Page 169 - AI for Good-Innovate for Impact Final Report 2024

AI for Good-Innovate for Impact

38�2� Use case description

38�2�1 Description

Facing a harsh work environment for steel workers and low level of automation, the steel 38-Huawei&XISC
manufacturing industry is in dire need of digital transformation. AI modeling for the industry
is vital in supporting steel production in the quest to make it safer, smarter and more efficient.

Intelligent steel manufacturing solutions employed by Xiangtan Steel have deployed an AI
training and application platform in this Chinese firm with an annual production capacity of
12 million tons of steel. The platform is based on Huawei’s Pangu models, and consists of
four foundation models: computer vision (CV), graph neural network (GNN), multimodal, and
natural language processing (NLP). For instance,Pangu CV model has undergone pre-training
on over 1 billion images and more than 100 terabytes of video data using the unsupervised
learning method. It could thus extract and store vast amounts of knowledge within its extensive
network, allowing it to represent intricate visual features with ease. Plus, Pangu AI models
have superior generalization performance compared to smaller models; when trained for one
scenario, it can also be applied to scenarios with a detection accuracy of over 23%, meaning
these models can be quickly deployed in other steel mills, with minimal need for repeated
training.
The Pangu AI Steel Model enables unsupervised self-learning, covering over 300 sub-scenarios
in steel operations. The smart charcoal blending solution consists of the intelligent metallurgical
coal/coking coal blending algorithm model and cloud storage and the coal blending system
of the customer. The solution summarizes and integrates data such as raw coking coal data,
process/working condition data, coke data and operation data. After the data is encrypted and
anonymized, it is uploaded to the cloud storage. Based on the mechanism and data, select a
model framework, train the model, and deploy and release the trained model in the form of API
to provide coke quality prediction, coke blending ratio optimization and other core functions.

UN Goals:
• SDG 8: Decent Work and Economic Growth,
• SDG 9: Industry, Innovation and Infrastructure

SDG 8 - Decent Work and Economic Growth: the system enhances the working conditions
in the iron and steel industry by replacing labor-intensive and unsafe tasks with AI-based
technologies. This not only improves the safety and health of the workers but also increases
efficiency and productivity, leading to economic growth. The use of AI for inventory and defect
detection reduces errors and increases the speed of operations, resulting in more steel billets
rolled per hour and a significant increase in annual revenue. Specifically, 1 more steel billet can
be rolled per mill every hour, resulting in 30,000 more tons rolled every year and an increase
in annual revenue by over 100 million. The reduction in device downtime and maintenance
costs also contributes to economic efficiency. In fact, the annual motor damage rate has been
reduced from 5% to 2%, and the number of spot check personnel in slabs and bars has been
reduced from 10 to 4.

SDG 9 - Industry, Innovation and Infrastructure: the collaboration between Xiangtan Iron
and Steel, China Mobile Hunan, and Huawei to develop the first 5G+AI smart plant is a clear
example of innovation in industry. The use of AI and 5G technology in the manufacturing

153