结构方程模型建立混凝土材料行为模式

The concrete mix design is a multi-objective optimization problem. It must meet the requirements of the strength, work degree as well as the economy. This project proposed a new concrete mix design methodology by three theoretical includes structural equation modeling (SEM), particle swarm algorithm (PBA) and multi-objective (MO) optimization design. SEM is a nonlinear model of the architecture with the optimization algorithm. It does not assuming the premise knowledge based model and establishes the multiple nonlinear self-organizing model. PBA is a non-linear optimization tool that integration swarms behavior patterns of bee algorithm (BA) and particle swarm optimization (PSO) inspired optimization process. Based on the strength, work degree, upper and lower limits, material proportion and volume of material limitations, etc. of the multi-objective concrete mix design theory. The PBA establishes a cost-optimized design of concrete mix.

混凝土配比设计可视为一个多目标最佳化问题, 它必须同时满足强度、工作度以及经济性的要求。本计划提出一个新的混凝土配比设计方法论, 其三大理论支柱是结构方程模型(SEM)、粒子蜂群演算法(PBA)及多目标(MO)配比优化设计。SEM是非线性模型的架构，可配合最佳化演算法，在不需假设模型之前提，建立多元非线性自组织模型，PBA是非线性模型的优化工具，是整合群智慧演算法方法中，蜂群演算法与粒子(鸟群)演算法群体行为模式的灵感优化过程；最后依据前述的成果，在考虑混凝土配比设计理论的强度、工作度、材料上下限、材料比例及材料体积等的多目标限制下，以粒子蜂群演算法建立最优化成本设计下的混凝土配比。

本研究旨在以基于遗传算法(genetic algorithm)的强化式学习(reinforced learning)解决二个实际的问题：(1)高性能混凝土(high performance concrete)强度行为建模；(2) 高性能混凝土工作度行为建模。研究结果显示，若用户要求的是可理解的HPC材料行为模型，遗传算法结合运算树(genetic algorithm of operation tree)是一个准确且能够产生自组织公式的HPC强度及坍度模型建构方法；若用户对于可理解的HPC材料行为模型要求度不高，仅在乎最后的结果是否准确，则倒传递网络(back-propagation networks)是一个更准确且快速的HPC强度及坍度模型建构方法。

内容获取失败，请点击重试