湖南大学硕士学位论文改进实数编码的遗传算法及其在结构损伤诊断中的应用姓名：林建雄申请学位级别：硕士专业：结构工程指导教师：易伟建20070601改进实数编码的遗传算法及其在结构损伤诊断中的应用 II摘 要 结构损伤的检测和识别是工程结构健康检测和检修的重要部分。 结构在服役的过程当中，往往由于腐蚀、疲劳、老化，或者由于受到冲击荷载、地震和风的影响而受到损伤。 这些不同原因造成的不同程度的损伤经过长期的积累必然导致结构发生破坏或使用性能降低。尽早的发现结构损伤，并及时检测和修复，不仅可以大大降低结构的维护、 维修费用， 还可以减少和避免不必要的生命财产损失。因此， 对工程结构进行实时的健康检测和诊断是目前国内外科技工作者和工程设计研究人员广泛关注和重点研究的领域之一。 结构发生损伤必然导致结构动力特性的改变， 基于结构动力特性的结构损伤识别是当前学术界和工程界研究的热点。本文的内容主要包括以下几点： (1) 首先对结构损伤识别的理论与方法进行了阐述，论述了损伤诊断的国内外研究现状，探讨了频率、振型等模态参数变化与损伤的内在联系，分类阐述了各种损伤诊断方式并分析了各种方法的特点。 (2) 对标准遗传算法的基本原理与实现技术进行了系统的研究，深入分析了其存在的缺陷与不足以及算法的可能改进途径， 在此基础上提出了自己的改进策略，定义了新的衡量种群多样性的指标，并将该指标引入到遗传算子中，使交叉和变异概率根据该指标自适应调整， 基于此提出了改进的基于实数编码的自适应并行遗传算法，通过对典型测试函数的数值实验，并与其他方法进行对比研究，表明本文提出的方法是一种有效的全局寻优算法。 (3) 应用基于改进的遗传算法的损伤诊断方法，采用数值仿真的方式来检验本文所提方法在损伤识别上的能力。 结果证明本文的方法不仅能够准确的识别损伤单元的位置及其损伤程度，而且具有较强的抗噪能力。 (4) 为了检验该方法在实际工程中的效用，本文基于已有的比例为1：3的四层钢筋混凝土空间框架模型的模态数据识别结构在不同工况下的物理参数。 不同于数值模拟可以准确判断识别结果是否正确，对工程实际的结构，其实际物理参数难于准确衡量， 所以通过对修正前后结构的分析模态和测量模态的相关性分析来衡量识别结果的正确与否。 (5) 某种损伤诊断的方法常常只适合特定问题的研究，为了测试本文方法对标准结构的损伤识别能力， 应用基于本文改进的遗传算法的损伤诊断方法，对由国际结构控制协会与美国土木工程学会（IASC-ASCE）提出的健康检测Benchmark结构进行了分析，结果表明，本文的方法能够准确的识别该框架结构的损伤。 (6) 应用基于本文改进遗传算法的损伤诊断方法识别梁结构的损伤，先后对硕士学位论文 III实验室预应力简支梁及试验铝制悬臂梁进行损伤诊断。结果显示本文的方法，能对梁结构的损伤进行准确的识别。 关键词：损伤诊断；遗传算法；自适应；种群多样性；数值模拟；Benchmark结构；参数识别； 改进实数编码的遗传算法及其在结构损伤诊断中的应用 IVAbstract The detection and identification of structural damage is a vital part of the monitoring and servicing of structural systems during their lifetime. Structural damage in normal service may include corrosion, fatigue, and aging, or it may be caused by impact loads, earthquakes, and wind. Damage in engineering structures is caused by natural disasters, the fatigue of the structures and so on. The accumulation of damage caused by these different reasons will lead the structure wreck or reduce the structure performance. If the damage can be identified as early as possible, and the structure can be repaired in time, it is obviously that a lot of maintenance costs can be saved, and unnecessary loss of human life and property can be avoided and reduced. So, the on-line health monitoring and diagnosis of engineering structures is a popular field which attract lots of attention from domestic and abroad researchers. It is absolutely that occurrence of damage leads to changes in the dynamic properties of the structure. Vibration-based damage identification techniques have received significant attention in academe and engineering fields. (1) First, the theory and methodology of the structural damage detection are expounded in the paper, the research status in quo from home and abroad is also introduced. In this paper, the inhesion relationship between modal parameters ,such as frequency、mode shape, and the damage is discussed. The existed damage detection methods are classified into a few categories and the characteristics of these methods are analyzed; (2) A systemic research about the basic theory and implement technique of the standard genetic algorithm is made, in succession, an in-depth study about the limitation and shortage of the standard genetic algorithm is made. And also the improve approach of the algorithm is analyzed. Based on these analyses an amelioration strategy is proposed, an index which scale the population diversity is defined, and this index is imported into the genetic operator, and it is used to instruct the crossover probability and mutation probability to tune self-adaptively. And an improved self-adaptive parallel genetic algorithm based on real coding is proposed, and some typical test functions are optimized with the proposed method, the results show that the algorithm proposed in this paper is an effective global 硕士学位论文 Voptimization algorithm. (3) This damage detection method based on the improved genetic algorithm proposed in this paper was employed to identify the damage of a frame model simulated with MATLAB. The results show that this method proposed in this paper can identify the location of the damaged elements and the extent of their damage，and this method is noise robust. It is also proved that we can use the limited modal parameters to identify the damage with a high precision. (4) In order prove the effectiveness of the proposed algorithm in practical engineering project, a set of parameter identifications was conducted on a four-story reinforced concrete frame structural model with the scale of 1:3 under the laboratory environment. Different from the numerical simulation, we cant accurately judge whether the identification result is right. Therefore, pertinence analysis between the analytical modal of the updated structure and the actual modal is used to judge whether the result is right. (5)This damage detection method based on the improved genetic algorithm proposed in this paper was employed to analyze Benchmark model structure proposed by the International Association for Structural Control and the American Society of Civil Engineers (IASC-ASCE) Task Group on Structural Health Monitoring. The results show that, the