基于FSOM神经网络的桥梁监测数据缺失重构算法

Reconstruction algorithm of bridge monitoring data missing based on FSOM neural network

在桥梁监测数据缺失重构过程中受到不同缺失数据之间相关性约束,容易出现删除或被忽略问题,导致数据缺失重构效果变差,针对该问题,设计基于FSOM神经网络的桥梁监测数据缺失重构算法。利用FSOM神经网络聚类分层处理桥梁监测数据。采用时空相关性分析方法,结合协方差评定两个缺失数据之间的线性相关性。计算时空相关数据与缺失数据之间相关权重,引入支持向量回归算法构建重构决策函数,通过FSOM神经网络更新权重,输出桥梁监测数据缺失重构结果。分析实验结果可知,该算法与实际重构数据存在最大为1 bit的误差,能够重构不同时间段内缺失桥梁的监测数据,缺失数据重构效果好。

In the process of bridge monitoring data loss reconstruction,it is constrained by the correlation between different missing data,which is easy to be deleted or ignored,resulting in poor data loss reconstruction effect.To solve this problem,a bridge monitoring data loss reconstruction algorithm based on FSOM neural network is designed.FSOM neural network is used to process bridge monitoring data by clustering and layering.The linear correlation between the two missing data was evaluated by using spatiotemporal correlation analysis and covariance.Calculate the correlation weight between spatiotemporal correlation data and missing data,introduce support vector regression algorithm to construct reconstruction decision function,update the weight through FSOM neural network,and output the reconstruction results of missing bridge monitoring data.The analysis of the experimental results shows that the maximum error between the algorithm and the actual reconstruction data is 1 bit.The analysis of the experimental results shows that the maximum error between the algorithm and the actual reconstruction data is 1 bit.The algorithm can reconstruct the missing bridge monitoring data in different time periods,and the reconstruction effect of data missing is good.