Abstract: The risk management of the construction on foundation pit of the metro in the soft clay area is challenging, not only affected by the complex geological conditions and construction environment, but also disturbed by engineering construction and external factors, making it difficult to obtain the actual data for predicting the deformation of foundation pit, thus resulting in the risk in construction and its surrounding environment. Based on the data analysis and the dynamic prediction theory of the wavelet transform and Elman neural network, a new time series wavelet Elman prediction model (WEPM) of the foundation pit deformation was proposed and was applied in the foundation pit engineering of the Wuhan metro. The actual deformation data was obtained based on the wavelet decomposition de-noising process of the previous measured data. Then the wavelet intelligence model was used to realize a rolling prediction on the deformation of foundation pit. The analysis results show that the WEPM has a high prediction accuracy and a high generalization performance, which was suitable for the time series prediction on the deformation of foundation pit under different conditions. The proposed model provides basis for deformation prediction, construction safety and accidents prevention of foundation pit construction in Wuhan.

Key words: foundation pit of metro, deformation prediction, wavelet transform, time series analysis