基于层次分析法和神经网络方法的建筑结构（基础）选型研究

【作者】 王经建；

【导师】 刘伯权；

【作者基本信息】 长安大学 ， 结构工程， 2004， 硕士

【摘要】 建筑结构的初步设计是一个涉及面广、综合性强的工作，需要多方面的知识和丰富的经验，目前国内外尚无理论对其进行充分的研究。尤其是西安地区这样的高地震烈度区，建筑结构初步设计的结构体系选择更为重要，它关系了整个建筑的安全、适用和经济。本文在借鉴了前人的研究成果的基础上，提出了在满足安全、适用的条件下，建筑结构经济性评价指标。根据这些指标收集了国内目前比较优秀的一些建筑结构的设计数据，作为专家经验库，充分运用神经计算具有分布式存储、并行处理和自适应学习的特点，显示运算快速、响应灵活、容错性良好的优越性，利用MATLAB语言构造了经济性评价神经网络系统，对上部结构、地下基础的经济性作了分析。又由于各种建筑的使用功能不同、各种人为因素对选型的影响，各种建筑结构对安全性、适用性、经济性和施工难以程度方面的侧重不同，会使得最终的选型结果不同(权重不同)；另一方面，有些影响结构选型因素又是模糊的，不可能用一个精确的量化指标来表达。因此本文运用模糊数学和层次分析方法，对多属性的问题进行了探讨。 实例应用表明，基于神经网络系统的经济性评价系统和多属性评判方法进行的结构体系选型是有效的。更多还原

【Abstract】 In the early stage of the design process, the design of a building is a complex work. It needs various knowledge and professional experience for the structure design. Until now there is not a theory analysis it adequately. Especially in a place where the high earthquake has to be considered in the design as xi’an, the structure forms selection is more important, because it connected with the safety, applicability and economy of the whole building. So using the former persons’ harvest, the economical evaluating index on the condition of safety and applicability is raised in this paper. And based on these indexes, the design data of some outstanding building has been collected as the expert databank. And the qualities of the ANN, storage dispersedly, management simultaneity, learn automatically, calculate quickly, response neatly, and so on, are adequately used in the research. Then using the MATLAB language, an economic evaluating artificial neural network system has been set up to analysis the economy of the upper structure and the basement. Because of the different using function and some artificial factors, different structure have different power to safety, applicability, economy and difficulty of construction. On the other side, some factors are fuzzy and cannot use an exact number to describe it. So in this paper, the fuzzy method and AHP have been used to deal with the multi- attribute problem. And the example using has been approved that both ANN and multi-attribute effectual process are effectual.更多还原