깊이별 분리 합성곱 신경망을 이용한 탄성파 속도 모델 구축
Seismic Velocity Modeling Building Using Depthwise Separable Convolutional Neural Network
0
조회
0
다운로드
1
인용
0
참고문헌
키워드
초록
정확한 속도 모델 구축은 석유 가스 탐사를 위한 탄성파 자료처리에서 가장 중요한 작업 중 하나이다. 최근 심층 신경망 기법이 지구물리학 분야에서 큰 인기를 얻으면서, 일반 합성곱 신경망을 이용하여 속도 모델을 예측하는 연구들이 출판되고 있다. 본 연구에서는 속도 모델 구축을 위해 깊이별분리 합성곱 층과 인코더-디코더 구조를 가진 신경망을 제안하였다. 이 신경망은 지도 학습 방식으로 훈련되며 시간 영역 파동장으로부터 P파 속도 모델을 예측한다. 이 신경망 구조의 핵심 부분인깊이별 분리 합성곱은 입력 채널 별로 독립적으로 공간 방향의 합성곱을 수행한다. 깊이별 분리 합성곱을 이용하면 일반 합성곱에 비해 모델 매개변수 수를 크게 줄이면서 신경망의 성능을 개선할수 있다. 훈련을 위해 생성한 합성 속도 모델들은 습곡, 단층 및 암염 구조 등 다양한 지질학적 특징을 포함한다. 깊이별 분리 합성곱을 이용한 신경망과 일반 합성곱을 이용한 신경망을 동일한 하이퍼파라미터 및 훈련 조건으로 비교 분석하였다. 실험 결과 깊이별 분리 합성곱 신경망이 일반 합성곱 신경망보다 속도 모델 구축 문제에서 더 효율적인 것으로 나타났다.
Abstract
The construction of an accurate velocity model is one of the most important tasks in seismic data processing for hydrocarbon exploration. Because deep neural networks have garnered significant attention in the field of geophysics recently, studies have been performed to predict velocity models using regular convolutional neural networks. Herein, we propose a neural network with depthwise separable convolutional layers and an encoder–decoder structure to construct a velocity model. This network is trained using a supervised learning approach, and we predict P-wave velocity models from time-domain wavefields. In this network structure, depthwise separable convolutions perform spatial-oriented convolutions independently for each input channel. These depthwise separable convolutions can improve network performance while significantly reducing the number of model parameters as compared with regular convolutions. Synthetic velocity models generated for training contain various geological features, including folds, faults, and salt-dome structures. We compare a network with depthwise separable convolutions and a network with regular convolutions based on the same training conditions and hyperparameters. Experiments demonstrate that the network with depthwise separable convolutions is more efficient than the network with regular convolutions for constructing a seismic velocity model.
참고문헌 (26)
- Baysal, E., "Reverse time migration", Geophysics, 48, pp.1514~1524 (1983)
- Bunks, C., "Multiscale seismic waveform inversion", Geophysics, 60, pp.1457~1473 (1995)
- Chollet, F., "Xception: Deep learning with depthwise separable convolutions", Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp.1251~1258 (2017)
- Chollet, F., "Deep Learning with Python", pp.261~263 (2017), Manning
- Gauthier, O., "Two-dimensional nonlinear inversion of seismic waveforms: Numerical results", Geophysics, 51, pp.1387~1403 (1986)
- Glorot, X., "Understanding the difficulty of training deep feedforward neural networks", Proceedings of Machine Learning Research, 9, pp.249~256 (2010)
- Hornik, K., "Universal approximation of an unknown mapping and its derivatives using multilayer feedforward networks", Neural Networks, 3, pp.551~560 (1990)
- Kaiser, L., "Depthwise separable convolutions for neural machine translation", Computation and Language, arXiv:1706.03059v2, ~ (2017)
- Keys, R. G., "Absorbing boundary conditions for acoustic media", Geophysics, 50, pp.892~902 (1985)
- Kingma, D.P., "Adam: A method for stochastic optimization", 3rd International Conference on Learning Representations, pp.1~15 (2015)
- Li, S., "Deep-learning inversion of seismic data", IEEE Transactions on Geoscience and Remote Sensing, 58, pp.2135~2149 (2020)
- Liu, B., "Deep-learning seismic full-waveform inversion for realistic structural models", Geophysics, 86, pp.R31~R44 (2021)
- Loffe, S., "Batch normalization: Accelerating deep network training by reducing internal covariate shift", Proceedings of the 32nd International Conference on Machine Learning, 37, pp.448~456 (2015)
- Mora, P., "Nonlinear two-dimensional elastic inversion of multioffset seismic data", Geophysics, 52, pp.1211~1228 (1987)
- Odena, A., "Deconvolution and checkerboard artifacts", Distill, 1(10), pp.e3~ (2016)
- Pratt, R.G., "Inverse theory applied to multi-source cross-hole tomography. Part 1: acoustic wave-equation method", Geophysical Prospecting, 38, pp.287~310 (1990)
- Pratt, R. G., "Gauss-Newton and full Newton methods in frequency-space seismic waveform inversion", Geophysical Journal International, 133, pp.341~362 (1998)
- Shin, C., "Waveform inversion in the Laplace domain", Geophysical Journal International, 173, pp.922~931 (2008)
- Tarantola, A., "Inversion of seismic reflection data in the acoustic approximation", Geophysics, 49, pp.1259~1266 (1984)
- Virieux, J., "An overview of full-waveform inversion in exploration geophysics", Geophysics, 74, pp.WCC1~WCC26 (2009)
- Wang, W., "Velocity model building in a crosswell acquisition geometry with image-trained artificial neural networks", Geophysics, 85, pp.U31~U46 (2020)
- Wang, Z., "Image quality assessment: from error visibility to structural similarity", IEEE Transactions on Image Processing, 13, pp.600~612 (2004)
- Wu, Y., "InversionNet: A real-time and accurate full waveform inversion with CNNs and continuous CRFs", Signal Processing, arXiv:1811.07875v2, ~ (2019)
- Yang, F., "Deep-learning inversion: A next-generation seismic velocity model building method", Geophysics, 84, pp.R583~R599 (2019)
- Zhang, J., "Nonlinear refraction and reflection travel time tomography", Journal of Geophysical Research Solid Earth, 103, pp.29743~29757 (1998)
- Zhang, Z., "Data-driven seismic waveform inversion: A study on the robustness and generalization", IEEE Transactions on Geoscience and Remote Sensing, 58, pp.6900~6913 (2020)