Том: 14106 , Год издания: 2023

Издатель: Springer International Publishing , Место издания: Berlin

This paper presents a novel approach to solving the inverse dynamic seismic problem in seismic monitoring for the viscoelastic medium. The proposed method offers a cost-effective alternative to Full Waveform Inversion by using a deep convolutional neural network Unet-type architecture with residual blocks to approximate an inverse problem operator that translates the change in seismic data into the change in velocity models. The operability of the proposed approach is demonstrated through a model example under the assumption that the distribution of the velocity model is known at the initial moment. Furthermore, the results of neural network prediction on a realistic sample with Gullfaks deposit indicate the practical applications of the proposed approach in seismic monitoring. The proposed approach shows significant potential for advancing the state-of-the-art in solving the inverse dynamic seismic problem for the viscoelastic medium, with potential implications for improving seismic monitoring techniques in industry and academia.