NISCO Deploys World's First Swarm-Intelligence Steel Model at Scale
Nanjing Iron Steel Co. integrates swarm intelligence to optimize complex manufacturing operations.
1Pioneering Swarm Intelligence
Nanjing Iron Steel Co. partnered with Huawei to launch the YuanYe Steel Large Model. This system utilizes a dual-drive architecture, combining probabilistic computing with scientific decision-making. By integrating metallurgical mechanism models with AI, the company successfully transitioned its core manufacturing operations from manual experience-based processes to a fully data-driven, intelligent framework.
2Infrastructure and Data Foundations
The deployment relies on Huawei’s Ascend, Kunpeng, and Cloud Stack technologies. NISCO aggregated millions of operational technology data points across 26 distinct business domains to create real-time digital twins. This robust foundation ensures that every stage of production, from ironmaking to quality inspection, is visible, precisely controlled, and capable of intelligent assessment.
3Proactive Ironmaking Optimization
YuanYe transforms blast furnace operations by predicting temperatures and health issues two hours in advance with over 90 percent accuracy. This shift from reactive remediation to proactive pre-control has increased prime-grade molten iron yield by more than 10 percent. Consequently, the facility significantly reduced unit energy consumption while maintaining stable, high-quality output.
4Precision in Steel Rolling
The rolling-force prediction model operates at millisecond speeds to manage complex conditions. This precision narrowed thickness tolerance to ±0.2 mm and reduced deviation to 1.9 percent, reaching world-leading standards. With the rolling-force hit rate jumping to 98 percent, NISCO supports an annual output of 2.5 million tons of high-specification steel plates.
5Automated Quality Inspection
The smart metallographic inspection system uses robotic synergy to achieve 95 percent accuracy. By automating sample grinding, polishing, and grading, the system reduced per-sample inspection time from 10 minutes to just 3–5 minutes. This 80 percent efficiency gain allows human operators to focus on complex anomaly analysis rather than repetitive manual tasks.
6Environmental and Energy Control
The S-Link Intelligent Agent manages gas and sulfur control by integrating 602 process parameters. It provides sulfur-risk warnings one hour in advance with 95 percent accuracy. By optimizing gas blending and emission quotas, the company slashed natural gas co-firing consumption by nearly 50 percent, reducing emission-related downtime by over 25 percent annually.
7Human-Centric AI Empowerment
NISCO has deployed over 300 AI models and 1,000 low-code agents developed by frontline employees. Management emphasizes that the goal is to unleash human potential rather than replace workers. By automating data-heavy tasks, the company empowers staff to focus on creative, high-value work, fostering a culture of inclusive digital transformation.
8Scalable Industrial Benchmarking
The initiative has shortened R&D cycles by 50 percent and reduced annual costs by over CNY100 million. As of October 2026, 25 business agents are active across the plants. NISCO aims to provide a replicable benchmark for the global steel industry, focusing on three-tier AI-native architecture to ensure long-term, scalable industrial growth.