WO2022078514A1 - 一种结构和波速随机布设的三维速度地质建模方法 - Google Patents

一种结构和波速随机布设的三维速度地质建模方法 Download PDF

Info

Publication number
WO2022078514A1
WO2022078514A1 PCT/CN2021/124210 CN2021124210W WO2022078514A1 WO 2022078514 A1 WO2022078514 A1 WO 2022078514A1 CN 2021124210 W CN2021124210 W CN 2021124210W WO 2022078514 A1 WO2022078514 A1 WO 2022078514A1
Authority
WO
WIPO (PCT)
Prior art keywords
model
velocity
dimensional
wave velocity
random
Prior art date
Application number
PCT/CN2021/124210
Other languages
English (en)
French (fr)
Inventor
蒋鹏
任玉晓
王其峰
左志武
许新骥
王凯
陈磊
马川义
曹帅
杨森林
王清扬
孟祥龙
Original Assignee
山东大学
山东高速集团有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 山东大学, 山东高速集团有限公司 filed Critical 山东大学
Priority to US18/031,693 priority Critical patent/US20230384470A1/en
Publication of WO2022078514A1 publication Critical patent/WO2022078514A1/zh

Links

Images

Classifications

    • GPHYSICS
    • G01MEASURING; TESTING
    • G01VGEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
    • G01V1/00Seismology; Seismic or acoustic prospecting or detecting
    • G01V1/28Processing seismic data, e.g. for interpretation or for event detection
    • G01V1/282Application of seismic models, synthetic seismograms
    • GPHYSICS
    • G01MEASURING; TESTING
    • G01VGEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
    • G01V20/00Geomodelling in general
    • GPHYSICS
    • G01MEASURING; TESTING
    • G01VGEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
    • G01V2210/00Details of seismic processing or analysis
    • G01V2210/60Analysis
    • G01V2210/62Physical property of subsurface
    • G01V2210/622Velocity, density or impedance
    • G01V2210/6222Velocity; travel time
    • GPHYSICS
    • G01MEASURING; TESTING
    • G01VGEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
    • G01V2210/00Details of seismic processing or analysis
    • G01V2210/60Analysis
    • G01V2210/64Geostructures, e.g. in 3D data cubes
    • G01V2210/642Faults
    • GPHYSICS
    • G01MEASURING; TESTING
    • G01VGEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
    • G01V2210/00Details of seismic processing or analysis
    • G01V2210/60Analysis
    • G01V2210/66Subsurface modeling

Definitions

  • the present disclosure belongs to the technical field of geophysical exploration, and relates to a three-dimensional velocity geological modeling method with random arrangement of structure and wave velocity.
  • geophysical exploration method is the current mainstream advance geological prediction method.
  • seismic method is widely used in petroleum exploration, coal field and metal mineral exploration, etc., and has broad application prospects.
  • seismic method is also the earliest and most widely used geophysical method. method. The main principle of the seismic method is based on wave field propagation.
  • the geophones located on the ground record the vibration information transmitted to the ground, and the seismic data is processed by imaging or inversion methods to obtain the distribution information of the subsurface medium.
  • the modeling of geological interfaces is an important step in geophysical exploration methods.
  • the method of building a wave velocity model based on deep learning is a popular method at present, and has achieved good results.
  • the current wave velocity inversion based on deep learning only proposes a simple two-dimensional velocity model design method, and there is still a lack of an automated three-dimensional wave velocity model construction method.
  • the deep learning method to construct the tunnel wave velocity model.
  • the current method only refers to the surface method to realize the construction of the two-dimensional geological wave velocity model. Therefore, we propose a set of seismic wave velocity model based on the deep learning method. The whole process of building.
  • the main process is:
  • the deep learning-based wave velocity model construction method is a data-driven algorithm. Its essence is to establish the mapping relationship from the geological wave velocity model to the observation data through a large amount of data. If a large amount of data cannot be obtained, the performance of the algorithm will also Great discount. Therefore, this method puts forward high requirements for data acquisition. Establishing a reasonable model and obtaining data through forward modeling is a commonly used method at present.
  • the existing velocity modeling methods lack tunnel modeling methods, mainly for ground detection modeling. Manual model building methods and two-dimensional batch velocity modeling are used. Modulo methods, these methods have the following problems:
  • the existing batch velocity modeling methods are mainly based on the establishment of two-dimensional velocity models, which are mainly two-dimensional simple layered or fault models, which are inconsistent with the actual geological conditions, and the model complexity is too low, so there is no rock mound model established.
  • the solution is not enough to simulate the actual geological situation, which directly leads to the poor effect of the neural network obtained by using the deep learning method in the face of more complex actual models.
  • Geological models are formed through geological movements and have great randomness. While regional, they also have geological history information. Therefore, if you want to generate a velocity model that can simulate real geological information through functions, it needs to conform to the general laws of strata and also There is enough randomness to avoid model duplication, which is more difficult.
  • the present disclosure proposes a three-dimensional velocity geological modeling method with random layout of structure and wave velocity.
  • the present disclosure aims to solve the problem of lack of training data sets for deep neural networks, and randomly builds three-dimensional velocity models in batches to solve the current problems in three-dimensional velocity models. Gaps in model building. The scale of the data set is increased, and the inversion effect of the deep learning method is effectively increased.
  • the present disclosure adopts the following technical solutions:
  • a three-dimensional velocity geological modeling method with random arrangement of structure and wave velocity comprising the following steps:
  • random wave velocity amplitude is carried out to realize 3D velocity modeling.
  • the specific process of determining the plane layered model according to the base point establishment equation includes:
  • the specific process of complicating the inclined layer of the plane layered model includes: establishing an equation based on the base point to determine the plane layered model, classifying different layer models, and for each point on the basis of the plane model The undulation function is established.
  • the undulation function is established.
  • the slope term for the curved surface is established to further complicate the sloped layer and construct the fold layer model of the curved surface in three-dimensional space.
  • the specific process includes:
  • a fold model is established based on the plane layered model, and the calculation formula is:
  • T i A i A i represent the period and amplitude respectively, and the values are randomly selected;
  • X ref Y ref is the coordinate of the base point, and the value of b 1 b 2 is randomly selected.
  • the specific process of establishing a three-dimensional fault-fold model includes:
  • the specific process of simulating the upward intrusion of salt domes in a geological body of a certain depth includes: the intrusion is fitted by a two-dimensional Gaussian function, the height of the vertical intrusion is defined by the amplitude, the magnitude is determined by the variance, and the strike is determined by The clockwise rotation angle is determined, and an affected area with a certain thickness is set.
  • the maximum intrusion height is at the bottom layer. The closer the affected area is to the surface, the smaller the impact, and the layers above the affected area remain unchanged, completing the addition of salt domes.
  • the formula for establishing a salt dome is:
  • G(X,Y) A exp(-(d 1 (XX ref ) 2 +d 3 (YY ref ) 2 +2d 2 (XX ref )(YY ref )))
  • A represents the height of the vertical invading salt dome, and the size of the salt dome is given by Control, the influence area of the salt dome is set to [A max +5, A max +15] where A max represents the maximum invasion height, in the influence area, the shallower the layer, the smaller the amplitude A of the corresponding Gaussian function, the upper part of the influence area The stratum remains unchanged.
  • a random wave velocity amplitude is performed, and the specific process of realizing three-dimensional velocity modeling includes:
  • the surface fold model is rotated 90° counterclockwise along the central axis to determine the geological conditions in the geological survey report before tunnel excavation, and the fault strike, layer thickness and wave velocity distribution are determined range, set the weight of modeling parameters in different ranges.
  • the three-dimensional velocity geological modeling method with random arrangement of structure and wave velocity further includes acquiring tunnel seismic records, using a convolutional neural network to perform feature extraction processing on the tunnel seismic records, and converting the location information of tunnel geophones into Added on additional channels to complete data encoding.
  • a convolutional neural network to decode the encoded data, and perform multi-objective learning.
  • the convolutional neural network and the fully connected neural network to process the decoding results of the decoder, respectively, the three-dimensional wave velocity model and 3D wave velocity modeling parameters.
  • the loss function used includes the wave velocity model loss function and the modeling parameter loss function, and the velocity model loss function is used to fit the real three-dimensional wave velocity model and network corresponding to the observation data.
  • Modeling a 3D wave velocity model; the modeling parameter loss function is used to fit the real 3D wave velocity model modeling parameters and network modeling parameters corresponding to the observation data.
  • a computer-readable storage medium stores a plurality of instructions, and the instructions are adapted to be loaded by a processor of a terminal device and execute the three-dimensional velocity geological modeling method with random arrangement of structure and wave velocity.
  • a terminal device comprising a processor and a computer-readable storage medium, where the processor is used to implement various instructions; the computer-readable storage medium is used to store a plurality of instructions, the instructions are suitable for being loaded by the processor and executing the described one A 3D velocity geological modeling method for random placement of structures and wave velocities.
  • the present disclosure deals with three-dimensional velocity modeling in geological modeling: considering that no three-dimensional velocity modeling method has been proposed yet, a three-dimensional velocity modeling method is proposed, and a velocity model that conforms to real geology is generated through function simulation. .
  • the present disclosure also proposes a batch modeling method.
  • the MATLAB software is used to write the algorithm, which greatly improves the modeling speed and forms a usable batch modeling method. , which greatly increases the data sets for 3D velocity inversion using deep learning methods, which can effectively improve the accuracy of deep learning methods for velocity inversion.
  • the present disclosure performs function fitting for salt domes, proposes the concept of influence layer, and reasonably simulates salt domes in deep geology, so that the modeling results are more accurate. close to real geology.
  • the present disclosure proposes a three-dimensional velocity parametric modeling method for batch tunnels, and establishes a three-dimensional wave velocity model that is highly consistent with the engineering geological conditions through the survey results in the early stage of tunnel construction.
  • the present disclosure proposes a new method for constructing a three-dimensional wave velocity model of a tunnel in view of the lack of an available deep learning tunneling 3D wave velocity modeling method, which adds the location information encoding of the geophone and the epicenter in the data, and uses a multi-task learning method to carry out network parameters. Optimization to effectively improve modeling accuracy.
  • Fig. 1 is the method flow chart of the present embodiment
  • Fig. 2 is the modeling flow chart of the layered wrinkle model of the present embodiment
  • Fig. 3 is the fault model modeling flow chart of the present embodiment
  • Fig. 4 is the salt dome model modeling flow chart of the present embodiment
  • Figure 5(a)-(c) are schematic diagrams of layered model, fault model and salt dome model respectively;
  • FIG. 6 is a flow chart of tunnel model modeling according to an embodiment of the present invention.
  • FIG. 7 is a flowchart of a deep learning three-dimensional wave velocity modeling method according to an embodiment of the present invention.
  • FIG. 8 is a schematic diagram of a tunnel wrinkle model according to an embodiment of the present invention.
  • FIG. 9 is a schematic diagram of a result of deep learning three-dimensional wave velocity modeling according to an embodiment of the present invention.
  • the 3D velocity modeling method includes the following steps:
  • step S1 a base point is determined in the three-dimensional space, an equation is established according to the base point to determine a plane layered model, and an undulation function is established for each point on the basis of the plane model.
  • the inclined term of further complicates the inclined layer, and establishes the fold layer model of the curved surface in three-dimensional space, as shown in Figure 2;
  • the size of the model in this embodiment is nx ⁇ ny ⁇ nz, and an x-layer layered model is generated.
  • a fold model is established based on the surface layer model, and the calculation formula is:
  • Step S2 based on the established three-dimensional surface fold layer model, establish an equation through the determined random reference point, determine the fault plane passing through the reference point, and then determine the displacement of each point in the global coordinate system through the rotation matrix, and establish a three-dimensional fault and fold model. ,As shown in Figure 3;
  • a velocity model containing salt domes is constructed based on the fold model.
  • the upward intrusion of salt domes is simulated in a geological body of a certain depth.
  • the intrusion is fitted by a two-dimensional Gaussian function, and the height of vertical intrusion is determined by the amplitude Definition, the size is determined by the variance ⁇ x , ⁇ y , and the strike is determined by the clockwise rotation angle ⁇ . Since the salt dome has little influence on the upper layer, a certain thickness of the influence area is set. The maximum intrusion height is at the bottom layer, and the closer the influence area is to the surface smaller, while the layers on the affected area remain the same, completing the addition of the salt dome.
  • G(X,Y) A exp(-(d 1 (XX ref ) 2 +d 3 (YY ref ) 2 +2d 2 (XX ref )(YY ref )))
  • A represents the height of the vertical invading salt dome, and the size of the salt dome is given by Control
  • the influence area of the salt dome is set to [A max +5, A max +15] where A max represents the maximum invasion height, in the influence area, the shallower the layer, the smaller the amplitude A of the corresponding Gaussian function, the upper part of the influence area
  • a max represents the maximum invasion height, in the influence area, the shallower the layer, the smaller the amplitude A of the corresponding Gaussian function, the upper part of the influence area
  • the stratum remains unchanged.
  • Figure 5(a)- Figure 5(c) they are the layered model, the fault model, and the salt dome model, respectively.
  • Step S4 based on the established model, randomly assign wave velocities to each layer.
  • the wave velocity assignment process is:
  • ⁇ v is the random increase speed value, which is 300 ⁇ 500m/s in this example.
  • the automatic construction of the three-dimensional wave velocity model can be realized. Further, the method can be extended to a deep learning-based construction method of the tunnel seismic wave velocity model, including:
  • the building module of the tunnel wave velocity model database is configured to generate a large number of three-dimensional wave velocity models in front of the tunnel based on the on-site geological exploration report to form the tunnel wave velocity model database;
  • Finite Difference Forward Modeling Module Acoustic Wave Equation, which can be expressed in the following form:
  • the elastic wave wave equation is simulated by the finite difference method, and the seismic wave amplitude information is received through the geophone arranged on the tunnel wall for the next step of deep learning modeling.
  • 3D seismic wave velocity building module based on deep learning It is configured to build a deep neural network for tunnel inversion.
  • the input of the network is the seismic observation data in the 3D observation mode, and the output is the predicted wave velocity model.
  • the fold model can be used as the three-dimensional wave velocity model of the tunnel by rotating 90° counterclockwise along the central axis.
  • the deep learning-based tunnel seismic wave velocity modeling method includes the following steps:
  • Step S1 establish a tunnel three-dimensional wave velocity model database
  • this embodiment uses the tunnel three-dimensional wave velocity modeling method proposed above, and randomly generates a three-dimensional velocity model with a size of 100m ⁇ 100m ⁇ 100m, including folds and faults, and the wave velocity model does not exceed three rocks.
  • the seismic wave velocity of the medium in front of the tunnel ranges from 2000m/s to 4000m/s.
  • the wave velocity of the tunnel surrounding rock is consistent with the wave velocity of the first layer of the wave velocity model.
  • Step S2 carry out the finite difference forward modeling, set the grid spacing to 1m, add 50 grids of sponge absorption boundary, and the final wave velocity model size during forward modeling of the wave field is [200, 200, 200].
  • 12 seismic sources and 12 geophones are arranged on the side walls of the tunnel. The hypocenter points are evenly distributed at 5m and 10m from the construction face, divided into three heights with a height difference of 1m. The geophones are located within the range of 20m to 60m from the construction face, with a spacing of 10m and at the same horizontal position.
  • the main frequency of the source is 100Hz
  • the unit time step is 0.1ms
  • the total time step is 2000 time steps. In this way, the tunnel seismic observation data corresponding to each model is generated.
  • a geological model can be built with other such parameters. Geophones and sources can also be replaced.
  • a geological wave velocity model in the database of this example is shown in Figure 8.
  • the batch database established in this example includes a total of 10,000 tunnel seismic wave velocity models, which are randomly divided into training set, verification set and test set according to the ratio of 8:1:1.
  • Step S3 build a tunnel wave velocity model to build a neural network
  • the input of the network is seismic observation data
  • input the network according to Batchsize 8
  • the size of the input data is [8, 1, 12, 2000]
  • the output is the three-dimensional tunnel wave velocity model and the three-dimensional velocity. modeling parameters.
  • the entire neural network consists of an encoder and a decoder.
  • the size of the processed feature vector is [8, 64, 12, 6], and this vector is input into the decoder for decoding.
  • the decoder that adds position coding information consists of 5 convolutional layers followed by two convolutional layers and 3 convolutional layers respectively. Layer fully connected layer, respectively output speed model and modeling parameters.
  • the wave velocity model After obtaining the wave velocity model of the corresponding data, compare the wave velocity model constructed by the neural network with the wave velocity model corresponding to the input seismic data, and use the least squares loss function to obtain the difference between the two wave velocity models. , and perform network backhaul to optimize network parameters.
  • the modeling parameters constructed by the neural network are compared with the modeling parameters corresponding to the input seismic data, and the least squares loss function and the minimum absolute error loss function are used to obtain them respectively. Difference, add the sum of the two loss functions and do a gradient back pass.
  • the weights of the loss functions generated by the two outputs are adjusted. As the training progresses, the weight of the speed model gradually increases.
  • m est represents the model wave velocity output by the deep neural network for tunnel wave velocity modeling
  • m tru represents the actual geological model in the tunnel wave velocity model database.
  • the Adam optimizer is used, the learning rate is kept constant at 5 ⁇ 10 -5 , the BatchSize is 8 in the network training stage, and the total number of iterations is 150 rounds.
  • step S4 the trained tunnel wave velocity construction neural network is used to test the wave velocity construction effect on the test set, and the test result is shown in FIG. 9 .
  • the test results show that the neural network constructed by the three-dimensional wave velocity of the tunnel can better construct the wave velocity in front of the tunnel.
  • the main network parameters and hardware conditions in this embodiment are: the calculation is implemented by NVIDIA RTX3090GPU*4. Use MATLAB and PYTHON for algorithm writing calculations.
  • a computer-readable storage medium stores a plurality of instructions, and the instructions are adapted to be loaded by a processor of a terminal device and execute the three-dimensional velocity geological modeling method with random arrangement of structure and wave velocity.
  • a terminal device comprising a processor and a computer-readable storage medium, where the processor is used to implement various instructions; the computer-readable storage medium is used to store a plurality of instructions, the instructions are suitable for being loaded by the processor and executing the described one A 3D velocity geological modeling method for random placement of structures and wave velocities.
  • embodiments of the present disclosure may be provided as a method, system, or computer program product. Accordingly, the present disclosure may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present disclosure may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein.
  • computer-usable storage media including, but not limited to, disk storage, CD-ROM, optical storage, etc.
  • These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory result in an article of manufacture comprising instruction means, the instructions
  • the apparatus implements the functions specified in the flow or flow of the flowcharts and/or the block or blocks of the block diagrams.

Landscapes

  • Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Engineering & Computer Science (AREA)
  • Remote Sensing (AREA)
  • General Life Sciences & Earth Sciences (AREA)
  • General Physics & Mathematics (AREA)
  • Geophysics (AREA)
  • Acoustics & Sound (AREA)
  • Environmental & Geological Engineering (AREA)
  • Geology (AREA)
  • Geophysics And Detection Of Objects (AREA)

Abstract

一种结构和波速随机布设的三维速度地质建模方法,在三维空间中确定基点,根据基点建立方程确定平面层状模型,将平面层状模型的倾斜层复杂化,构建三维空间上曲面的褶皱层模型;基于三维曲面褶皱层模型,结合随机参考点的断层面,以及各点在全局坐标系中的位移,建立三维断层褶皱模型;基于三维断层褶皱模型,构造含有盐丘的速度模型,在一定深度的地质体中将盐丘向上侵入进行模拟;根据已经设定完成的层状类别,按照设定的速度范围,和每层地质之间的波速差值范围,进行随机的波速幅值,实现三维速度建模;三维速度地质建模方法提高了深度学习方法用于地球物理反演时的模型数据量,提高了深度学习方法反演效果。

Description

一种结构和波速随机布设的三维速度地质建模方法 技术领域
本公开属于地球物理勘探技术领域,涉及一种结构和波速随机布设的三维速度地质建模方法。
背景技术
本部分的陈述仅仅是提供了与本公开相关的背景技术信息,不必然构成在先技术。
在隧道开挖过程中存在突水突泥,塌方等地质灾害,诱发的TBM卡机等工程事故给隧道施工带来重大挑战构造导致的施工风险,造成人员伤亡及经济损失,因此超前地质预报是隧道施工的关键步骤,地球物理勘探方法是目前主流的超前地质预报方法。地震法作为最常用的地球物理勘探方法之一,被广泛应用于石油勘探和煤田、金属矿藏探测等,具有广阔的应用前景,在隧道超前预报中,地震法也是应用最早,最广泛的地球物理方法。地震法的主要原理基于波场传播,将多个检波器布置于地表,通过多次激发人工震源产生波场并在地下介质中传播,当遇到地下介质波阻抗变化产生反射或折射返回地面,位于地面的检波器记录传播至地面的震动信息,通过成像或反演方法处理地震数据,以获取地下介质的分布信息。
其中对地质界面进行建模是地球物理勘探方法的一个重要步骤。据发明人了解,基于深度学习的波速模型构建方法是当前较为热门的方法,并取得了较好的效果。然而目前基于深度学习的波速反演都仅提出简单的二维速度模型设计方法,尚缺少自动化的三维波速模型构建方法,同时对于利用深度学习和探测数据进行隧道前方波速模型构建的方法,尚处于技术空白,存在隧道三维观测系统难编码、三维波速模型参数巨大
然而在隧道环境下,目前使用深度学习方法进行隧道波速模型构建仍是空白,当前方法仅参考地表方式实现了二维地质波速模型构建,由此我们提出了一套基于深度学习方法进行地震波速模型构建的全流程。
主要流程为:
1.确定具体参数,建立大量隧道三维波速模型;
2.进行三维正演模拟,获得地震数据;
3.使用深度神经网络对学习地震数据与隧道模型波速及参数的映射关系,获得波速模型构建网络。
基于深度学习的波速模型构建方法是基于数据驱动的一种算法,本质是通过大量的数据 来建立从地质波速模型到观测数据的映射关系,如果不能获得大量的数据,那么该算法的性能也会大打折扣。因此该方法对于数据的获取提出了很高的要求。建立合理模型,通过正演模拟获得数据是目前常用的一种方法,现有的速度建模方法缺少隧道建模方法主要是针对地面探测建模,采用手工模型建立方法以及二维的批量速度建模方法,这些方法存在以下问题:
第一,建模成本太高;
在传统建模方法中,建立复杂速度模型通常依赖于对地球某个地下区域的地震勘探数据进行地质解释的专业知识,进而构建该地区的速度模型。在模型构建过程中,极大的人工工作量和成本会导致大量的标记数据集无法用于训练深度学习反演网络。
第二,模型复杂度低;
现有的批量速度建模方法主要以二维速度模型建立为主,建立的主要是二维的简单层状或断层模型,与实际的地质情况不符,模型复杂度太低,没有建立岩丘模型的方案,不足以模拟实际地质情况,直接导致使用深度学习方法获得的神经网络在面对较复杂实际模型时效果不佳。
实现速度模型建立的主要困难有以下两个方面:
1)需要合理的算法以及函数来随机建立合理的速度模型;
地质模型是经过地质运动形成的,具有很大的随机性,区域性的同时,也有地质历史信息,因此想要通过函数来生成能够模拟真实地质信息的速度模型需要符合地层的一般规律同时也要有足够的随机性以避免模型重复,存在较大难度。
2)在保证模型复杂度的前提下难以实现快速建模;
建立模型时需要对参数进行随机选择,在合理范围内生成速度模型,需要保证模型复杂度就对于参数选择提出了较高的要求,而进一步实现快速的随机速度建模难度就进一步加大。
在深度学习建模方面,现有方法主要是针对二维情况下的地面以及隧道建模,对于三维隧道深度学习建模方法,目前尚未有人提出。
发明内容
本公开为了解决上述问题,提出了一种结构和波速随机布设的三维速度地质建模方法,本公开针对缺失深度神经网络缺少训练数据集的问题,随机批量建立三维速度模型以解决目前在三维速度模型建立方面的空白。提高了数据集规模,有效增加了深度学习方法的反演效果。
根据一些实施例,本公开采用如下技术方案:
一种结构和波速随机布设的三维速度地质建模方法,包括以下步骤:
在三维空间中确定基点,根据基点建立方程确定平面层状模型,将平面层状模型的倾斜层复杂化,构建三维空间上曲面的褶皱层模型;
基于三维曲面褶皱层模型,结合随机参考点的断层面,以及各点在全局坐标系中的位移,建立三维断层褶皱模型;
基于三维断层褶皱模型,构造含有盐丘的速度模型,在一定深度的地质体中将盐丘向上侵入进行模拟;
根据已经设定完成的层状类别,按照设定的速度范围,和每层地质之间的波速差值范围,进行随机的波速幅值,实现三维速度建模。
作为可选择的实施方式,根据基点建立方程确定平面层状模型的具体过程包括:
根据基点(X ref,Y ref,Z ref)(X ref,Y ref,Z ref)建立方程确定平面层状模型,计算公式:
Figure PCTCN2021124210-appb-000001
其中
Figure PCTCN2021124210-appb-000002
代表倾角。
作为可选择的实施方式,将平面层状模型的倾斜层复杂化的具体过程包括:根据基点建立方程确定平面层状模型,对不同层模型进行类别划分,在平面模型的基础上针对每一个点建立起伏函数,通过调整起伏函数中三角函数的周期和振幅,再建立针对曲面的倾斜项,进一步复杂倾斜层,构建三维空间上曲面的褶皱层模型。
作为进一步限定,具体过程包括:
基于平面层状模型建立褶皱模型,计算公式:
Figure PCTCN2021124210-appb-000003
其中T iA iA i分别代表周期与振幅,值随机选取;
建立针对曲面的倾斜项公式为:D(X,Y)=b 1(X-X ref)+b 2(Y-Y ref)
其中X ref Y ref为基点坐标,b 1 b 2的值随机选取。
作为可选择的实施方式,建立三维断层褶皱模型的具体过程包括:
在褶皱模型中加入断层的公式为:
c 1(X-X ref)+c 2(Y-Y ref)+c 3(Z-Z ref)=0
其中c 1 c 2 c 3是通过旋转矩阵计算得到的:
Figure PCTCN2021124210-appb-000004
旋转矩阵
Figure PCTCN2021124210-appb-000005
其中φθ均随机在[0,2π]中取值,对于随机的d x d y,在全局坐标中D X D Y D Z
Figure PCTCN2021124210-appb-000006
作为可选择的实施方式,在一定深度的地质体中将盐丘向上侵入进行模拟的具体过程包括:侵入通过二维高斯函数拟合,垂直侵入的高度由振幅定义,大小由方差确定,走向由顺时针旋转角确定,设置一定厚度的影响区,最大侵入高度在底层,在影响区内越靠近地表影响越小,而影响区上的层保持不变,完成对盐丘的添加。
作为进一步的限定,建立盐丘的公式为:
G(X,Y)=A exp(-(d 1(X-X ref) 2+d 3(Y-Y ref) 2+2d 2(X-X ref)(Y-Y ref)))
其中
Figure PCTCN2021124210-appb-000007
A代表垂直入侵盐丘的高度,盐丘的大小由
Figure PCTCN2021124210-appb-000008
控制,盐丘的影响区域设置为[A max+5,A max+15]其中A max代表最大入侵高度,在影响区内,层越浅,对应的高斯函数的振幅A越小,影响区域上方的地层保持不变。
作为可选择的实施方式,进行随机的波速幅值,实现三维速度建模的具体过程包括:
根据层数n随机生成n+1个元素的向量V';
对向量V'累加生成向量V 1
取随机速度基准值M,M∈[x 1,x 2],x 1,x 2为速度上下界;
取V 1最后一个元素v end
速度赋值为V=(V 1/v end)·M;
对盐丘随机取速度
Figure PCTCN2021124210-appb-000009
其中Δv为随机增加速度值。
作为可选择的实施方式,在三维速度建模的过程中,将地表褶皱模型沿中心轴逆时针旋转90°,确定隧道开挖前地质勘察报告中的地质情况拟定断层走向、层厚和波速分布的范围,设定建模参数在不同范围的权重。
作为可选择的实施方式,所述结构和波速随机布设的三维速度地质建模方法,还包括获取隧道地震记录,使用卷积神经网络对隧道地震记录进行特征提取处理,并将隧道检波器位置信息增加在额外的通道上,完成数据编码。
作为可选择的实施方式,使用卷积神经网络对编码数据进行解码,并进行多目标学习,通过对解码器解码结果分别使用卷积神经网络和全连接神经网络进行处理,分别得到三维波速模型以及三维波速建模参数。
作为可选择的实施方式,在构建波速模型的过程中,使用的损失函数包括波速模型损失函数及建模参数损失函数,速度模型损失函数用于拟合与观测数据对应的真实三维波速模型与网络建模三维波速模型;建模参数损失函数用于拟合与观测数据对应的真实三维波速模型建模参数与网络建模参数。
一种计算机可读存储介质,其中存储有多条指令,所述指令适于由终端设备的处理器加载并执行所述的一种结构和波速随机布设的三维速度地质建模方法。
一种终端设备,包括处理器和计算机可读存储介质,处理器用于实现各指令;计算机可读存储介质用于存储多条指令,所述指令适于由处理器加载并执行所述的一种结构和波速随机布设的三维速度地质建模方法。
与现有技术相比,本公开的有益效果为:
本公开针对地质建模中的三维速度建模,进行了处理:考虑到目前尚未有提出三维速度建模方法,提出了一种三维速度建模方法,通过函数模拟,生成符合真实地质的速度模型。
同时,本公开还提出了批量建模方法,考虑到原有地质建模方法存在不能批量建模的问题,使用MATLAB软件编写算法,大幅度提高了建模速度,形成了可用的批量建模方法,使得使用深度学习方法进行三维速度反演的数据集大幅度增加,可以有效提高深度学习方法进行速度反演的准确性。
本公开针对传统建模中未进行盐丘模拟的问题,进行了针对盐丘的函数拟合,提出了 影响层的概念,对深层地质中的盐丘进行了合理的模拟,使得建模结果更接近于真实地质。
本公开提出批量隧道三维速度参数化建模方法,通过隧道施工前期勘察结果,建立与工程地质情况吻合度较高的三维波速模型。
本公开针对尚无可用的深度学习隧道三维波速建模方法,提出一种新的隧道三维波速模型构建方法,在数据中增加检波器与震源点位置信息编码,并使用多任务学习方式进行网络参数优化,有效提高建模精度。
附图说明
构成本公开的一部分的说明书附图用来提供对本公开的进一步理解,本公开的示意性实施例及其说明用于解释本公开,并不构成对本公开的不当限定。
图1是本实施例的方法流程图;
图2是本实施例的层状褶皱模型建模流程图;
图3是本实施例的断层模型建模流程图;
图4是本实施例的盐丘模型建模流程图;
图5(a)-(c)分别为层状模型,断层模型,盐丘模型示意图;
图6是本发明实施例的隧道模型建模流程图;
图7是本发明实施例的深度学习三维波速建模方法流程图;
图8是本发明实施例的隧道褶皱模型示意图;
图9是本发明实施例的深度学习三维波速建模结果示意图。
具体实施方式
下面结合附图与实施例对本公开作进一步说明。
应该指出,以下详细说明都是例示性的,旨在对本公开提供进一步的说明。除非另有指明,本文使用的所有技术和科学术语具有与本公开所属技术领域的普通技术人员通常理解的相同含义。
需要注意的是,这里所使用的术语仅是为了描述具体实施方式,而非意图限制根据本公开的示例性实施方式。如在这里所使用的,除非上下文另外明确指出,否则单数形式也意图包括复数形式,此外,还应当理解的是,当在本说明书中使用术语“包含”和/或“包括”时,其指明存在特征、步骤、操作、器件、组件和/或它们的组合。
三维速度建模方法,如图1所示,包括以下步骤:
步骤S1,在三维空间中确定基点,根据基点建立方程确定平面层状模型,在平面模型 的基础上针对每一个点建立起伏函数,通过调整起伏函数中三角函数的周期和振幅,再建立针对曲面的倾斜项,进一步复杂倾斜层,建立三维空间上曲面的褶皱层模型,如图2所示;
本实施例的模型大小为nx×ny×nz,生成x层层状模型。
本实例中根据基点(X ref,Y ref,Z ref)建立方程确定平面层状模型,计算公式:
Figure PCTCN2021124210-appb-000010
本实例中基于面层状模型建立褶皱模型,计算公式:
Figure PCTCN2021124210-appb-000011
本实例中建立针对曲面的倾斜项公式为:D(X,Y)=b 1(X-X ref)+b 2(Y-Y ref)
步骤S2,基于建立好的三维曲面褶皱层模型,通过确定的随机参考点建立方程,确定通过参考点的断层面,再通过旋转矩阵确定各点在全局坐标系中的位移,建立三维断层褶皱模型,如图3所示;
本实施例中对于建立断层模型的公式为:
c 1(X-X ref)+c 2(Y-Y ref)+c 3(Z-Z ref)=0
其中c 1 c 2 c 3是通过旋转矩阵计算得到的:
Figure PCTCN2021124210-appb-000012
旋转矩阵
Figure PCTCN2021124210-appb-000013
其中φθ均随机在[0,2π]中取值
步骤S3,基于褶皱模型构造含有盐丘的速度模型,如图4所示,在一定深度的地质体中将盐丘向上侵入进行模拟,侵入通过二维高斯函数拟合,垂直侵入的高度由振幅定义,大小由方差σ x,σ y确定,走向由顺时针旋转角θ确定,由于盐丘对上层影响小,设置一定厚度的影响区,最大侵入高度在底层,在影响区内越靠近地表影响越小,而影响区上的层保持不变,完成了对盐丘的添加。
本实例中建立盐丘的公式为:
G(X,Y)=A exp(-(d 1(X-X ref) 2+d 3(Y-Y ref) 2+2d 2(X-X ref)(Y-Y ref)))
其中
Figure PCTCN2021124210-appb-000014
A代表垂直入侵盐丘的高度,盐丘的大小由
Figure PCTCN2021124210-appb-000015
控制,盐丘的影响区域设置为[A max+5,A max+15]其中A max代表最大入侵高度,在影响区内,层越浅,对应的高斯函数的振幅A越小,影响区域上方的地层保持不变。如图5(a)-图5(c)所示分别为层状模型,断层模型,盐丘模型。
步骤S4,基于建立好的模型,对各层随机赋值波速。
波速赋值流程为:
1.根据层数n随机生成n+1个元素的向量V'
2.对向量V'累加生成向量V 1
3.取随机速度基准值M,M∈[x 1,x 2](x 1,x 2为速度上下界,本例中取M∈[2000m/s,4000m/s])
4.取V 1最后一个元素v end
5.速度赋值为V=(V 1/v end)·M
6.对盐丘随机取速度
Figure PCTCN2021124210-appb-000016
其中Δv为随机增加速度值,在本例中取300~500m/s。
基于以上方案,可以实现对三维波速模型的自动化构建,进一步的,该方法可以拓展到一种基于深度学习的隧道地震波波速模型构建方法,包括:
隧道波速模型数据库构建模块,被配置为基于现场地质勘探报告大量生成隧道前方三维波速模型,构成隧道波速模型数据库;
有限差分正演模块:声波波动方程,可以表示成如下形式:
Figure PCTCN2021124210-appb-000017
使用有限差分法对弹性波波动方程进行正演模拟,通过隧道壁上布置的检波器接收地震波振幅信息进行下一步深度学习建模。
基于深度学习的三维地震波波速构建模块:被配置为构建隧道反演深度神经网络,网络的输入为三维观测方式下的地震观测数据,输出为预测波速模型。
基于以上地表三维波速模型的构建,其褶皱模型通过沿中心轴逆时针旋转90°,即可作为隧道三维波速模型。
本实施例的隧道模型如图6所示。具体地,如图7所示,基于深度学习的隧道地震波速建模方法,包含以下步骤:
步骤S1,建立隧道三维波速模型数据库,本实施例使用前文提出的隧道三维波速建模方法,随机生成尺寸为100m×100m×100m的三维速度模型,包含褶皱及断层,波速模型不超过三个岩性分界面,隧道前方介质地震波速范围为2000m/s~4000m/s。隧道围岩波速与波速模型第一层介质波速一致。
步骤S2,进行有限差分正演模拟,设定网格间距为1m,增设海绵吸收边界50个网格,最终波场正演模拟时的波速模型大小为[200,200,200]。隧道边墙上分别布置了12个地震震源和12个检波器。震源点平均分布在离施工掌子面5m及10m位置,分三个高度,高差1m。检波器位于离施工掌子面20m到60m范围内,间距为10m,在同一水平位置。震源主频为100Hz,单位时间步为0.1ms,总时间步长为2000时间步。以此生成各模型对应的隧道地震观测数据。
在其他实例中,可以通过其他此参数建立地质模型。检波器以及震源也可以进行替换。
本实例数据库中的一个地质波速模型如图8所示。
本实例建立的批量数据库共包括10000个隧道地震波速模型,按照8:1:1的比例随机分为训练集、验证集和测试集。
步骤S3,构建隧道波速模型构建神经网络,网络的输入为地震观测数据,按照Batchsize=8输入网络,输入数据的大小为[8,1,12,2000],输出为三维隧道波速模型及三维速度建模参数。整个神经网络由一个编码器,以及一个解码器组成。将地震观测信息输入编码器模块,由6层卷积神经网络组成的全局特征编码器,编码器将输入转换为包含波速模型信息的特征向量,输出的特征向量大小为[8,63,12,6],在特征向量中增加一个通道,添加对应地震数据的检波器以及震源信息,即检波器位置信息(x n,y n,z n,x,y,z)其中x n,y n,z n代表第n个震源点坐标,x,y,z代表检波器坐标。处理后的特征向量大小为[8,64,12,6]并将此向量输入解码器中进行解码,增加位置编码信息的解码器由5层卷积层后分别接两层卷积层和3层全连接层,分别输出速度模型以及建模参数。
对于波速模型,在获得对应的数据的波速模型之后,将神经网络构建的波速模型与输入地震数据所对应的波速模型进行比对,使用最小二乘损失函数求取两个波速模型之间的差 异,并进行网络回传,优化网络参数。
对于建模参数,在获得建模参数之后,将神经网络构建的建模参数与输入地震数据所对应的建模参数进行比对,分别使用最小二乘损失函数与最小化绝对误差损失函数求取差异,将两个损失函数之和相加并进行梯度回传。
由于存在两个输出对网络进行优化,因此在网络学习过程中,对两个输出产生的损失函数的权重进行调整,在网络训练初期,建模参数的权重较大,速度模型权重较小,随着训练进行,速度模型权重逐渐增大。
其中最小二乘损失函数可以表示为
L m=∥m est-m tru2
最小化绝对误差损失函数可以表示为:
L m=∥m est-m tru1
其中,m est表示隧道波速建模深度神经网络输出的模型波速,m tru表示隧道波速模型数据库中的实际地质模型。
在S3阶段的网络训练过程中,采用Adam优化器,学习率采用5×10 -5保持不变,在网络训练阶段的BatchSize为8,总迭代轮数为150轮。
步骤S4,将训练好的隧道波速构建神经网络在测试集上测试波速构建效果,测试结果如图9所示。测试结果说明隧道三维波速构建神经网络可以较好的构建出隧道前方波速。
本实施例中主要网络参数和硬件条件为:计算采用NVIDIA RTX3090GPU*4实现。使用MATLAB及PYTHON进行算法编写计算。
还提供以下产品实施例:
一种计算机可读存储介质,其中存储有多条指令,所述指令适于由终端设备的处理器加载并执行所述的一种结构和波速随机布设的三维速度地质建模方法。
一种终端设备,包括处理器和计算机可读存储介质,处理器用于实现各指令;计算机可读存储介质用于存储多条指令,所述指令适于由处理器加载并执行所述的一种结构和波速随机布设的三维速度地质建模方法。
本领域内的技术人员应明白,本公开的实施例可提供为方法、系统、或计算机程序产品。因此,本公开可采用完全硬件实施例、完全软件实施例、或结合软件和硬件方面的实施例的形式。而且,本公开可采用在一个或多个其中包含有计算机可用程序代码的计算机可用 存储介质(包括但不限于磁盘存储器、CD-ROM、光学存储器等)上实施的计算机程序产品的形式。
本公开是参照根据本公开实施例的方法、设备(系统)、和计算机程序产品的流程图和/或方框图来描述的。应理解可由计算机程序指令实现流程图和/或方框图中的每一流程和/或方框、以及流程图和/或方框图中的流程和/或方框的结合。可提供这些计算机程序指令到通用计算机、专用计算机、嵌入式处理机或其他可编程数据处理设备的处理器以产生一个机器,使得通过计算机或其他可编程数据处理设备的处理器执行的指令产生用于实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能的装置。
这些计算机程序指令也可存储在能引导计算机或其他可编程数据处理设备以特定方式工作的计算机可读存储器中,使得存储在该计算机可读存储器中的指令产生包括指令装置的制造品,该指令装置实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能。
这些计算机程序指令也可装载到计算机或其他可编程数据处理设备上,使得在计算机或其他可编程设备上执行一系列操作步骤以产生计算机实现的处理,从而在计算机或其他可编程设备上执行的指令提供用于实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能的步骤。
以上所述仅为本公开的优选实施例而已,并不用于限制本公开,对于本领域的技术人员来说,本公开可以有各种更改和变化。凡在本公开的精神和原则之内,所作的任何修改、等同替换、改进等,均应包含在本公开的保护范围之内。

Claims (13)

  1. 一种结构和波速随机布设的三维速度地质建模方法,其特征是:包括以下步骤:
    在三维空间中确定基点,根据基点建立方程确定平面层状模型,将平面层状模型的倾斜层复杂化,构建三维空间上曲面的褶皱层模型;
    基于三维曲面褶皱层模型,结合随机参考点的断层面,以及各点在全局坐标系中的位移,建立三维断层褶皱模型;
    基于三维断层褶皱模型,构造含有盐丘的速度模型,在一定深度的地质体中将盐丘向上侵入进行模拟;
    根据已经设定完成的层状类别,按照设定的速度范围,和每层地质之间的波速差值范围,进行随机的波速幅值,实现三维速度建模;
    所述将平面层状模型的倾斜层复杂化的具体过程包括:根据基点建立方程确定平面层状模型,对不同层模型进行类别划分,在平面模型的基础上针对每一个点建立起伏函数,通过调整起伏函数中三角函数的周期和振幅,再建立针对曲面的倾斜项,进一步复杂倾斜层,构建三维空间上曲面的褶皱层模型。
  2. 如权利要求1所述的一种结构和波速随机布设的三维速度地质建模方法,其特征是:根据基点建立方程确定平面层状模型的具体过程包括:
    根据基点(X ref,Y ref,Z ref)建立方程确定平面层状模型,计算公式:
    Figure PCTCN2021124210-appb-100001
    其中
    Figure PCTCN2021124210-appb-100002
    代表倾角。
  3. 如权利要求1所述的一种结构和波速随机布设的三维速度地质建模方法,其特征是:具体过程包括:
    基于平面层状模型建立褶皱模型,计算公式:
    Figure PCTCN2021124210-appb-100003
    其中T i A i分别代表周期与振幅,值随机选取;
    建立针对曲面的倾斜项公式为:D(X,Y)=b 1(X-X ref)+b 2(Y-Y ref)
    其中X ref Y ref为基点坐标,b 1 b 2的值随机选取。
  4. 如权利要求1所述的一种结构和波速随机布设的三维速度地质建模方法,其特征是: 建立三维断层褶皱模型的具体过程包括:
    在褶皱模型中加入断层的公式为:
    c 1(X-X ref)+c 2(Y-Y ref)+c 3(Z-Z ref)=0
    其中c 1 c 2 c 3是通过旋转矩阵计算得到的:
    Figure PCTCN2021124210-appb-100004
    旋转矩阵
    Figure PCTCN2021124210-appb-100005
    其中φθ均随机在[0,2π]中取值,对于随机的d x d y,在全局坐标中D X D Y D Z有:
    Figure PCTCN2021124210-appb-100006
  5. 如权利要求1所述的一种结构和波速随机布设的三维速度地质建模方法,其特征是:在一定深度的地质体中将盐丘向上侵入进行模拟的具体过程包括:侵入通过二维高斯函数拟合,垂直侵入的高度由振幅定义,大小由方差确定,走向由顺时针旋转角确定,设置一定厚度的影响区,最大侵入高度在底层,在影响区内越靠近地表影响越小,而影响区上的层保持不变,完成对盐丘的添加。
  6. 如权利要求5所述的一种结构和波速随机布设的三维速度地质建模方法,其特征是:建立盐丘的公式为:
    G(X,Y)=A exp(-(d 1(X-X ref) 2+d 3(Y-Y ref) 2+2d 2(X-X ref)(Y-Y ref)))
    其中
    Figure PCTCN2021124210-appb-100007
    A代表垂直入侵盐丘的高度,盐丘的大小由
    Figure PCTCN2021124210-appb-100008
    控制,盐丘的影响区域设置为[A max+5,A max+15]其中A max代表最大入侵高度,在影响区内,层越浅,对应的高斯函数的振幅A越小,影响区域上方的地层保持不变。
  7. 如权利要求1所述的一种结构和波速随机布设的三维速度地质建模方法,其特征是:进行随机的波速幅值,实现三维速度建模的具体过程包括:
    根据层数n随机生成n+1个元素的向量V';
    对向量V'累加生成向量V 1
    取随机速度基准值M,M∈[x 1,x 2],x 1,x 2为速度上下界;
    取V 1最后一个元素v end
    速度赋值为V=(V 1/v end)·M;
    对盐丘随机取速度V 盐丘盐丘V∈[M,M+Δv],其中Δv为随机增加速度值。
  8. 如权利要求1所述的一种结构和波速随机布设的三维速度地质建模方法,其特征是:在三维速度建模的过程中,将地表褶皱模型沿中心轴逆时针旋转90°,确定隧道开挖前地质勘察报告中的地质情况拟定断层走向、层厚和波速分布的范围,设定建模参数在不同范围的权重。
  9. 如权利要求1所述的一种结构和波速随机布设的三维速度地质建模方法,其特征是:所述结构和波速随机布设的三维速度地质建模方法,还包括获取隧道地震记录,使用卷积神经网络对隧道地震记录进行特征提取处理,并将隧道检波器位置信息增加在额外的通道上,完成数据编码。
  10. 如权利要求9所述的一种结构和波速随机布设的三维速度地质建模方法,其特征是:使用卷积神经网络对编码数据进行解码,并进行多目标学习,通过对解码器解码结果分别使用卷积神经网络和全连接神经网络进行处理,分别得到三维波速模型以及三维波速建模参数。
  11. 如权利要求1所述的一种结构和波速随机布设的三维速度地质建模方法,其特征是:在构建波速模型的过程中,使用的损失函数包括波速模型损失函数及建模参数损失函数,速度模型损失函数用于拟合与观测数据对应的真实三维波速模型与网络建模三维波速模型;建模参数损失函数用于拟合与观测数据对应的真实三维波速模型建模参数与网络建模参数。
  12. 一种计算机可读存储介质,其特征是:其中存储有多条指令,所述指令适于由终端设备的处理器加载并执行权利要求1-11中任一项所述的一种结构和波速随机布设的三维速度地质建模方法。
  13. 一种终端设备,其特征是:包括处理器和计算机可读存储介质,处理器用于实现各指令;计算机可读存储介质用于存储多条指令,所述指令适于由处理器加载并执行权利要求1-11中任一项所述的一种结构和波速随机布设的三维速度地质建模方法。
PCT/CN2021/124210 2020-10-15 2021-10-15 一种结构和波速随机布设的三维速度地质建模方法 WO2022078514A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US18/031,693 US20230384470A1 (en) 2020-10-15 2021-10-15 Method for three-dimensional velocity geological modeling with structures and velocities randomly arranged

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202011105341.7A CN112415582B (zh) 2020-10-15 2020-10-15 一种结构和波速随机布设的三维速度地质建模方法
CN202011105341.7 2020-10-15

Publications (1)

Publication Number Publication Date
WO2022078514A1 true WO2022078514A1 (zh) 2022-04-21

Family

ID=74854582

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/124210 WO2022078514A1 (zh) 2020-10-15 2021-10-15 一种结构和波速随机布设的三维速度地质建模方法

Country Status (3)

Country Link
US (1) US20230384470A1 (zh)
CN (1) CN112415582B (zh)
WO (1) WO2022078514A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2024045285A1 (zh) * 2022-08-31 2024-03-07 中国科学技术大学 基于多源异构数据的地质构造建模方法

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112415582B (zh) * 2020-10-15 2022-07-01 山东大学 一种结构和波速随机布设的三维速度地质建模方法
CN114236095B (zh) * 2021-12-02 2024-03-19 山东高速集团四川乐宜公路有限公司 一种山区高速公路沿线降雨诱发滑坡分区分级预警方法
CN117574518B (zh) * 2024-01-15 2024-04-09 国网湖北省电力有限公司经济技术研究院 一种抽水蓄能电站地下厂房三维地质模型建模方法及系统

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101582173A (zh) * 2009-06-24 2009-11-18 中国石油集团川庆钻探工程有限公司 复杂地质构造块状模型构建方法
WO2014126959A2 (en) * 2013-02-18 2014-08-21 Saudi Arabian Oil Company Systems, methods, and computer-readable media for modeling complex wellbores in field-scale reservoir simulation
CN106934860A (zh) * 2017-01-12 2017-07-07 天津大学 一种基于t样条的三维地质建模方法
CN110297272A (zh) * 2019-07-05 2019-10-01 中南大学 三维速度模型生成方法、装置、设备及存储介质
CN112415582A (zh) * 2020-10-15 2021-02-26 山东大学 一种结构和波速随机布设的三维速度地质建模方法

Family Cites Families (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7523024B2 (en) * 2002-05-17 2009-04-21 Schlumberger Technology Corporation Modeling geologic objects in faulted formations
CN101021568A (zh) * 2007-02-07 2007-08-22 匡斌 基于最大能量旅行时计算的三维积分叠前深度偏移方法
US20130223187A1 (en) * 2011-11-11 2013-08-29 International Geophysical Company, Inc. Geological Structure Contour Modeling and Imaging
CN103646423B (zh) * 2013-12-24 2016-06-08 中国科学院地质与地球物理研究所 一种三维地质建模的方法及装置
CN108693572B (zh) * 2018-03-12 2020-07-03 太原理工大学 一种基于三维建模的地应力场反演方法
CN109242968A (zh) * 2018-08-24 2019-01-18 电子科技大学 一种基于多属性超体素图割的河道三维建模方法
CN110007343B (zh) * 2019-04-18 2021-01-29 中国矿业大学 一种区域地质构造复杂程度的评价方法
CN110632653A (zh) * 2019-08-16 2019-12-31 中国石油天然气股份有限公司 盐构造分析方法、装置及系统
CN111077566B (zh) * 2019-12-10 2021-02-09 成都理工大学 一种基于矩阵分解的双程波叠前深度偏移的方法
CN111562611B (zh) * 2020-04-08 2021-07-06 山东大学 基于波动方程驱动的半监督深度学习地震数据反演方法

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101582173A (zh) * 2009-06-24 2009-11-18 中国石油集团川庆钻探工程有限公司 复杂地质构造块状模型构建方法
WO2014126959A2 (en) * 2013-02-18 2014-08-21 Saudi Arabian Oil Company Systems, methods, and computer-readable media for modeling complex wellbores in field-scale reservoir simulation
CN106934860A (zh) * 2017-01-12 2017-07-07 天津大学 一种基于t样条的三维地质建模方法
CN110297272A (zh) * 2019-07-05 2019-10-01 中南大学 三维速度模型生成方法、装置、设备及存储介质
CN112415582A (zh) * 2020-10-15 2021-02-26 山东大学 一种结构和波速随机布设的三维速度地质建模方法

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
JIANG XIANYI: "The Study of Seismic Acquisition Method Based on Modeling of 2D and 3D Complex Geologic Structure", CHINESE DOCTORAL DISSERTATIONS & MASTER'S THESES FULL-TEXT DATABASE, GRADUATE SCHOOL OF PEKING UNION MEDICAL COLLEGE, CN, no. 4, 15 December 2004 (2004-12-15), CN , XP055921570, ISSN: 1671-6779 *
WU, Q. XU, H. ZOU, X.: "An effective method for 3D geological modeling with multi-source data integration", COMPUTERS & GEOSCIENCES, PERGAMON, AMSTERDAM, NL, vol. 31, no. 1, 1 February 2005 (2005-02-01), AMSTERDAM, NL , pages 35 - 43, XP004684743, ISSN: 0098-3004, DOI: 10.1016/j.cageo.2004.09.005 *

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2024045285A1 (zh) * 2022-08-31 2024-03-07 中国科学技术大学 基于多源异构数据的地质构造建模方法

Also Published As

Publication number Publication date
CN112415582B (zh) 2022-07-01
US20230384470A1 (en) 2023-11-30
CN112415582A (zh) 2021-02-26

Similar Documents

Publication Publication Date Title
WO2022078514A1 (zh) 一种结构和波速随机布设的三维速度地质建模方法
US10331817B1 (en) Systems and methods for modeling faults in the subsurface
CN104914465B (zh) 火山岩裂缝定量预测方法及装置
CN104866653B (zh) 一种获取地下三维密度结构的方法
US10466388B2 (en) System and method for editing geological models by switching between volume-based models and surface-based structural models augmented with stratigraphic fiber bundles
CN113919196A (zh) 储层三维应力场模拟方法、模拟系统、终端、存储介质
US11042676B2 (en) Representing structural uncertainty in a mesh representing a geological environment
CN110031896A (zh) 基于多点地质统计学先验信息的地震随机反演方法及装置
AU2013326843A1 (en) Analyzing microseismic data from a fracture treatment
CN106154334A (zh) 基于网格搜索的井下微地震事件实时反演定位方法
CN103797381A (zh) 绘制地质特征
CN104749617A (zh) 一种多尺度裂缝储层正演模型建立方法
EA008733B1 (ru) Способ сейсморазведочных построений в сложных геологических формациях
WO2016001697A1 (en) Systems and methods for geologic surface reconstruction using implicit functions
CN114676644B (zh) 一种基于机器学习约束的密度突变界面反演方法及系统
CN107817524B (zh) 三维地震层析成像的方法和装置
CN113221228B (zh) 水电站地下洞群围岩力学参数反演方法
CN110705168A (zh) 构造应力场的模拟方法
Wang et al. Multi-source data integration and multi-scale modeling framework for progressive prediction of complex geological interfaces in tunneling
CN105301638A (zh) 一种提取风化层底界面的方法和装置
CN109358364B (zh) 一种建立地下暗河储集体地质模型的方法、装置及系统
CN114371504B (zh) 一种地震震中位置确定方法、装置、设备及可读存储介质
CN113987095A (zh) 一种面向地下非法采矿识别的gis时空数据模型
Hajian et al. Prior applications of neural networks in geophysics
CN108897061A (zh) 一种确定储层砂体比例的方法、装置及系统

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 21879555

Country of ref document: EP

Kind code of ref document: A1

WWE Wipo information: entry into national phase

Ref document number: 18031693

Country of ref document: US

NENP Non-entry into the national phase

Ref country code: DE

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 21/09/2023)