Acoustic impedance inversion base on dual learning

Abstract

Acoustic impedance inversion is an effective way to predict oil and gas reservoirs, but the acoustic impedance inversion based on traditional convolution neural network is limited by the number of labeled data. In order to solve this problem of insufficient labeled data in acoustic impedance inversion, we proposed an acoustic impedance inversion method base on dual learning. This method can be used for impedance inversion under the constraint of the small number of labeled data, and can obtain accurate inversion results.