Chinalco's Digital Aluminum Journey: From Traditional Workshops to Intelligent Cloud Control

From manual calculations in high-temperature workshops to intelligent, cloud-based AI predictions and scheduling—Aluminum Corporation of China (Chinalco for short) and Huawei are leading this data-driven revolution and reshaping the aluminum industry. The Kun'an Model, jointly released by both parties at the end of 2024, revolutionized traditional production methods and provided a practical blueprint for the global non-ferrous metal industry's intelligent transformation, drawing on China's expertise.

Digital transformation in the aluminum industry: A crucial step to new heights

Aluminum, the most abundant metal in the Earth's crust, known for its lightweight and corrosion-resistant properties, serves as a cornerstone of modern industrial society, finding use in diverse applications from beverage cans and new energy vehicles to high-voltage power grids and aerospace materials. As the world's largest producer and consumer of aluminum, China has maintained this dominant position for many years.

As a leading enterprise in China's non-ferrous metal industry, Chinalco's industrial chain encompasses over 20 non-ferrous metal elements such as aluminum, copper, lead, zinc, gallium, and germanium, making it the world's largest supplier of alumina and electrolytic aluminum. However, despite its position as an industry leader with operations in over 20 countries and regions, Chinalco still faces unique technological barriers and complex management challenges inherent to the non-ferrous metal industry.

The smelting and processing of non-ferrous metals involve intricate chemical reactions and physical processes, demanding precise control over process parameters. Even slight errors can lead to quality defects, increased energy consumption, and even environmental pollution. Additionally, frequent manual operations in high-temperature, high-magnetic, corrosive environments, coupled with repetitive precision testing, introduce significant safety risks and quality instability.

Furthermore, Chinalco's comprehensive industry chain layout, spanning mineral exploration, extraction, smelting, finished product processing, and supply chain, results in highly dispersed data. The lack of a specialized visual data management system prevents the efficient integration and correlation of this vast amount of data, resulting in silos that hinder the full realization of its value.

Within Chinalco, most of its business units and subsidiaries opt to independently establish data centers during the digitalization process. This decentralized model makes it challenging to coordinate globally and facilitate efficient resource sharing. What's even more critical is that these standalone data centers generally lack a dual-active mechanism or robust disaster recovery capabilities. In the event of sudden failures or natural disasters, Chinalco's stable operations are at significant risk from data loss and service interruptions.

Kun'an Model: Reconstructing the entire production chain with data and AI

In December 2024, built on Huawei Cloud Stack hybrid cloud, Chinalco unveiled the Kun'an Model. After jointly establishing a unified computing center, data platform, and AI development platform, and utilizing a cloud-edge collaborative architecture, Chinalco is now responsible for training L1 and L2 models. After model training is complete, subsidiaries can request additional computing power and freely access data models. This framework enables the rapid deployment and replication of models across Chinalco, while ensuring compliance for on-premises data. The Kun'an Model is currently applied to 18 high-potential scenarios across eight business domains: mining, smelting, processing, secure production, supply chain, business management, scientific research, and smart services.

Zhisheng Aluminum Oxide Model: Creating an automated control system that leverages data to overcome production process delays

Aluminum oxide production is hindered by inefficient data detection and complex processes. To address this, Guangxi Huasheng utilized the Kun'an Model and released the industry's first dedicated AI model—Zhisheng Aluminum Oxide. Deployed across three critical stages—evaporation, dissolution, and sedimentation—the model enables minute-level prediction of key metrics. Leveraging Chinalco's powerful computing center, Guangxi Huasheng has developed an end-to-end, intelligent management process that encompasses everything from data preprocessing and feature engineering, to model-based predictions and real-time production control. According to preliminary estimates, Guangxi Huasheng's main control operations' workload has been reduced by 85%, while sampling and testing efficiency has improved by 30%. Furthermore, the dissolution rate has increased by 0.2%. The solution has also reduced, steam consumption has reduced by 1%, alkali consumption has reduced by 1 g/L, and the annual cost has reduced by over CNY10 million(≈USD1.4 million).

Green Aluminum AI: Equipping electrolytic aluminum production with 'smart valves' to enable precise control and decision-making in auxiliary operations In aluminum electrolysis production, challenges such as frequent manual interventions, difficult data collection, and low accuracy of mechanistic models have long plagued the entire industry. Yunnan Aluminum Co., Ltd. developed the first AI model for the industry—Green Aluminum AI. The model is used to predict and automatically adjust both aluminum output and the addition of alumina and aluminum fluoride during electrolysis. This has greatly improved the safety, continuity, and stability of production. It is estimated that the model, since its deployment at Yunnan Aluminium Wenshan's 500 kA electrolytic cells, has reduced costs by over CNY50 million(≈USD7 million) annually for the company.

Cloud-based Aluminum Production Scheduling: Intelligent production planning and scheduling helps reduce costs and improve efficiency In terms of aluminum processing, Chinalco Ruimin developed an AI model for intelligent production planning and scheduling. During cold rolling, the model develops an optimal, minute-level production plan by instantly calculating hundreds of millions of variables. As a result, the on-time delivery rate has increased from 94.1% to over 98%, generating annual savings worth millions of CNY. In the event of unexpected demand shifts, the model can quickly predict the impact and provide actionable data for production plan adjustments.

Metal Vision: Delivering expert-level power to every operation

In aluminum alloy research, metallographic analysis is essential to gain insight into its internal structure and infer its performance. In Chinalco, more than 150,000 images need to be analyzed every year, with manual labelling of a single photo taking 3 to 5 minutes. The repetitive nature of this task makes it difficult to ensure consistency. In response, Chinalco Materials Research Institute developed an AI model for the analysis of metal microstructures. Capable of batch-processing aluminum alloy microstructures, the model quickly measures the area and size distribution of microstructural features, achieving 99.5% accuracy and boosting analysis efficiency by over 90%.

From model advancement to ecological leadership: Chinalco's intelligent blueprint

In the next three years, Chinalco's Kun'an Model aims to achieve its goal by: constructing 100P of hybrid cloud computing power, accumulating 100 billion production data points, building 1000 industry-specific application scenarios, and assembling a 1000-person team. This foundation will serve as a lever to catalyze a full-chain digital and intelligent transformation across the non-ferrous metals industry.

Looking ahead, Chinalco will continue to harness data and AI as dual drivers, strengthening data governance, building high-quality datasets, pioneering high-value application scenarios, and expanding the transformative reach of the Kun'an Model across the industry.