摘要
为解决桥梁结构健康监测中长期累积的海量监测数据处理问题,采用数据挖掘中的时间序列分析方法,利用自回归移动平均(ARMA)技术对桥梁历史静态监测量进行分析,直接得出关于其过去行为的有关结论,进而推断其未来发展趋势。与基于因果关系的结构式模型预测法不同,ARMA无需明确模型结构或边界条件,而是直接从统计学的角度预测变量未来发展情况。采用该技术对西安白蛇峪大桥应变监测数据进行分析,试验结果显示出较高的预测精度;采用该技术对重庆偏岩子大桥的应变、挠度、裂缝、倾斜监测量进行预测,试验统计结果表明,ARMA单步预测误差小于10%的置信度在97%以上,在工程实际中具有可实用性,可为桥梁结构的安全预警提供重要参考依据。
In order to process the enormous amount of monitoring data accumulated during me- dium or long-term bridge structure health monitoring, the time series analysis methods for data i- dentification are adopted and the autoregressive moving average （ARMA） technique was used, to analyze the existing static monitoring data of the bridge, and based on which the related conclu- sions concerning the load bearing behavior of the bridge in the past can be drawn directly, and in turn the future condition of the bridge can be deduced. Different from the structural model predic- tion method based on causal relation, ARMA can, with out clear model structure or boundary con- dition, directly predict the future condition of variables from the view of statistics. The technique set f.orth has been used for the analysis of the stress monitoring data of Baisheyu Bridge in Xiran. The statistic data of the test show that the errors of ARMA single step prediction less than 10% bring about the confidence level above 97 %. The technique is practical in actual engineering projects, and can prorde reference data for security warning of bridge structure.
出处
《世界桥梁》
北大核心
2015年第3期44-48,共5页
World Bridges
基金
交通运输部西部交通建设科技项目(2013-364-740-600)
重庆市科委新产品创新青年科技人才培养项目(cstc2013kjrc-qnrc30001)
作者简介
作者筒介：唐浩（1983-），男，高级工程师，2004年毕业于西安交通大学机械工程及自动化专业，获学士学位，2010年毕业于西安交通大学机械制造及自动化专业，获博士学位（E—mail：tanghaol@cmhk．com）。
