实验室地震机器学习预测研究

地震是一种危害极强的自然灾害，给人类社会带来了巨大的生命财产损失。由于地震的检测难度大，地震成因复杂且难以获取足够的数据，数值模拟便成为了近年来广泛采用的研究地震预测的重要手段。如何通过合理建模提高地震预测能力，是地球物理学领域的关键科学问题。黏滑现象被认为是地震的重要机理。相比天然地震，在实验室环境中进行断层泥黏滑试验所产生的实验室地震事件具有数据量大、可控性好等优势，可为地震研究提供大量详尽数据。机器学习算法可从数据特征中挖掘变量间的复杂映射关系，为地震预测研究提供了新的思路。本文主要从数据特征筛选、融合模型构建和实验室多模态特征提取等角度开展了实验室地震人工智能预测研究，主要工作和研究成果如下：

(1) 利用耦合有限元-离散元法 (FDEM) 模拟实验室地震数据训练机器学习模型，提出了基于集成学习的关键空间孕震特征筛选方法。该方法利用各特征对预测目标的信息增益变化大小进行权重赋值和排序，从大量特征中筛选出真正对地震预测有重要影响的关键特征。结果表明，该方法可以得到信息冗余较小的特征子集，降低特征维度，提高预测效率。并且不同类型特征对预测结果的贡献存在显著差异，合理利用特征重要性差异能够对数据采集方法的改进提供重要依据。

(2) 采用机器学习可解释性模型对关键孕震特征进行可解释性分析。机器学习可解释性模型可以量化各特征对模型输出的贡献。通过计算每个特征在每个时间节点上对预测结果的贡献，定量解析不同类型特征对预测结果在时间和空间维度上的影响。结果显示，剪切方向上的位移特征的影响分布范围最大，其在黏滞加载阶段的贡献尤为明显；剪切方向上的速度特征则主要在滑动阶段产生影响。该研究解析各特征对预测结果的贡献，为特征选择和模型优化提供定量依据，指导了关键特征的识别和监测点布置的优势位置，亦为理解断层泥黏滑过程的孕震机制提供技术手段。

(3) 借助无监督学习的关键时间序列孕震特征提取方法，获得蕴含关键信息的数据聚类。该方法对较高特征维度的数据集进行先降维后聚类的特征工程，得到了蕴含差异化特征表达的数据聚类。通过比较各聚类数据集上的训练效果，可以识别对地震预测更关键的时间序列。结果表明，通过该方法获得的特征数据聚类可以显著提升后续模型的预测性能。与直接使用原始高维特征相比，该方法可以自动识别信息冗余较小的关键特征组合，降低时序特征输入训练时的特征长度。为实验室地震预测任务提供了更优化的特征空间，提升了机器学习的训练效率。

(4) 通过融合多个对不同序列提取能力有差异的深度学习模型，构建实验室地震预测融合机器学习模型。该模型整合提取时间序列短期规律、全局依赖和空间特征的模型架构，进一步提升对复杂时序特征的建模能力。结果表明，相比单一模型，融合模型可以显著提高对不同长度序列的预测准确性，能够在不同试验条件下的测试集中均取得较好的效果，体现强泛化能力。

(5) 构建融合多模态特征的实验室地震预测模型，利用注意力机制学习不同模态之间的内在联系，实现时序和视觉模态特征动态融合。融合多模态特征模型可以提取时序和视觉模态的特征表达，为机器学习模型理解预测任务提供多角度特征。多模态特征的综合利用可显著增强预测性能和泛化能力。结果表明，相比于基于单一时序模态特征训练的模型，多模态特征的加入可以显著提升模型对长序列的预测表现。除此之外，可以在保持大部分模型参数不变的情况下微调训练模型，实现融合多模态特征模型在不同试验条件下的高效迁移。

Earthquakes are highly destructive natural disasters that result in significant loss of life and property. Detecting earthquakes and understanding their complex causes pose challenges due to limited data availability. In recent years, numerical simulation has emerged as an important approach in earthquake prediction research. The critical scientific issue in the field of geophysics is how to improve earthquake prediction capabilities through effective modeling. The stick-slip phenomenon, a fundamental mechanism in earthquakes, is extensively studied through laboratory earthquake experiments. These experiments offer advantages such as abundant and detailed data, as well as precise control over variables. Machine learning algorithms have proven valuable in analyzing the intricate relationships between variables and providing new insights for earthquake prediction research. This paper systematically applies machine learning approaches to conduct a series of innovative studies on the laboratory earthquake problem from the perspective of features selection, fusion model construction and multimodalities features extraction. The major achievements are listed as follows:

(1) A feature selection method based on the ensemble model is proposed by training with laboratory earthquake friction dynamics data simulated by the combined finite-discrete element method (FDEM). This method can obtain feature subsets with less redundant information and reduce dimensions to improve prediction efficiency. Moreover, features have significant differences in their contributions to prediction results. Utilizing the differences in feature importance can provide important guidance for improving data acquisition methods.

(2) Machine learning interpretable models are adopted to conduct explainable analysis on selected features. The interpretable models can quantify the contribution of each feature to model predicted outputs. By calculating the contribution of each feature at every time step, the impacts of different types of features on earthquake prediction results in temporal and spatial dimensions are quantitatively analyzed. The analysis results show that the displacement features in the shearing direction have wide range distribution and their contribution is particularly significant during the stick stage. On the other hand, the velocity features in the shearing direction mainly affect the slip stage. This study analyzes the contribution of each feature to the prediction results, provides basis for feature selection and model optimization. It also guides the identification of key features and the optimal location of monitoring sensors setup, while providing technical means for understanding the seismic mechanism of fault stick-slip shearing.

(3) By applying the unsupervised learning method of key time series earthquake feature extraction, data clustering that contains important information can be obtained. This approach involves feature engineering on high-dimensional datasets by reducing dimensionality and then clustering the data, resulting in distinct feature expressions within the data clusters. By comparing the training effects on each clustered dataset, the time series features that are more critical to earthquake prediction can be identified. Results demonstrate that the data clustering obtained through this method significantly enhances the predictive performance of subsequent models. Compared with the model trained with original high-dimensional dataset, this approach can automatically recognize feature combinations with minimal redundancy, reducing the length of time series features during training. It provides a more optimized feature space for laboratory earthquake prediction tasks and improves the training efficiency of machine learning.

(4) A laboratory earthquake fusion model is constructed by integrating multiple deep learning models with different sequence extraction capabilities. The fusion model combines models that extract short-term memory, global dependencies, and spatial features of time series, further improving the modeling capability on complex time series features. Results show that the fusion model can significantly improve the prediction performance of sequences with different lengths compared to a single deep learning model. Moreover, the fusion model performs effectively on test sets under various experimental conditions, demonstrating its robust generalization capability.

(5) A multimodalities fusion model is built to learn the intrinsic connection between different modalities through attention mechanisms. The multimodalities fusion model can simultaneously learn the feature expressions of temporal and visual modalities. By extracting feature representations from both temporal and visual data, multimodalities fusion model provides multiple perspectives for machine learning models to understand prediction tasks. Utilizing these multimodal features significantly improves the predictive performance and generalization ability of the model. The results demonstrate that compared to models trained solely on temporal data, multimodal features greatly enhance the model's ability to predict long sequences. Additionally, it is possible to fine-tune the training model while keeping most of the model parameters unchanged, enabling efficient transfer of the multimodal feature fusion model across various experimental conditions.

[

[1] 刘传正, 陈春利. 中国地质灾害成因分析 [J]. 地质论评, 2020, 66(05): 1334-48.

[2] 陈运泰 . 地 震 预 测 : 回顾与展望 [J]. 中 国 科 学 ( 地 球 科 学 ), 2009, 39(12): 1633-58.

[3] JOHNSON P A, JIA X. Nonlinear dynamics, granular media and dynamic earthquake triggering [J]. Nature, 2005, 437(7060): 871 -4.

[4] BRACE W, BYERLEE J. Stick-slip as a mechanism for earthquakes [J]. Science, 1966, 153(3739): 990 -2.

[5] BYERLEE J D, BRACE W. Stick slip, stable sliding, and earthquakes—effect of rock type, pressure, strain rate, and stiffness [J]. Journal of Geophysical Research, 1968, 73(18): 6031 -7.

[6] KANAMORI H, BRODSKY E E. The physics of earthquakes [J]. Reports on progress in physics, 2004, 67(8): 1429.

[7] 吕征. 含颗粒物模拟断层粘滑运动机制的实验研究 [D]; 清华大学, 2019.

[8] 宋义敏, 马少鹏, 杨小彬, et al. 断层黏滑动态变形过程的实验研究 [J]. 地球物理学报, 2012, 55(1): 171-9.

[9] ZHUO Y-Q, LIU P, CHEN S, et al. Laboratory Observations of Tremor-Like Events Generated During Preslip [J]. Geophys Res Lett, 2018, 45(14): 6926-34.

[10] BERGEN K J, JOHNSON P A, DE HOOP M V, BEROZA G C. Machine learning for data -driven discovery in solid Earth geoscience [J]. Science, 2019, 363(6433).

[11] JOHNSON P A, FERDOWSI B, KAPROTH B M, et al. Acoustic emission and microslip precursors to stick -slip failure in sheared granular material [J]. Geophys Res Lett, 2013, 40(21): 5627 -31.

[12] RIVIÈRE J, LV Z, JOHNSON P A, MARONE C. Evolution of b -value during the seismic cycle: Insights from laboratory experiments on simulated faults [J]. Earth and Planetary Science Letters, 2018, 482: 407-13.

[13] 马瑾. 从“是否存在有助于预报的地震先兆”说起 [J]. 科学通报, 2016, 61(Z1): 409-14.

[14] JOHNSON P A, ROUET-LEDUC B, PYRAK-NOLTE L J, et al. Laboratory earthquake forecasting: A machine learning competition [J]. Proceedings of the National Academy of Sciences, 2021, 118(5): e2011362118.

[15] GAO K, EUSER B J, ROUGIER E, et al. Modeling of Stick -Slip Behavior in Sheared Granular Fault Gouge Using the Combined Finite-Discrete Element Method [J]. J Geophys Res Solid Earth, 2018, 123: 5774– 92.

[16] ROUET-LEDUC B, HULBERT C, LUBBERS N, et al. Machine Learning Predicts Laboratory Earthquakes [J]. Geophysical Research Letters, 2017, 44(18): 9276-82.

[17] JASPERSON H, BOLTON D C, JOHNSON P, et al. Attention Network Forecasts Time-to-Failure in Laboratory Shear Experiments [J]. J Geophys Res Solid Earth, 2021, 126(11): e2021JB022195.

[18] MCBECK J A, AIKEN J M, MATHIESEN J, et al. Deformation Precursors to Catastrophic Failure in Rocks [J]. Geophys Res Lett, 2020, 47(24): e2020GL090255.

[19] SHREEDHARAN S, BOLTON D C, RIVIÈRE J, MARONE C. Machine Learning Predicts the Timing and Shear Stress Evolution of Lab Earthquakes Using Active Seismic Monitoring of Fault Zone Processes [J]. J Geophys Res Solid Earth, 2021, 126(7): e2020JB021588.

[20] CHAIPORNKAEW L, ELSTON H, COOKE M, et al. Predicting Off-Fault Deformation From Experimental Strike -Slip Fault Images Using Convolutional Neural Networks [J]. Geophys Res Lett, 2022, 49(2): e2021GL096854.

[21] HAYES G P, EARLE P S, BENZ H M, et al. National earthquake information center strategic plan, 2019 –23 [R]: US Geological Survey, 2019.

[22] 杨旭, 李永华, 盖增喜. 机器学习在地震学中的应用进展 [J]. 地球与行星物理论评, 2021, 52(01): 76-88.

[23] MIGNAN A, BROCCARDO M. Neural Network Applications in Earthquake Prediction (1994-2019): Meta-Analytic and Statistical Insights on Their Limitations [J]. Seismol Res Lett, 2020, 91(4): 2330-42.

[24] BEROZA G C, SEGOU M, MOSTAFA MOUSAVI S. Machine learning and earthquake forecasting—next steps [J]. Nature Communications, 2021, 12(1): 4761.

[25] VAN KLAVEREN S, VASCONCELOS I, NIEMEIJER A. Predicting laboratory earthquakes with machine learning [J]. arXiv preprint arXiv:201106669, 2020.

[26] ASIM K M, MARTÍNEZ-ÁLVAREZ F, BASIT A, IQBAL T. Earthquake magnitude prediction in Hindukush region using machine learning techniques [J]. Natural Hazards, 2016, 85(1): 471-86.

[27] RAMIREZ JR J, MEYER F G. Machine Learning for Seismic Signal Processing: Phase Classification on a Manifold [Z]. 2011 10th International Conference on Machine Learning and Applications and Workshops. 2011: 382 -8.10.1109/icmla.2011.91

[28] BIANCO M J, GERSTOFT P, OLSEN K B, LIN F C. High -resolution seismic tomography of Long Beach, CA using machine learning [J]. Sci Rep, 2019, 9(1): 14987.

[29] BAI T, TAHMASEBI P. Efficient and data -driven prediction of water breakthrough in subsurface systems using deep long short term memory machine learning [J]. Computational Geosciences, 2020, 25(1): 285-97.

[30] DEVRIES P M R, VIEGAS F, WATTENBERG M, MEADE B J. Deep learning of aftershock patterns following large earthquakes [J]. Nature, 2018, 560(7720): 632 -4.

[31] BREGMAN Y, RABIN N. Aftershock Identification Using Diffusion Maps [J]. Seismological Research Letters, 2018, 90(2A): 539-45.

[32] MEIER M A, ROSS Z E, RAMACHANDRAN A, et al. Reliable Real ‐ Time Seismic Signal/Noise Discrimination With Machine Learning [J]. Journal of Geophysical Research: Solid Earth, 2019, 124(1): 788-800.

[33] LI Z, MEIER M-A, HAUKSSON E, et al. Machine Learning Seismic Wave Discrimination: Application to Earthquake Early Warning [J]. Geophysical Research Letters, 2018, 45(10): 4773 -9.

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

[35] ROUET-LEDUC B, HULBERT C, JOHNSON P A. Continuous chatter of the Cascadia subduction zone revealed by machine learning [J]. Nature Geoscience, 2018, 12(1): 75 -9.

[36] JOHNSON C W, JOHNSON P A. Learning the Low Frequency Earthquake Activity on the Central San Andreas Fault [J]. Geophysical Research Letters, 2021, 48(13).

[37] ALLEN R M, MELGAR D. Earthquake Early Warning: Advances, Scientific Challenges, and Societal Needs [J]. Annual Review of Earth and Planetary Sciences, 2019, 47(1): 361 -88.

[38] JORDAN T H. Earthquake predictability, brick by brick [J]. Seismological Research Letters, 2006, 77(1): 3 -6.

[39] OGATA Y, ZHUANG J. Space–time ETAS models and an improved extension [J]. Tectonophysics, 2006, 413(1 -2): 13-23.

[40] ZHUANG J, OGATA Y, VERE‐JONES D. Analyzing earthquake clustering features by using stochastic reconstru ction [J]. Journal of Geophysical Research: Solid Earth, 2004, 109(B5).

[41] ZHUANG J, OGATA Y, VERE-JONES D. Stochastic declustering of space-time earthquake occurrences [J]. Journal of the American Statistical Association, 2002, 97(458): 369 -80.

[42] MARZOCCHI W, ZHUANG J. Statistics between mainshocks and foreshocks in Italy and Southern California [J]. Geophysical Research Letters, 2011, 38(9).

[43] AKI K. A probabilistic synthesis of precursory phenomena [J]. Earthquake prediction: an international review, 1981, 4: 566 -74.

[44] UTSU T. Probabilities associated with earthquake prediction and their relationships [J]. Earthq Predict Res, 1983, 2: 105 -14.

[45] DING X, WANG X, WANG L, DOU A. Study on the development of seismic disaster prediction of lifeline systems based on ESRI ArcGIS engine 9; proceedings of the 2007 IEEE International Geoscience and Remote Sensing Symposium, F, 2007 [C]. IEEE.

[46] NISHIMURA T, FUJIWARA S, MURAKAMI M, et al. Fault model of the 2005 Fukuoka-ken Seiho-oki earthquake estimated from coseismic deformation observed by GPS and InSAR [J]. Earth, planets and space, 2006, 58(1): 51 -6.

[47] MORIKAWA N, FUJIWARA H. A new ground motion prediction equation for Japan applicable up to M9 mega -earthquake [J]. Journal of Disaster Research, 2013, 8(5): 878 -88.

[48] KAGAN Y Y, JACKSON D D. Long-term earthquake clustering [J]. Geophysical Journal International, 1991, 104(1): 117 -33.

[49] FIELD E H, GUPTA N, GUPTA V, et al. Hazard calculations for the WGCEP-2002 earthquake forecast using OpenSHA and distributed object technologies [J]. Seismological Research Letters, 2005, 76(2): 161-7.

[50] HUBERT ‐ FERRARI A, SUPPE J, VAN DER WOERD J, et al. Irregular earthquake cycle along the southern Tianshan front, Aksu area, China [J]. Journal of Geophysical Research: Solid Earth, 2005, 110(B6).

[51] DAËRON M, AVOUAC J P, CHARREAU J. Modeling the shortening history of a fault tip fold using structural and geomorphic records of deformation [J]. Journal of Geophysical Research: Solid Earth, 2007, 112(B3).

[52] 王芃, 邵志刚, 刘琦, et al. 基于多学科物理观测的地震概率预测方法在川滇地区的应用 [J]. 地球物理学报, 2019, 62(9): 3448-63.

[53] UTSU T, OGATA Y. The centenary of the Omori formula for a decay law of aftershock activity [J]. Journal of Physics of the Earth, 1995, 43(1): 1-33.

[54] SHI Y-L, LIU J, VERE-JONES D, et al. Application of mechanical and statistical models to the study of seismicity of synthetic earthquakes and the prediction of natural ones [J]. Acta Seismologica Sinica, 1998, 11: 421-30.

[55] 蒋长胜, 庄建仓. 基于时-空 ETAS 模型给出的川滇地区背景地震活动和强震潜在危险区 [J]. 地球物理学报, 2010, 53(2): 305-17.

[56] AGNEW D C, JONES L M. Prediction probabilities from foreshocks [J]. Journal of Geophysical Research: Solid Earth, 1991, 96(B7): 11959-71.

[57] PANAKKAT A, ADELI H. Recurrent neural network for approximate earthquake time and location prediction using multiple seismicity indicators [J]. Computer ‐ Aided Civil and Infrastructure Engineering, 2009, 24(4): 280 -92.

[58] MORALES-ESTEBAN A, MARTÍNEZ-ÁLVAREZ F, REYES J. Earthquake prediction in seismogenic areas of the Iberian Peninsula based on computational intelligence [J]. Tectonophysics, 2013, 593: 121-34.

[59] ASENCIO-CORTÉS G, MARTÍNEZ-ÁLVAREZ F, MORALES ESTEBAN A, REYES J. A sensitivity study of seismicity indicators in supervised learning to improve earthquake prediction [J]. Knowledge-Based Systems, 2016, 101: 15-30.

[60] ASIM K M, IDRIS A, IQBAL T, MARTÍNEZ-ÁLVAREZ F. Earthquake prediction model using support vector regressor and hybrid neural networks [J]. PloS one, 2018, 13(7): e0199004.

[61] BRYKOV M N, PETRYSHYNETS I, PRUNCU C I, et al. Machine Learning Modelling and Feature Engineering in Seismology Experiment [J]. Sensors (Basel), 2020, 20(15).

[62] CORBI F, BEDFORD J, SANDRI L, et al. Predicting Imminence of Analog Megathrust Earthquakes With Machine Learning: Implications for Monitoring Subduction Zones [J]. Geophysical Research Letters, 2020, 47(7).

[63] DEVRIES P M, VIÉGAS F, WATTENBERG M, MEADE B J. Deep learning of aftershock patterns following large earthquakes [J]. Nature, 2018, 560(7720): 632-4.

[64] NIKSARLIOGLU S, KULAHCI F. An artificial neural network model for earthquake prediction and relations between environmental parameters and earthquakes [J]. International Journal of Geological and Environmental Engineering, 2013, 7(2): 87-90.

[65] SURATGAR A A, SETOUDEH F, SALEMI A H, NEGARESTANI A. Magnitude of earthquake prediction using neural network; proceedings of the 2008 Fourth International Conference on Natural Computation, F, 2008 [C]. IEEE.

[66] CORBI F, SANDRI L, BEDFORD J, et al. Machine learning can predict the timing and size of analog earthquakes [J]. Geophysical Research Letters, 2019, 46(3): 1303 -11.

[67] NICOLIS O, PLAZA F, SALAS R. Prediction of intensity and location of seismic events using deep learning [J]. Spatial Statistics, 2021, 42: 100442.

[68] ASIM K M, IDRIS A, IQBAL T, MARTÍNEZ-ÁLVAREZ F. Seismic indicators based earthquake predictor system using Genetic Programming and AdaBoost classification [J]. Soil Dynamics and Earthquake Engineering, 2018, 111: 1 -7.

[69] LI R, LU X, LI S, et al. DLEP: A deep learning model for earthquake prediction; proceedings of the 2020 International Joint Conference on Neural Networks (IJCNN), F, 2020 [C]. IEEE.

[70] KAIL R, BURNAEV E, ZAYTSEV A. Recurrent convolutional neural networks help to predict location of earthquakes [J]. IEEE Geoscience and Remote Sensing Letters, 2021, 19: 1 -5.

[71] HUANG J, WANG X, ZHAO Y, et al. LARGE EARTHQUAKE MAGNITUDE PREDICTION IN TAIWAN BASED ON DEEP LEARNING NEURAL NETWORK [J]. Neural Network World, 2018, (2).

[72] BORATE P, RIVIÈRE J, MARONE C, et al. Using a physics informed neural network and fault zone acoustic monitoring to predict lab earthquakes [J]. Nature Communications, 2023, 14(1): 3693.

[73] NIEMEIJER A, MARONE C, ELSWORTH D. Frictional strength and strain weakening in simulated fault gouge: Competition between geometrical weakening and chemical strengthening [J]. Journal of Geophysical Research, 2010, 115(B10).

[74] GELLER D A, ECKE R E, DAHMEN K A, BACKHAUS S. Stick -slip behavior in a continuum-granular experiment [J]. Phys Rev E Stat Nonlin Soft Matter Phys, 2015, 92(6): 060201.

[75] TINTI E, SCUDERI M M, SCOGNAMIGLIO L, et al. On the evolution of elastic properties during laboratory stick ‐ slip experiments spanning the transition from slow slip to dynamic rupture [J]. Journal of Geophysical Research: Solid Earth, 2016, 121(12): 8569-94.

[76] LEEMAN J R, SAFFER D M, SCUDERI M M, MARONE C. Laboratory observations of slow earthquakes and the spectrum of tectonic fault slip modes [J]. Nat Commun, 2016, 7: 11104.

[77] BOLTON D C, SHOKOUHI P, ROUET ‐ LEDUC B, et al. Characterizing Acoustic Signals and Searching for Precursors during the Laboratory Seismic Cycle Using Unsupervised Machine Learning [J]. Seismological Research Letters, 2019, 90(3): 1088-98.

[78] SHREEDHARAN S, BOLTON D C, RIVIÈRE J, MARONE C. Machine Learning Predicts the Timing and Shear Stress Evolution of Lab Earthquakes Using Active Seismic Monitoring of Fault Zone Processes [J]. Journal of Geophysical Research: Solid Earth, 2021, 126(7).

[79] JASPERSON H, BOLTON D C, JOHNSON P, et al. Attention Network Forecasts Time ‐ to ‐ Failure in Laboratory Shear Experiments [J]. Journal of Geophysical Research: Solid Earth, 2021, 126(11).

[80] ROUET-LEDUC B, HULBERT C, BOLTON D C, et al. Estimating Fault Friction From Seismic Signals in the Laboratory [J]. Geophysical Research Letters, 2018, 45(3): 1321 -9.

[81] GAO K, EUSER B J, ROUGIER E, et al. Modeling of Stick -Slip Behavior in Sheared Granular Fault Gouge Using the Combined Finite-Discrete Element Method [J]. Journal of Geophysical Research: Solid Earth, 2018, 123(7): 5774-92.

[82] LUBBERS N, BOLTON D C, MOHD‐YUSOF J, et al. Earthquake Catalog ‐ Based Machine Learning Identification of Laboratory Fault States and the Effects of Magnitude of Completeness [J]. Geophysical Research Letters, 2018, 45(24).

[83] KHOSRAVIKIA F, CLAYTON P. Machine learning in ground motion prediction [J]. Computers & Geosciences, 2021, 148.

[84] WANG K, JOHNSON C W, BENNETT K C, JOHNSON P A. Predicting Future Laboratory Fault Friction Through Deep Learning Transformer Models [J]. Geophysical Research Letters, 2022, 49(19).

[85] ZHAO Q, GLASER S D. Relocating Acoustic Emission in Rocks with Unknown Velocity Structure with Machine Learning [J]. Rock Mechanics and Rock Engineering, 2019, 53(5): 2053 -61.

[86] XIE Y, EBAD SICHANI M, PADGETT J E, DESROCHES R. The promise of implementing machine learning in earthquake engineering: A state-of-the-art review [J]. Earthquake Spectra, 2020, 36(4): 1769-801.

[87] AHARONOV E, SPARKS D. Stick-slip motion in simulated granular layers [J]. Journal of Geophysical Research: Solid Earth, 2004, 109(B9).

[88] FERDOWSI B, GRIFFA M, GUYER R A, et al. Three -dimensional discrete element modeling of triggered slip in sheared granular media [J]. Phys Rev E Stat Nonlin Soft Matter Phys, 2014, 89(4): 042204.

[89] DRATT M, KATTERFELD A. Coupling of FEM and DEM simulations to consider dynamic deformations under particle load [J]. Granular Matter, 2017, 19(3).

[90] GAO K, GUYER R A, ROUGIER E, JOHNSON P A. Plate motion in sheared granular fault system [J]. Earth Planet Sc Lett, 2020, 548: 116481.

[91] GAO K, GUYER R, ROUGIER E, et al. From Stress Chains to Acoustic Emission [J]. Phys Rev Lett, 2019, 123(4): 048003.

[92] REN C X, DOROSTKAR O, ROUET‐LEDUC B, et al. Machine Learning Reveals the State of Intermittent Frictional Dynamics in a Sheared Granular Fault [J]. Geophysical Research Letters, 2019, 46(13): 7395-403.

[93] MA G, MEI J, GAO K, et al. Machine learning bridges microslips and slip avalanches of sheared granular gouges [J]. Earth and Planetary Science Letters, 2022, 579.

[94] GAO K, GUYER R A, ROUGIER E, JOHNSON P A. Plate motion in sheared granular fault system [J]. Earth and Planetary Science Letters, 2020, 548.

[95] JOHNSON P A, ROUET-LEDUC B, PYRAK-NOLTE L J, et al. Laboratory earthquake forecasting: A machine learning competition [J]. Proc Natl Acad Sci U S A, 2021, 118(5).

[96] MUNJIZA A, LEI Z, DIVIC V, PEROS B. Fracture and fragmentation of thin shells using the combined finite -discrete element method [J]. International Journal for Numerical Methods in Engineering, 2013, 95(6): 478 -98.

[97] WANG K, JOHNSON C W, BENNETT K C, JOHNSON P A. Predicting fault slip via transfer learning [J]. Nature communications, 2021, 12(1): 1-11.

[98] HE K, ZHANG X, REN S, SUN J. Deep residual learning for image recognition; proceedings of the Proceedings of the IEEE conference on computer vision and pattern recognition, F, 2016 [C].

[99] POLIKAR R. Ensemble learning [M]. Ensemble machine learning. Springer. 2012: 1-34.

[100] DIETTERICH T G. Ensemble methods in machine learning; proceedings of the International workshop on multiple classifier systems, F, 2000 [C]. Springer.

[101] CHEN T, HE T, BENESTY M, et al. Xgboost: extreme gradient boosting [J]. R package version 04 -2, 2015, 1(4): 1-4.

[102] KE G, MENG Q, FINLEY T, et al. LightGBM: a highly efficient gradient boosting decision tree [Z]. Proceedings of the 31st International Conference on Neural Information Processing Systems. Long Beach, California, USA; Curran Associates Inc. 2017: 3149–57

[103] MUNJIZA A. Discrete elements in transient dynamics of fractured media [D]; Swansea University, 1992.

[104] TRUGMAN D T, MCBREARTY I W, BOLTON D C, et al. The Spatiotemporal Evolution of Granular Microslip Precursors to Laboratory Earthquakes [J]. Geophysical Research Letters, 2020, 47(16).

[105] GUTENBERG B, RICHTER C F. Earthquake magnitude, intensity, energy, and acceleration: (Second paper) [J]. Bulletin of the seismological society of America, 1956, 46(2): 105 -45.

[106] LUNDBERG S M, ERION G G, LEE S-I. Consistent individualized feature attribution for tree ensembles [J]. arXiv preprint arXiv:180203888, 2018.

[107] REJEB S, DUVEAU C, REBAFKA T. Self-Organizing Maps for Exploration of Partially Observed Data and Imputation of Missing Values [J]. arXiv preprint arXiv:220207963, 2022.

[108] PELLEG D, MOORE A W. X-means: Extending k-means with efficient estimation of the number of clusters; proceedings of the Icml, F, 2000 [C].

[109] BAI S, KOLTER J Z, KOLTUN V. An empirical evaluation of generic convolutional and recurrent networks for sequence modeling [J]. arXiv preprint arXiv:180301271, 2018.

[110] HOCHREITER S, SCHMIDHUBER J. Long short-term memory [J]. Neural computation, 1997, 9(8): 1735-80.

[111] VASWANI A, SHAZEER N, PARMAR N, et al. Attention is all you need [J]. Advances in neural information processing systems, 2017, 30.

[112] BADRINARAYANAN V, KENDALL A, CIPOLLA R. Segnet: A deep convolutional encoder-decoder architecture for image segmentation [J]. IEEE transactions on pattern analysis and machine intelligence, 2017, 39(12): 2481 -95.

[113] SENGUPTA A, YE Y, WANG R, et al. Going deeper in spiking neural networks: VGG and residual architectures [J]. Frontiers in neuroscience, 2019, 13: 95.

[114] AL-JAWFI R. Handwriting Arabic character recognition LeNet using neural network [J]. Int Arab J Inf Technol, 2009, 6(3): 304 -9.

[115] IANDOLA F N, HAN S, MOSKEWICZ M W, et al. SqueezeNet: AlexNet-level accuracy with 50x fewer parameters and< 0.5 MB model size [J]. arXiv preprint arXiv:160207360, 2016.

[116] LIN M, CHEN Q, YAN S. Network in network [J]. arXiv preprint arXiv:13124400, 2013.

[117] MOUSAVI S M, BEROZA G C. A Machine ‐Learning Approach for Earthquake Magnitude Estimation [J]. Geophysica l Research Letters, 2020, 47(1).

[118] LAURENTI L, TINTI E, GALASSO F, et al. Deep learning for laboratory earthquake prediction and autoregressive forecasting of fault zone stress [J]. Earth and Planetary Science Letters, 2022, 598.

[119] SHREEDHARAN S, BOLTON D C, RIVIÈRE J, MARONE C. Preseismic Fault Creep and Elastic Wave Amplitude Precursors Scale With Lab Earthquake Magnitude for the Continuum of Tectonic Failure Modes [J]. Geophysical Research Letters, 2020, 47(8).

[120] REN C X, PELTIER A, FERRAZZINI V, et al. Machine Learning Reveals the Seismic Signature of Eruptive Behavior at Piton de la Fournaise Volcano [J]. Geophys Res Lett, 2020, 47(3): e2019GL085523.

[121] BERGSTRA J, BENGIO Y. Random search for hyper-parameter optimization [J]. Journal of machine learning research, 2012, 13(2).

[122] MARTINEZ-CANTIN R. BayesOpt: a Bayesian optimization library for nonlinear optimization, experimental design and bandits [J]. J Mach Learn Res, 2014, 15(1): 3735 -9.

[123] VICTORIA A H, MARAGATHAM G. Automatic tuning of hyperparameters using Bayesian optimization [J]. Evolving Systems, 2021, 12(1): 217-23.

[124] BORATE P, RIVIERE J, MARONE C, et al. Using a physics informed neural network and fault zone acoustic monitoring to predict lab earthquakes [J]. Nat Commun, 2023, 14(1): 3693.

[125] KEMNA K B, ROTH M P, WACHE R M, et al. Small Magnitude Events Highlight the Correlation Between Hydraulic Fracturing Injection Parameters, Geological Factors, and Earthquake Occurrence [J]. Geophysical Research Letters, 2022, 49(21).

[126] LI Y E, O ’ MALLEY D, BEROZA G, et al. Machine Learning Developments and Applications in Solid‐Earth Geosciences: Fad or Future? [J]. Journal of Geophysical Research: Solid Earth, 2023, 128(1).

[127] VAN HOUDT G, MOSQUERA C, NÁPOLES G. A review on the long short-term memory model [J]. Artificial Intelligence Review, 2020, 53(8): 5929-55.

[128] BEROZA G C, SEGOU M, MOSTAFA MOUSAVI S. Machine learning and earthquake forecasting -next steps [J]. Nat Commun, 2021, 12(1): 4761.

[129] WANG K, JOHNSON C W, BENNETT K C, JOHNSON P A. Predicting fault slip via transfer learning [J]. Nat Commun, 2021, 12(1): 7319.

[130] GAO P, JIANG Z, YOU H, et al. Dynamic fusion with intra -and inter-modality attention flow for visual question answering; proceedings of the Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, F, 2019 [C].

[131] HOULSBY N, GIURGIU A, JASTRZEBSKI S, et al. Parameter efficient transfer learning for NLP; proceedings of the International Conference on Machine Learning, F, 2019 [C]. PMLR.