Concrete as the second-most consumed material globally after water plays a vital role in modern construction and infrastructure development due to its superior mechanical properties, longevity, and versatility. Conventional methodologies for concrete design predominantly rely on empirical approaches, often resulting in extended cycles for material development and suboptimal performance outcomes in practical applications. The advent of the Materials Genome Initiative (MGI), launched in the United States in 2011, highlights a need for data-driven approaches to material design. Concurrently, significant advancements in Artificial Intelligence (AI) have new avenues for enhancing material discovery and characterization. It is thus necessary to establish a robust and comprehensive database capable of accommodating the extensive, heterogeneous, and multi-scale data associated with various concrete properties and performance metrics. This study was to develop an innovative AI-powered intelligent design framework specifically tailored for a concrete material performance gene database. This framework could promise efficient data management and intelligent analytical capabilities and facilitate advanced material development and optimization.

This framework system integrated the information on concrete material composition, mix parameters, environmental factors, etc., and extracted the topological structure and coupling relationships between gene features to establish a concrete material knowledge graph based on conventional relational databases, realizing the visualization of gene feature relationships. This work designed a "gene importance-driven hierarchical density clustering" algorithm for quantitative analysis of concrete material gene features, extracting gene structure with different action relationships, via addressing the complex influencing factors and multi-scale coupling characteristics of concrete material performance. On this basis, this work also designed a closed-loop reasoning path from feature selection, model prediction, and contribution quantification for typical concrete material performance to reduce the cycle of traditional test methods, lower experimental costs, and achieve rapid evaluation and iterative optimization of concrete performance. This could provide a scientific theoretical basis for material selection via the performance prediction model as a core model and combining with multi-objective optimization algorithms. The framework could give a systematic technical framework and theoretical support for intelligent material research via implementing the whole-process design from concrete material data management to intelligent development, promoting the deep integration and innovative development of material science research and civil engineering decision-making.

The implementation of the AI-enabled framework demonstrates substantial enhancements in the accuracy and reliability of concrete performance predictions. The algorithms effectively uncover intricate and complex relationships between various material constituents, and their resultant influence on the critical performance indicators, such as compressive strength, durability, and workability. The unique integration of multi-source data within the database facilitates enhances feature extraction and visual representation of critical metrics, effectively addressing the intricate coupling effects that often complicate conventional testing methodologies. Through rigorous testing, validation, and benchmarking against existing standards, the framework exhibits exceptionally high accuracy and reliability in predicting concrete properties. This robust performance ultimately supports the creation of optimized mix designs, propelling advancements in concrete technology and application.

This study designed the AI-powered Intelligent Design of Concrete Material Gene Database, providing a systematic and technically sound foundation for concrete material research, and significantly enhancing the integration of material science principles and informed engineering decision-making. This study established a high-quality concrete gene database that could facilitate rapid evaluations and predictions via effectively leveraging advanced AI technologies. Furthermore, this study supported innovative advancements in concrete material design, paving a way for more efficient, sustainable practices in construction. Future research endeavors could broaden the system's capabilities, focusing on cross-disciplinary applications and the extension of the database to include low-carbon, environmentally-friendly material options.