面向宏观性能演化机理的水泥水化动力学反向建模研究

Computer modeling and simulation for cement hydration have been proven to be an effective tool for understanding the mechanism of cement hydration, predicting the hydration process, researching the origin of macroscopic performance, and improving the design of high performance cement material. However, due to the extreme complexity of cement hydration mechanism and paste microstructure, there’s no hydration kinetics model, which reveals the dynamic functional relationship between macroscopic performance and microstructure to aid the understanding of developmental mechanism of macroscopic performance. Scientists failed to image three dimensional solid microstructure containing information of phases for supervising modeling, failed to construct feature variables describing characteristics of microstructure precisely, and also failed to establish hydration kinetic equations reflecting complex changing of real structure and performance. Guided by multimodal cement data, this project will reversely build hydration kinetics model of dynamic relationships between macroscopic performance and microstructure using intelligent method. The project first synthesizes three dimensional sold microstructure image containing information of phases using 2D-3D registration, neural network classification, and other technologies. Then, the project automatically generates feature variables for complex microstructure to describe its characteristics precisely using gene expression programming and other intelligent computing methods. In addition, the project will reversely extract hydration kinetic equations fusing macroscopic performance and microscopic features using interactive hybrid evolutionary computation under the environment of GPGPU clusters.

水泥水化过程的计算机建模与模拟是揭示水泥水化机理、预测水化过程、研究宏观性能起源及辅助高性能水泥材料设计的有效手段。然而，由于水泥水化机理和泥浆微观结构的极端复杂性，目前尚不存在完整揭示水化过程中宏观性能与微观结构动态函数关系的水化动力学系统来帮助理解宏观性能的演化机理，人们未能清晰成像含物相信息的三维实体微元来指导建模，未能构造对微观结构特性进行精确描述的特征变量组合，也没能建立反映真实结构与性能复杂变化的水化动力学方程。本课题力求以多模态水泥数据为导向，通过智能算法反向建立宏观性能与微观结构动态关系的水化动力学模型。课题首先通过跨维配准和神经网络分类等技术合成含物相信息的三维实体微观结构复合图像，然后利用基因表达式编程等智能计算方法为复杂的微观结构自动生成特征变量来精确描述微结构特性，并在GPGPU机群环境下使用交互式混合进化计算反向萃取宏观性能和微观特征融合的水化动力学方程。

水泥及其基材料是应用广泛的重要基础材料。水泥水化过程的计算机建模与模拟是揭示水泥水化机理、预测水化过程、研究宏观性能起源及辅助高性能水泥材料设计的有效手段。本课题力求以多模态水泥数据为导向，通过智能算法反向建立可以反映宏观性能演化机理的水化动力学模型。在基金项目的支持下，课题组在《IEEE Transactions on Cybernetics》、《IEEE Transactions on Fuzzy Systems》、《IEEE Transactions on Industrial Informatics》、《Pattern Recognition》、AAAI等刊物与会议上发表了相关学术论文31篇，获得2项软件著作权，授权发明专利1项。基于真实水泥数据，初步探索了面向水化过程的水泥微观结构构建；研究了水泥微观结构特征变量的独立抽取技术；研究了基于进化计算的水泥动力学方程快速发现；研发了一个支持多种智能模型快速集成的水泥真实微观结构演化仿真系统。针对课题中所涉及到的反向演化、图像合成、函数生成、神经网络、参数优化、结构分析等问题，进行了深入的理论和方法研究。在所获得的理论研究成果的应用方面，除了计划研究外，还进行了成果应用的扩充和调整，结合部分典型的应用领域，增加了理论方法的应用研究。