基于BP神经网络的大跨高空连廊应变监测数据恢复  

Strain Monitoring Data Restoring of Large-span Steel Skybridge Based on BP Neural Network

作　　者：赵昕[1] 贾京[2] 郑毅敏[1] 

机构地区：[1]同济大学建筑设计研究院,上海200092 [2]同济大学土木工程学院,上海200092

出　　处：《建筑科学与工程学报》2009年第1期101-106,共6页Journal of Architecture and Civil Engineering

基　　金：上海青年科技启明星计划项目(07QB14044)

摘　　要：为了解决大跨高空连廊吊装阶段性态监测中应变监测数据存在缺失的问题,利用BP神经网络进行数据恢复。首先基于相关性分析,选择与数据缺失监测点应变值相关性最强的5个监测点作为参考点;然后利用未缺失时间段内待恢复监测点和参考点的应变数据进行建模和检验,一半数据用来建立BP神经网络模型,一半数据用来进行模型的检验;最后利用建立的模型对缺失的数据进行恢复,得到了完整的应变监测数据。利用得到的恢复数据与参考点数据在缺失段内的相关系数对数据恢复的效果进行了评价。结果表明:该方法可有效地恢复缺失的应变监测数据。In order to solve the problem of strain monitoring data absence in performance monitoring of large-span steel skybridge, the data restoring was carried out by using BP neural networks. Firstly, based on correlation analysis, five reference points which were most correlative with the data missing points were obtained, then the data of both reference points and data missing point in the stage were simulated and verified. All data were separated into two subsets： one for training the BP neural network model, and the other for validating the model. The missing data were restored by using the trained BP neural networks. Finally, the integrated strain monitoring data were gained, and the correlation coefficients of the data missing point and each reference point were calculated; comparing the correlation coefficients, the performance of data restoring was evaluated. The results show that this method can restore the missing strain monitoring data effectively.

关 键 词：性态监测 相关性分析 BP神经网络 数据恢复 数据缺失 参考点 高空连廊 

分 类 号：TU317[建筑科学—结构工程]