基于数据关联度的监测数据预测方法

Prediction Method of Monitoring Data Based on Data Association Degree

为解决病态数据对损伤识别与状态评估的不利影响,提出基于数据关联度的监测数据预测方法。通过BP神经网络建立多通道数据间的关联度模型,以数据间的关联度对出现病态数据通道的数据进行预测和修正,并通过实测数据进行验证。研究表明考虑多通道数据间关联度的预测值比单通道的预测值具有更高精度,能够满足工程应用要求。

In order to solve the adverse influence caused by ill-conditioned data on damage identification and condition assessment,the monitoring data prediction method based on data correlation was proposed.The correlation degree between multi-channels data was established by BP neural network,and the ill-conditioned data channel was predicted and corrected by the correlation degree between the data,which was verified by the measured data.The results indicate that the prediction value achieved using BP neural network prediction method considering the correlation degree between multi-channel data has higher accuracy than the prediction value of single channel,which can meet the engineering requirement.