An inversion of surface waves dispersion curves for shear-wave velocity profiles is a key step of the method of multichannel analysis of seismic surface waves and is an ill-posed problem. It leads to the necessity of selecting the most likely of the restored velocity model parameters. Evaluation of the likelihood of these restored velocity models is a promising goal. Our research is focused on the inversion of surface wave dispersion curve with an artificial neural network (ANN) for 1D velocity model parameters and its estimating uncertainty ranges using an adapted Markov chain Monte Carlo algorithm. The main advantage of using the ANN for the inversion of dispersion curves is the high computational efficiency of a large amount of data inversion, which is necessary for the building of a statistically significant sample. Field data processing results indicate the promising prospects of the proposed uncertainty estimation algorithm.
