Genetic Algorithm Devleopment for Multi-Objective Optimization of Seismic Mixed-Constructued Structures With/ Without Control of Systems

本课题研究的是智能工程中经常出现的一个基本问题，即涉及多变量和荷载条件的建筑和民用基础设施系统的优化设计。优化设计是指，除其他外，最大限度地延长使用寿命，安全性和功能性，同时尽量减少材料使用，人工，维护等。在多目标优化设计过程中，一个关键步骤和要求是对复杂的优化公式给出一致、稳定的解的计算算法。在本项目中，将开发一种创新的优化技术，即遗传算法，用于地震条件下的民用结构应用。该算法基于达尔文的适者生存策略，通过消除不适合的特征，并通过繁殖、重组（交叉）和突变进行随机交换。重点介绍了n父均匀交叉算子的构造和并行化，以实现全局和鲁棒最优解的新兴计算技术。本研究还开展了其他研究工作，包括混合建筑材料的建筑系统结构建模，建筑结构的多目标函数和约束的制定，基于静态和地震荷载的优化设计标准的开发，有/没有控制系统，以及评估和改进当前的优化设计算法。* * *

9520804 Cheng This project deals with a fundamental problem which often arises in intelligent engineering aimed at optimum design of buildings and civil infrastructure systems where multiple variables and loading conditions are involved. Optimum design means, among other things, to maximize service life, safety and functionality while minimizing material use, labor, maintenance,etc. A critical step and requirement in the process of multi-objective optimal design is a computational algorithm which could give consistent, stable solutions to the complex optimization formulation. In this project, an innovative optimization technique, namely, a genetic algorithm will be developed for civil structural applications under seismic condition. This algorithm is based on the Darwinian survival-of-the-fittest strategy by eliminating unfit characteristics and using random exchange through reproduction, recombination (cross over), and mutation. Emphasis is placed on formulation of n-parent uniform cross- over operator and parallelization to achieve an emerging computing technique for both global and robust optimum solutions. Other research efforts also undertaken as a part of this research include structural modeling of building systems with mixed construction materials, formulation of multi- objective functions and constraints for building structures, development od optimum based design criteria for static and earthquake load with/without control systems, and assessment and improvement of current optimum design algorithms. ***