From Expertise to Intelligence: NISCO Forges a New Future with AI-Driven Digital Transformation

Iconic national construction projects demand raw materials that can perform under extreme conditions. This fact has remained true across China, from the robust hull of China's first domestic luxury cruise liner, to the deep-sea structures required for gas hydrate extraction, to the critical safety components of clean energy power plants or the monumental bridges of the Sichuan-Xizang Railway. In each of these projects, Nanjing Iron & Steel Co., Ltd. (NISCO) has played a key role. Its precision-engineered specialty steels guarantee the final projects can withstand daunting real-world conditions, from icy oceans to high-altitude plateaus.

An industry in need of a breakthrough

The global steel industry faces two major structural challenges. First, traditional production models are overly dependent on manual labor and the experience of seasoned experts. The business is also increasingly unsustainable. A case in point is the traditional coke blending process which heavily relies on trial-and-error and the intuition of experienced workers who oversee the process with large shovels. This leads to inaccurate coal ratios which drive up cost per metric ton of coke. On the other hand, stringent green development mandates are intensifying pressure across the industry. Traditional energy-intensive production approaches now face a critical choice: transform or be eliminated.

As the industry confronts the new demands of modernization, a pressing question emerges: How can steelmakers pivot from pursuing scale to sustainably improving quality and efficiency? NISCO has decided digital and intelligent transformation will help them achieve this goal, using the "soft power" of digital and intelligent capabilities to augment their "hard power" in steel.

Huawei Cloud and NISCO: Intelligent assistance for experienced steel-makers

As a leading player in China's specialty steel sector, NISCO possesses an annual production capacity of over 10 million tons. Also leading in digital and intelligent transformation in the steel industry, the company initiated comprehensive data governance in 2021 to build a tailored data management system for large steel enterprises. This transformation allowed NISCO to achieve a core data asset ingestion rate of over 90% in its centralized data lake. In early 2024, NISCO launched a three-year action plan for AI model adoption, accelerating its intelligent transformation. On June 21, 2025, the company launched the YuanyeSteel Large Model in collaboration with Huawei, marking a new milestone in Yuanye’s digital and intelligent transformation.

Currently, NISCO uses a two-tier cloud-edge model training and inference system, built on Huawei Cloud Stack's large-model hybrid cloud, which spans its group center and production lines. This mechanism enables efficient data flow and iterative algorithm upgrades, forming a self-improving intelligent loop that continuously evolves and learns during operation. The company has already deployed 20 intelligent application scenarios across four major business domains: R&D and design, production and manufacturing, marketing and service, and operations management. This system not only provides the computing power required for efficient training of scenario-specific large models in the steel industry, but also enables lightweight inference deployment at the edge, driving real-time application and continuous optimization of AI in production environments. These deployments have helped NISCO undergo a fundamental shift from mechanization and automation to intelligent manufacturing, and from isolated technological breakthroughs to integrated, system-wide coordination. This evolution marks a qualitative transition from experience-based to intelligent steel-making. The company is now advancing plans for an AI-powered super factory that has high-end production, digital and intelligent capabilities, sustainability, cross-functional integration, and a global footprint.

Carbon-manganese low-temperature steel R&D: Carbon-manganese low-temperature steel is a key material used in high-value vessels such as liquefied gas carriers. During material development, excess carbon increases brittleness, while too much manganese compromises weldability. The strength and toughness of this material are critical to the structural integrity of many types of heavy machinery, and traditional R&D approaches often struggle finding the correct balance of elements due to the complexity of compositional interactions. This means R&D often resorts to costly, experience-dependent trial and error. With the help of Huawei Cloud Stack's large-model hybrid cloud platform, NISCO has integrated metallurgical principles with predictive and optimization models. By applying big data analytics to composition and process parameters, the company can now accurately forecast the mechanical properties of different compositions of carbon-manganese low-temperature steel, monitor production in real time, and fine-tune processes accordingly. This has resulted in a 1.5% or higher improvement in product qualification rates and a fundamental transformation in specialty steel R&D.

Coke blending: Coke blending is a process that impacts both production efficiency and final steel quality. Traditional approaches rely heavily on experience and small-scale oven tests, often leading to inconsistent quality and high operating costs. By implementing an intelligent blending algorithm featuring multi-objective optimization, NISCO can now dynamically calculate the optimal blend in 1–2 minutes, compared to the previous 1–2 days. This has reduced the cost per metric ton of coke by CNY5–10, transforming a process hampered by delayed feedback into one that features real-time optimization.

Intelligent rolling: The quality of high-end steel plates also depends on the precise control of force during the rolling process. Accurate prediction of rolling force is challenging due to the strong nonlinearity, multivariable coupling, and rapidly changing operating conditions. To tackle this challenge, NISCO collaborated with Huawei Cloud Stack to co-develop an AI-powered rolling force prediction engine that integrates knowledge and data. By enhancing training data with expert knowledge and extreme-condition scenario information, the engine enables fast modeling and high-precision predictions in small-sample scenarios.

Steel quality inspection: Traditional inspection approaches in steel production depend heavily on human expertise, and is often constrained by visual limitations and environmental conditions. It can be difficult for inspectors to gather samples, and the wide variety of possible defect types and large dimensional variability in sampling lots can further complicate inspection. The intelligent computer vision (CV) model has brought "eagle-eyed" capabilities to inspection, enabling highly accurate metallographic and macrographic grading and leaving virtually no anomaly undetected. For example, intelligent metallographic analysis combines computer vision with robotic automation. During the analysis process, the robotic unit prepares, etches, and captures images of its samples, while AI algorithms analyze the metallographic images to automatically classify structure types and grades. Using this tool, a single operator can now process 240 samples in an 8-hour shift, covering grinding, polishing, etching, inspection, and report generation. Previously, an operator could only process 60 samples in the same timeframe. This has resulted in a significant increase in overall inspection efficiency and a 99.5% defect detection rate, significantly reducing both inspection costs and the task's reliance on human experience.

Intelligent Q&A: NISCO has additionally implemented a dual-mode AI system that integrates structured knowledge bases with dynamic data engines. This architecture delivers accurate responses and visual traceability in intelligent equipment management scenarios such as intelligent Q&A, which has resulted in a remarkable increase in operational efficiency. Moving forward, the company aims to extend AI-assisted Q&A coverage to all business functions, further advancing semantic indexing and SQL generation capabilities. These efforts will help accelerate the conversion of data into business insights, strengthening AI-driven support across R&D, production, and operations.

"Intelligent R&D is the crown jewel of AI in the steel industry," said Chen Linheng, Director of NISCO's New Material Research Institute. With their "AI + Materials R&D" framework, the company is continuing to shift from reliance on experience-based trial and error to a hybrid model powered by data and domain knowledge, already achieving significant improvements in R&D efficiency and quality.

Steel meets cloud: Forging a new narrative of industrial progress

Today, NISCO's operations are supported by 138 AI application models, 456 automated reporting systems, and 1,467 visual dashboards, bringing visibility and transparency to its iron-making processes. With their AI-driven management, the company has achieved measurable gains in efficiency and cost savings. Performance data shows a noticeable reduction in operational bottlenecks: Hot metal ladle turnover has risen by 5%, temperature loss has dropped by 15°C, and production cost per metric ton of hot metal has decreased by CNY200.

"Digital and intelligent transformation is the essential pathway for the steel industry to thrive in the new industrial age," said Sun Maojie, NISCO's CIO. Through years of practice, NISCO has developed distinctive achievements in industrial digitalization and digital industrialization, effectively overcoming the industry-wide challenge of increasing output without proportional gains in efficiency.

Looking ahead, as this model of transformation expands from pioneering cases to the entire industry, other "Made in China" initiatives are expected to make a real leap from "competing on scale" to "competing on quality". This transition may even help the global manufacturing community develop their own style of "NISCO Intelligence."