, energy consumption per ton of steel has decreased by 10%. In the future, more than 60 new innovative applications will be implemented to help XISC reduce costs, improve quality, and enhance efficiency.