基于神经网络的大跨度空间钢结构应力实测缺失数据修复方法研究

Missing data reconstruction in stress monitoring of steel spatial structures using neural network techniques

结构健康监测技术正被广泛应用于大跨度空间钢结构.而在长期监测的实践中,由于设备故障、信号中断等各种原因,数据丢失问题不可避免.数据的缺失可能对进一步结构分析和安全评估的准确性造成影响.根据某体育场钢结构屋盖长期实测应力和温度数据的特点,基于神经网络建立测点应力数据间、应力与温度数据间的相关关系模型,从而达到修复缺失数据的目的,并对该方法的适用性进行了研究.

Structural health monitoring systems have been widely applied in large-span spatial structures.In long-term monitoring practice,data missing problem is inevitable due to sensor fault,communication failure,energy interruption and other factors.It may bring about incorrectness in the subsequent data analysis and structural safety assessment.Referring to the characteristics of the stress and temperature data,this paper proposed a method to recover the missing data by establishing correlation between stress and temparature data based on neural network.The applicability of this method was also discussed.