Rapid seismic domain transfer: Seismic velocity inversion and modeling using deep generative neural networks

Lukas Mosser,Wouter Kimman,Jesper Dramsch,Steve Purves,Alfredo De la Fuente,Graham Ganssle

DOI: https://doi.org/10.3997/2214-4609.201800734

2018-05-23

Abstract:Traditional physics-based approaches to infer sub-surface properties such as full-waveform inversion or reflectivity inversion are time-consuming and computationally expensive. We present a deep-learning technique that eliminates the need for these computationally complex methods by posing the problem as one of domain transfer. Our solution is based on a deep convolutional generative adversarial network and dramatically reduces computation time. Training based on two different types of synthetic data produced a neural network that generates realistic velocity models when applied to a real dataset. The system's ability to generalize means it is robust against the inherent occurrence of velocity errors and artifacts in both training and test datasets.

Geophysics,Computer Vision and Pattern Recognition

What problem does this paper attempt to address?

The problem that this paper attempts to solve is the large computational time - consuming and resource - consuming issues in traditional physics - based methods when inferring underground geological structures. Specifically, although traditional full - waveform inversion (FWI) or reflectivity inversion methods can achieve satisfactory results in practice, these methods require a large amount of computational resources and there is no guarantee of global convergence, which makes them inapplicable in cases of time constraints or limited computational resources.

To solve these problems, the author proposes a data - driven method using deep generative neural networks, modeling the seismic forward and inverse processes as domain - transformation problems. This method not only significantly reduces the computational time, but also, by training neural networks based on two different types of synthetic data, can generate realistic velocity models when applied to real - data sets. In addition, the generalization ability of this system makes it robust to the velocity errors and artifacts inherent in the training and test data sets.

Specifically in terms of technology, the paper adopts deep convolutional generative adversarial networks (DCGAN) and uses cycle - consistency loss to ensure that the generated data statistically represents and is restricted by the training data set, thus making this architecture suitable for seismic inversion tasks. In this way, the paper shows how to utilize the latest advances in the fields of deep learning and computer vision to improve the efficiency and accuracy of seismic data processing.