Inversion of Surface Waves Dispersion Curves Using Artificial Neural Network [Электронный ресурс]
(ИНГГ СО РАН)
(ИНГГ СО РАН)
дата публикации: 2021
The present study is concerned with the inversion of the surface waves dispersion curves for the shear-wave velocity profiles. That is a key step of a multichannel analysis of surface waves, which is one of the popular near surface seismic exploration methods. The last modifications of the local search methods (e.g. Occam's inversion) provide the smooth subsurface model that does not allow to restore the layer borders. The global search methods are the computationally expensive. We propose to use fully-connected artificial neural network (ANN) for reconstructing the values of the velocities and thicknesses for each layer of the model. The synthetic tests illustrate that the inversion accuracy of ANN is equal to global search, while ANN is more robust in the case of noisy data. The field data processing showed the better coherence of ANN results with the borehole data than of considered global search methods. The advantage of the ANN is that it provides the dispersion curves inversion with the accuracy similar to the global search methods, while requires much less computational resources (time) for large amount of surface wave data.
первоисточник: 82nd EAGE Annual Conference and Exhibition. European Association of Geoscientists Engineers (Amsterdam, The Netherlands, 8-11 June 2020)