Recently, a team led by Dr. Ye Mao and Academician Liu Zhongmin from the Dalian Institute of Chemical Physics, Chinese Academy of Sciences, in collaboration with Huawei Technologies Co., Ltd., along with the School of Software at Dalian University of Technology and the Yulin Zhongke Clean Energy Innovation Institute, has developed the Advanced Intelligent Chemical Engineering Model 1.0. This model was unveiled at the Huawei "Ascend AI Framework Summit 2024".
Built upon the Huawei Ascend AI infrastructure, the Advanced Intelligent Chemical Engineering Model facilitates rapid retrieval of chemical knowledge and enables autonomous design and optimization of chemical process engineering. It is poised to shorten the R&D cycle of chemical processes, potentially expediting the transition of laboratory achievements into industrial applications.
Chemical industry is a crucial foundational sector of the national economy, with products spanning across various domains like clothing, food, housing, and transportation. However, under the traditional paradigm, the development of new technologies in the chemical industry faces multiple challenges. Due to the lack of precise theoretical descriptions for multi-scale complex chemical processes, new technology development relies on step-by-step amplification. This often results in R&D cycles of over a decade, accompanied by high costs and significant market risks. Artificial intelligence technologies, represented by grand models, offer a data-driven new paradigm for chemical technology development, potentially altering the long-standing reliance of China's chemical design field on foreign industrial software.
Drawing upon the abundant chemical application scenarios and massive R&D data from Dalian Institute of Chemical Physics (DICP), the research team has developed an Intelligent Chemical Engineering Grand Model, laying an important foundation for establishing a new paradigm in chemical R&D and achieving domestic independence in chemical industrial software.
Simultaneously, the team has developed a Chemical Data Processing Platform centered around the Intelligent Chemical Engineering Grand Model. This platform enables efficient processing of heterogeneous chemical data from multiple sources for pre-training large models. Additionally, a framework for constructing knowledge graphs in the chemical domain has been developed, enabling automatic annotation and structured processing of chemical domain knowledge based on large models.
The Intelligent Chemical Engineering Grand Model not only facilitates rapid and accurate retrieval of chemical knowledge through interactive question-answering methods but also preliminarily realizes the automatic generation, simulation, and feedback optimization of chemical process flows. Furthermore, the research team has developed a fully automated high-throughput catalytic reaction robotic testing device. By coupling with the Intelligent Chemical Engineering Grand Model, this device explores the catalytic reaction mechanisms, achieving automatic generation of reaction kinetics for processes such as methanol-to-propylene.
