【作者】 韩玲；

【导师】 杜修力；

【作者基本信息】 北京工业大学 ， 结构工程， 2004， 硕士

【摘要】 摘 要遗传算法是近些年来发展比较完善的一种全局寻优方法，鲁棒性能好，适用性强、精度高，尤其适用于损伤指标常常对损伤敏感性差、观测数据噪声大的土木工程结构的反演分析。对于复杂大自由度系统的反演分析，遗传算法计算量巨大，关键在于进化计算中包含大量正演分析。减少反演分析中的正演计算次数，是扩大遗传算法使用范围的有效途径。神经网络由于其大规模并行处理、容错性、自组织、自适应能力和联想功能强的特点，已成为解决很多问题的重要手段，广泛应用于各工程领域。本文提出将神经元网络模型用于遗传进化中适应度函数的预测，实现了遗传代内算法的并行性。本文将此算法称为经验遗传算法。六个经典算例表明：本文提出的经验遗传算法计算效率高，大大减少了正演计算工作量，对大型复杂问题的反演分析有重要的意义。损伤识别指标的确定是结构健康监测的关键，通过研究分析及比较，本文选择柔度矩阵作为损伤识别的指标，提出将经验遗传算法应用于高架桥结构损伤识别的思路。更多还原

【Abstract】 Abstrcat Genetic Algorithms is a global optimization method perfectly developed in theseyears, which is robust、widely used and has high accuracy. It often used to solve civilengineering problem, especially for conversion analysis with low sensitive damageindex and inspective dates affected by big noise. In the complex conversion analysiswith multi-degrees of freedom, genetic algorithm has so many calculation counts. Thepoint is that evolution algorithm has a great deal positive analysis. Reduce the positivecalculation counts is an effective method to enlarge the range of GA’s using. Due to ANN’s large scale concurrent ability 、 the fault tolerance 、self-organization、self-adapting and associational function, it was an important meansto solve problems in many project fields. An algorithm is proposed in this paper whichmakes an organic assemblage between GA and ANN. It performs the parallelcomputation in the evolutional population. The six classical functions show the EGAis more effective than GA. The counts are reduced. It is significant for the complexproblems. We call it empirical genetic algorithms. The definition of damage index is the key in the structure health supervision.Through comparing and research, flexibility matrix, a damage indetification index, isused in this paper for diagnosis damage location and severity. Based on the theoryproposed in this paper, an ideas is proposed in this paper——apply EGA to detect theviaduct damage.更多还原