基于免疫遗传算法的钢筋混凝土框剪结构优化研究

发布时间：2018-08-25 10:24

【摘要】：人工免疫系统模拟了生物免疫系统中与信息处理有关的一些基本概念和机理,其包含两个分支——免疫克隆选择算法和人工进化算法。免疫克隆选择算法模拟了人体自身的免疫系统,与标准遗传算法相比,该算法注重变异算子,在搜索过程中保留了更多局部信息。人工进化算法模拟了人工获得免疫,在一般进化算法的流程中添加了免疫算子。免疫算子利用实际问题的特征信息或先验知识,对种群进化进行局部调整,起到避免种群出现退化或出现无效操作的作用,改善种群的搜索方向。笔者通过对比研究多种仿生算法,得出免疫遗传算法是最适合对框剪结构进行布局优化的结论。本文的主要研究内容和取得的成果包括:(1)推导了框剪结构的抗侧、抗扭刚度矩阵,并通过分析不同优化变量对结构刚度的灵敏度,得出了剪力墙的长度和距形心的距离是影响结构抗扭、抗侧刚度的主要因素的结论。为优化设计模型的建立提供了理论基础。(2)建立了钢筋混凝土框剪结构的优化设计模型。框剪结构的优化目标是结构中梁、柱、墙的造价之和。根据优化变量对结构刚度的灵敏度分析,提出将优化变量分为2级,首先确定墙体的布局再进行截面尺寸优化的思路。第1级优化变量为剪力墙的长度,剪力墙距形心的距离,以及墙、柱的混凝土强度等级。第2级优化变量为框架梁、框架柱、剪力墙的构件尺寸。与优化变量相对应的,结构的约束条件也分为2级,第1级为结构的整体指标约束,第2级为构件的承载力约束和构造约束。第1级优化通过免疫算法对种群进行进化来实现。第2级优化通过网格搜索法来完成。不论是免疫遗传算法还是网格搜索法都适用于离散型变量,可以使结构的优化结果满足模数制要求。(3)在免疫遗传算法中设计了3种疫苗:墙体布置疫苗、混凝土强度等级疫苗和整体约束疫苗。墙体布置疫苗和混凝土强度等级疫苗主要起到将设计习惯和先验知识引入到算法中,令所有个体所对应的结构都具可行性,弥补标准遗传算法在搜索时的盲目性的缺陷。整体约束疫苗的作用是优化算法的搜索方向,提高种群的整体适应度,提高算法的效率。(4)编写了钢筋混凝土框架-剪力墙结构优化设计软件SWOD。软件采用VC++编程语言,采用面向对象的编程方法。软件可以导入PKPM的数据文件STRUSTRU.SAT和LOAD.SAT,保证了输入数据的准确性,并且使平面复杂结构的优化设计成为可能。结构优化过程中需要用到的内力分析、结构设计过程全部由SWOD完成。SWOD只需用户输入少量的设计参数,大多参数根据规范进行设定,使用简便。SWOD的设计利用了面向对象编程方法的多态性,使程序有较高的灵活性,若程序需求发生改变,对原有代码的修改量很小。(5)SWOD采用CSC矩阵存储策略和基于超节点的LDL求解器,适用于大型稀疏矩阵求解。CSC矩阵存储策略显著降低了结构对计算机存储空间的需求,基于超节点的LDL求解器能够高效完成结构的动力分析和静力分析,使大型框剪结构优化设计成为可能。(6)分析对比了5个算例,分别从SWOD的内力分析和结构设计的正确性,网格搜索法的有效性,对比三种设计模式的结构优化结果,多层框剪结构的优化效果以及高层框剪结构优化效果等几方面,证明了将免疫遗传算法用于框剪结构优化的有效性和SWOD软件的使用范围的广泛。根据本文免疫遗传算法编写的SWOD软件得出的框剪结构优化结果满足各项设计规范要求,引入多项设计习惯,可以提供给设计人员可行的初步设计方案。
[Abstract]:Artificial immune system simulates some basic concepts and mechanisms related to information processing in biological immune system, which includes two branches: immune clonal selection algorithm and artificial evolution algorithm. Artificial evolutionary algorithm simulates artificial immunity and adds an immune operator to the general evolutionary algorithm. The immune operator makes use of the characteristic information or prior knowledge of the actual problem to adjust the population evolution locally, which can avoid the degeneration or invalid operation of the population. The main contents and achievements of this paper include: (1) The lateral and torsional stiffness matrices of frame-shear structures are derived, and the stiffness matrices of frame-shear structures with different optimization variables are analyzed. It is concluded that the length of the shear wall and the distance from the center of the shear wall are the main factors affecting the torsion resistance and lateral stiffness of the structure. The theoretical basis is provided for the establishment of the optimal design model. (2) The optimal design model of reinforced concrete frame-shear structure is established. According to the sensitivity analysis of the optimization variables to the structural stiffness, the optimization variables are divided into two stages. The first stage is the length of the shear wall, the distance between the shear wall and the center of the shape, and the concrete strength grade of the wall and the column. The second stage is the frame beam and the frame. Corresponding to the optimization variables, the constraints of the structure are also divided into two levels: the first level is the overall index constraints of the structure, the second level is the bearing capacity constraints and structural constraints of the members. Epidemic genetic algorithm or grid search method are both suitable for discrete variables, which can make the structure optimization result meet the modulus requirements. (3) Three vaccines are designed in immune genetic algorithm: wall layout vaccine, concrete strength class vaccine and whole constraint vaccine. The design habit and prior knowledge are introduced into the algorithm to make the corresponding structures of all individuals feasible and make up for the blindness of the standard genetic algorithm in the search process. The software SWOD adopts VC++ programming language and object-oriented programming method.The software can import PKPM data files STRUSTRU.SAT and LOAD.SAT to ensure the accuracy of input data and make the optimization design of complex planar structures possible. SWOD only needs users to input a small number of design parameters, and most parameters are set according to the specifications. The design of SWOD makes use of the polymorphism of object-oriented programming method to make the program more flexible. If the program needs to be changed, the modification of the original code is very small. (5) SWOD uses CS. C matrix storage strategy and LDL solver based on super-node are suitable for large-scale sparse matrix solution. CSC matrix storage strategy significantly reduces the demand for computer storage space. LDL solver based on super-node can efficiently complete the dynamic and static analysis of the structure, making the optimization design of large-scale frame-shear structure possible. Five examples are analyzed and compared. From the internal force analysis of SWOD and the correctness of structural design, the effectiveness of grid search method, the optimization results of three design modes, the optimization effect of multi-storey frame-shear structure and the optimization effect of high-rise frame-shear structure, it is proved that the immune genetic algorithm is suitable for the optimization of frame-shear structure. The results of optimization of frame-shear structure based on SWOD software compiled by immune genetic algorithm in this paper can meet the requirements of various design specifications. The introduction of a number of design habits can provide designers with feasible preliminary design schemes.
【学位授予单位】：上海大学
【学位级别】：博士
【学位授予年份】：2015
【分类号】：TU398.2

【参考文献】