《计算机应用》唯一官方网站 ›› 2024, Vol. 44 ›› Issue (7): 2310-2318.DOI: 10.11772/j.issn.1001-9081.2023070915
• 前沿与综合应用 • 上一篇
收稿日期:
2023-07-11
修回日期:
2023-10-02
接受日期:
2023-10-13
发布日期:
2023-10-26
出版日期:
2024-07-10
通讯作者:
张志禹
作者简介:
赵亦群(1998—),男,河南安阳人,硕士研究生,主要研究方向:深度学习、地震信号处理;基金资助:
Yiqun ZHAO, Zhiyu ZHANG(), Xue DONG
Received:
2023-07-11
Revised:
2023-10-02
Accepted:
2023-10-13
Online:
2023-10-26
Published:
2024-07-10
Contact:
Zhiyu ZHANG
About author:
ZHAO Yiqun, born in 1998, M. S. candidate. His research interests include deep learning, seismic signal processing.Supported by:
摘要:
针对目前利用物理信息神经网络计算旅行时只是应用在各向同性介质上、在远离震源时误差较大和效率低等问题,而有限差分法、试射法和弯曲法等方法在多震源、高密度网格上计算成本高等问题,提出一种密集残差物理信息神经网络计算各向异性介质旅行时的方法。首先推导了各向异性因式分解后的程函方程作为损失函数项;其次引入局部自适应反正切函数为激活函数和L-BFGS-B(Limited-memory Broyden-Fletcher-Goldfarb-Shanno-B)作为优化器;最后在网络中采用分段式训练的方式,先训练深层密集残差网络,然后冻结其参数,再训练具有物理意义的浅层密集残差网络,从而评估网络得到旅行时。实验结果表明,所提方法在均匀速度模型下的旅行时最大绝对误差达到了0.015 8 μs,其他速度模型下平均绝对误差平均下降了两个数量级,在效率方面也平均提高了1倍,明显优于快速扫描法。
中图分类号:
赵亦群, 张志禹, 董雪. 基于密集残差物理信息神经网络的各向异性旅行时计算方法[J]. 计算机应用, 2024, 44(7): 2310-2318.
Yiqun ZHAO, Zhiyu ZHANG, Xue DONG. Anisotropic travel time computation method based on dense residual connection physical information neural networks[J]. Journal of Computer Applications, 2024, 44(7): 2310-2318.
模型 | 模型大小/(km×km) | 震源/km | 速度/(km·s-1) | 网格大小 | 优化器激活函数 | 迭代次数 |
---|---|---|---|---|---|---|
均匀速度模型 | 2×2 | (0.5,1.0) | 2.0 | 101×101 | Adam/l-atan | 2 000 |
速度渐变模型 | 2×2 | (1,0.5.0) | 2.0~3.2 | 101×101 | L-BFGS-B/l-atan | 2 000 |
Marmousi模型 | 2×2 | (1.0,1.0) | 1.5~4.5 | 101×101 | L-BFGS-B/l-atan | 3 000 |
TTI模型 | 2×2 | (1.5,1.0) | 2.0~6.0 | 101×101 | L-BFGS-B/l-atan | 2 000 |
VTI模型 | 10×10 | (3.0,0.0) | 1.5~4.5 | 201×201 | L-BFGS-B/l-atan | 2 500 |
表1 各模型参数
Tab. 1 Parameters for various models
模型 | 模型大小/(km×km) | 震源/km | 速度/(km·s-1) | 网格大小 | 优化器激活函数 | 迭代次数 |
---|---|---|---|---|---|---|
均匀速度模型 | 2×2 | (0.5,1.0) | 2.0 | 101×101 | Adam/l-atan | 2 000 |
速度渐变模型 | 2×2 | (1,0.5.0) | 2.0~3.2 | 101×101 | L-BFGS-B/l-atan | 2 000 |
Marmousi模型 | 2×2 | (1.0,1.0) | 1.5~4.5 | 101×101 | L-BFGS-B/l-atan | 3 000 |
TTI模型 | 2×2 | (1.5,1.0) | 2.0~6.0 | 101×101 | L-BFGS-B/l-atan | 2 000 |
VTI模型 | 10×10 | (3.0,0.0) | 1.5~4.5 | 201×201 | L-BFGS-B/l-atan | 2 500 |
网络(训练点) | 平均绝对误差/s | 平均相对误差/% | 最大绝对误差/s | 训练时间/min |
---|---|---|---|---|
DRC-PINN(25%) | 3.52E-07 | 6.48E-5 | 7.69E-07 | 3.65 |
PINN(25%) | 1.15E-05 | 8.49E-4 | 4.58E-05 | 8.31 |
DRC-PINN(50%) | 9.35E-09 | 1.06E-6 | 2.69E-08 | 5.87 |
PINN(50%) | 7.27E-06 | 3.16E-5 | 1.56E-05 | 14.38 |
DRC-PINN(75%) | 8.61E-09 | 9.83E-7 | 1.58E-08 | 9.39 |
表2 PINN改进前后不同训练点数性能对比
Tab. 2 Performance comparison before and after improvement of PINN with different training points
网络(训练点) | 平均绝对误差/s | 平均相对误差/% | 最大绝对误差/s | 训练时间/min |
---|---|---|---|---|
DRC-PINN(25%) | 3.52E-07 | 6.48E-5 | 7.69E-07 | 3.65 |
PINN(25%) | 1.15E-05 | 8.49E-4 | 4.58E-05 | 8.31 |
DRC-PINN(50%) | 9.35E-09 | 1.06E-6 | 2.69E-08 | 5.87 |
PINN(50%) | 7.27E-06 | 3.16E-5 | 1.56E-05 | 14.38 |
DRC-PINN(75%) | 8.61E-09 | 9.83E-7 | 1.58E-08 | 9.39 |
1 | ALEXANDROV D, WAHEED U, EISNER L. Microseismic location error due to eikonal traveltime calculation [J]. Applied Sciences, 2021, 11(3): 982. |
2 | ZHOU J, SHEN X, QIU Y, et al. Microseismic location in hardrock metal mines by machine learning models based on hyperparameter optimization using Bayesian optimizer [J]. Rock Mechanics and Rock Engineering, 2023, 56: 8771-8788. |
3 | CLAUSOLLES N, COLLON P, IRAKARAMA M, et al. Stochastic velocity modeling for assessment of imaging uncertainty during seismic migration: application to salt bodies [J]. Interpretation, 2023, 11(2): T361-T378. |
4 | ZHANG X, LI R, CUI L, et al. Least squares reverse time migration imaging with illumination preconditioned based on improved PRP conjugate gradients [J]. Scientific Reports, 2023, 13: 13623. |
5 | 张晓丹,张志禹.地震波逆时偏移成像与哈夫曼编码的应用研究[J].计算机工程与应用, 2013, 49(10): 22-24, 75. |
ZHANG X D, ZHANG Z Y. Research on seismic wave reverse time migration imaging and Huffman coding [J]. Computer Engineering and Applications, 2013, 49(10): 22-24, 75. | |
6 | 任志明,戴雪,包乾宗. VSP直达波和反射波波动方程走时联合反演[J].地球物理学报, 2023, 66(9): 3816-3827. |
REN Z M, DAI X, BAO Q Z. Joint wave-equation traveltime inversion of direct and reflected waves for VSP data [J]. Chinese Journal of Geophysics, 2023, 66(9): 3816-3827. | |
7 | ZHANG Y, BACHMAYR M, DE SIENA L. Total variation regularization for first-break travel time inversion [C]// Proceedings of the 84th EAGE Annual Conference & Exhibition. [S.l.]: European Association of Geoscientists & Engineers, 2023: 1-5. |
8 | FAN Y, YING L. Solving traveltime tomography with deep learning [J]. Communications in Mathematics and Statistics, 2023, 11: 3-19. |
9 | QIAN J, SYMES W W. Finite-difference quasi-P traveltimes for anisotropic media [J]. Geophysics, 2002, 67(1): 147-155. |
10 | QIAN J, SYMES W W. Paraxial eikonal solvers for anisotropic quasi-P travel times [J]. Journal of Computational Physics, 2001, 173(1): 256-278. |
11 | FOMEL S, LUO S, ZHAO H. Fast sweeping method for the factored eikonal equation [J]. Journal of Computational Physics, 2009, 228(17): 6440-6455. |
12 | LUO S, QIAN J. Fast sweeping methods for factored anisotropic eikonal equations: multiplicative and additive factors [J]. Journal of Scientific Computing, 2012, 52: 360-382. |
13 | RAISSI M, PERDIKARIS P, KARNIADAKIS G E. Physics-informed neural networks: a deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations [J]. Journal of Computational Physics, 2019, 378: 686-707. |
14 | GUO Y, CAO X, LIU B, et al. Solving partial differential equations using deep learning and physical constraints [J]. Applied Sciences, 2020, 10(17): 5917. |
15 | MOSELEY B, NISSEN-MEYER T, MARKHAM A. Deep learning for fast simulation of seismic waves in complex media [J]. Solid Earth, 2020, 11(4): 1527-1549. |
16 | WAHEED U BIN, HAGHIGHAT E, ALKHALIFAH T, et al. PINNeik: eikonal solution using physics-informed neural networks [J]. Computers & Geosciences, 2021, 155: 104833. |
17 | SONG C, ALKHALIFAH T, WAHEED U BIN. Solving the frequency-domain acoustic VTI wave equation using physics-informed neural networks [J]. Geophysical Journal International, 2021, 225(2): 846-859. |
18 | 王海涛,王新超,朱颖.微调残差物理神经网络建模和参数整定方法[J].计算机应用, 2022, 42(S2): 175-179. |
WANG H T, WANG X C, ZHU Y. Fine-tuning residual physics-informed neural network for physical modeling and parameter tuning [J]. Journal of Computer Applications, 2022, 42(S2): 175-179. | |
19 | ZHANG Y, TIAN Y, KONG Y, et al. Residual dense network for image super-resolution [C]// Proceedings of the 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2018: 2472-2481. |
20 | YANG L, LIU J, YAN R, et al. Spline adaptive filter with arctangent-momentum strategy for nonlinear system identification [J]. Signal Processing, 2019, 164: 99-109. |
21 | HRGA T, POVH J. Solving SDP relaxations of Max-Cut problem with large number of hypermetric inequalities by L-BFGS-B [J]. Optimization Letters, 2023, 17: 1201-1213. |
22 | HE K, ZHANG X, REN S, et al. Deep residual learning for image recognition [C]// Proceedings of the 2016 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2016: 770-778. |
23 | 廖万平,王兴建,李卿武,等.基于Marmousi模型的CFS-NPML分数阶粘声波正演模拟[J].物探化探计算技术, 2021, 43(6): 683-689. |
LIAO W P, WANG X J, LI Q W, et al. CFS-NPML fractional order based on Marmousi model Viscoacoustic forward modeling [J]. Computing Techniques for Geophysical and Geochemical Exploration, 2021, 43(6): 683-689. | |
24 | FEHLER M, KELIHER P J. SEAM Phase 1: Challenges of Subsalt Imaging in Tertiary Basins, with Emphasis on Deepwater Gulf of Mexico [M]. Tulsa, Oklahoma: Society of Exploration Geophysicists, 2011. |
[1] | 李顺勇, 李师毅, 胥瑞, 赵兴旺. 基于自注意力融合的不完整多视图聚类算法[J]. 《计算机应用》唯一官方网站, 2024, 44(9): 2696-2703. |
[2] | 潘烨新, 杨哲. 基于多级特征双向融合的小目标检测优化模型[J]. 《计算机应用》唯一官方网站, 2024, 44(9): 2871-2877. |
[3] | 秦璟, 秦志光, 李发礼, 彭悦恒. 基于概率稀疏自注意力神经网络的重性抑郁疾患诊断[J]. 《计算机应用》唯一官方网站, 2024, 44(9): 2970-2974. |
[4] | 王熙源, 张战成, 徐少康, 张宝成, 罗晓清, 胡伏原. 面向手术导航3D/2D配准的无监督跨域迁移网络[J]. 《计算机应用》唯一官方网站, 2024, 44(9): 2911-2918. |
[5] | 黄云川, 江永全, 黄骏涛, 杨燕. 基于元图同构网络的分子毒性预测[J]. 《计算机应用》唯一官方网站, 2024, 44(9): 2964-2969. |
[6] | 刘禹含, 吉根林, 张红苹. 基于骨架图与混合注意力的视频行人异常检测方法[J]. 《计算机应用》唯一官方网站, 2024, 44(8): 2551-2557. |
[7] | 顾焰杰, 张英俊, 刘晓倩, 周围, 孙威. 基于时空多图融合的交通流量预测[J]. 《计算机应用》唯一官方网站, 2024, 44(8): 2618-2625. |
[8] | 石乾宏, 杨燕, 江永全, 欧阳小草, 范武波, 陈强, 姜涛, 李媛. 面向空气质量预测的多粒度突变拟合网络[J]. 《计算机应用》唯一官方网站, 2024, 44(8): 2643-2650. |
[9] | 徐松, 张文博, 王一帆. 基于时空信息的轻量视频显著性目标检测网络[J]. 《计算机应用》唯一官方网站, 2024, 44(7): 2192-2199. |
[10] | 孙逊, 冯睿锋, 陈彦如. 基于深度与实例分割融合的单目3D目标检测方法[J]. 《计算机应用》唯一官方网站, 2024, 44(7): 2208-2215. |
[11] | 吴筝, 程志友, 汪真天, 汪传建, 王胜, 许辉. 基于深度学习的患者麻醉复苏过程中的头部运动幅度分类方法[J]. 《计算机应用》唯一官方网站, 2024, 44(7): 2258-2263. |
[12] | 李欢欢, 黄添强, 丁雪梅, 罗海峰, 黄丽清. 基于多尺度时空图卷积网络的交通出行需求预测[J]. 《计算机应用》唯一官方网站, 2024, 44(7): 2065-2072. |
[13] | 张郅, 李欣, 叶乃夫, 胡凯茜. 基于暗知识保护的模型窃取防御技术DKP[J]. 《计算机应用》唯一官方网站, 2024, 44(7): 2080-2086. |
[14] | 赵雅娟, 孟繁军, 徐行健. 在线教育学习者知识追踪综述[J]. 《计算机应用》唯一官方网站, 2024, 44(6): 1683-1698. |
[15] | 刘源泂, 何茂征, 黄益斌, 钱程. 基于ResNet50和改进注意力机制的船舶识别模型[J]. 《计算机应用》唯一官方网站, 2024, 44(6): 1935-1941. |
阅读次数 | ||||||
全文 |
|
|||||
摘要 |
|
|||||