【Ten Years of Rapid Development in Chemical Equipment】From 2021 to 2024, Dalian Institute of Chemical Physics released the Intelligent Chemical Engineering Large Model 2.0

Thread Content

Safety first. Prevention first – let’s give praise and support to the achievements made in China’s chemical technology and equipment sector. **********************【Ten Years of Rapid Development in Chemical Equipment】A continuously updated summary post is available; feel free to join the discussions at https://bbs.hcbbs.com/thread-3576046-1-1.html (Source: Haichuan Chemical Industry Forum Website) ***************** On November 27th, the Dalian Institute of Chemical Physics under the Chinese Academy of Sciences unveiled the Intelligent Chemical Engineering Model 2.0 at the “Conference on Utilizing Artificial Intelligence to Facilitate Supply-Demand Matching in New-Type Industrialization and to Promote the Development of the Petrochemical and Chemical Industries.” This model was developed through collaboration between Researcher Ye Mao from the Department of Low-Carbon Catalysis and Engineering at the Dalian Institute of Chemical Physics, Academician Liu Zhongmin’s team, and other relevant organizations; it will play a key role in enabling the intelligent transformation of the chemical industry. 　　The chemical industry has long been constrained by traditional models; the development of new technologies relies on incremental scaling, resulting in long cycles and high investment costs ； It also faces issues such as insufficient security capabilities, low process efficiency, and high energy consumption and carbon emissions. Therefore, the chemical industry urgently needs a paradigm shift. 　　To accelerate the transformation toward lower carbon emissions, greater intelligence, and higher sophistication, the team led by Ye Mao at the Dalian Institute of Chemical Physics has been working since 2016 on applying artificial intelligence technologies to the development of chemical technologies. In March this year, the team, in collaboration with Huawei Technologies Co., Ltd., Dalian University of Technology, and Yulin Zhongke Clean Energy Innovation Institute, released the Intelligent Chemical Engineering Large Model 1.0, which already possesses capabilities such as chemical knowledge retrieval, autonomous design and optimization of processes, and automatic generation of reaction kinetics data. 　　Building on the Intelligent Chemical Engineering Large Model 1.0, a team from the Dalian Institute of Chemical Physics, in collaboration with iFlytek Co., Ltd. and the Documentation and Information Center of the Chinese Academy of Sciences, further developed the Intelligent Chemical Engineering Large Model 2.0 using iFlytek’s Xinghuo Cognitive Large Model as a foundation. The Intelligent Chemical Engineering Large Model 2.0 features 70 billion parameters and an 8K context length. It was developed using 20 billion tokens of unsupervised data from the chemical engineering field and 1 billion tokens of supervised fine-tuning data, through secondary pre-training and supervised fine-tuning on domestically developed computing platforms. It demonstrates outstanding capabilities in various specialized fields such as chemical engineering fundamentals, industrial catalysis, fluid simulation, chemical safety, and chemical separation. 　　Leveraging the Intelligent Chemical Engineering Large Model 2.0, the team, in collaboration with Huawei Technologies Co., Ltd., Dalian University of Technology, Yulin Zhongke Clean Energy Innovation Institute, Shenzhen University, and others, has developed four intelligent platforms for catalytic reactions, process development, pilot-scale scaling up, and plant optimization. These platforms include an intelligent robot-based system for catalytic reaction experiments, an agent for automatic generation of catalytic reaction kinetics models, an agent for rapid analysis of the physical properties of multifunctional catalysts, an agent for recognition and analysis of PFD/P&ID diagrams, an agent for automatic generation and optimization of chemical process flows, an agent for automatic 3D modeling of plants, a digital twin system for the DMTO process for producing olefins from methanol, and a system for intelligent simulation, modeling, and optimization of DMTO plants. Together, they cover all the key aspects of chemical technology development and application, including catalyst evaluation and characterization, process development, pilot-scale scaling up, and plant operation. 　　The results and applications released this time are expected to help drive a transformation in industry paradigms, steering the chemical industry toward lower carbon emissions, higher sophistication, and greater intelligence.