Distributed Acoustic Sensing (DAS) is a technology that enables continuous, real-time measurements along the entire length of a fiber optic cable. The low-frequency band of DAS can be used to analyze hydraulic fracture geometry and growth. In this study, the low-frequency strain waterfall plots with their corresponding pumping curves were analyzed to obtain information on fracture azimuth, propagation speed, number of fractures created in each stage, and re-stimulation of pre-existing fractures. We also use a simple geomechanical model to predict fracture growth rates while accounting for changes in treatment parameters. As expected, the hydraulic fractures principally propagate perpendicular to the treated well, that is, parallel to the direction of maximum horizontal stress. During many stages, multiple frac hits are visible indicating that multiple parallel fractures are created and/or re-opened. Secondary fractures deviate towards the heel of the well, likely due to the cumulative stress shadow caused by previous and current stages. The presence of heart-shaped tips reveals that some stress and/or material barrier is overcome by the hydraulic fracture. The lobes of the heart are best explained by the shear stresses at 45-degree angles from the fracture tip instead of the tensile stresses directly ahead of the tip. Antennas ahead of the fracture hits indicate the re-opening of pre-existing fractures. Tails in the waterfall plots provide information on the continued opening, closing, and interaction of the hydraulic fractures within the fracture domain and stage domain corridors. Analysis of the low-frequency DAS plots thus provides in-depth insights into the rock deformation and rock-fluid interaction processes occurring close to the observation well.
分布式声学传感（DAS）是一种能够沿光缆全长进行连续、实时测量的技术。DAS 的低频波段可用于分析水力断裂的几何形状和生长情况。在本研究中，我们分析了低频应变瀑布图及其相应的泵送曲线，以获取有关裂缝方位角、传播速度、每个阶段产生的裂缝数量以及对原有裂缝的再刺激等信息。我们还使用一个简单的地质力学模型来预测裂缝增长率，同时考虑到处理参数的变化。不出所料，水力压裂主要是垂直于处理过的油井传播，即平行于最大水平应力方向。在许多阶段，可以看到多个压裂点，这表明产生和/或重新打开了多条平行裂缝。次生裂缝向井跟方向偏离，这可能是由于前一阶段和当前阶段造成的累积应力阴影。心形顶端的出现表明水力压裂克服了某些应力和/或材料障碍。心形裂片的最佳解释是与裂缝尖端成 45 度角的剪应力，而不是尖端正前方的拉应力。断裂点前方的天线表明先前存在的断裂重新打开。瀑布图中的尾部提供了断裂域和阶段域走廊内水力断裂持续打开、关闭和相互作用的信息。因此，分析低频 DAS 图可以深入了解观察井附近发生的岩石变形和岩流相互作用过程。
{"title":"Interpretation of low-frequency distributed acoustic sensing data based on geomechanical models","authors":"Ana Karen Ortega Perez, M. van der Baan","doi":"10.1190/geo2023-0348.1","DOIUrl":"https://doi.org/10.1190/geo2023-0348.1","url":null,"abstract":"Distributed Acoustic Sensing (DAS) is a technology that enables continuous, real-time measurements along the entire length of a fiber optic cable. The low-frequency band of DAS can be used to analyze hydraulic fracture geometry and growth. In this study, the low-frequency strain waterfall plots with their corresponding pumping curves were analyzed to obtain information on fracture azimuth, propagation speed, number of fractures created in each stage, and re-stimulation of pre-existing fractures. We also use a simple geomechanical model to predict fracture growth rates while accounting for changes in treatment parameters. As expected, the hydraulic fractures principally propagate perpendicular to the treated well, that is, parallel to the direction of maximum horizontal stress. During many stages, multiple frac hits are visible indicating that multiple parallel fractures are created and/or re-opened. Secondary fractures deviate towards the heel of the well, likely due to the cumulative stress shadow caused by previous and current stages. The presence of heart-shaped tips reveals that some stress and/or material barrier is overcome by the hydraulic fracture. The lobes of the heart are best explained by the shear stresses at 45-degree angles from the fracture tip instead of the tensile stresses directly ahead of the tip. Antennas ahead of the fracture hits indicate the re-opening of pre-existing fractures. Tails in the waterfall plots provide information on the continued opening, closing, and interaction of the hydraulic fractures within the fracture domain and stage domain corridors. Analysis of the low-frequency DAS plots thus provides in-depth insights into the rock deformation and rock-fluid interaction processes occurring close to the observation well.","PeriodicalId":55102,"journal":{"name":"Geophysics","volume":"141 36","pages":""},"PeriodicalIF":3.3,"publicationDate":"2023-12-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"138953420","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
D. Colombo, E. Turkoglu, E. Sandoval-Curiel, Taqi Alyousuf
We develop a recursive, self-supervised machine learning inversion for fast and accurate full waveform inversion of land seismic data. Machine learning generalization is enhanced by using virtual super gathers of field data for training. These are obtained from midpoint-offset sorting and stacking after applying surface-consistent corrections from the decomposition of the transmitted wavefield. The procedure implements reinforcement learning concepts by adopting an inversion agent to interact with the environment and explore the model space under a data misfit optimization policy. The generated parameter distributions and related forward responses are used as new training samples for supervised learning. The active learning paradigm is further embedded in the procedure where queries on data diversity and uncertainty are used to generate fully informative reduced sets for training. The procedure is recursive. At each cycle, the physics-based inversion is coupled to the machine learning predictions via penalty terms that promote a long-term data misfit reduction. The resulting self-supervised, active learning, physics-driven deep learning inversion generalizes well with field data. The method is applied to perform full waveform inversion of a complex land seismic dataset characterized by transcurrent faulting and related structures. High signal-to-noise virtual super gathers are inverted with a 1.5D Laplace-Fourier full waveform inversion scheme. The active learning inversion procedure utilizes a small fraction of data for training while achieving sharper velocity reconstructions and a lower data misfit when compared to previous results. Active learning full waveform inversion is highly generalizable and effective for land seismic velocity model building and for other inversion scenarios.
{"title":"Self-Supervised, Active Learning Seismic Full Waveform Inversion","authors":"D. Colombo, E. Turkoglu, E. Sandoval-Curiel, Taqi Alyousuf","doi":"10.1190/geo2023-0308.1","DOIUrl":"https://doi.org/10.1190/geo2023-0308.1","url":null,"abstract":"We develop a recursive, self-supervised machine learning inversion for fast and accurate full waveform inversion of land seismic data. Machine learning generalization is enhanced by using virtual super gathers of field data for training. These are obtained from midpoint-offset sorting and stacking after applying surface-consistent corrections from the decomposition of the transmitted wavefield. The procedure implements reinforcement learning concepts by adopting an inversion agent to interact with the environment and explore the model space under a data misfit optimization policy. The generated parameter distributions and related forward responses are used as new training samples for supervised learning. The active learning paradigm is further embedded in the procedure where queries on data diversity and uncertainty are used to generate fully informative reduced sets for training. The procedure is recursive. At each cycle, the physics-based inversion is coupled to the machine learning predictions via penalty terms that promote a long-term data misfit reduction. The resulting self-supervised, active learning, physics-driven deep learning inversion generalizes well with field data. The method is applied to perform full waveform inversion of a complex land seismic dataset characterized by transcurrent faulting and related structures. High signal-to-noise virtual super gathers are inverted with a 1.5D Laplace-Fourier full waveform inversion scheme. The active learning inversion procedure utilizes a small fraction of data for training while achieving sharper velocity reconstructions and a lower data misfit when compared to previous results. Active learning full waveform inversion is highly generalizable and effective for land seismic velocity model building and for other inversion scenarios.","PeriodicalId":55102,"journal":{"name":"Geophysics","volume":"126 40","pages":""},"PeriodicalIF":3.3,"publicationDate":"2023-12-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"138953655","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Traditional gradient-based inversion methods usually suffer from the problems of falling into local minima and relying heavily on initial guesses. Deep learning methods have received increasing attention due to their excellent nonlinear fitting ability. However, given the recent application of deep learning methods in the field of MT inversion, there are currently challenges associated with achieving high inversion resolution and extracting sufficient features. We develop a neural network model (called MT2DInv-Unet) based on the deformable convolution for two-dimensional MT inversion to approximate the nonlinear mapping from the MT response data to the resistivity model. The deformable convolution is achieved by adding an additional offset to each sample point of the conventional convolution operation, which extracts hidden relationships and allows the flexible adjustment of the size and shape of the feature region. Meanwhile, we design the network structure with multi-scale residual blocks, which effectively extract the multi-scale features of the MT response data. This design not only enhances the network performance but also alleviates issues such as vanishing gradients and network degradation. The results of synthetic models show that the proposed network inversion method has stable convergence, good robustness and generalization performance, and performs better than the fully convolutional neural network (FCN) and U-Net network. Finally, the inversion results of field data show that MT2DInv-Unet can effectively obtain a reliable underground resistivity structure, and has a good application prospect in MT inversion.
{"title":"MT2DInv-Unet: A two-dimensional magnetotelluric inversion method based on deep learning technology","authors":"Kejia Pan, Weiwei Ling, Jiajing Zhang, Xin Zhong, Zhengyong Ren, Shuanggui Hu, Dongdong He, Jingtian Tang","doi":"10.1190/geo2023-0004.1","DOIUrl":"https://doi.org/10.1190/geo2023-0004.1","url":null,"abstract":"Traditional gradient-based inversion methods usually suffer from the problems of falling into local minima and relying heavily on initial guesses. Deep learning methods have received increasing attention due to their excellent nonlinear fitting ability. However, given the recent application of deep learning methods in the field of MT inversion, there are currently challenges associated with achieving high inversion resolution and extracting sufficient features. We develop a neural network model (called MT2DInv-Unet) based on the deformable convolution for two-dimensional MT inversion to approximate the nonlinear mapping from the MT response data to the resistivity model. The deformable convolution is achieved by adding an additional offset to each sample point of the conventional convolution operation, which extracts hidden relationships and allows the flexible adjustment of the size and shape of the feature region. Meanwhile, we design the network structure with multi-scale residual blocks, which effectively extract the multi-scale features of the MT response data. This design not only enhances the network performance but also alleviates issues such as vanishing gradients and network degradation. The results of synthetic models show that the proposed network inversion method has stable convergence, good robustness and generalization performance, and performs better than the fully convolutional neural network (FCN) and U-Net network. Finally, the inversion results of field data show that MT2DInv-Unet can effectively obtain a reliable underground resistivity structure, and has a good application prospect in MT inversion.","PeriodicalId":55102,"journal":{"name":"Geophysics","volume":"120 10","pages":""},"PeriodicalIF":3.3,"publicationDate":"2023-12-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"138953762","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Geofluid discrimination and permeability prediction are indispensable steps in reservoir evaluation. From the perspective of pre-stack seismic inversion, predicting fluid indicators is an effective method for obtaining fluid properties directly from seismic data. In contrast, the direct prediction of permeability from observed seismic gathers is constrained by the difficulty in establishing a link between permeability and elastic parameters. However, we show that the pore structure parameters in seismic petrophysical theory are highly related to permeability, providing a new solution for predicting permeability using seismic data. Therefore, the correlation between the shear flexibility factor and permeability is first verified based on logging curves and laboratory data, and the results demonstrate that the shear flexibility factor can give an indicator of reservoir permeability. Secondly, an approximate reflection coefficient equation is derived for the direct characterization of the shear flexibility factor. In the proposed equation, a novel fluid indicator, expressed as the ratio of Russells fluid indicator to the square of the shear flexibility factor, is defined for the simultaneous prediction of fluid types and permeability. With the validated response of the novel fluid indicator to geofluid types, we achieve simultaneous predictions of fluid types and reservoir permeability characteristics from pre-stack seismic data, employing a boundary-constrained Bayesian inversion strategy. The model tests and the application on field data from a clastic reservoir confirm the effectiveness and applicability of the method.
{"title":"Simultaneous prediction of geofluid and permeability of reservoirs in pre-stack seismic inversion","authors":"Wenqiang Yang, Zhaoyun Zong, Qianhao Sun","doi":"10.1190/geo2023-0218.1","DOIUrl":"https://doi.org/10.1190/geo2023-0218.1","url":null,"abstract":"Geofluid discrimination and permeability prediction are indispensable steps in reservoir evaluation. From the perspective of pre-stack seismic inversion, predicting fluid indicators is an effective method for obtaining fluid properties directly from seismic data. In contrast, the direct prediction of permeability from observed seismic gathers is constrained by the difficulty in establishing a link between permeability and elastic parameters. However, we show that the pore structure parameters in seismic petrophysical theory are highly related to permeability, providing a new solution for predicting permeability using seismic data. Therefore, the correlation between the shear flexibility factor and permeability is first verified based on logging curves and laboratory data, and the results demonstrate that the shear flexibility factor can give an indicator of reservoir permeability. Secondly, an approximate reflection coefficient equation is derived for the direct characterization of the shear flexibility factor. In the proposed equation, a novel fluid indicator, expressed as the ratio of Russells fluid indicator to the square of the shear flexibility factor, is defined for the simultaneous prediction of fluid types and permeability. With the validated response of the novel fluid indicator to geofluid types, we achieve simultaneous predictions of fluid types and reservoir permeability characteristics from pre-stack seismic data, employing a boundary-constrained Bayesian inversion strategy. The model tests and the application on field data from a clastic reservoir confirm the effectiveness and applicability of the method.","PeriodicalId":55102,"journal":{"name":"Geophysics","volume":"23 6","pages":""},"PeriodicalIF":3.3,"publicationDate":"2023-12-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139002787","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Environmental calibration of logging curves is critical to petrophysical interpretation and sweet spot characterization. Wellbore failure frequently occurs in clay-rich (shalely) rocks during drilling, leading to biased logging interpretation and uncertainty. To reduce the biased correction or erroneous decision-making in the interpreter-dominated logging curve calibration process, we develop an empirically-informed CNN (EiCNN) logging curve correction strategy to calibrate the borehole failure-induced logging curve abnormity more accurately. The EiCNN method, together with high-quality logging curves as labeled samples, provides a nonlinear mapping between input logging curves and calibrations for the distorted curves. The EiCNN method completely alleviates biased correction or decision-making by the interpreter-dominated method. It has strong generalization ability, using many empirically interpreted high-quality data as input samples. The field validation wells demonstrate that the EiCNN model can precisely correct the distorted logging curves of mudstone segments with a correlation coefficient of >0.95. Moreover, the validation and test wells illustrate that the EiCNN method is capable of precisely correcting logging curves of interlayer mudstone, implying that the EiCNN method, to a certain degree, can also accurately perform environmental correction of logging curves from thin mudstone layers.
{"title":"Empirically-informed CNN model for logging curve calibration","authors":"Xinyu Hu, Hui Li, Hao Zhang, Baohai Wu, Li Ma, Xiaogang Wen, Jinghuai Gao","doi":"10.1190/geo2022-0696.1","DOIUrl":"https://doi.org/10.1190/geo2022-0696.1","url":null,"abstract":"Environmental calibration of logging curves is critical to petrophysical interpretation and sweet spot characterization. Wellbore failure frequently occurs in clay-rich (shalely) rocks during drilling, leading to biased logging interpretation and uncertainty. To reduce the biased correction or erroneous decision-making in the interpreter-dominated logging curve calibration process, we develop an empirically-informed CNN (EiCNN) logging curve correction strategy to calibrate the borehole failure-induced logging curve abnormity more accurately. The EiCNN method, together with high-quality logging curves as labeled samples, provides a nonlinear mapping between input logging curves and calibrations for the distorted curves. The EiCNN method completely alleviates biased correction or decision-making by the interpreter-dominated method. It has strong generalization ability, using many empirically interpreted high-quality data as input samples. The field validation wells demonstrate that the EiCNN model can precisely correct the distorted logging curves of mudstone segments with a correlation coefficient of >0.95. Moreover, the validation and test wells illustrate that the EiCNN method is capable of precisely correcting logging curves of interlayer mudstone, implying that the EiCNN method, to a certain degree, can also accurately perform environmental correction of logging curves from thin mudstone layers.","PeriodicalId":55102,"journal":{"name":"Geophysics","volume":"5 4","pages":""},"PeriodicalIF":3.3,"publicationDate":"2023-12-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"138971157","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Jing Li, Chang Zhang, Sherif Hanafy, Han Yu, Lige Bai
Wave equation dispersion (WD) inversion techniques for surface waves have proven to be a robust way to invert for the shear-wave (S-wave) velocity model. Unlike 1D dispersion curve inversion, the proposed WD method obviates the need for a layered model assumption and reduces the susceptibility to cycle-skipping issues in surface wave full waveform inversion (FWI). Previous WD inversion experiments conducted on Rayleigh and Love waves have highlighted that inverting Love waves yields better stability due to their independence from the P-wave velocity model. Nevertheless, Rayleigh waves possess the advantage of greater penetration depth compared to Love waves with similar wavelengths. Therefore, combining the two types of surface waves is a feasible way to improve the accuracy of S-velocity tomograms. In light of this, we propose a novel approach: a joint WD inversion encompassing both Rayleigh and Love waves. This innovative technique adjusts the weighting of individual WD gradients using the sensitivity factor of an equivalent layered model, offering a significant advancement in subsurface characterization. Synthetic model tests demonstrate that the joint WD inversion method can generate a more accurate S-velocity model, particularly in the presence of complex low-velocity layers (LVL) or high-velocity layers (HVL), when compared to individual wave WD inversion techniques. Simultaneously, the results of field tests validate the effectiveness of the proposed joint WD inversion strategy in producing a more dependable S-wave velocity distribution that aligns closely with the actual geological structure.
用于面波的波方程频散（WD）反演技术已被证明是反演剪切波（S 波）速度模型的可靠方法。与一维频散曲线反演不同，所提出的 WD 方法无需分层模型假设，并降低了面波全波形反演（FWI）中周期跳跃问题的敏感性。之前对雷利波和洛夫波进行的 WD 反演实验表明，由于洛夫波与 P 波速度模型无关，因此反演洛夫波具有更好的稳定性。然而，与波长相似的洛夫波相比，瑞利波具有穿透深度更大的优势。因此，将这两种面波结合起来是提高 S-速度层析成像精度的可行方法。有鉴于此，我们提出了一种新方法：包含瑞利波和爱波的联合 WD 反演。这项创新技术利用等效分层模型的灵敏度系数调整单个 WD 梯度的权重，为地下特征描述提供了重大进展。合成模型试验表明，与单独的波WD反演技术相比，联合WD反演方法可以生成更精确的S-速度模型，尤其是在存在复杂的低速层（LVL）或高速层（HVL）的情况下。同时，现场测试结果验证了所提出的联合 WD 反演策略在生成更可靠的 S 波速度分布方面的有效性，该速度分布与实际地质结构非常吻合。
{"title":"Joint wave-equation inversion of Rayleigh- and Love- dispersion curves","authors":"Jing Li, Chang Zhang, Sherif Hanafy, Han Yu, Lige Bai","doi":"10.1190/geo2023-0070.1","DOIUrl":"https://doi.org/10.1190/geo2023-0070.1","url":null,"abstract":"Wave equation dispersion (WD) inversion techniques for surface waves have proven to be a robust way to invert for the shear-wave (S-wave) velocity model. Unlike 1D dispersion curve inversion, the proposed WD method obviates the need for a layered model assumption and reduces the susceptibility to cycle-skipping issues in surface wave full waveform inversion (FWI). Previous WD inversion experiments conducted on Rayleigh and Love waves have highlighted that inverting Love waves yields better stability due to their independence from the P-wave velocity model. Nevertheless, Rayleigh waves possess the advantage of greater penetration depth compared to Love waves with similar wavelengths. Therefore, combining the two types of surface waves is a feasible way to improve the accuracy of S-velocity tomograms. In light of this, we propose a novel approach: a joint WD inversion encompassing both Rayleigh and Love waves. This innovative technique adjusts the weighting of individual WD gradients using the sensitivity factor of an equivalent layered model, offering a significant advancement in subsurface characterization. Synthetic model tests demonstrate that the joint WD inversion method can generate a more accurate S-velocity model, particularly in the presence of complex low-velocity layers (LVL) or high-velocity layers (HVL), when compared to individual wave WD inversion techniques. Simultaneously, the results of field tests validate the effectiveness of the proposed joint WD inversion strategy in producing a more dependable S-wave velocity distribution that aligns closely with the actual geological structure.","PeriodicalId":55102,"journal":{"name":"Geophysics","volume":"14 14","pages":""},"PeriodicalIF":3.3,"publicationDate":"2023-12-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"138972168","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Carlos A. M. Assis, Hervé Chauris, F. Audebert, Paul Williamson
Inversion velocity analysis (IVA) is an image domain method built upon the spatial scale separation of the model. Accordingly, the IVA method is performed with an iterative process composed of two minimization steps consisting of migration (inner loop) and tomography (outer loop), respectively, with each step accounting for its Hessian or not. The migration part provides the common image gathers (CIGs) with extension in the horizontal subsurface offset. Then, the differential semblance optimization (DSO) misfit measures the focusing of the events in the CIGs which indicates the quality of the velocity model. Commonly, the velocity updates are obtained from the DSO gradient. IVA is a modified version where the approximate inverse replaces the adjoint of the inner loop process: in that case, the migration Hessian is approximately diagonal in the high-frequency regime. In this work, we report the implementation of the tomographic Hessian (i.e., the second derivative of the DSO misfit with respect to the background model) for the estimation of the background velocity model. We apply the second-order adjoint-state method to obtain the application of the tomographic Hessian on a vector. Then, we use the truncated-Newton method to obtain the update directions by computing approximately the application of the inverse of the tomographic Hessian on the descent direction. We also make a theoretical comparison between the tomography in the IVA and full-waveform inversion contexts. Two numerical examples are used to compare, in terms of geophysical results and computational costs, the truncated-Newton method with different gradient-based optimization methods applied to IVA. A small model allows us to evaluate the eigenvalues of the tomographic Hessian which explains the large damping needed in the truncated-Newton case.
{"title":"Investigating Hessian-based inversion velocity analysis","authors":"Carlos A. M. Assis, Hervé Chauris, F. Audebert, Paul Williamson","doi":"10.1190/geo2022-0689.1","DOIUrl":"https://doi.org/10.1190/geo2022-0689.1","url":null,"abstract":"Inversion velocity analysis (IVA) is an image domain method built upon the spatial scale separation of the model. Accordingly, the IVA method is performed with an iterative process composed of two minimization steps consisting of migration (inner loop) and tomography (outer loop), respectively, with each step accounting for its Hessian or not. The migration part provides the common image gathers (CIGs) with extension in the horizontal subsurface offset. Then, the differential semblance optimization (DSO) misfit measures the focusing of the events in the CIGs which indicates the quality of the velocity model. Commonly, the velocity updates are obtained from the DSO gradient. IVA is a modified version where the approximate inverse replaces the adjoint of the inner loop process: in that case, the migration Hessian is approximately diagonal in the high-frequency regime. In this work, we report the implementation of the tomographic Hessian (i.e., the second derivative of the DSO misfit with respect to the background model) for the estimation of the background velocity model. We apply the second-order adjoint-state method to obtain the application of the tomographic Hessian on a vector. Then, we use the truncated-Newton method to obtain the update directions by computing approximately the application of the inverse of the tomographic Hessian on the descent direction. We also make a theoretical comparison between the tomography in the IVA and full-waveform inversion contexts. Two numerical examples are used to compare, in terms of geophysical results and computational costs, the truncated-Newton method with different gradient-based optimization methods applied to IVA. A small model allows us to evaluate the eigenvalues of the tomographic Hessian which explains the large damping needed in the truncated-Newton case.","PeriodicalId":55102,"journal":{"name":"Geophysics","volume":"33 2","pages":""},"PeriodicalIF":3.3,"publicationDate":"2023-12-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139005649","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Teng Long, Xuan Qin, Q. Wei, Luanxiao Zhao, Yang Wang, Feng Chen, Michael T. Myers, Yingcai Zheng, De-Hua Han
Understanding the elastic and attenuation signatures of shales is of considerable interest for unconventional reservoir characterization and sealing capacity evaluation for CO2 sequestration and nuclear waste disposal. We conducted laboratory measurements on seven shale samples at seismic frequencies (2100 Hz) to study the effects of clay-bound water (CBW) on their wave dispersion and attenuation signatures. With Nuclear Magnetic Resonance (NMR) and helium porosimeter, the volume of CBW in the shale samples is quantified. The forced-oscillation measurement reveals that Youngs modulus exhibits a continuous dispersion trend from 2 to 100 Hz. The extensional attenuation [Formula: see text] shows a weak frequency- and pressure-dependence on effective pressure ranging from 5 to 35 MPa. The magnitude of extensional attenuation shows a positive correlation with CBW, with an R-square value of 0.89. It is found that 4% of CBW in the rock frame causes roughly a 5% modulus increase from 2 to 100 Hz. We adopt a constant Q model for assigning frequency-dependent bulk and shear moduli to the CBW in the rock physics modeling, which can fit the experimental data of modulus dispersion and attenuation well, indicating that both the bulk and shear moduli of CBW in shales might behave viscoelastically.
{"title":"QUANTIFYING THE INFLUENCE OF CLAY-BOUND WATER ON WAVE DISPERSION AND ATTENUATION SIGNATURES OF SHALE: AN EXPERIMENTAL STUDY","authors":"Teng Long, Xuan Qin, Q. Wei, Luanxiao Zhao, Yang Wang, Feng Chen, Michael T. Myers, Yingcai Zheng, De-Hua Han","doi":"10.1190/geo2022-0646.1","DOIUrl":"https://doi.org/10.1190/geo2022-0646.1","url":null,"abstract":"Understanding the elastic and attenuation signatures of shales is of considerable interest for unconventional reservoir characterization and sealing capacity evaluation for CO2 sequestration and nuclear waste disposal. We conducted laboratory measurements on seven shale samples at seismic frequencies (2100 Hz) to study the effects of clay-bound water (CBW) on their wave dispersion and attenuation signatures. With Nuclear Magnetic Resonance (NMR) and helium porosimeter, the volume of CBW in the shale samples is quantified. The forced-oscillation measurement reveals that Youngs modulus exhibits a continuous dispersion trend from 2 to 100 Hz. The extensional attenuation [Formula: see text] shows a weak frequency- and pressure-dependence on effective pressure ranging from 5 to 35 MPa. The magnitude of extensional attenuation shows a positive correlation with CBW, with an R-square value of 0.89. It is found that 4% of CBW in the rock frame causes roughly a 5% modulus increase from 2 to 100 Hz. We adopt a constant Q model for assigning frequency-dependent bulk and shear moduli to the CBW in the rock physics modeling, which can fit the experimental data of modulus dispersion and attenuation well, indicating that both the bulk and shear moduli of CBW in shales might behave viscoelastically.","PeriodicalId":55102,"journal":{"name":"Geophysics","volume":"48 1","pages":""},"PeriodicalIF":3.3,"publicationDate":"2023-12-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139004216","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Deep learning is prevalent in many fields and attempts have been made to use it in non-bidirectional mapping problems, such as seismic inversion. These non-bidirectional mapping problems have two special issues, that is, insufficient labels and uncertainty of solution. Therefore, current deep learning structures are not suitable for handling this kind of problem. A distinctive knowledge embedded close-looped (KECL) deep learning framework is proposed, tuned to the characteristic of seismic inverse problem. The KECL deep learning framework is composed of a reservoir parameter generator (RPG) and a reservoir parameter updater (RPU). The former half loop is RPG, which takes seismic data as input to generate the initial reservoir parameters. The latter loop is RPU, which takes the initial parameters as input to output synthetic seismic data. Through the training by well data, the difference between field seismic data and synthetic seismic data modelled by the RPU is used to optimize the RPG and RPU. In this deep learning framework, knowledge of the Robinson convolutional model is embedded to address the problem of insufficient labels. Furthermore, semi-supervised learning is used as prior information to reduce the uncertainty of solution. After training, with the help of prior geological information data, the RPU is used to update the initial reservoir parameters generated by RPG for final reservoir parameter inversion. Numerical models and field data are used to test the feasibility of the proposed deep learning framework. We found that intelligent inversion results using data from one well to train the KECL network are consistent with results using multiple well data. Experiments demonstrate that it is adapted to situations in which insufficient well data are available and is able to achieve reliable intelligent inversion.
{"title":"A Knowledge-embedded Close-looped Deep Learning Framework for Intelligent Inversion of Multi-solution Problems","authors":"Fanchang Zhang, Lei Zhu, Xunyong Xu","doi":"10.1190/geo2023-0334.1","DOIUrl":"https://doi.org/10.1190/geo2023-0334.1","url":null,"abstract":"Deep learning is prevalent in many fields and attempts have been made to use it in non-bidirectional mapping problems, such as seismic inversion. These non-bidirectional mapping problems have two special issues, that is, insufficient labels and uncertainty of solution. Therefore, current deep learning structures are not suitable for handling this kind of problem. A distinctive knowledge embedded close-looped (KECL) deep learning framework is proposed, tuned to the characteristic of seismic inverse problem. The KECL deep learning framework is composed of a reservoir parameter generator (RPG) and a reservoir parameter updater (RPU). The former half loop is RPG, which takes seismic data as input to generate the initial reservoir parameters. The latter loop is RPU, which takes the initial parameters as input to output synthetic seismic data. Through the training by well data, the difference between field seismic data and synthetic seismic data modelled by the RPU is used to optimize the RPG and RPU. In this deep learning framework, knowledge of the Robinson convolutional model is embedded to address the problem of insufficient labels. Furthermore, semi-supervised learning is used as prior information to reduce the uncertainty of solution. After training, with the help of prior geological information data, the RPU is used to update the initial reservoir parameters generated by RPG for final reservoir parameter inversion. Numerical models and field data are used to test the feasibility of the proposed deep learning framework. We found that intelligent inversion results using data from one well to train the KECL network are consistent with results using multiple well data. Experiments demonstrate that it is adapted to situations in which insufficient well data are available and is able to achieve reliable intelligent inversion.","PeriodicalId":55102,"journal":{"name":"Geophysics","volume":"180 S459","pages":""},"PeriodicalIF":3.3,"publicationDate":"2023-12-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139006463","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
In general, velocity anisotropy in shale media has been widely observed in lab and field work, which means that disregarding this characteristic can lead to inaccurate imaging locations when data are imaged with reverse time migration (RTM). Wavefields simulated with the conventional coupled pseudo-acoustic wave equation may introduce shear wave noise and this equation is only valid in transversely isotropic media (TI, [Formula: see text]). Certain decoupled qP-wave equations require the use of the pseudo-spectral method, which makes them computationally inefficient. To address these issues, we propose a new pure qP acoustic wave equation based on the acoustic assumption, which can be solved more efficiently using the finite difference method. This equation can also be used in the forward modeling process of RTM in tilted transverse isotropic (TTI) media. First, we perform a Taylor expansion of the root term in the pure qP-wave dispersion relation. This leads to an anisotropic dispersion relation that is decomposed into an elliptical anisotropic background factor and a circular correction factor. Second, we obtain the pure qP-wave equation in TTI media without a pseudo-differential operator. The new equation can be efficiently solved using finite difference methods and can be applied to RTM in TTI media with strong anisotropy. The proposed method shows greater tolerance to numerical errors and is better suited for strong anisotropy, as compared to previously published methods. Numerical examples show the high kinematic and phase accuracy of the proposed pure qP-wave equation along with its stability in TTI media characterized by ([Formula: see text]). By utilizing a sag model and an overthrust TTI model, we demonstrate the efficiency and accuracy of the proposed TTI RTM.
{"title":"Efficient reverse time migration method in TTI media based on a pure pseudo-acoustic wave equation","authors":"Jiale Han, Jianping Huang, Yi Shen, Jidong Yang, X. Mu, Liang Chen","doi":"10.1190/geo2023-0302.1","DOIUrl":"https://doi.org/10.1190/geo2023-0302.1","url":null,"abstract":"In general, velocity anisotropy in shale media has been widely observed in lab and field work, which means that disregarding this characteristic can lead to inaccurate imaging locations when data are imaged with reverse time migration (RTM). Wavefields simulated with the conventional coupled pseudo-acoustic wave equation may introduce shear wave noise and this equation is only valid in transversely isotropic media (TI, [Formula: see text]). Certain decoupled qP-wave equations require the use of the pseudo-spectral method, which makes them computationally inefficient. To address these issues, we propose a new pure qP acoustic wave equation based on the acoustic assumption, which can be solved more efficiently using the finite difference method. This equation can also be used in the forward modeling process of RTM in tilted transverse isotropic (TTI) media. First, we perform a Taylor expansion of the root term in the pure qP-wave dispersion relation. This leads to an anisotropic dispersion relation that is decomposed into an elliptical anisotropic background factor and a circular correction factor. Second, we obtain the pure qP-wave equation in TTI media without a pseudo-differential operator. The new equation can be efficiently solved using finite difference methods and can be applied to RTM in TTI media with strong anisotropy. The proposed method shows greater tolerance to numerical errors and is better suited for strong anisotropy, as compared to previously published methods. Numerical examples show the high kinematic and phase accuracy of the proposed pure qP-wave equation along with its stability in TTI media characterized by ([Formula: see text]). By utilizing a sag model and an overthrust TTI model, we demonstrate the efficiency and accuracy of the proposed TTI RTM.","PeriodicalId":55102,"journal":{"name":"Geophysics","volume":"34 4","pages":""},"PeriodicalIF":3.3,"publicationDate":"2023-12-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139006028","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：