基于CEEMDAN-NAR-ARIMA组合模型的桥梁结构健康监测应变预测  

Strain Prediction of Bridge Structural Health Monitoring based on CEEMDAN-NAR-ARIMA Combination Model

作　　者：朱利明 卓静超 邢世玲 ZHU Li-ming;ZHUO Jing-chao;XING Shi-ling(College of Transportation Engineeriing,Nanjing Tech University,Nanjing 210000 China)

机构地区：[1]南京工业大学交通运输工程学院,南京210000

出　　处：《科学技术与工程》2020年第4期1639-1644,共6页Science Technology and Engineering

摘　　要：提出了一种基于CEEMDAN-NAR-ARIMA的组合模型用以预测桥梁结构健康监测(structural health monitoring,SHM)应变数据。针对经典时间序列理论对模态混叠的数据无法有效预测的问题,采用了带自适应噪声的完全集成经验模态分解(complete ensemble empirical mode decomposition with adaptive noise,CEEMDAN)方法对桥梁SHM应变监测数据进行了分解,并使用排列熵(permutation entropy,PE)方法将分解得到的大量分量按照随机程度进行分类重组形成数个新数据序列,最后结合非线性自回归(nonlinear auto regressive,NAR)动态神经网络和求和自回归移动平均(autoregressive integrated moving average,ARIMA)模型分别对每个新数据序列进行预测并叠加得到了最终的预测值。将该方法用于上海市某座斜拉桥的SHM应变数据预测,结果表明,该方法相比于经典时间序列理论提高了预测的准确性,具有良好的工程应用价值。A combination model based on CEEMDAN-NAR-ARIMA was proposed to predict the strain data of bridge structural health monitoring(SHM).Generally,the classical time series theory can not predict the modal aliasing data effectively.In order to tackle this problem,the complete ensemble empirical mode decomposition with adaptive noise(CEEMDAN)method was used to decompose the strain monitoring data of the bridge SHM.To deal with a large number of components after CEEMDAN,the permutation entropy(PE)method was used to form a number of new data sequences according to the random degree.Finally,combined with nonlinear auto regressive(NAR)dynamic neural network and autoregressive integrated moving average(ARIMA)model,each new data sequence was predicted and the final prediction was obtained.The method was used to predict the SHM strain data of a cable-stayed bridge in Shanghai.Results show that the proposed method is more accurate than the classical time series theory and has great engineering application value.

关 键 词：结构健康监测 时间序列分析 CEEMDAN 应变预测 

分 类 号：U446[建筑科学—桥梁与隧道工程]