Application of the technique to ellipitical markers deformed by pressure-solution

Compression and extension axes are deduced from quartz deformation lamellae in a quartzite and a graywacke folded into an asymetrical syncline. Deformation lamellae fabrics in the two sandstones are distinctly different. In the graywacke, regardless of bedding orientation or position on the fold, compression axes are normal or nearly normal to the axial planar rough cleavage. Extension axes generally lie in the cleavage plane, parallel to dip. In most quartzite samples, compression axes are parallel or subparallel to bedding, at high angles to the fold axis and extension axes are normal to bedding. Two samples from the very base of the formation indicate compression parallel to the fold axis with extension parallel to bedding, at high angles to the fold axis. One of these two shows both patterns. The lamellae fabric geometry in these two samples suggests the presence of a neutral surface in the quartzite. The lamellae-derived compression and extension axes are in good agreement with the buckling behavior of a viscous layer (quartzite) embedded in a less viscous medium (graywacke and shale below and shale and carbonate above).     

Structural Setting and Kinematic Analysis of the Halaban Region,Eastern Arabian Shield,Saudi Arabia

The vorticity analysis technique was applied to measure the different lithological units,such as schist,metagranite and metavolcano-sedimentary rocks,which are present in the Halaban region.This work aims to interpret the relationship between the different lithologies and the tectonic setting,in order to elucidate the nature of kinematic analysis in the Halaban region.The kinematic analyses were applied to feldspar porphyroclasts,quartz and hornblende for twentysix samples.The kinematic vorticity number (W_m) for deformed rocks in the study area ranged from～0.6 to 0.9.The direction of the long axes for finite strain data (X axes) revealed a WNW trend with shallow dipping.The direction of the short axes for finite strain data (Z axes) were represented by vertical with associated horizontal foliation.The results of the kinematic vorticity and strain analyses are characterized by simple shear with different degrees of deformation in the Halaban region.Furthermore,our finite strain data shows no significant volume change during deformation.The subhorizontal foliation was synchronized with thrusting and deformation.Furthermore,throughout the overlying nappes,the same attitudes of tectonic contacts are observable,the nappes in the orogens being formed from simple shear deformation.     

石英Ti/Ge-P:基于机器学习的矿床类型判别新图解SCIEI北大核心CSCD

石英的微量元素记录了石英生长的物理化学条件。通过微量元素对石英原岩进行分类的研究历史已久,经典工作是在以微量元素为坐标轴的图解上绘制各类型石英的分布范围,以区分石英类型。经典图解包括Rusk(2012)提出用于区分三种矿床类型石英的Al-Ti二元图解,和Schr9n et al.(1988)提出的用于判别不同岩浆岩类型石英的Ti-Al-Ge三元图解。越来越多的研究表明,上述图解不能满足对更多石英类型进行分类的需求,同时也出现与部分已知产状类型的石英微量元素判别相矛盾的情况。随着石英原位微区测试方法的成熟,高精度石英微量元素数据逐渐丰富为系统开展机器学习提供了大数据基础,为石英微量元素研究提供了新的角度和可能性。本研究运用机器学习分类方法对石英微量元素进行精确数学分析,提出Ti/Ge-P图解为石英成因研究提出新的地球化学指标。本文同时测试了六种经典机器学习分类算法,提高Ti/Ge-P图解在石英成因分类研究上的精度。此Ti/Ge-P图解适用于多种矿床研究,包括但不局限于斑岩型矿床、矽卡岩型矿床、浅成低温热液型矿床、卡林型矿床以及造山型矿床中的石英。这项工作是大数据技术与机器学习技术在地球化学研究中的积极探索。     

基于PCA-GEP算法的边坡稳定性预测

提出一种基于主成分分析的基因表达式程序设计算法,并将其用于边坡稳定性预测。该算法先采用主成分分析法对样本数据进行预处理,有效地减少预测模型的输入量,消除输入数据间的相关性,再将得到的新样本数据输入基因表达式,构建边坡稳定性的预测模型。利用该预测模型对82个危险圆弧破坏边坡实例中的71个实例进行学习,对另外11个实例进行预测,取得了较好的效果。在保留传统的以误差值作为评判模型优劣标准的同时,引入AIC信息准则法,分别对v-SVR算法和GA-BP网络算法和PCA-GEP算法三种预测模型进行比较分析,结果表明,运用该算法可以获得更优的预测模型,其预测结果比v-SVR算法和GA-BP网络等其他算法得到的结果具有更高的预测精度。工程实例计算表明,该方法是合理、可行的。     

Lattice-preferred orientation of olivine found in diamond-bearing garnet peridotites in Finsch,South Africa and implications for seismic anisotropy

Seismic anisotropy in the upper mantle provides important constraints on mantle dynamics, continental evolution and global tectonics and is believed to be produced by the flow-induced lattice-preferred orientation (LPO) of olivine. Recent experimental studies at high pressure and temperature have suggested that the LPO of olivine is affected by pressure in addition to water and stress. However, there has been no report yet for the pressure-induced LPO of natural olivine because samples from the deep upper mantle are rare and often unsuitable for study due to ambiguous foliation and lineation. Here we show evidence of the pressure-induced LPO of natural olivine in diamond-bearing garnet peridotites from Finsch, South Africa. We found that the [010] axes of olivine are aligned subnormal to foliation and that the [001] axes are aligned subparallel to lineation, which is known as B-type LPO of olivine. The equilibrium pressure of the samples, as estimated using geobarometer, was greater than 4 GPa, indicating that the samples originated from a depth greater than ∼120 km. In addition, FTIR spectroscopy of the olivine showed that the samples are dry, with a water content of less than 90 ± 20 ppm H/Si (5.5 ± 1.2 ppm wt. H₂O). These data suggest that the samples are the first natural examples of olivine displaying B-type LPOs produced due to high pressure under dry condition. Our data indicate that the trench-parallel seismic anisotropy observed in many subduction zones in and below subducting slabs at depths greater than ∼90 km under dry condition may be attributed to the pressure-induced olivine fabrics (B-type LPO) and may be interpreted as the entrainment of the sub-lithospheric mantle in the direction of subduction rather than anomalous trench-parallel flow.     

3种经典机器学习算法在火山岩测井岩性识别中的对比

岩性识别一直是火山岩油气勘探中的重要课题,基于测井数据的岩性识别也逐渐成为火山岩研究的需要,大数据背景下的机器学习算法为测井岩性识别提供了一个新方向。为提高某研究区火山岩岩性识别符合率,本文采用K近邻、支持向量机和自适应增强3种经典机器学习算法,对研究区内的粗面岩、非致密粗面岩、辉绿岩、辉长岩、玄武岩和非致密玄武岩等6类中基性火山岩进行岩性识别。从常规测井系列中优选对研究区岩性敏感的自然伽马、声波时差、补偿中子、深侧向电阻率和补偿密度等5种测井参数作为岩性识别模型的输入向量,从研究区内5口有岩心样品或薄片鉴定资料的目标层中选取测井数据点1 440个,其中960个作为训练样本,其余480个作为测试样本。以识别符合率和时间作为评价指标,对3种算法的识别结果进行对比分析,实验表明：自适应增强算法的分类准确率最高,6类岩性平均识别符合率达到82.10%;支持向量机算法表现良好,平均识别符合率为81.04%;K近邻算法平均识别符合率为76.04%。     

Characterization of U.S. lignites by pyrolysis mass spectrometry and multivariate analysis

Sixteen Texas (Gulf Province) lignite samples and six Montana and Wyoming (Northern Great Plains Province) lignitic coals were obtained from the Pennsylvania State University Coal Bank and analyzed in triplicate by pyrolysis mass spectrometry (Py-MS) using Curie-point pyrolysis (equilibrium temp. 610°C) in combination with low-voltage (12 eV) electron ionization. The spectra obtained were evaluated by means of factor analysis, followed by discriminant analysis using only factors with eigenvalue ⩾ 1 and regarding each set of triplicate spectra as a separate category. The discriminant analysis results showed a definite separation between lignites from the two provinces as well as some clustering of samples from the same seam field or region. Six additional lignite samples obtained from an independent source and representing other regions of the Gulf Province were found to cluster with the Texas lignite samples when treated as “unknowns” in the discriminant analysis procedure.Chemical interpretation of the spectral differences underlying the clustering behavior of the lignite samples in the discriminant analysis procedure was attempted using a newly developed, unsupervised numerical extraction method for chemical components in complex spectra. This procedure, the Variance Diagram (VARDIA) technique, revealed the presence of six major chemical component axes. Examination of the spectral patterns corresponding to these component axes showed a softwood lignin-like component (high in Northern Great Plains lignitic coals) and an alipathic (algal?) hydrocarbon component (high in Gulf lignites) to be primarily responsible for the differences between the two provinces. In addition, two biomarker patterns, namely a terpenoid resin-like component and an unknown component, were shown to be highly characteristic for the Northern Great Plains and Gulf Province samples, respectively. Two other component axes were found to consist largely of sulfur-containing ion series, one of which appeared to represent an obvious marine influence on the South Texas region of the Gulf Province.Furthermore, a set of seventeen conventional coal parameters, including petrographic, ultimate and proximate analysis data as well as sulfur content, calorific value and vitrinite reflectance, obtained from the Pennsylvania State University coal data bank on all twenty-two samples, was also submitted to factor analysis. Comparison of the scores of the first two factors from this set with the scores of the first two discriminant functions of the Py-MS data set revealed an overall similarity in clustering behavior of the samples from the two provinces. Subsequently, canomical variate analysis was used to rotate both the conventional and Py-MS data sets to a common set of vectors describing the correlating (“overlapping”) portions of both data sets. Examination of the first two pairs of canonical variate functions revealed strong correlations between the conventional data and the Py-MS data, e.g., with regard to alipathic vs. aromatic or hydrocarbon vs. heteroatomic tendencies, as well as sulfur containing moieties. This enabled a final, tentative synthesis of all the data into several highly simplified schemes relating compositional aspects to depositional environments.     

Deformation microstructures of olivine and chlorite in chlorite peridotites from Almklovdalen in the Western Gneiss Region,southwest Norway,and implications for seismic anisotropy

Chlorite peridotites from Almklovdalen in southwest Norway were studied to understand the deformation processes and seismic anisotropy in the upper mantle. The lattice preferred orientation (LPO) of olivine and chlorite was determined using electron backscattered diffraction (EBSD)/scanning electron microscopy. A sample with abundant garnet showed [100] axes of olivine aligned sub-parallel to lineation, and [010] axes aligned subnormal to foliation: A-type LPO. Samples rich in chlorite showed different olivine LPOs. Two samples showed [001] axes aligned sub-parallel to lineation, and [010] axes aligned subnormal to foliation: B-type LPO. Two other samples showed [100] axes aligned sub-parallel to lineation, and [001] axes aligned subnormal to foliation: E-type LPO. Chlorite showed a strong LPO characterized by [001] axes aligned subnormal to foliation with a weak girdle subnormal to lineation. Fourier transform infrared (FTIR) spectroscopy of the specimens revealed that the olivines with A-type LPO contain a small amount (170 ppm H/Si) of water. In contrast, the olivines with B-type LPOs contain a large amount (340 ppm H/Si) of water.

The seismic anisotropy of the olivine and chlorite was calculated. Olivine showed Vp anisotropy of up to 3.8% and a maximum Vs anisotropy of up to 2.7%. However, the chlorite showed a much stronger Vp anisotropy, up to 21.1%, and a maximum Vs anisotropy of up to 31.7%. A sample with a mixture of 25% of olivine and 75% of chlorite can produce a Vp anisotropy of 14.2% and a maximum Vs anisotropy of 22.9%. Because chlorite has a wide stability field at high pressure and high temperature in the subduction zone, the strong LPO of chlorite can be a source of the observed trench-normal or trench-parallel seismic anisotropy in the mantle wedge as well as in subducting slabs depending on the dipping angle of slab in a subduction zone where chlorite is stable.     

Spiral and staircase inclusion trail axes within garnet and staurolite porphyroblasts from schists of the Bolton Syncline, Connecticut: timing of porphyroblast growth and the effects of fold development

In the Littleton Formation, garnet porphyroblasts preserve three generations of growth that occurred before formation of the Bolton Syncline. Inclusion trails of foliations overgrown by these porphyroblasts are always truncated by the matrix foliation suggesting that garnet growth predated the matrix foliation. In contrast, many staurolite porphyroblasts grew synchronously with formation of the Bolton Syncline. However, local rim overgrowths of the matrix foliation suggest that some staurolite porphyroblasts continued to grow after development of the fold during younger crenulation producing deformations. The axes of curvature or intersection of foliations defined by inclusion trails inside the garnet porphyroblasts lie oblique to the axial plane of the Bolton Syncline but do not change orientation across it. This suggests the garnets were not rotated during the subsequent deformation associated with fold development or during even younger crenulation events. Three samples also contain a different set of axes defined by curvature of inclusion trails in the cores of garnet porphyroblasts suggesting a protracted history of garnet growth. Foliation intersection axes in staurolite porphyroblasts are consistently orientated close to the trend of the axial plane of the Bolton Syncline on both limbs of the fold. In contrast, axes defined by curvature or intersection of foliations in the rims of staurolite porphyroblasts in two samples exhibit a different trend. This phase of staurolite growth is associated with a crenulation producing deformation that postdated formation of the Bolton Syncline. Measurement of foliation intersection axes defined by inclusion trails in both garnet and staurolite porphyroblasts has enabled the timing of growth relative to one another and to the development of the Bolton Syncline to be distinguished in rocks where other approaches have not been successful. Consistent orientation of foliation intersection axes across a range of younger structures suggests that the porphyroblasts did not rotate relative to geographical coordinates during subsequent ductile deformation. Foliation intersection axes in porphyroblasts are thus useful for correlating phases of porphyroblastic growth in this region.     

The Effect of finite strain and deformation history of HALABAN area,eastern Arabian shield,Saudi Arabia

The present study aims to evaluate a relationship between the mineralogy and structural analysis in the Halaban area and to document the tectonic evolution of Halaban and Al Amar faults. The collected samples were taken from deformed granitiods rocks (such as granite, gneisses and tonalite), metasedimentary, metavolcanic, metagabbro and carbonate rocks are trend to NE-SW with low dip angle in the Halaban area. These samples were 8 from granite, 14 metagabbro, 6 metavolcanics, 5 tonalite, 6 metasedimentary, 10 gneisses and 8 carbonate rocks. Our results are described for the different axial ratios of deformed rocks as the following: XZ sections range from 1.10 to 4.60 in the Fry method and range from 1.70 to 2.71 in the Rf/? method. YZ sections range from 1.10 to 3.34 in the Fry method and range from 1.62 to 2.63 in the Rf/Phi method. In addition, XY sections range from 1 to 3.51 in the Fry method and range from 1 to 1.27 in the Rf/? method for deformed granite rocks, metasedimentry rocks, and metagabbro. The stretch axes for measured samples in the X direction axes (S_X) variety from 1.06 to 2.53 in the Fry method and vary from 1.20 to 1.45 in the Rf/? method. The values of the Y direction axes (S_Y) vary from 0.72 to 1.43 in the Fry method, which indicates contraction and extension in this direction and vary from 1.13 to 1.37 in the Rf/? method which indicates extension in this direction. Furthermore, the Z direction axes (SZ) varies from 0.09 to 0.89 in the Fry method and from 0.52 to 0.71 in the Rf/? method. The stretches axes in the Z direction (SZ) show a vertical shortening about 11% to 91% in the Fry method and show vertical shortening about 29% to 48% in the Rf/? method. The studied rock units are generally affected by brittle-ductile shear zones, which are sub-parallel to parallel NW or NNW trend. It assumed that different rock types of have similar deformation behavior. Based on these results, it is concluded that the finite strain is accumulated during the metamorphism after that was started the deformation by thrusting activity. The contacts between the different rock types were deformed during thrusting under semi-brittle to ductile deformation conditions by simple shear. A component of vertical shortening is also involved causing subhorizontal foliation in the Halaban area.     

The Geothermal Pulse of Hot-Spot and Its Dispersion Process Over China-India-Burma and Change of Short-Term Climate

THE GEOTHERMAL PULSE OF HOT-SPOT AND ITS DISPERSION PROCESS OVER CHINA-INDIA-BURMA AND CHANGE OF SHORT-TERM CLIMATE     

梯级水库优化调度的有后效性动态规划模型及应用

建立了梯级水库在洪水期间发电调度的优化模型,由于含有河道洪水演进方程,该模型成为一类有后效性的动态规划模型.提出了两种新的解法——多维动态规划近似解法与有后效性动态规划逐次逼近算法.实例研究表明:这两种解法可行,结果合理,特别是逐次逼近算法计算更快速,是求解这类问题的有效的方法.     

矿物光率体主轴-晶轴-弹性张量的数学转换和计算机实现的方法

以往是通过手工转图由光率体主轴获取晶轴和应力主轴的,把手工操作变为计算机运算和成图,是一个重要的附加课题．主要介绍了计算机实现轴与轴以及坐标系和投影面之间的各种转换方法,具体步骤有：（１）把费氏台下读数转换为地理坐标系下的数据;（２）已知任意两个光率体主轴求第三个光率体主轴;（３）投影面的旋转;（４）由光率体主轴获取晶轴;（５）把弹性张量从与晶轴相关的坐标系转到地理坐标系中．由此编制的处理程序和绘图软件已在应用中．     

Till kinematics in the Stargard drumlin field,NW Poland constrained by microstructural proxies

Deciphering the relationships between streamlined subglacial landforms and deposits therein helps to constrain the dynamics of past ice sheets. Here we present microstructural analyses of till from the largest drumlin field in the North European Lowland generated by a major palaeo-ice stream of the Scandinavian Ice Sheet. Data derived from thin sections and X-ray microtomography (μCT) reveal multiple microstructures including microshears, circular structures, grain stacks and crushed grains. Long axes of micro-clasts measured in thin sections are directionally clustered and have low, down-ice-oriented dip angles. Some samples reveal subordinate clustering oriented at high angles to the main direction. The μCT data show low clustering strengths of clasts and no systematic fabric pattern in the entire set of samples. There is no correspondence between the clast orientations and the directions of drumlin long axes, suggesting that the till is antecedent in relation to the drumlin-shaping process. The observations indicate spatially and temporally variable, shallow-seated brittle and ductile deformation controlled by porewater pressure fluctuations and generated time-transgressively parallel to till accretion. Consistently with previous studies, the data suggest that the Stargard drumlins are erosional remnants of a pre-existing terrain shaped by material removal from the inter-drumlin areas.     

大型抽水试验最优化设计研究

抽水试验最优化设计,是指针对一个具体的地下水系统优选出一个抽水试验设计,它使抽水试验的费用最小化,而试验所获资料能满足识别参数所要求的可靠性,它可以概化为一个非线性的混合整数规划问题,其决策变量是确定抽水井和观测井的位置及井数、抽水量和观测频率等,采用搜索算法求解了这一优化问题,得到了理论最优解,并进行了实例演算。     

Environmental gradients in carbonate sediments and rocks detected by correspondence analysis: examples from the Recent of Norway and the Dinantian of southwest England

Continuous sedimentary gradients are only crudely expressed by standard facies and microfacies methods which are more appropriate to situations where changes occur in relatively discrete steps. In carbonate sediments and rocks, continuous gradients are often represented by the arrangement of component grain types in a relay, that is, a systematic shifting of the relative importance of the components. Subdivision of such relays into (micro)facies can only be arbitrary. Correspondence analysis is shown to be useful for detecting, isolating and describing relays. Particular use is made of the arch effect in which samples and components from data sets with a strong unidimensional structure (a relay) plot in the form of an arch in the plane of the first two factor axes. A relay index, indicating the position of samples in the relay, can be extracted from the analysis and plotted on maps and stratigraphic logs to reveal details of the sedimentary gradient in areal and/or stratigraphic context. Examples are given from: (i) Recent shallow-marine carbonate sediments from northern Norway, illustrating a relatively simple depositional setting where surface sediments are viewed in plan; and (ii) Lower Carboniferous carbonates of southwest England, representing a more complex regional study of a particular stratigraphic interval viewed in cross-section. In both examples the relays can be related to identifiable environmental gradients.     

基于粒子群-高斯过程回归耦合算法的滑坡位移时序分析预测智能模型

高斯过程回归(GPR)学习机有着容易实现、超参数自适应获取及预测输出具有概率意义等优点。通常采用共轭梯度法获取GPR超参数,但其存在优化效果对初值依赖性太强,迭代次数难以确定,易陷入局部最优的缺点。改用粒子群优化 (PSO)算法进行最优超参数搜索,形成粒子群-高斯过程回归耦合算法(PSO-GPR)。将该算法引入三峡永久船闸高边坡、卧龙寺新滑坡、链子崖滑坡3个不同的典型滑坡变形时序分析中,对每个滑坡分别采用稳态核及一种新式神经网络（NN）、平方指数（SE）、有理二次型（RQ）3种单一核函数进行外推预报测试。工程应用表明,基于3种不同单一核函数的粒子群-高斯过程回归算法（PSO-GPR）均能完全适应不同滑坡时序分析,其中以NN核函数外推预测效果最佳,平均相对误差分别为6.37%、7.62%、1.07%,从而改善了在进行不同滑坡变形时序分析时采用单一核函数的核机器外推能力存在较大差异性的问题,提高了单一核函数对不同数据类型的兼容性     

初始群体飘移遗传算法用于解混和整数非线性规划问题

目前对混合整数非线性规划问题做的工作甚少，这里在研究了混舍整数非线性规划和遗传算法各自特点的基础上，提出初始群体飘移思想和变尺度整型细密网格技术，构造了一种求解混合整数非线性规划的新的遗传算法～初始群体飘移遗传算法。经理论分析和数值试验表明：该算法对大范围、多峰、非光滑非线性规划问题有较好的全局求解能力，在解的精度、稳定性和收敛速度方面均优于一般的算法。     

Deformation fabrics of glaucophane schists and implications for seismic anisotropy: the importance of lattice preferred orientation of phengite

ABSTRACT

Strong seismic anisotropy is observed in many subduction zones. This effect is attributed partly to subducting oceanic crust that is transformed into blueschist facies rocks. Because blueschist facies constituents such as glaucophane, epidote, and phengite show strong anisotropic elasticity, seismic anisotropy in subducting oceanic crust can be attributed to the lattice preferred orientation (LPO) of these minerals. We studied the deformation fabrics and seismic properties of phengite-rich epidote–glaucophane schists from the Franciscan Complex of Ring Mountain, California. The samples are composed mainly of glaucophane, epidote, and phengite. Some samples contain abundant phengite, the maximum being 40%. The LPOs of glaucophane showed that the [001] axes are aligned subparallel to lineation, and both (110) poles and [100] axes are aligned subnormal to foliation. The epidote [001] axes are aligned subnormal to foliation, with both (110) and (010) poles aligned subparallel to lineation. The LPOs of phengite are characterized by the maxima of [001] axes subnormal to foliation, and both (110) and (010) poles and [100] axes are aligned in a girdle subparallel to foliation. The phengite showed substantially strong seismic anisotropy (AV_P = 42%, max.AV_S = 37%). The glaucophane schist with abundant phengite showed significantly stronger seismic anisotropy (AV_P = 30%, max.AV_S = 23%) than the epidote–glaucophane schist (AV_P = 13%, max.AV_S = 9%). When the subduction angle of phengite-rich glaucophane schist is considered, the polarization direction of the fast S-waves for vertically propagating S-waves changed to a nearly trench-parallel direction for the subduction angle of 45?60°, and the S-wave anisotropy became stronger for vertically propagating S-waves with increasing subduction angles. Our data showed that phengite-rich blueschist facies rock can therefore contribute to the strong trench-parallel seismic anisotropy occurring at the subducting oceanic crust and at the slab–mantle interface in many subduction zones.     

A novel type of neural networks for feature engineering of geological data:Case studies of coal and gas hydrate-bearing sediments

The nature of the measured data varies among different disciplines of geosciences.In rock engineering,features of data play a leading role in determining the feasible methods of its proper manipulation.The present study focuses on resolving one of the major deficiencies of conventional neural networks(NNs) in dealing with rock engineering data.Herein,since the samples are obtained from hundreds of meters below the surface with the utmost difficulty,the number of samples is always limited.Meanwhile,the experimental analysis of these samples may result in many repetitive values and 0 s.However,conventional neural networks are incapable of making robust models in the presence of such data.On the other hand,these networks strongly depend on the initial weights and bias values for making reliable predictions.With this in mind,the current research introduces a novel kind of neural network processing framework for the geological that does not suffer from the limitations of the conventional NNs.The introduced single-data-based feature engineering network extracts all the information wrapped in every single data point without being affected by the other points.This method,being completely different from the conventional NNs,re-arranges all the basic elements of the neuron model into a new structure.Therefore,its mathematical calculations were performed from the very beginning.Moreover,the corresponding programming codes were developed in MATLAB and Python since they could not be found in any common programming software at the time being.This new kind of network was first evaluated through computer-based simulations of rock cracks in the 3 DEC environment.After the model's reliability was confirmed,it was adopted in two case studies for estimating respectively tensile strength and shear strength of real rock samples.These samples were coal core samples from the Southern Qinshui Basin of China,and gas hydrate-bearing sediment(GHBS) samples from the Nankai Trough of Japan.The coal samples used in the experiments underwent nuclear magnetic resonance(NMR) measurements,and Scanning Electron Microscopy(SEM) imaging to investigate their original micro and macro fractures.Once done with these experiments,measurement of the rock mechanical properties,including tensile strength,was performed using a rock mechanical test system.However,the shear strength of GHBS samples was acquired through triaxial and direct shear tests.According to the obtained result,the new network structure outperformed the conventional neural networks in both cases of simulation-based and case study estimations of the tensile and shear strength.Even though the proposed approach of the current study originally aimed at resolving the issue of having a limited dataset,its unique properties would also be applied to larger datasets from other subsurface measurements.