【摘要】 本文建立了基于模态曲率法和人工神经网络技术相结合的、适用于大跨度空间网格结构的损伤定位新方法,即首先应用模态曲率法判断结构是否发生损伤并识别发生损伤的局部结构,然后对发生损伤的局部结构利用人工神经网络技术识别损伤的准确位置。通过分析和比较发现,以模态曲率为基础的损伤参数比较适合于大跨度空间网格结构的损伤定位,三种以模态曲率为基础的损伤定位参数按有效性进行排序,从低到高依次为模态曲率、模态曲率差、模态曲率变化率;针对天津奥林匹克中心体育场大跨度悬挑管桁结构进行了不同损伤状况的数值模拟,验证了所建立的损伤定位方法的适用性和有效性。研究结果表明:利用模态曲率变化率识别损伤发生的大致位置,当单榀桁架发生损伤时,识别的准确率达到100%,当多榀桁架同时发生损伤时,识别的准确率达93.7%;采用人工神经网络技术识别损伤桁架的准确损伤位置时,在无测量噪声影响下,损伤定位的准确率达到97.0%,且测量噪声对损伤定位准确率的影响很大。更多还原

【Abstract】 In this paper,based on the combination of the mode curvature method and the artificial neural network technology,a new method of damage localization for the long-span spatial lattice structures is established,in which the decision of whether damage will occur or not and the identification of damaged partial structure are to be carried out by using the mode curvature method firstly,and then the accurate damage location in the damaged partial structure is recognized using the artificial neural network technology.Through the investigation and comparison,the parameters suitable for the damage localization of the long-span spatial lattice structures are determined,and the different damage cases for a long-span cantilever tubular truss structure of the Tianjin Olympic Center Stadium is numerically simulated to verify the applicability and effectiveness of the established damage localization method.The results show that during the approximate damage location is identified using the varying ratio of mode curvature method,the identification accuracy can reach 100% when the damage occur only in truss,and 93.7% when the damage occur in multiple trusses simultaneously.While the accurate damage location of the damaged truss is identified using the artificial neural network technology,the accuracy for damage location can reach 97.0% in the case of without test noise,it indicates that the influence of test noise on the accuracy of damage location is significant.更多还原