CN113156538A - Magnetotelluric-seismic wave first arrival time joint inversion method - Google Patents
Magnetotelluric-seismic wave first arrival time joint inversion method Download PDFInfo
- Publication number
- CN113156538A CN113156538A CN202110436707.7A CN202110436707A CN113156538A CN 113156538 A CN113156538 A CN 113156538A CN 202110436707 A CN202110436707 A CN 202110436707A CN 113156538 A CN113156538 A CN 113156538A
- Authority
- CN
- China
- Prior art keywords
- image
- resistivity
- inversion
- magnetotelluric
- model
- Prior art date
- Legal status (The legal status 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 status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V11/00—Prospecting or detecting by methods combining techniques covered by two or more of main groups G01V1/00 - G01V9/00
- G01V11/007—Prospecting or detecting by methods combining techniques covered by two or more of main groups G01V1/00 - G01V9/00 using the seismo-electric effect
Landscapes
- Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Life Sciences & Earth Sciences (AREA)
- General Physics & Mathematics (AREA)
- Geophysics (AREA)
- Geophysics And Detection Of Objects (AREA)
Abstract
A magnetotelluric-seismic wave first arrival time joint inversion method comprises the following steps: constructing a training set, wherein the training set comprises a resistivity image and a speed image; respectively carrying out combined training on the resistivity image and the velocity image through a neural network to obtain nonlinear mapping from the inversion result resistivity and the velocity image to the resistivity and the velocity image of the real model; and acquiring a reference model for magnetotelluric-seismic wave first arrival time inversion according to the nonlinear mapping from the inversion result to the real model, and updating the resistivity and the speed simultaneously under the constraint of the reference model so as to perform magnetotelluric-seismic wave first arrival time joint inversion. This application has following advantage: 1. the resistivity and speed model obtained by inversion has better structural similarity and more accurate boundary; 2. jumping the target function out of a local minimum value; 3. has high generalization ability. The method and the device realize the joint inversion of the magnetotelluric-seismic wave in the first arrival time.
Description
Technical Field
The method relates to the technical field of geophysical inversion imaging, in particular to a deep learning magnetotelluric-seismic wave first arrival time joint inversion method based on attribute fusion.
Background
Geological models have a variety of geophysical properties and may be estimated using different detection methods in order to solve for subsurface structures. For example, magnetotelluric data may be used to infer subsurface resistivity distributions, seismic first arrival data may be used to infer subsurface velocity fields. Due to the difference of the two methods in the resolution and sensitivity of the underground structure, the magnetotelluric method and the seismic wave first arrival method are respectively used for carrying out inversion on the same measuring area, and the obtained results may be different. This presents many challenges to geophysical data interpretation.
Because the electromagnetic and seismic wave data contain complementary information of the underground structure, compared with single inversion, the joint inversion of the two data can further limit the model search space, so that the resistivity and the speed obtained by inversion have more similar structures, and the cross validation of the two models is completed. There are two main frameworks for current joint inversion: one based on petrophysical relationships and the other based on structural similarity. In the first framework, the correlation between two attributes is established by empirical equations. We can find a direct relationship between velocity and resistivity, or convert resistivity and velocity into porosity and fluid saturation by Archie or the Waxman-Smits equation and the Gassmann equation. Much work has been done based on this approach and has shown good performance. However, the construction of the relationship depends on predefined coefficients in the empirical equation, which need to be selected from different situations. Furthermore, this approach does not take into account the statistical correlation between velocity and resistivity. The second framework assumes that the inverted models of the different survey methods should have a common boundary, and by applying structural similarity constraints to the objective function, the velocity to be inverted and the resistivity structure are made to be consistent. Such constraints include cross-gradients, interactive regularization, and the like. For a cross-gradient, the boundary of velocity and resistivity (if present) will point to the same location and direction, but it does not force the same structure everywhere. For interactive regularization, the boundaries of one physical model will be imposed according to the boundaries of one of the physical models. Joint inversion based on structural similarity cannot utilize a priori resistivity-velocity statistical rule to constrain the inversion process. In addition, the two frameworks cannot directly establish the connection between the model of the inversion output and the real model, and cannot help the inversion to jump out the local minimum value.
In the prior art, joint inversion is performed based on a mapping method:
1. the mapping method uses a neural network, the mapping from the resistivity image of a real model to a speed image or from the speed image to the resistivity image is trained, the corresponding relation between the structure of two modes and the physical numerical value is mainly fused, wherein, the input and the output of the neural network can only be one of the resistivity or the speed, only the real model is needed when a training set is constructed, the independent inversion is not needed, and two neural networks are needed to be trained;
2. the mapping method adopts a single-channel CNN, and input-output is resistivity-speed or speed-resistivity;
3. the mapping method directly relates to the mutual mapping relation of resistivity and speed, and the resistivity and speed images need to be alternately updated every iteration; after a resistivity (velocity) image is generated by each iteration through the mapping method, the image is placed into one of the two trained neural networks to obtain a reference velocity (resistivity) image which is used as a reference for the next round of independent velocity (resistivity) iteration;
the method is a novel method for assisting magnetotelluric and seismic wave first arrival time joint inversion by utilizing deep learning. The nonlinear mapping relation between an inversion result and a real model is learned by utilizing a deep neural network, so that the resistivity and speed model obtained by inversion has higher structural similarity and physical property corresponding relation, the structure of the real model is more accurately described and is closer to a global minimum value, and the geophysical interpretation is better facilitated.
Disclosure of Invention
The present application is directed to solving, at least to some extent, one of the technical problems in the related art.
Therefore, the first purpose of the present application is to provide a magnetotelluric-seismic wave first-arrival joint inversion method, so as to implement the method.
The first is the joint inversion of magnetotelluric-seismic waves at first arrival.
A third object of the present application is to propose a computer device.
In order to achieve the above object, a first embodiment of the present application provides a magnetotelluric-seismic wave first arrival joint inversion method, including:
constructing a training set, wherein the training set comprises a resistivity image of an inversion result, a speed image of the inversion result, a resistivity image of a real model and a speed image of the real model;
performing combined training on the resistivity image and the speed image through a neural network to obtain nonlinear mapping from the resistivity image of the inversion result and the speed image of the inversion result to the resistivity image of the real model and the speed image of the real model;
acquiring a reference model for magnetotelluric-seismic wave first arrival time inversion according to the nonlinear mapping from the resistivity image of the inversion result and the velocity image of the inversion result to the resistivity image of the real model and the velocity image of the real model; and
and under the constraint of a reference model, simultaneously updating the resistivity image and the velocity image so as to carry out joint inversion of magnetotelluric-seismic wave first arrival time.
The method of the embodiment of the application is a magnetotelluric-seismic wave first arrival joint inversion method, which is mainly realized based on an attribute fusion method and comprises the following steps:
the attribute fusion method adopts a deep neural network to train nonlinear mapping from a resistivity image and a velocity image which are given by independent inversion to a resistivity image and a velocity image of a real model so as to automatically extract prior information which is blended in a two-mode independent inversion process. The output and the input of the neural network both have resistivity images and velocity images, and when a training set is constructed, each case needs to be independently inverted once, and only one neural network needs to be trained. The input and output of the neural network can be not only resistivity and speed images, but also images obtained by performing mathematical transformation on the resistivity and speed images;
the attribute fusion method adopts a double-channel-double-channel CNN to respectively input and output resistivity and speed;
the attribute fusion method does not directly relate to the mutual mapping relation of resistivity and speed, so that the resistivity and speed images are updated at the same time in each iteration;
and in each iteration of the optimization process, the attribute fusion method simultaneously inputs the resistivity and the velocity image generated in the last iteration into the previously trained neural network, and simultaneously obtains a reference resistivity and velocity image as a reference for the independent optimization of the resistivity and the velocity in the next iteration.
Compared with the prior art, the application has the following advantages:
1. the resistivity and speed model obtained by inversion has better structural similarity and more accurate boundary;
2. jumping the target function out of a local minimum value;
3. has high generalization ability.
In the embodiment of the application, when the training set is constructed, the resistivity image and the velocity image are separately inverted, and the training set is constructed by using the final resistivity image and the final velocity image obtained after inversion and the real resistivity image and the real velocity image.
In an embodiment of the present application, the neural network is a convolutional neural network, wherein the convolutional neural network is a two-channel-two-channel CNN.
In the embodiment of the present application, in the joint inversion of the first arrival of magnetotelluric-seismic waves by using the reference model, an objective function of the joint inversion is:
wherein rho is the resistivity to be inverted, s is the slowness (reciprocal of speed) to be inverted, D is the interval to be inverted, F is a forward operator for solving the magnetotelluric problem, G is a forward operator for solving the seismic wave problem, and D isobsFor magnetotelluric measurement data, tobsFor seismic first-arrival time measurement data, prefAnd srefFor the nonlinear mapping output of the current model input to the neural network, R (ρ) and R(s) are regular terms describing the smoothness of the model, and α, λ, γ are coefficients for adjusting the weights.
In the embodiment of the application, the target functional is minimized through an iterative method.
In order to achieve the above object, a second aspect of the present application provides a joint inversion of magnetotelluric-seismic first arrival time, including:
the construction module is used for constructing a training set, wherein the training set comprises a resistivity image of an inversion result, a speed image of the inversion result, a resistivity image of a real model and a speed image of the real model;
the training module is used for respectively carrying out combined training on the resistivity image and the speed image through a neural network so as to obtain nonlinear mapping from the resistivity image of the inversion result and the speed image of the inversion result to the resistivity image of the real model and the speed image of the real model;
the acquisition module is used for acquiring a reference model for magnetotelluric-seismic wave first arrival time inversion according to the nonlinear mapping from the resistivity image of the inversion result and the velocity image of the inversion result to the resistivity image of the real model and the velocity image of the real model; and
and the inversion module is used for updating the resistivity image and the velocity image simultaneously under the constraint of the reference model so as to carry out joint inversion of magnetotelluric-seismic wave in first arrival.
The method of the embodiment of the application is realized by an attribute-based fusion method, and compared with the prior art, the method has the following advantages:
1. the resistivity and speed model obtained by inversion has better structural similarity and more accurate boundary;
2. jumping the target function out of a local minimum value;
3. has high generalization ability.
In the embodiment of the application, in the construction module, when the training set is constructed, the resistivity image and the velocity image are separately inverted, and the final resistivity image and the final velocity image obtained after inversion, and the real resistivity image and the real velocity image are used for constructing the training set.
In the embodiment of the present application, in the joint inversion of the first arrival of magnetotelluric-seismic waves by using the reference model, an objective function of the joint inversion is:
wherein rho is the resistivity to be inverted, s is the slowness (reciprocal of speed) to be inverted, D is the interval to be inverted, F is a forward operator for solving the magnetotelluric problem, G is a forward operator for solving the seismic wave problem, and D isobsFor magnetotelluric measurement data, tobsFor seismic first-arrival time measurement data, prefAnd srefFor the nonlinear mapping output of the current model input to the neural network, R (ρ) and R(s) are regular terms describing the smoothness of the model, and α, λ, γ are coefficients for adjusting the weights.
In an embodiment of the present application, the apparatus further comprises a processing module configured to minimize the target functional using an iterative approach.
To achieve the above object, an embodiment of a third aspect of the present application provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, where the processor executes the computer program to implement the methods described in the embodiments of the first aspect and the embodiments of the second aspect of the present application.
Additional aspects and advantages of the present application will be set forth in part in the description which follows and, in part, will be obvious from the description, or may be learned by practice of the present application.
Drawings
The foregoing and/or additional aspects and advantages of the present application will become apparent and readily appreciated from the following description of the embodiments, taken in conjunction with the accompanying drawings of which:
fig. 1 is a schematic flowchart of a magnetotelluric-seismic wave first arrival joint inversion method according to an embodiment of the present disclosure;
fig. 2 is a graph for verifying a simulation reconstructed resistivity and velocity model constructed by the present application, where the left graph is a simulation reconstructed resistivity model and the right graph is a simulation reconstructed velocity model;
FIG. 3 is a set of examples in a training set provided by an embodiment of the present application, with the left graph showing an input resistivity model and the right graph showing an input velocity model;
fig. 4 is another set of examples in the training set provided in the embodiment of the present application, where the left graph is a resistivity output label corresponding to the resistivity model input in fig. 3, and the right graph is a velocity output label corresponding to the velocity model input in fig. 3;
FIG. 5 is a joint inversion result of simulation data obtained by the method of the present application according to an embodiment of the present application; and
FIG. 6 is a graph of the results of an individual inversion of simulation data provided by an embodiment of the present application but not by the method described herein.
Detailed Description
Reference will now be made in detail to embodiments of the present application, examples of which are illustrated in the accompanying drawings, wherein like or similar reference numerals refer to the same or similar elements or elements having the same or similar function throughout. The embodiments described below with reference to the drawings are exemplary and intended to be used for explaining the present application and should not be construed as limiting the present application.
A magnetotelluric-seismic wave first-arrival time joint inversion method and device according to an embodiment of the present application are described below with reference to the accompanying drawings.
Fig. 1 is a schematic flowchart of a magnetotelluric-seismic wave first arrival joint inversion method according to an embodiment of the present disclosure.
To solve this problem, an embodiment of the present application provides a magnetotelluric-seismic wave first-arrival joint inversion method to implement magnetotelluric-seismic wave first-arrival joint inversion, as shown in fig. 1, the method includes the following steps:
constructing a training set, wherein the training set comprises a resistivity image of an inversion result, a speed image of the inversion result, a resistivity image of a real model and a speed image of the real model;
performing combined training on the resistivity image and the speed image through a neural network to obtain nonlinear mapping from the resistivity image of the inversion result and the speed image of the inversion result to the resistivity image of the real model and the speed image of the real model;
acquiring a reference model for magnetotelluric-seismic wave first arrival time inversion according to the nonlinear mapping from the resistivity image of the inversion result and the velocity image of the inversion result to the resistivity image of the real model and the velocity image of the real model; and
and under the constraint of a reference model, simultaneously updating the resistivity image and the velocity image so as to carry out joint inversion of magnetotelluric-seismic wave first arrival time.
The method of the embodiment of the application is a magnetotelluric-seismic wave first arrival joint inversion method, which is mainly realized based on an attribute fusion method and comprises the following steps:
the attribute fusion method adopts a deep neural network to train nonlinear mapping from a resistivity image and a velocity image which are given by independent inversion to a resistivity image and a velocity image of a real model so as to automatically extract prior information which is blended in a two-mode independent inversion process. The output and the input of the neural network both have resistivity images and velocity images, and when a training set is constructed, each case needs to be independently inverted once, and only one neural network needs to be trained. The input and output of the neural network can be not only resistivity and speed images, but also images obtained by performing mathematical transformation on the resistivity and speed images;
the attribute fusion method adopts a double-channel-double-channel CNN to respectively input and output resistivity and speed;
the attribute fusion method does not directly relate to the mutual mapping relation of resistivity and speed, so that the resistivity and speed images are updated at the same time in each iteration;
and in each iteration of the optimization process, the attribute fusion method simultaneously inputs the resistivity and the velocity image generated in the last iteration into the previously trained neural network, and simultaneously obtains a reference resistivity and velocity image as a reference for the independent optimization of the resistivity and the velocity in the next iteration.
Compared with the prior art, the application has the following advantages:
1. the resistivity and speed model obtained by inversion has better structural similarity and more accurate boundary;
2. jumping the target function out of a local minimum value;
3. has high generalization ability.
In the embodiment of the application, when the training set is constructed, the resistivity image and the velocity image are separately inverted, and the training set is constructed by using the final resistivity image and the final velocity image obtained after inversion and the real resistivity image and the real velocity image.
In an embodiment of the present application, the neural network is a convolutional neural network, wherein the convolutional neural network is a two-channel-two-channel CNN.
In the embodiment of the present application, in the joint inversion of the first arrival of magnetotelluric-seismic waves by using the reference model, an objective function of the joint inversion is:
wherein rho is the resistivity to be inverted, s is the slowness (reciprocal of speed) to be inverted, D is the interval to be inverted, F is a forward operator for solving the magnetotelluric problem, G is a forward operator for solving the seismic wave problem, and D isobsFor magnetotelluric measurement data, tobsFor seismic first-arrival time measurement data, prefAnd srefFor the nonlinear mapping output of the current model input to the neural network, R (ρ) and R(s) are regular terms describing the smoothness of the model, and α, λ, γ are coefficients for adjusting the weights.
In the embodiment of the application, the target functional is minimized through an iterative method.
In order to achieve the above object, a second aspect of the present application provides a joint inversion of magnetotelluric-seismic first arrival time, including:
the construction module is used for constructing a training set, wherein the training set comprises a resistivity image of an inversion result, a speed image of the inversion result, a resistivity image of a real model and a speed image of the real model;
the training module is used for respectively carrying out combined training on the resistivity image and the speed image through a neural network so as to obtain nonlinear mapping from the resistivity image of the inversion result and the speed image of the inversion result to the resistivity image of the real model and the speed image of the real model;
the acquisition module is used for acquiring a reference model for magnetotelluric-seismic wave first arrival time inversion according to the nonlinear mapping from the resistivity image of the inversion result and the velocity image of the inversion result to the resistivity image of the real model and the velocity image of the real model; and
and the inversion module is used for updating the resistivity image and the velocity image simultaneously under the constraint of the reference model so as to carry out joint inversion of magnetotelluric-seismic wave in first arrival.
The method of the embodiment of the application is realized by an attribute-based fusion method, and compared with the prior art, the method has the following advantages:
1. the resistivity and speed model obtained by inversion has better structural similarity and more accurate boundary;
2. jumping the target function out of a local minimum value;
3. has high generalization ability.
In the embodiment of the application, in the construction module, when the training set is constructed, the resistivity image and the velocity image are separately inverted, and the final resistivity image and the final velocity image obtained after inversion, and the real resistivity image and the real velocity image are used for constructing the training set.
In the embodiment of the present application, in the joint inversion of the first arrival of magnetotelluric-seismic waves by using the reference model, an objective function of the joint inversion is:
wherein rho is the resistivity to be inverted, s is the slowness (reciprocal of speed) to be inverted, D is the interval to be inverted, F is a forward operator for solving the magnetotelluric problem, G is a forward operator for solving the seismic wave problem, and D isobsFor magnetotelluric measurement data, tobsFor seismic first-arrival time measurement data, prefAnd srefFor the nonlinear mapping output of the current model input to the neural network, R (ρ) and R(s) are regular terms describing the smoothness of the model, and α, λ, γ are coefficients for adjusting the weights.
In an embodiment of the present application, the apparatus further comprises a processing module configured to minimize the target functional using an iterative approach.
In order to make the present application more comprehensible to those skilled in the art, the present application provides another magnetotelluric-seismic wave first-arrival joint inversion method.
The method and apparatus of embodiments of the present application are described below with reference to the drawings.
Fig. 1 is a schematic flowchart of a magnetotelluric-seismic wave first arrival joint inversion method according to an embodiment of the present disclosure;
a deep learning magnetotelluric-seismic wave first arrival time joint inversion method based on attribute fusion is disclosed, wherein an objective function has the following form:
where ρ isref，sref＝Nρ，s(ρ0，s0) Rho is the resistivity to be inverted, s is the slowness (reciprocal of velocity) to be inverted, F is the forward operator for solving the magnetotelluric problem, G is the forward operator for solving the seismic wave problem, dobsFor magnetotelluric measurement data, tobsIs the first arrival of seismic wavesThe data is measured. N is a radical ofρ，sA non-linear mapping from the current resistivity and velocity model to the reference resistivity and velocity model is done for the trained neural network. Rho0And s0For the current resistivity and velocity model, ρrefAnd srefAnd outputting the current model through the nonlinear mapping of the neural network. R (rho) and R(s) are regular terms describing the smoothness degree of the model, and alpha, lambda and gamma are coefficients for adjusting the terms;
in the embodiment of the present application, the target functional L is minimized by the iterative methodJoint(ρ，s)。
In the embodiment of the application, the resistivity and the velocity field model (or the model obtained by transforming the resistivity and the velocity field model) which are finally output by inversion are used as input, and the real resistivity and the velocity field model (or the model obtained by transforming the resistivity and the velocity field model) are used as a training set of output. Transformation herein refers to mathematical operations performed on a resistivity or velocity field model for the purpose of better extracting training set information, facilitating training, and the like, including but not limited to taking gradients, resampling, and the like.
In the embodiment of the application, the input is the inversion final resistivity and velocity field, and the output is the neural network N of the mapped resistivity and velocity fieldρ，sIncluding various configurations.
In the embodiment of the present application, the regularization terms R (ρ), R(s) include various forms.
In the embodiment of the present application, the method for minimizing the target functional includes various minimization methods.
Further, when R (ρ), R(s) use the smoothest constraint, the objective function of the joint inversion is specifically written as
Wherein P isref，sref＝Nρ，s(ρ0，s0)，
Respectively represent waterAnd D is an interval to be inverted.
Further, resistivity ρ and velocity s are updated simultaneously using gauss-newton method, and at the k-th iteration, the objective function becomes
Where ρ isk，ref，sk，ref＝Nρ，s(ρk-1，sk-1)
Further, the above equation is made to have perturbation quantities Δ ρ, Δ s of resistivity and velocity equal to 0, respectively, and thus the obtained results are obtained
And
wherein, in the formula Jρ，JsThe jacobian matrix of magnetotelluric and seismic positive problems, respectively.
Further, the model to be inverted is updated according to the following two formulas.
ρk＝ρk-1+vρΔρk
sk＝sk-1+νsΔsk
Wherein v isρ，νsFor updating the step length, the step length is obtained by a linear searching mode.
Fig. 2 is a simulation reconstructed resistivity and velocity model constructed by the magnetotelluric-seismic wave first arrival joint inversion method for verifying the attribute fusion method, where the left graph is a simulation reconstructed resistivity model and the right graph is a simulation reconstructed velocity model, as shown in fig. 2, the theoretical data is generated by the simulation reconstructed resistivity and velocity model, and 5% of noise is applied to the theoretical data;
FIG. 3 is a set of examples in a training set of the present application provided by an embodiment of the present application, with the left graph being an input resistivity model and the right graph being an input velocity model; fig. 4 is another set of examples in a training set of the present application provided by an embodiment of the present application, where the left graph is a resistivity output label corresponding to the resistivity model input in fig. 3, and the right graph is a velocity output label corresponding to the velocity model input in fig. 3; as shown in fig. 3 and 4, the embodiment of the present application separately inverts the model and the real model by the resistivity-velocity trained by the training set.
FIG. 5 is a joint inversion result of simulation data obtained by the method of the present application provided in the embodiments of the present application, and FIG. 6 is a separate inversion result of simulation data obtained by a method other than the method of the present application provided in the embodiments of the present application; as shown in fig. 5 and fig. 6, it can be seen that the resistivity and velocity models in the results of the joint inversion of the simulation data obtained by the method according to the embodiment of the present application have better structural similarity and more accurate boundaries.
It should be noted that the explanation of the embodiment of the magnetotelluric-seismic wave first-arrival joint inversion method is also applicable to the magnetotelluric-seismic wave first-arrival joint inversion of the embodiment, and is not repeated here.
In order to implement the foregoing embodiments, the present application further provides a computer device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor executes the computer program.
In the description herein, reference to the description of the term "one embodiment," "some embodiments," "an example," "a specific example," or "some examples," etc., means that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the application. In this specification, the schematic representations of the terms used above are not necessarily intended to refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples. Furthermore, various embodiments or examples and features of different embodiments or examples described in this specification can be combined and combined by one skilled in the art without contradiction.
Furthermore, the terms "first", "second" and "first" are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one such feature. In the description of the present application, "plurality" means at least two, e.g., two, three, etc., unless specifically limited otherwise.
Any process or method descriptions in flow charts or otherwise described herein may be understood as representing modules, segments, or portions of code which include one or more executable instructions for implementing steps of a custom logic function or process, and alternate implementations are included within the scope of the preferred embodiment of the present application in which functions may be executed out of order from that shown or discussed, including substantially concurrently or in reverse order, depending on the functionality involved, as would be understood by those reasonably skilled in the art of the present application.
The logic and/or steps represented in the flowcharts or otherwise described herein, e.g., an ordered listing of executable instructions that can be considered to implement logical functions, can be embodied in any computer-readable medium for use by or in connection with an instruction execution system, apparatus, or device, such as a computer-based system, processor-containing system, or other system that can fetch the instructions from the instruction execution system, apparatus, or device and execute the instructions. For the purposes of this description, a "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of the computer-readable medium would include the following: an electrical connection (electronic device) having one or more wires, a portable computer diskette (magnetic device), a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). Additionally, the computer-readable medium could even be paper or another suitable medium upon which the program is printed, as the program can be electronically captured, via for instance optical scanning of the paper or other medium, then compiled, interpreted or otherwise processed in a suitable manner if necessary, and then stored in a computer memory.
It should be understood that portions of the present application may be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, the various steps or methods may be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. If implemented in hardware, as in another embodiment, any one or combination of the following techniques, which are known in the art, may be used: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application specific integrated circuit having an appropriate combinational logic gate circuit, a Programmable Gate Array (PGA), a Field Programmable Gate Array (FPGA), or the like.
It will be understood by those skilled in the art that all or part of the steps carried by the method for implementing the above embodiments may be implemented by hardware related to instructions of a program, which may be stored in a computer readable storage medium, and when the program is executed, the program includes one or a combination of the steps of the method embodiments.
In addition, functional units in the embodiments of the present application may be integrated into one processing module, or each unit may exist alone physically, or two or more units are integrated into one module. The integrated module can be realized in a hardware mode, and can also be realized in a software functional module mode. The integrated module, if implemented in the form of a software functional module and sold or used as a stand-alone product, may also be stored in a computer readable storage medium.
The storage medium mentioned above may be a read-only memory, a magnetic or optical disk, etc. Although embodiments of the present application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present application, and that variations, modifications, substitutions and alterations may be made to the above embodiments by those of ordinary skill in the art within the scope of the present application.
Claims (10)
1. A magnetotelluric-seismic wave first arrival time joint inversion method is characterized by comprising the following steps:
constructing a training set, wherein the training set comprises a resistivity image of an inversion result, a speed image of the inversion result, a resistivity image of a real model and a speed image of the real model;
performing combined training on the resistivity image and the speed image through a neural network to obtain nonlinear mapping from the resistivity image of the inversion result and the speed image of the inversion result to the resistivity image of the real model and the speed image of the real model;
acquiring a reference model for magnetotelluric-seismic wave first arrival time inversion according to the nonlinear mapping from the resistivity image of the inversion result and the velocity image of the inversion result to the resistivity image of the real model and the velocity image of the real model; and
and under the constraint of a reference model, simultaneously updating the resistivity image and the velocity image so as to carry out joint inversion of magnetotelluric-seismic wave first arrival time.
2. The method of claim 1, wherein the resistivity image and the velocity image are separately inverted during construction of the training set, and the final resistivity image and the final velocity image obtained by inversion are used to construct the training set with the true resistivity image and the velocity image.
3. The method of claim 1, wherein the neural network is a convolutional neural network, wherein the convolutional neural network is a two-channel-to-two-channel CNN.
4. The method according to any one of claims 1 to 3, wherein in the joint inversion of the magnetotelluric-seismic wave first arrival times by means of the reference model, an objective function of the joint inversion is:
wherein rho is the resistivity to be inverted, s is the slowness (reciprocal of speed) to be inverted, D is the interval to be inverted, F is a forward operator for solving the magnetotelluric problem, G is a forward operator for solving the seismic wave problem, and D isobsFor magnetotelluric measurement data, tobsFor seismic first-arrival time measurement data, prefAnd srefFor the nonlinear mapping output of the current model via input to the neural network, R (ρ) and R(s) are regular terms describing the smoothness of the model,α, λ, γ are coefficients for adjusting the weights.
5. The method of claim 4, wherein the target functional is minimized by an iterative approach.
6. A magnetotelluric-seismic wave first arrival joint inversion method is characterized by comprising the following steps,
the construction module is used for constructing a training set, wherein the training set comprises a resistivity image of an inversion result, a speed image of the inversion result, a resistivity image of a real model and a speed image of the real model;
the training module is used for respectively carrying out combined training on the resistivity image and the speed image through a neural network so as to obtain nonlinear mapping from the resistivity image of the inversion result and the speed image of the inversion result to the resistivity image of the real model and the speed image of the real model;
the acquisition module is used for acquiring a reference model for magnetotelluric-seismic wave first arrival time inversion according to the nonlinear mapping from the resistivity image of the inversion result and the velocity image of the inversion result to the resistivity image of the real model and the velocity image of the real model; and
and the inversion module is used for updating the resistivity image and the velocity image simultaneously under the constraint of the reference model so as to carry out joint inversion of magnetotelluric-seismic wave in first arrival.
7. The apparatus of claim 6, wherein in the construction module, the resistivity image and the velocity image are separately inverted during construction of the training set, and the final resistivity image and the final velocity image obtained after inversion are used for construction of the training set together with the true resistivity image and the velocity image.
8. The apparatus of claim 6, wherein in the joint inversion of the magnetotelluric-seismic wave first arrival times through the reference model, an objective function of the joint inversion is:
wherein rho is the resistivity to be inverted, s is the slowness (reciprocal of speed) to be inverted, D is the interval to be inverted, F is a forward operator for solving the magnetotelluric problem, G is a forward operator for solving the seismic wave problem, and D isobsFor magnetotelluric measurement data, tobsFor seismic first-arrival time measurement data, prefAnd srefFor the nonlinear mapping output of the current model input to the neural network, R (ρ) and R(s) are regular terms describing the smoothness of the model, and α, λ, γ are coefficients for adjusting the weights.
9. The apparatus of claim 8, further comprising a processing module to minimize the target functional using an iterative approach.
10. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the method of any one of claims 1-5 when executing the computer program.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110436707.7A CN113156538A (en) | 2021-04-22 | 2021-04-22 | Magnetotelluric-seismic wave first arrival time joint inversion method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110436707.7A CN113156538A (en) | 2021-04-22 | 2021-04-22 | Magnetotelluric-seismic wave first arrival time joint inversion method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113156538A true CN113156538A (en) | 2021-07-23 |
Family
ID=76869429
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110436707.7A Pending CN113156538A (en) | 2021-04-22 | 2021-04-22 | Magnetotelluric-seismic wave first arrival time joint inversion method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113156538A (en) |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080059075A1 (en) * | 2006-09-04 | 2008-03-06 | Daniele Colombo | Methods and apparatus for geophysical exploration via joint inversion |
| US20110166840A1 (en) * | 2008-09-24 | 2011-07-07 | Green Kenneth E | Systems and Methods For Subsurface Electromagnetic Mapping |
| CN108680964A (en) * | 2018-03-30 | 2018-10-19 | 吉林大学 | A kind of normalization weight magnetoelectricity shake joint inversion method based on structural constraint |
| CN110927806A (en) * | 2019-10-29 | 2020-03-27 | 清华大学 | Magnetotelluric inversion method and magnetotelluric inversion system based on supervised descent method |
| US20200183041A1 (en) * | 2018-12-11 | 2020-06-11 | Exxonmobil Research And Engineering Company | Machine learning-augmented geophysical inversion |
| CN112308206A (en) * | 2020-09-14 | 2021-02-02 | 清华大学 | Magnetotelluric-seismic wave first arrival time joint inversion method based on deep learning |
| CN112379462A (en) * | 2020-10-30 | 2021-02-19 | 中国石油天然气集团有限公司 | Electromagnetic seismic data joint processing method and device |
-
2021
- 2021-04-22 CN CN202110436707.7A patent/CN113156538A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080059075A1 (en) * | 2006-09-04 | 2008-03-06 | Daniele Colombo | Methods and apparatus for geophysical exploration via joint inversion |
| US20110166840A1 (en) * | 2008-09-24 | 2011-07-07 | Green Kenneth E | Systems and Methods For Subsurface Electromagnetic Mapping |
| CN108680964A (en) * | 2018-03-30 | 2018-10-19 | 吉林大学 | A kind of normalization weight magnetoelectricity shake joint inversion method based on structural constraint |
| US20200183041A1 (en) * | 2018-12-11 | 2020-06-11 | Exxonmobil Research And Engineering Company | Machine learning-augmented geophysical inversion |
| CN110927806A (en) * | 2019-10-29 | 2020-03-27 | 清华大学 | Magnetotelluric inversion method and magnetotelluric inversion system based on supervised descent method |
| CN112308206A (en) * | 2020-09-14 | 2021-02-02 | 清华大学 | Magnetotelluric-seismic wave first arrival time joint inversion method based on deep learning |
| CN112379462A (en) * | 2020-10-30 | 2021-02-19 | 中国石油天然气集团有限公司 | Electromagnetic seismic data joint processing method and device |
Non-Patent Citations (1)
| Title |
|---|
| RUI GUO等: "Joint Inversion of Audio-Magnetotelluric and Seismic Travel Time Data With Deep Learning Constraint", 《IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING》 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11668853B2 (en) | Petrophysical inversion with machine learning-based geologic priors | |
| Khuc et al. | Structural identification using computer vision–based bridge health monitoring | |
| Narazaki et al. | Efficient development of vision-based dense three-dimensional displacement measurement algorithms using physics-based graphics models | |
| CN111596366B (en) | Wave impedance inversion method based on seismic signal optimization processing | |
| KR102219561B1 (en) | Unsupervised stereo matching apparatus and method using confidential correspondence consistency | |
| CN110610486A (en) | Monocular image depth estimation method and device | |
| Zhang et al. | Complex image background segmentation for cable force estimation of urban bridges with drone‐captured video and deep learning | |
| Chamoin et al. | Synergies between the constitutive relation error concept and PGD model reduction for simplified V&V procedures | |
| CN111580163B (en) | Full waveform inversion method and system based on non-monotonic search technology | |
| CN109073772A (en) | The displacement between seismic image is determined using light stream | |
| US10677948B2 (en) | Context based bounded hydrocarbon formation identification | |
| Ceruso et al. | Relative multiscale deep depth from focus | |
| CN116558789A (en) | Fault diagnosis method, device, equipment and storage medium for power transformer | |
| CN108508481B (en) | A kind of method, apparatus and system of longitudinal wave converted wave seismic data time match | |
| CN112308206B (en) | Magnetotelluric-seismic wave first arrival time joint inversion method based on deep learning | |
| CN116127314B (en) | Method for predicting underground density based on self-adaptive multi-scale deep learning network | |
| CN113156538A (en) | Magnetotelluric-seismic wave first arrival time joint inversion method | |
| CN115754007A (en) | Damage detection method based on acoustic emission technology and tomography technology | |
| RU2714612C1 (en) | Method of identifying nonlinear systems | |
| CN114155354A (en) | Capacitance tomography reconstruction method and device based on graph convolution network | |
| Yang et al. | Building near-surface velocity models by integrating the first-arrival traveltime tomography and supervised deep learning | |
| Bodnariuc et al. | Adaptive dictionary-based spatio-temporal flow estimation for Echo PIV | |
| CN117607957B (en) | Seismic wave travel time solving method and system based on equivalent slowness rapid propulsion method | |
| RU2730367C1 (en) | Method of calibrating intelligent sensors | |
| Jennings | Automatic interpretative seismic image-focusing analysis |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |