基于BP神经网络与BOLL通道的结构监测数据识别修复

Research on Monitoring Data Repair Based on BP Neural Network and BOLL Channel Method

通过BOLL通道法巧妙地引入第三方数据,与原始数据互相验证分析,以此构建上下轨线,求取偏离值得到改进型BOLL通道法进行相关的非均匀温度场异常温度识别过滤,再结合IDW算法验证对非均匀温度场缺失温度数据的补全效果,并且基于BP神经网络模型提出两种应力修复方法,对应力监测数据进行修复,然后对修复效果进行研究,验证方法的实用价值.研究结果表明:改进型BOLL通道法在处理非均匀温度场方面具有独特的区域自适应性,在去除异常温度数据的同时较好保留了非均匀温度场研究的突变温度细节;IDW算法综合考虑了测点温度的非均匀性,对于温度场缺失温度数据的补全具有较好效果;同时基于BP神经网络构建温度相关性、多测点相关性的应力修复方法,对非均匀温度场监测中的应力数据进行修复,其结果具有相当的可靠度.

As the monitoring technology is widely used in large span space structures,the lack of monitoring data caused by equipment failure and signal interruption is gradually highlighted,so it is urgent to corresponding data processing repair methods to ensure the integrity of monitoring data for subsequent scientific research,in the monitoring data of heterogeneous temperature field,but traditional temperature identification method is too rough,temperature repair and related stress repair methods are not studied enough.Therefore,BOLL channel method was adopted to introduce the third party data and raw data to build upper and lower rail lines,seeking the deviation to improve the BOLL channel method for heterogeneous temperature field anomalous temperature identification and filtering,in combined with IDW algorithm to verify the completion effect of the heterogeneous temperature field missing temperature;and based on machine learning related neural network,two stress repair methods were put forward,the stress monitoring data were repaired,and the repair effect was studied,which verifies practical value of the method.The experimental results showed that the proposed BOLL channel method has unique regional adaptability in dealing with the heterogeneous temperature field,which better retains the details of the mutation temperature studied while removing the abnormal temperature value.The IDW algorithm comprehensively considers the inhomogeneity of the measurement point temperature of the temperature field.The stress repair method based on temperature and stress correlation and multi-measurement point correlation based on BP neural network is quite reliable in stress data repair in heterogeneous temperature field monitoring,which deserves further study in this respect.