How NISCO revolutionized the steelmaking industry with its in-house YuanYe AI model

For years, factories worldwide have used sensor technology, helping advance IoT and revolutionize how we produce and distribute goods and services. But now a new era is upon us – the AI era – which is fueled by data collected by the advanced sensor technology used in factories.

AI can serve as a smart assistant in a factory, monitoring dashboards and boosting productivity. It needs to understand on-site data, such as temperature, pressure, images, and sound, but also integrate metallurgical principles (the science and engineering of metal alloys), process rules, and equipment conditions to make judgements.

Only in this way can the analysis results be transformed into on-site warnings and operational recommendations that can be properly used.

Technology in action

Nanjing Iron & Steel Co., Ltd (NISCO) worked with Huawei and other ecosystem partners to commercially deploy the world’s first steel model powered by Swarm Intelligence – the Yuan Ye Steel Large Model (YuanYe).

The steel manufacturer built a robust digital and intelligent foundation by adopting industry-leading products such as Ascend, Kunpeng, Huawei Cloud Stack, and Pangu models. By embedding AI into every stage of its production, it has forged green and high-end AI-powered super steel plants.

With sensing data aggregated from millions of operational technology data points across 26 business domains, NISCO has achieved real-time digital twin mapping within its intelligent operations center. Laying solid data and scenario foundations for YuanYe.

It has enabled the entire steelmaking process to become not only visible, but also capable of precise control and intelligent assessment.

Now, 100% of NISCO's key equipment is connected and digitally controlled, shortening its new-product R&D cycles by 50% and reducing annual costs by over CNY 100 million (roughly just over $15m). With YuanYe, NISCO has set a new benchmark for digital and intelligent steel transformation.

YuanYe model

Much of this is underpinned by its AI model YuanYe – developed in collaboration with Huawei – which is described as a ‘swarm intelligence’. YuanYe fuses multiple AI models with the steel industry’s metallurgical mechanism models to create a drive-dual, multi-model cluster architecture. This architecture comprises a ‘right brain ’ for handling probabilistic computing, and a ‘left brain’ that deals in precise and scientific decision-making.

Observation is the first part of the solution; machine vision is used to identify flames, plate shapes, and metallographic structures. Sonar sensors monitor reactions inside the furnace, while other sensors continuously collect data on temperatures, pressure, vibrations, and the condition of the equipment. These scattered signals are then aggregated into a unified whole to form a continuous perception of what is happening on-site in real-time.

Predictive models also help to identify trends from historical and real-time data, so that they can make early judgments about the temperature of the furnace, the performance of the product, or whether there are any abnormal risks. It helps to think of this as a navigation system; not only does it tell you where you are, but it also reminds you of where possible traffic jams may occur, or whether you need to make an adjustment to avoid trouble ahead. The early warning system allows employees time to verify potential faults and find solutions, rather than waiting for the trouble to happen.

While the AI system is advanced and groundbreaking, it is not here to replace humans; For NISCO, employees are responsible for verifying whether the data is credible, whether the working conditions are within the model's applicable range, determining the cause of abnormalities, handling special steel grades and equipment failures, and making decisions on key operations. The execution results are then fed back into the system for calibrating the models and rules.

"Digital and intelligent transformation is inherently human-centric,” said Li Jinyan, the general manager of the Digital and Intelligent Integrated Development Department and Director of the AI Research Institute at NISCO.

“Its purpose is not to replace people, but to unleash human potential. Our goal is to lift employees out of the endless sea of data, empowering them to direct their creative energy toward work of higher value."

The blending of raw materials, production schedules, and energy management is compared under multiple variables – quality, cost, delivery time, and so on – while agents also understand business tasks, use different models and rules, and complete cross-process collaborations.

The end of the process is the formation of a closed loop: from data entering the system to models identifying the status, and the generation of warnings and/or solutions. Humans are a part of the loop, verifying alerts and confirming next steps. All on-site results are then fed back into the system. For AI to be used in production in this way, NISCO believes that it must not only calculate accurately, but also ensure that the operators know when to adapt recommendations and intervene with a manual fix.

"The real value of industrial AI can only be realized when it expands from single-point applications to comprehensive integration with production, operations, and management, which can unlock sustained, scalable empowerment," added Cheng Yongfu, the senior industrial solution expert at Huawei's Steel & Non-ferrous BU.

To date, NISCO has deployed more than 300 AI models across its plants, alongside more than 1,000 low-code agents and lightweight applications built independently by frontline employees.