At present, artificial intelligence is widely and deeply integrated with various industries and fields in the economy and society, and has become an important engine to drive the industry to upgrade iteratively and reshape the development pattern. China Building Materials Group actively embraces this historic opportunity, fully deploys the deep integration of artificial intelligence technology and building materials industry, and opens a new journey of transformation and upgrading from "manufacturing" to "intelligent manufacturing". We officially launched a series of publicity reports on "AI + Action", focusing on front-line practice, excavating benchmarking cases, and striving to create a strong atmosphere of "understanding AI, using AI and creating AI", so as to accelerate the popularization and deep empowerment of AI technology within the group.
This issue goes into the AI innovation practice of China Building Materials Research Institute Limited (hereinafter referred to as "China Building Materials Research Institute").
AI enables the whole chain of building materials, Digital Intelligence Reconstructs the New Ecology
of Industry 
foundation of AI R & D
is the core cornerstone of AI R & D. In view of the challenges faced by research and development in the field of building materials, such as unclear cross-scale structure-activity relationship, difficult multi-field coupling and multi-source heterogeneous data barriers, the General Institute of Building Materials of China has innovatively designed a "one subject, multi-disciplinary" data base architecture of "one lake and one warehouse" to systematically promote the data aggregation and management of inorganic non-metallic materials. By overcoming the key technologies of automatic extraction, processing and fusion of multi-modal heterogeneous data, the General Institute of Building Materials of China has created a database of building materials covering composition, parameters, performance and carbon emissions, and constructed a multi-level and full-chain data system of "acquisition-integration-knowledge-use" to achieve 100% accurate and automatic recognition of heterogeneous characters. It has created a knowledge map of more than 2000 knowledge nodes and more than 1000 relationships.
Up to now, the material database has gathered a large amount of resources, including 41651 basic data of materials and more than 750 million data of building materials covering 110 production lines. As well as the establishment of the system of more than 500000 literature data and more than 50000 cement
Forging
"Algorithm-Model-Agent" R & D engine-driven
material intelligent design
relies on deep industry accumulation, and China Building Materials Institute has constructed a core technology matrix from "Algorithm, Model" to "Big Model, Agent". Focusing on the key links of preparation and service of low-carbon building materials, the five-element collaborative algorithm of "discipline-formulation-process-equipment-verification" is innovatively established, which deeply couples multi-disciplinary knowledge, formula process, equipment working conditions and experimental verification, and forms a complete closed-loop iteration mechanism of "prediction-design-simulation-verification-learning", so as to promote material research and development from "trial and error" to "system optimization".
Based on this theoretical framework, China Building Materials Institute has developed a forward performance prediction and component reverse design model, which uses deep neural network fusion and graph neural network technology to achieve cross-scale correlation mapping of "composition-structure-performance". The prediction error of hydration thermodynamics is less than 4. The model system has successfully designed 38 low-carbon clinker systems and 72 low-carbon compositions, and found 2 new low-carbon clinker phases with high activity, which meet the needs of multi-objective collaborative research and development. In addition, the energy-quality coupling model of high temperature calcination and particle crushing process is developed to realize multi-field and multi-phase accurate simulation under new equipment and new energy fuel conditions, so that the temperature prediction error of core equipment is less than 3%, and the pressure loss is reduced by more than 15%.
Enabling the
construction of an intelligent research and development platform
for low-carbon building materials to achieve independent experimental closed-loop
, China Building Materials Institute integrates data, algorithms and computing resources, and initiates a digital research and development platform for low-carbon building materials. Based on the self-developed CBMA model, material data base, scientific modeling tools, academic agents and automated experimental equipment,
China Building Materials Institute has deeply integrated 75 years of material discipline accumulation with cutting-edge digital technology. Build the first "AI + Intelligent Laboratory" with 100% localization of core hardware and software in the building materials industry. The laboratory realizes an autonomous and complete experimental closed-loop from scientific problem analysis, literature intelligent analysis, experimental scheme design to experimental automation execution, and constructs a new research and development form with data and AI as the core and multi-agent efficient collaboration, which breaks the space-time barrier of traditional material research.
To
promote the application of full-scene industry and release the value
of digitalization transformation, China Building Materials Institute will deeply bind AI technology with industrial demand, and form a full-process digitalization solution covering many fields. It has been applied on a large scale in the research and development of special materials, low-carbon manufacturing, resource recycling and other scenarios.
浙公网安备33010802003254号