| In response to the demand for ground motion fields and ground motion time series in post-earthquake emergency response to destructive earthquakes,this paper develops a method for predicting the Fourier amplitude spectrum of ground motion based on the observational data of existing station networks,and realises the prediction of ground motion time series for regions without observational data after earthquakes.The following are the details of this paper:(1)A model for predicting the amplitude spectrum of ground motion by applying deep neural networks was constructed.The model takes the station spacing and the ground vibration amplitude spectrum of each observation station as the input layer and the ground vibration amplitude spectrum of the target station as the output,where the hidden layer is set to a 5-layer structure.The model is divided into a training set and a test set in the ratio of 8:2.To improve the efficiency and generalization capability of the model,the best model and the optimal hyperparameters are selected by the cross-validation set method.The reliability and rationality of the model were verified,and the results showed that the predicted amplitude spectra were in good agreement with the measured records in terms of amplitude magnitude,remarkable frequency and variation trend.(2)A ground motion simulation method applying strong vibration observation records is developed.The method uses a deep neural network approach to predict the ground motion amplitude spectrum and an equivalent group velocity phase spectrum based model to predict the ground motion timescale at unobserved record sites.The results based on this paper’s method,the kriging interpolation method and the inverse distance weight interpolation method are analysed for two perspectives:single target stations for a single seismic event and multiple target stations for multiple seismic events.The results show that the ground motion peaks,amplitude spectra and response spectra of the ground motion time courses obtained by the method in this paper are in good agreement with the observed records.(3)The method of this paper was applied to three Chinese earthquake instances to construct the ground motion waveform fields,and the obtained waveform fields were applied to the generation of instrumental seismic intensity fields and compared with the seismic intensity maps released by the China Seismological Bureau,further verifying the reliability and reasonableness of the method of this paper. |
