| 摘要:针对桥梁健康监测系统所采集的实时数据具有不完备性，严重影响桥梁的安全评估，提出基于支持向量回归（Support Vector Regression，SVR）算法的桥梁健康监测系统缺失数据实时在线预测方法。首先，分析实测数据具有时序、非线性和周期性等特点，利用变量的自相关和变量间的相关性重新构造支持向量回归模型的输入样本维数；在此基础上，根据样本在线更新的特点，采用序列最小优化算法对支持向量回归模型中的拉格朗日乘子进行实时更新，解决高精度在线填补的需求；最后，从实际问题出发，实现了支持向量回归模型的在线和离线自适应预测模式。通过对桥梁实测数据进行在线模式和离线模式预测对比，结果表明在线模式以样本更新的方式能够获得对将来值更高的预测精度。 |
| 关键词: 支持向量回归 在线预测 桥梁健康监测系统 缺失数据 序列最小优化算法 |
|
| Study on Online Missing Data Imputation in Bridge Health Monitoring System Based on Support Vector Regression |
|
|
| Abstract:The real-time datasets collected by the bridge health monitoring system are incomplete，and they seriously affect the safety assessment of the bridge. In this paper, the real-time online prediction model of missing data in bridge health monitoring system based on support vector regression (SVR) was proposed. The measured data has the characteristics of temporality, nonlinearity and periodicity, the input sample dimension in support vector regression model was firstly reconstructed according to the autocorrelation of variables and the correlation among variables. In order to improve the accuracy and efficiency of the prediction model for the missing datasets of bridge, the Lagrange multipliers in the support vector regression model are updated in real-time by using the sequential minimum optimization(SMO) algorithm according to the updated samples. Finally, the online adaptive prediction model of support vector regression model is put forward and realized in the practical problems. The measured missing data of bridge are predicted by online model and off-line model, the experimental results show that the online model can achieve higher prediction accuracy in the future by updating the samples. |
| Key words: Support Vector Regression online prediction bridge health monitoring system missing data Sequential minimal optimization(SMO) algorithm |