Acoustic velocity inversion based on convolutional autoencoder embedded with Fourier neural operator

Authors

LI Chen, School of Mathematics and Statistics, Xi’an Jiaotong University, Xi’an 710049, ChinaFollow
ZHAO Haixia, School of Mathematics and Statistics, Xi’an Jiaotong University, Xi’an 710049, ChinaFollow
BAI Zhaowei, School of Mathematics and Statistics, Xi’an Jiaotong University, Xi’an 710049, China
HAO Yufan, School of Mathematics and Statistics, Xi’an Jiaotong University, Xi’an 710049, China

Abstract

Background Seismic wave inversion serves as an effective method for obtaining the characteristics of structures, lithology, and physical properties of subsurface media using the arrival times, amplitude, and waveforms of seismic waves. Seismic inversion methods based on wave equations iteratively update model parameters using forward modeling. This generally involves extensive numerical simulations and optimization calculations, requiring large quantities of computational resources and time. In recent years, neural operators for deep learning, represented by the Fourier neural operator (FNO), have gained widespread attention. However, the original FNO structure fails to effectively learn the wavefield information with sharp changes in geological structures in the seismic wave inversion of complex media, leading to low accuracy of inversion results. Objective and Methods To enhance the accuracy and generalization performance of FNO in learning seismic wavefield information under complex geological models, this study developed a novel acoustic velocity inversion method—Convolutional autoencoder embedded with Fourier neural operator (CAE-FNO), which utilized an encoder for feature extraction and performed efficient training based on FNO to effectively capture the fine features of the seismic wavefield and improved prediction accuracy. During the network training, the CAE-FNO method progressively reduced the size of the Fourier mode, thus effectively reducing the number of network parameters while enhancing the generalization capability of the network. Results and Conclusions The numerical experiments on homogeneous, heterogeneous, layered, and Marmousi2 models demonstrate that the CAE-FNO method exhibited significantly higher inversion accuracy than FNO and its variants UFNO and UNO. The experiments on the homogeneous model revealed that the velocity inversion results of the CAE-FNO method had a relative error of 1.3% and those of UFNO, UNO, and FNO exhibited relative errors of 1.7%, 2.3%, and up to 10.1%, respectively. In the experiments on the heterogeneous model, CAE-FNO yielded accurate inversion results of geological structures and velocity change positions, whereas UFNO and UNO exhibited higher errors for zones with sharp velocity fluctuations. During the experiments on the layered model, CAE-FNO clearly distinguished minor velocity changes between layers, while FNO failed. For both smooth zones and zones with abrupt changes in the Marmousi2 model, CAE-FNO exhibited higher accuracy in capturing irregular interfaces with velocity changes than UFNO and UNO, while FNO failed to effectively handle the abrupt changes in velocity and detail changes in these zones. Therefore, the CAE-FNO method, demonstrating small loss functions and high accuracy, enjoys advantages in the inversion of complex media, providing a novel research philosophy for seismic inversion.

Keywords

seismic wave inversion, Fourier neural operator, convolutional autoencoder, deep learning, data-driven

DOI

10.12363/issn.1001-1986.24.02.0136

Recommended Citation

LI Chen, ZHAO Haixia, BAI Zhaowei, et al. (2024) "Acoustic velocity inversion based on convolutional autoencoder embedded with Fourier neural operator," Coal Geology & Exploration: Vol. 52: Iss. 11, Article 12.
DOI: 10.12363/issn.1001-1986.24.02.0136
Available at: https://cge.researchcommons.org/journal/vol52/iss11/12

Reference

[1] 杨培杰. 砂泥岩储层孔隙度和含水饱和度同步反演[J]. 地球物理学报，2018，61(2)：673−682.

YANG Peijie. Porosity and water saturation simultaneous inversion for sand-mudstone reservoir[J]. Chinese Journal of Geophysics，2018，61(2)：673−682.

[2] 韩宏伟，程远锋，张云银，等. 储层物性的地震预测技术综述[J]. 地球物理学进展，2021，36(2)：595−610.

HAN Hongwei，CHENG Yuanfeng，ZHANG Yunyin，et al. Review of seismic prediction of reservoir geophysical properties[J]. Progress in Geophysics，2021，36(2)：595−610.

[3] CHOPRA S，MARFURT K J. Seismic attributes—a historical perspective[J]. Geophysics，2005，70(5)：3SO−28SO.

[4] VIRIEUX J，OPERTO S. An overview of full-waveform inversion in exploration geophysics[J]. Geophysics，2009，74(6)：WCC127.

[5] ZELT C A，SMITH R B. Seismic traveltime inversion for 2-D crustal velocity structure[J]. Geophysical Journal International，1992，108(1)：16−34.

[6] RASHT-BEHESHT M，HUBER C，SHUKLA K，et al. Physics-informed neural networks (PINNs) for wave propagation and full waveform inversions[J]. Journal of Geophysical Research：Solid Earth，2022，127(5)：e2021JB023120.

[7] DWIVEDI V，PARASHAR N，SRINIVASAN B. Distributed physics informed neural network for data-efficient solution to partial differential equations[EB/OL]. 2019：1907.08967. https：//arxiv. org/abs/1907.08967v1

[8] MENG Xuhui，LI Zhen，ZHANG Dongkun，et al. PPINN：Parareal physics-informed neural network for time-dependent PDEs[J]. Computer Methods in Applied Mechanics and Engineering，2020，370：113250.

[9] WU Chenxi，ZHU Min，TAN Qinyang，et al. A comprehensive study of non-adaptive and residual-based adaptive sampling for physics-informed neural networks[J]. Computer Methods in Applied Mechanics and Engineering，2023，403：115671.

[10] CHEN Yanlai，KOOHY S. GPT-PINN. Generative pre-trained physics-informed neural networks toward non-intrusive meta-learning of parametric PDEs[J]. Finite Elements in Analysis and Design，2024，228：104047.

[11] SONG Chao，LIU Yang，ZHAO Pengfei，et al. Simulating multicomponent elastic seismic wavefield using deep learning[J]. IEEE Geoscience and Remote Sensing Letters，2023，20：3001105.

[12] WU Yanqi，AGHAMIRY H S，OPERTO S，et al. Helmholtz-equation solution in nonsmooth media by a physics-informed neural network incorporating quadratic terms and a perfectly matching layer condition[J]. Geophysics，2023，88(4)：T185−T202.

[13] SONG Chao，ALKHALIFAH T，WAHEED U B. Solving the frequency-domain acoustic VTI wave equation using physics-informed neural networks[J]. Geophysical Journal International，2021，225(2)：846−859.

[14] SONG Chao，ALKHALIFAH T A. Wavefield reconstruction inversion via physics-informed neural networks[J]. IEEE Transactions on Geoscience and Remote Sensing，2022，60：5908012.

[15] SHAKEEL M，NISHIDA K，ITOYAMA K，et al. 3D convolution recurrent neural networks for multi-label earthquake magnitude classification[J]. Applied Sciences，2022，12(4)：2195.

[16] BERHICH A，BELOUADHA F Z，KABBAJ M I. LSTM-based models for earthquake prediction[C]//Proceedings of the 3rd International Conference on Networking，Information Systems & Security. Marrakech Morocco. ACM，2020.

[17] ZHENG Qiqi，WEI Chao，YAN Xinfei，et al. Seismic elastic parameter inversion via a FCRN and GRU hybrid network with multi-task learning[J]. Applied Sciences，2023，13(18)：10519.

[18] MOUSAVI S M，ELLSWORTH W L，ZHU Weiqiang，et al. Earthquake transformer-an attentive deep-learning model for simultaneous earthquake detection and phase picking[J]. Nature Communications，2020，11(1)：3952.

[19] WANG Tiantong，TRUGMAN D，LIN Youzuo. SeismoGen：Seismic waveform synthesis using GAN with application to seismic data augmentation[J]. Journal of Geophysical Research：Solid Earth，2021，126(4)：e2020JB020077.

[20] LI Zongyi，KOVACHKI N，AZIZZADENESHELI K，et al. Neural operator：Graph kernel network for partial differential equations[J]. ArXiv e-Prints，2020：arXiv：2003.03485.

[21] LU Lu，JIN Pengzhan，PANG Guofei，et al. Learning nonlinear operators via DeepONet based on the universal approximation theorem of operators[J]. Nature Machine Intelligence，2021，3(3)：218−229.

[22] CAI Shengze，WANG Zhicheng，LU Lu，et al. DeepM\&Mnet：Inferring the electroconvection multiphysics fields based on operator approximation by neural networks[J]. Journal of Computational Physics，2021，436：110296.

[23] CAO Qianying，GOSWAMI S，KARNIADAKIS G. LNO：Laplace neural operator for solving differential equations[J]. Nat. Mac. Intell.，2023，6：631−640.

[24] LI Z，KOVACHKI N，AZIZZADENESHELI K，et al. Fourier neural operator for parametric partial differential equations[J]. arXiv preprint arXiv：201008895，2020.

[25] LI Bian，WANG Hanchen，FENG Shihang，et al. Solving seismic wave equations on variable velocity models with Fourier neural operator[J]. IEEE Transactions on Geoscience and Remote Sensing，2023，61：4508618.

[26] SONG Chao，WANG Yanghua. High-frequency wavefield extrapolation using the Fourier neural operator[J]. Journal of Geophysics and Engineering，2022，19(2)：269−282.

[27] 潘代英，何清龙. 基于改进的傅里叶神经算子数值求解频率域声波方程[J]. 运筹与模糊学，2023，13(6)：5978−5990.

PAN Daiying，HE Qinglong. Numerical solution of frequency-domain acoustic wave equation based on improved Fourier neural operator[J]. Operations Research and Fuzziology，2023，13(6)：5978−5990.

[28] YANG Yan，GAO A F，AZIZZADENESHELI K，et al. Rapid seismic waveform modeling and inversion with neural operators[J]. IEEE Transactions on Geoscience and Remote Sensing，2023，61：5906712.

[29] WEN Gege，LI Zongyi，AZIZZADENESHELI K，et al. U-FNO—An enhanced Fourier neural operator-based deep-learning model for multiphase flow[J]. Advances in Water Resources，2022，163：104180.

[30] RAHMAN M A，ROSS Z E，AZIZZADENESHELI K. U-NO：U-shaped neural operators[J]. ArXiv e-Prints，2022：arXiv：2204.11127.

[31] KONUK T，SHRAGGE J. Physics-guided deep learning using Fourier neural operators for solving the acoustic VTI wave equation[C]//82nd EAGE Annual Conference & Exhibition. Amsterdam，The Netherlands，European Association of Geoscientists & Engineers，2021.

[32] LEHMANN F，GATTI F，BERTIN M，et al. Fourier neural operator surrogate model to predict 3D seismic waves propagation[J]. ArXiv e-Prints，2023：arXiv：2304.10242.

[33] LEHMANN F，GATTI F，BERTIN M，et al. 3D elastic wave propagation with a Factorized Fourier Neural Operator (F-FNO)[J]. Computer Methods in Applied Mechanics and Engineering，2024，420：116718.

[34] LEHMANN F，GATTI F，BERTIN M，et al. Seismic hazard analysis with a Factorized Fourier Neural Operator (F-FNO) surrogate model enhanced by transfer learning[C]//proceedings of the NeurIPS 2023 AI for Science Workshop，F，2023.

[35] ZHANG Yijie，ZHU Xueyu，GAO Jinghuai. Parameter estimation of acoustic wave equations using hidden physics models[J]. IEEE Transactions on Geoscience and Remote Sensing，2020，58(7)：4629−4639.

[36] TARANTOLA A. Inversion of seismic reflection data in the acoustic approximation[J]. Geophysics，1984，49(8)：1259.