Abstract: Temperature and strain measured data of the steel grid roof structure at Zhuhai port of Hong Kong-Zhu-hai-Macao bridge were taken as the actual engineering background，and the nonlinear relationship between tempera-ture and strain was simulated by back propagation (BP) neural network to realize temperature effect prediction of long-span spatial steel structure． The strain monitored of the structure under the action of temperature load and wind load was separated to realize the independent determination of the dominant working condition of the loads of the member． Finally，a neural network program based on database and Matlab interface was developed，and it was fur-ther embedded into building information model (BIM) software in the form of a plug-in． The information obtained from the analysis of the monitoring data was integrated into the BIM three-dimensional model，so as to guide the monitoring personnel to conduct maintenance on the structure． The neural network model constructed in this study can well fit the non-linear relationship between temperature and strain of long-span spatial steel structure． Based on the secondary development of BIM technology，the prediction of temperature effect and the identification of wind load were realized，and the research on the analysis of monitoring data was further developed，which can provide reference for similar projects．

Key words: steel grid roof structure, BP neural network, BIM technology, temperature effect prediction, identifi-cation of wind load