China Aluminum Research Institute Dataset Selected as the First Batch of Excellent Achievements by SASAC 2025-05-20 08:16:20 Source: China Nonferrous Metals News
During the 8th Digital China Summit, the State-owned Assets Supervision and Administration Commission of the State Council released the first batch of 30 outstanding achievements in the construction of high-quality data sets in the artificial intelligence industry of central enterprises in more than 10 industries. The "Aluminum Alloy Material Metallographic Structure Image Dataset" submitted by China Aluminum Material Application Research Institute Co., Ltd. (hereinafter referred to as "China Aluminum Material Institute"), recommended by China Aluminum Corporation, was successfully selected as one of the five benchmark achievements in the field of industrial manufacturing.
In response to the long-standing industry pain points of metallographic analysis, such as reliance on manual experience, low detection efficiency, and insufficient result stability, China Aluminum Material Institute focuses on aluminum alloy research and development and production scenarios, and systematically builds high-quality data sets. The company's scientific research team has collected a large number of microstructure image samples of key aluminum processing processes. After strict screening by aluminum alloy material metallographic analysis experts with more than 10 years of experience, key microstructure features such as looseness and compounds are annotated at the pixel level, forming a high-quality data set that can be directly used for AI model training.
The data has a solid foundation, and selects tissue image samples covering mainstream aluminum alloy materials such as 2 series, 3 series, 5 series, 6 series, and 7 series. The accuracy rate after expert annotation is over 99%. The metallographic intelligent identification and analysis model trained based on this data set has been put into use within Chinalco, with a microstructure identification accuracy of 99% and a detection efficiency 10 times higher than that of traditional methods.
Through the closed-loop mechanism of "data collection-model training-application feedback", the team continuously accumulates annotated data and realizes model iteration and upgrade; the metallographic analysis intelligent system developed simultaneously, Metal Smart Eye, deeply integrates expert experience with AI algorithms, effectively reduces errors caused by differences in personnel experience, and provides stable and reliable data support for the performance evaluation of aluminum alloy materials.
In the future, Chinalco Materials Research Institute will further deepen the application of AI technology in the field of microstructure analysis, strengthen multi-source data integration, and continue to promote the implementation verification of high-quality data sets and iterative optimization of high-stability large models, providing strong support for more enterprises and scientific research teams; in conjunction with relevant enterprises, it will gather various high-quality data elements in the industry across regions and levels, drive all parties to deepen data collaborative sharing, promote the efficient and safe development and utilization of various data resources, and promote the construction of a multi-data resource library, providing strong support for the digital transformation of my country's non-ferrous metals industry and Chinalco.
