CN113807020A - Magnetotelluric inversion method based on deep learning constraint - Google Patents
Magnetotelluric inversion method based on deep learning constraint Download PDFInfo
- Publication number
- CN113807020A CN113807020A CN202111143591.4A CN202111143591A CN113807020A CN 113807020 A CN113807020 A CN 113807020A CN 202111143591 A CN202111143591 A CN 202111143591A CN 113807020 A CN113807020 A CN 113807020A
- Authority
- CN
- China
- Prior art keywords
- model
- sample set
- data
- neural network
- geoelectric
- 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.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/23—Design optimisation, verification or simulation using finite element methods [FEM] or finite difference methods [FDM]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/30—Assessment of water resources
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- Biophysics (AREA)
- Computer Hardware Design (AREA)
- Mathematical Physics (AREA)
- Geometry (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Geophysics And Detection Of Objects (AREA)
Abstract
The invention discloses a magnetotelluric inversion method based on deep learning constraint, which comprises the steps of firstly, manufacturing a geoelectric model data set and a magnetotelluric model forward modeling response data set for neural network training; normalizing the manufactured data set and proportionally dividing the data set into a training set, a verification set and a test set; sending the preprocessed training set into a neural network for training to obtain a network model; and calculating a mapping model by using the actual observation data to be inverted by using the trained network model, and inverting the mapping model serving as an initial model by using a traditional optimization method to obtain a final inversion result. The method is suitable for magnetotelluric inversion under complex geological structure conditions, and can obtain a geoelectric model closer to a real situation. The fineness degree of the inversion result in the aspects of stratum and fault depiction is far higher than that of the inversion of the traditional uniform half-space initial model.
Description
Technical Field
The invention relates to the field of geological technology research, in particular to a magnetotelluric inversion method based on deep learning constraint.
Background
The Magnetotelluric (MT) depth measurement is a deep detection method for deducing underground resistivity structure by inversion by measuring the earth electromagnetic field change on the earth surface, and is an important method for researching the electrical properties and distribution characteristics of the earth internal structure and rock stratum. Typical magnetotelluric inversion methods at present include Bostick inversion method (Bostick), nonlinear conjugate gradient inversion method (NLCG), ocam inversion method (Occam), and the like. For a simpler geoelectric model, the conventional inversion method can invert the position and size of an abnormal body in the geoelectric model based on a uniform half-space initial model. However, the actual geological structure is often complex, the conventional magnetotelluric inversion method has strong multi-solution, and the actual underground model structure is often difficult to obtain by taking uniform half-space or one-dimensional inversion as an initial model for inversion. The conventional magnetotelluric inversion method has strong dependence on an initial model, so that the establishment of a good initial model is very important. The method for establishing the initial model at present mainly utilizes the prior exploration result and the prior knowledge of people on the exploration area to establish the initial model through manual explanation. However, this method has the following disadvantages: the subjective factor of modeling is strong, and the quality of the model is closely related to the quality of a modeler.
The method for manually establishing the initial model has strong subjective factors and is not beneficial to the inversion of the complex geological structure model by the conventional optimization method. The electromagnetic observation data and the earth-electricity model can be learned through deep learning, and mapping of the electromagnetic observation data to the earth-electricity model is established. The neural network does not consider a real electromagnetic physics formula (Maxwell equation set), but only learns the mapping relation between electromagnetic observation data and a geoelectric model, so that the geologic body of the generated model and the real model has better similarity on the whole structure, but the forward response of the geologic body is obviously different from the actual observation result. Therefore, the invention aims to combine the conventional magnetotelluric inversion method with deep learning and solve the problems of high dependence on an initial model, high inversion difficulty of a complex model and low inversion resolution of the conventional electromagnetic inversion.
Disclosure of Invention
The invention aims to provide a magnetotelluric inversion method based on deep learning constraint. By using the method, the inversion of the complex geological model can be realized, and the describing fineness of the inversion result on the stratum and the fault is far higher than that of the inversion result of the traditional uniform half-space initial model.
In order to achieve the purpose, the invention is implemented according to the following technical scheme:
firstly, a geoelectric model data set and a magnetotelluric model forward modeling response data set for neural network training are prepared. And normalizing the manufactured data set, dividing the data set into a training set, a verification set and a test set according to a proportion, and sending the preprocessed training set into a neural network for training to obtain a network model. And calculating a mapping model by using the trained network model for the actual observation data to be inverted, and inverting the inversion result as an initial model by using a traditional optimization method to obtain a final inversion result.
The method comprises the following specific steps:
1. determining a detection area, and establishing a similar geoelectric model sample set of the area according to existing exploration data and priori knowledge of the detection area. The geoelectric model sample set is required to conform to the characteristics of the geological structure of the detection region, but is not required to be completely identical to the geological structure under the detection region, and some differences exist among the geoelectric models so that the learnt neural network model has better generalization capability.
2. The forward response of the geoelectric model sample set S1 is calculated using a conventional finite element method to obtain a forward response sample set S2.
3. Taking the logarithm of the resistivity of the geoelectric model sample set S1 and the forward response sample set S2 by taking the base 10 as a whole, and counting the maximum value rho in the geoelectric model sample set and the forward response sample set after taking the logarithmmaxAnd minimum value ρminThen, mapping the resistivity X of the input forward modeling sample data to 0-1 as input data X to be transmitted into the neural network, namely:
for the phase angle with small span of the data range in the forward response sample set, only the maximum and minimum values of the phase angle need to be counted, and a normalization formula is used for mapping to 0-1 to be used as input data to be transmitted into the neural network.
4. The sample data set S2 is used as input data of the neural network, the geoelectric model sample set S1 is used as a reference object of the output predicted value of the neural network model, and the geoelectric model sample set and the forward response sample set jointly form a training sample. And dividing experience according to the deep learning data, and dividing the ground electric model sample set and the forward response sample set after pretreatment into a training set, a verification set and a test set according to the training experience of the neural network.
5. And establishing a neural network model by combining the magnetotelluric inversion characteristics. The magnetotelluric inversion-oriented neural network model is out of the protection scope of the invention. The generated neural network model is suitable for magnetotelluric inversion, and a Sigmoid activation function, a SmoothL1Loss function and an Adam optimizer are used for establishing a magnetotelluric inversion-oriented neural network structure. The input data of the input layer of the neural network model is a forward modeling response sample set of the magnetotelluric model, and the output data of the output layer is geoelectric model prediction parameters.
6. And transmitting the training set and the verification set into a neural network model, and training the network model and determining model weights by the neural network according to the geoelectric model sample set and the forward response sample set in the training set. And calculating the predicted geoelectricity model parameters of the forward modeling response sample set in the verification set by using the network model, and continuously optimizing the neural network model parameters according to the error between the predicted geoelectricity model parameters and the parameters of the real geoelectricity model sample set in the verification set so as to fit the mapping relation between the geoelectricity model sample set and the forward modeling response sample set. When the loss value between the predicted value of the forward response sample set and the actual value of the geoelectric model sample set in the verification set is reduced to a threshold value and begins to oscillate, an ideal network model M is trained.
7. Normalizing actual observation data to be inverted according to a formula (1) in the step3, inputting the normalized data into a network model M to obtain mapping model output, and finally reflecting mapping model data y back to the resistivity range in the step3, namely:
8. and discretizing the mapping model after the inverse mapping into a grid model, and taking the discretized grid model as an initial model to perform inversion by using a traditional optimization method.
The invention has the beneficial effects that:
compared with the prior art, the magnetotelluric inversion method based on the deep learning constraint is suitable for magnetotelluric inversion under complex geological structure conditions, and can obtain a geoelectric model closer to a real situation. The fineness degree of the inversion result in the aspects of stratum and fault depiction is far higher than that of the inversion of the traditional uniform half-space initial model.
Drawings
FIG. 1 is a flow chart of a magnetotelluric inversion method based on deep learning constraints.
FIG. 2 is a diagram of a magnetotelluric inversion-oriented neural network architecture.
FIG. 3 is an input to a magnetotelluric inversion neural network, where a is the apparent resistivity of the forward response and b is the impedance phase of the forward response.
FIG. 4 is the output of a magnetotelluric inversion-oriented neural network.
FIG. 5 is a discretized mesh model for magnetotelluric inversion neural network output.
FIG. 6 is an inversion result of a magnetotelluric inversion method based on a depth learning constraint.
FIG. 7 shows the inversion results of a conventional uniform half-space initial model.
FIG. 8 is a flowchart of an example magnetotelluric inversion based on deep learning constraints.
Detailed Description
The invention will be further described with reference to the drawings and specific embodiments, which are illustrative of the invention and are not to be construed as limiting the invention.
As shown in fig. 1-8: the magnetotelluric inversion method based on deep learning constraint comprises the following steps:
step 1: in the embodiment, the geoelectric actual measurement data of a local region of a certain plateau in southwest China is used, and a geoelectric model sample set of the local region is established according to the actual measurement data. According to the geoelectric model sample set, calculating a forward response sample set of the geoelectric model sample set by using a MARE2DEM forward method: apparent resistivity and impedance phase.
Step 2: maximum value rho of logarithmic resistivity in statistical geoelectric model sample set S2maxAnd minimum value ρminAnd carrying out normalization treatment, wherein the normalization formula is as follows:
and counting the maximum and minimum values of the phase angles in the forward response sample set, and mapping the data to 0-1 by using a linear function normalization formula.
And Step3, taking the normalized geoelectric model sample set and the forward response sample set as training sample sets, and according to the ratio of 8: 1: the proportion of 1 is divided into a training set, a verification set and a test set.
Step4, establishing a magnetotelluric inversion-oriented neural network structure based on the characteristics of magnetotelluric inversion: in the embodiment, a D-LinkNet framework with a good magnetotelluric inversion effect in a U-type network is selected, and a hop-level link structure in the D-LinkNet structure is deleted; using 2-channel convolutional layers for inputting apparent resistivity and impedance phase data in a forward response sample set; using Sigmoid activation function, SmoothL1Loss function, Adam optimizer to facilitate better training and convergence of neural networks.
Step 5: and (4) inputting the training set obtained in the Step (Step 3) and the verification set into the magnetotelluric inversion-oriented neural network model established in the Step (4) for training. Each round of training input data was 32 batches with a training period of 100 epochs. Setting the initial learning rate to 0.01, if the training loss value is not reduced in three consecutive rounds, automatically reducing the learning rate to 0.2 times of the original learning rate until the loss value is reduced to a minimum value, which indicates that a more ideal model M is trained.
Step 6: and (4) processing the forward response sample set in the test set according to a normalization formula, and then transmitting the forward response sample set into the model M trained in the Step5 to obtain mapping model output. The mapping model is back mapped to the resistivity range of Step2 statistics, and the mapping formula is as follows:
step 7: and discretizing the mapping model into a grid model, and taking the discretized grid model as an initial model to perform inversion by using an NLCG inversion method so as to obtain a final magnetotelluric inversion result.
According to the method, the magnetotelluric inversion result based on the depth learning constraint is used as an initial model to be inverted by using a traditional inversion method, the electrical matching degree between the inversion result and a design model is high, and the fineness of the inversion result in the aspect of stratum and fault depiction is far higher than that of the traditional uniform half-space initial model inversion.
The technical solution of the present invention is not limited to the limitations of the above specific embodiments, and all technical modifications made according to the technical solution of the present invention fall within the protection scope of the present invention.
Claims (5)
1. A magnetotelluric inversion method based on deep learning constraint is characterized in that: the method comprises the following steps:
s1: firstly, making a geoelectric model data set and a magnetotelluric model forward modeling response data set for neural network training;
s2: normalizing the manufactured data set and proportionally dividing the data set into a training set, a verification set and a test set;
s3: sending the preprocessed training set into a neural network for training to obtain a network model;
s4: and calculating a mapping model by using the actual observation data to be inverted by using the trained network model, and inverting the mapping model serving as an initial model by using a traditional optimization method to obtain a final inversion result.
2. The deep learning constraint-based magnetotelluric inversion method of claim 1, wherein: the step S1 specifically includes the following steps:
s101: determining a detection area, and establishing a similar geoelectric model sample set of the area according to existing exploration data and priori knowledge of the detection area;
s102: calculating the forward response of the geoelectricity model sample set data by using a traditional finite element method to obtain a forward response sample set of the geoelectricity model;
s103: taking logarithm on the basis of 10 uniformly according to the resistivity of the geoelectric model sample set data and the forward response sample set, counting the geoelectric model sample set after taking logarithm, and taking the maximum value rho in the forward response sample setmaxAnd minimum value ρminThen, mapping the resistivity X of the input forward response sample data to 0-1 as input data X to be transmitted into the neural network, namely:
for the phase angle with small span of the data range in the forward response sample set, only the maximum and minimum values of the phase angle need to be counted, and a normalization formula is used for mapping to 0-1 to be used as input data to be transmitted into the neural network.
3. The deep learning constraint-based magnetotelluric inversion method of claim 2, wherein: the step S2 specifically includes:
and the forward response sample set is used as input data of the neural network, the geoelectric model sample set is used as a reference object of the output predicted value of the neural network model, and the geoelectric model sample set and the forward response sample set jointly form a training sample. And dividing experience according to the deep learning data, and dividing the ground electric model sample set and the forward response sample set after pretreatment into a training set, a verification set and a test set according to the training experience of the neural network.
4. The deep learning constraint-based magnetotelluric inversion method of claim 3, wherein: the step S3 is specifically a step of:
s301: establishing a neural network model by combining the magnetotelluric inversion characteristics, wherein the generated neural network model is suitable for magnetotelluric inversion, a Sigmoid activation function, a Smooth L1Loss function and an Adam optimizer are used for establishing a magnetotelluric inversion-oriented neural network structure, input data of an input layer of the neural network model is a forward response sample set of the magnetotelluric model, and output data of an output layer is geoelectric model prediction parameters;
s302: introducing a training set and a verification set into a neural network model, training the network model and determining model weights by the neural network according to a geoelectric model sample set and a forward response sample set in the training set, calculating predicted geoelectric model parameters of the forward response sample set in the verification set by using the network model, continuously optimizing the neural network model parameters according to errors between the predicted geoelectric model parameters and the parameters of a real geoelectric model sample set in the verification set so as to fit a mapping relation between the geoelectric model sample set and the forward response sample set, and representing that an ideal network model M is trained when loss values between predicted values of the forward response sample set in the verification set and real values of the geoelectric model sample set in the verification set are reduced to a threshold value and begin to oscillate.
5. The deep learning constraint-based magnetotelluric inversion method of claim 4, wherein: the step S4 specifically includes:
s401: normalizing actual observation data to be inverted according to formula (1) in step S103, inputting the normalized data into a network model M to obtain mapping model output, and finally reflecting mapping model data y back to the resistivity range in step S103, that is:
s402: and discretizing the mapping model after the inverse mapping into a grid model, and taking the discretized grid model as an initial model to perform inversion by using a traditional optimization method.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111143591.4A CN113807020B (en) | 2021-09-28 | 2021-09-28 | Magnetotelluric inversion method based on deep learning constraint |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111143591.4A CN113807020B (en) | 2021-09-28 | 2021-09-28 | Magnetotelluric inversion method based on deep learning constraint |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113807020A true CN113807020A (en) | 2021-12-17 |
| CN113807020B CN113807020B (en) | 2023-04-07 |
Family
ID=78938708
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111143591.4A Active CN113807020B (en) | 2021-09-28 | 2021-09-28 | Magnetotelluric inversion method based on deep learning constraint |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113807020B (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114254566A (en) * | 2021-12-27 | 2022-03-29 | 厦门大学 | Neural network geological detection inversion method based on 2.5-dimensional mixed spectral element method |
| CN114386464A (en) * | 2022-01-12 | 2022-04-22 | 中国科学院地质与地球物理研究所 | Deep learning extraction method for transient electromagnetic excitation information |
| CN114781254A (en) * | 2022-04-14 | 2022-07-22 | 成都理工大学 | Electromagnetic exploration inversion model construction method and device and storage medium |
| CN116595706A (en) * | 2023-02-28 | 2023-08-15 | 南方科技大学 | Method, electronic equipment and storage medium for inverting underground structure based on width learning |
| CN117371330A (en) * | 2023-10-30 | 2024-01-09 | 重庆大学 | Magnetotelluric two-dimensional deep learning inversion method based on traditional inversion guidance |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20180329104A1 (en) * | 2017-05-09 | 2018-11-15 | Pgs Geophysical As | Determining sea water resistivity |
| CN110058317A (en) * | 2019-05-10 | 2019-07-26 | 成都理工大学 | Aviation transient electromagnetic data and aviation magnetotelluric data joint inversion method |
| CN110058316A (en) * | 2019-05-10 | 2019-07-26 | 成都理工大学 | A kind of electromagnetic sounding constraint inversion method based on resistivity principle of equivalence |
| CN110968826A (en) * | 2019-10-11 | 2020-04-07 | 重庆大学 | Magnetotelluric deep neural network inversion method based on spatial mapping technology |
| CN111126591A (en) * | 2019-10-11 | 2020-05-08 | 重庆大学 | Magnetotelluric deep neural network inversion method based on space constraint technology |
| CN111812732A (en) * | 2020-06-29 | 2020-10-23 | 中铁二院工程集团有限责任公司 | Magnetotelluric nonlinear inversion method based on convolutional neural network |
| CN113158571A (en) * | 2021-04-26 | 2021-07-23 | 中国科学院地质与地球物理研究所 | Magnetotelluric inversion method based on full convolution neural network |
-
2021
- 2021-09-28 CN CN202111143591.4A patent/CN113807020B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20180329104A1 (en) * | 2017-05-09 | 2018-11-15 | Pgs Geophysical As | Determining sea water resistivity |
| CN110058317A (en) * | 2019-05-10 | 2019-07-26 | 成都理工大学 | Aviation transient electromagnetic data and aviation magnetotelluric data joint inversion method |
| CN110058316A (en) * | 2019-05-10 | 2019-07-26 | 成都理工大学 | A kind of electromagnetic sounding constraint inversion method based on resistivity principle of equivalence |
| CN110968826A (en) * | 2019-10-11 | 2020-04-07 | 重庆大学 | Magnetotelluric deep neural network inversion method based on spatial mapping technology |
| CN111126591A (en) * | 2019-10-11 | 2020-05-08 | 重庆大学 | Magnetotelluric deep neural network inversion method based on space constraint technology |
| CN111812732A (en) * | 2020-06-29 | 2020-10-23 | 中铁二院工程集团有限责任公司 | Magnetotelluric nonlinear inversion method based on convolutional neural network |
| CN113158571A (en) * | 2021-04-26 | 2021-07-23 | 中国科学院地质与地球物理研究所 | Magnetotelluric inversion method based on full convolution neural network |
Non-Patent Citations (2)
| Title |
|---|
| KEELEY EDWARDS等: "Stored Grain Inventory Management Using Neural-Network-Based Parametric Electromagnetic Inversion" * |
| 廖晓龙 等: "基于卷积神经网络的大地电磁反演" * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114254566A (en) * | 2021-12-27 | 2022-03-29 | 厦门大学 | Neural network geological detection inversion method based on 2.5-dimensional mixed spectral element method |
| CN114254566B (en) * | 2021-12-27 | 2024-06-04 | 厦门大学 | Neural network geological exploration inversion method based on 2.5-dimensional mixed spectral element method |
| CN114386464A (en) * | 2022-01-12 | 2022-04-22 | 中国科学院地质与地球物理研究所 | Deep learning extraction method for transient electromagnetic excitation information |
| CN114781254A (en) * | 2022-04-14 | 2022-07-22 | 成都理工大学 | Electromagnetic exploration inversion model construction method and device and storage medium |
| CN116595706A (en) * | 2023-02-28 | 2023-08-15 | 南方科技大学 | Method, electronic equipment and storage medium for inverting underground structure based on width learning |
| CN117371330A (en) * | 2023-10-30 | 2024-01-09 | 重庆大学 | Magnetotelluric two-dimensional deep learning inversion method based on traditional inversion guidance |
| CN117371330B (en) * | 2023-10-30 | 2024-06-14 | 重庆大学 | Magnetotelluric two-dimensional deep learning inversion method based on traditional inversion guidance |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113807020B (en) | 2023-04-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113807020B (en) | Magnetotelluric inversion method based on deep learning constraint | |
| CN110968826B (en) | Magnetotelluric deep neural network inversion method based on spatial mapping technology | |
| CN111126591B (en) | Magnetotelluric deep neural network inversion method based on space constraint technology | |
| Li et al. | Hybrid memetic pretrained factor analysis-based deep belief networks for transient electromagnetic inversion | |
| Smith et al. | Magnetotelluric inversion for minimum structure | |
| WO2022227206A1 (en) | Fully convolutional neural network-based magnetotelluric inversion method | |
| Nadiri et al. | Bayesian artificial intelligence model averaging for hydraulic conductivity estimation | |
| US11782183B2 (en) | Magnetotelluric inversion method based on fully convolutional neural network | |
| CN108897042A (en) | Content of organic matter earthquake prediction method and device | |
| Li et al. | An improved extreme learning machine algorithm for transient electromagnetic nonlinear inversion | |
| Feng et al. | Resistivity-depth imaging with the airborne transient electromagnetic method based on an artificial neural network | |
| Li et al. | Using wavelet packet denoising and a regularized ELM algorithm based on the LOO approach for transient electromagnetic inversion | |
| CN112468229B (en) | Atmospheric turbulence channel fading parameter estimation method based on mixed distribution model | |
| Dai et al. | Nonlinear inversion for electrical resistivity tomography based on chaotic DE-BP algorithm | |
| CN117390961A (en) | Method for solving electromagnetic response of ground model by using physical information neural network | |
| Wang et al. | Deep velocity generator: A plug-in network for FWI enhancement | |
| CN109239771A (en) | A kind of elastic wave imaging method based on non-homogeneous background media | |
| Zhdanov et al. | Regularized focusing inversion of marine CSEM data using minimum vertical support stabilizer | |
| Ma et al. | Improved seismic impedance inversion based on uncertainty analysis | |
| CN117371330B (en) | Magnetotelluric two-dimensional deep learning inversion method based on traditional inversion guidance | |
| Hu et al. | Spectral level prediction model of ocean ambient noise based on GA-LM-BP neural network | |
| Yang et al. | A wireless propagation model based on artificial intelligence technology | |
| CN115856882B (en) | Intelligent inversion and imaging method for multi-polarization geological radar data | |
| Zhao et al. | A new method to measure the dielectric constant in complex electromagnetic environment with black-box modeling regression method | |
| Tang et al. | A Deep learning estimation for probing Depth of Transient Electromagnetic Observation |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |