2024-10-12 BY
IN Industry trends

Aluminum, lightweight and tough, is an indispensable basic raw material in modern industry. Due to its light texture and excellent chemical corrosion resistance, aluminum can occupy a “place” in many industrial fields, from aerospace to transportation equipment, from construction and packaging materials to chemical equipment. While playing a crucial role, the production of electrolytic aluminum also faces problems such as high energy consumption, low level of refined management in production and operation, and urgent need to improve the professional skills of talents.

Production safety cannot be ignored. Therefore, building a digitalized and intelligent production management model through digital and intelligent technologies has become a trend to promote the green and low-carbon development of the industry.

In this context, Inspur Cloud has developed the “Knowledge Industry Model” for the industrial sector. With advantages such as industrial data analysis, industrial knowledge reasoning, and industrial code generation, the digital empowerment of the aluminum industry has enabled it to move towards “new” and “green” development. It has successfully helped a Fortune 500 aluminum industry leader achieve digital production and refined operations, paving the way for high-quality development of the aluminum industry.

Aggregate global data to activate the “source power” of aluminum industry
Aluminum electrolysis production is a process with significant delay, multivariate coupling, and nonlinearity. Due to the high temperature and strong magnetic field environment inside the aluminum electrolytic cell, the real-time material stacking status and electrochemical reaction process inside are difficult to directly measure through sensor devices, resulting in the inability to accurately describe the production status qualitatively and quantitatively, which has become a core challenge restricting the improvement of electrolytic cell production efficiency, stability, and service life.

In response to the “data blind spots” in the electrolysis process, Zhiye Big Model has created the Intelligent Aluminum Data Capability Center. On the one hand, it realizes the rapid assetization of massive, multi-source, and heterogeneous data across the entire domain, reducing data extraction from days to seconds. Based on the collection and aggregation of historical data on electrolysis cell production such as molecular ratio, cell temperature, average voltage, current efficiency, two-level, power consumption, and furnace bottom pressure drop, it provides efficient and easy-to-use data services to achieve data-driven refined operation; On the other hand, by collecting data assets to extract knowledge of the electrolysis industry, the monitoring of electrolytic cell conditions can be transformed from manual duty to machine active learning. Real time monitoring of core indicators such as temperature, average voltage, current efficiency, and furnace bottom pressure drop of the electrolytic cell can be achieved, and a real-time evaluation model for the operation status of the electrolytic cell can be established. Root cause analysis of common abnormal conditions can be conducted and effective solution guidance can be provided. Hidden knowledge can be fully explored and transformed into valuable intellectual property rights, continuously helping enterprises innovate and develop.
At present, in the application process of the aluminum industry, the “Zhiye Big Model” has gathered data collection from 32 aluminum factories and 4000+equipment, completing data integration covering various process sections of electrolytic aluminum, alumina, and thermoelectric sectors, achieving a 5% improvement in various numerical collection accuracy, and providing strong support for equipment diagnosis and warning, and assisting intelligent decision-making.
Developing segmentation models to shape a strong production engine
Building a data model is the foundation for conducting data analysis. Based on the most complete and comprehensive data of the aluminum industry, Inspur Cloud has developed targeted multi class segmented scenario models, which can not only empower various production links and capture the digital transformation of core manufacturing links, but also build a high-level industry dataset, effectively optimizing the entire industry ecosystem and promoting industry transformation and upgrading.
The “Zhiye Big Model” regards each electrolytic cell as a complete lifecycle process, carries out various algorithm research and development, dynamically generates electrolytic cell plan aluminum output evaluation, fluoride salt addition recommendation, automatic adjustment of alumina feeding interval time, automatic adjustment of anode voltage and other process algorithms. A production control and process optimization model for the entire lifecycle of aluminum production based on the “Knowledge Industry Big Model” has reduced the abnormal fluctuation rate of electrolytic cell classification by 13%, significantly improving the accuracy of cell control decisions.

The prerequisite for the application of industry big models is first to form the basic training set of the big model. Inspur Yunzhou is also developing a large-scale model for the aluminum electrolysis industry. By sorting out basic data such as aluminum industry mechanism knowledge, expert experience, operating procedures, industry standards, policies and regulations, a knowledge text set and a question and answer set are constructed, forming an aluminum industry knowledge base based on industrial large-scale model technology. Among them, through multimodal services, it is also possible to achieve voice interaction between users and the system, assisting enterprises in carrying out employee training and evaluation work, and applicable to scenarios such as job knowledge learning, equipment maintenance guidance, and solution recommendation.
Building a new coordinate for digital twin innovation quality control
Currently, the development of digital twin technology has entered the “fast lane”, and a good “chemical reaction” continues to emerge between digital twin and industrial development. Inspur Cloud Island takes advantage of the situation and explores the application of cutting-edge technologies such as industrial digital twins. Based on the “knowledge industry big model”, it constructs a “digital industry factory” for the industry that is similar to the physical aluminum industry, and realizes, deepens, and optimizes digital transformation and upgrading.
Under the guidance of data, Inspur Yunzhou focuses on creating a scenario based concept for factory management by utilizing digital twins and VR to achieve full process transparency in production. The digital twin scenes in the factory and workshop are displayed on a large screen to achieve visualization of factory production. The personnel used in the workshop production process, electricity consumption, gas consumption, aluminum output, working conditions of each tank, and the operation of changing poles to produce aluminum can all be displayed uniformly.
By constructing this aluminum electrolysis simulation twin that integrates physical dynamic models and virtual simulations, the “Zhiye Big Model” provides a multifunctional digital twin service that integrates plant operation management, government visit experience, collaborative scheduling, park virtual simulation, emergency linkage command, and major event support. Currently, the digital twin construction of 12 square kilometers of electrolytic aluminum and alumina plant areas has been completed, achieving real-time perception, dynamic analysis, and elastic control of more than 6000 devices and over 3000 core production indicators. Based on the obtained production data, production personnel can quickly grasp the production situation and optimize it in a timely manner, ultimately achieving a 1.5% optimization of single slot aluminum output and an average annual reduction of 100 million kWh in power consumption.
Only by truly rooted within the industry, delving into practical application scenarios, and promoting trustworthy and free flow of business based data, can the value of big models be truly demonstrated, thereby empowering thousands of industries. The successful practice of the “Zhiye Big Model” in the aluminum industry has created a typical demonstration effect, which is conducive to the promotion and application of the big model in the industrial field.