A method, device and equipment for fast Bayesian inversion of airborne transient electromagnetic

Abstract

The present invention provides a method, device and equipment for rapid Bayesian inversion of airborne transient electromagnetic, which relates to the technical field of geophysical signal processing and analysis, wherein the method for rapid Bayesian inversion of airborne transient electromagnetic comprises: obtaining a resistivity model and a transmitting coil height; using the resistivity model and the transmitting coil height as input data and electromagnetic response data as output data, training a deep learning network model to obtain a forward simulation model; coupling the forward simulation model to a parallel Bayesian inversion framework to generate a target network model; and performing resistivity inversion and quantitative uncertainty analysis on the airborne transient electromagnetic data to be analyzed based on the target network model to obtain an analysis result. By integrating the deep learning forward simulator into the Bayesian inversion framework, rapid Bayesian inversion and quantitative uncertainty estimation of airborne transient electromagnetic data can be achieved, greatly improving the inversion efficiency.

Description

Technical Field

[0001] The present invention relates to the technical field of geophysical signal processing and analysis, and in particular to an airborne transient electromagnetic fast Bayesian inversion method, device and equipment. Background Art

[0002] Transient electromagnetic method is a geophysical method that uses an ungrounded return line or a grounded line source to emit a pulsed magnetic field underground, and uses a coil or a grounding electrode to observe the secondary induced eddy current field caused in the underground medium during the interval of the pulsed magnetic field, thereby detecting the medium. By observing the change law of the secondary induced eddy current field over time, the resistivity characteristics of the geology can be studied, thereby solving related geological problems. Inversion of underground medium resistivity by secondary induced eddy current field is one of the important steps in transient electromagnetic exploration. Transient electromagnetic inversion mainly reconstructs unknown geological model parameters (such as thickness and resistivity, etc.) that cannot be directly observed from a set of transient electromagnetic response data. It is highly nonlinear and non-unique. Traditional inversion methods can usually only obtain a single inversion result, and detailed uncertainty analysis of the estimated model parameters cannot be performed. The lack of quantitative uncertainty estimation limits the understanding of the inversion results, which may lead to misjudgment in actual engineering applications.

[0003] Bayesian theory provides a probabilistic framework for quantifying parameter uncertainty in inversion problems. The Bayesian inversion algorithm is a Monte Carlo search method that can be used for both optimization and uncertainty estimation. In Bayesian inversion, sampling-based methods are usually used to sample from the posterior probability density to obtain asymptotically unbiased samples. However, sampling-based methods are computationally intensive and often require hundreds of thousands of iterations to obtain meaningful results, resulting in a very time-consuming inversion process and extremely high computational costs. Summary of the invention

[0004] The purpose of the present invention is to solve the problem that the inversion process is time-consuming and the calculation cost is extremely high in Bayesian inversion. The present invention provides an airborne transient electromagnetic fast Bayesian inversion method, device and equipment.

[0005] The technical solution of the embodiment of the present application is implemented as follows:

[0006] A first aspect of an embodiment of the present application provides an airborne transient electromagnetic fast Bayesian inversion method, comprising:

[0007] Obtaining a resistivity model and a transmitting coil height; the resistivity model represents the resistivity distribution of the underground medium, and the transmitting coil height represents the flight altitude of the airborne transient electromagnetic system;

[0008] The resistivity model and the transmitting coil height are used as input data, and the electromagnetic response data are used as output data to train the deep learning network model to obtain a forward simulation model;

[0009] Coupling the forward simulation model to a parallel Bayesian inversion framework to generate a target network model;

[0010] Based on the target network model, resistivity inversion and quantitative uncertainty analysis are performed on the airborne transient electromagnetic data to be analyzed to obtain analysis results.

[0011] Optionally, the resistivity model and the transmitting coil height are used as input data, and the electromagnetic response data are used as output data to train the deep learning network model to obtain the forward simulation model, including:

[0012] Normalizing the resistivity model, the transmitting coil height and the electromagnetic response data to generate normalized data;

[0013] Dividing the normalized data into low magnetic moment excitation data and high magnetic moment excitation data based on a preset magnetic field excitation intensity threshold;

[0014] The deep learning network model is trained using the low magnetic moment excitation data and the high magnetic moment excitation data respectively to obtain a forward simulation model.

[0015] Optionally, the method of respectively using the low magnetic moment excitation data and the high magnetic moment excitation data to train a deep learning network model to obtain a forward simulation model further includes:

[0016] The low magnetic moment excitation data is used to train the deep learning network model, and the mean absolute error is used as the loss function of the deep learning network model training for iterative training until the iteration stop condition is met to obtain a forward simulation model.

[0017] Optionally, the method of respectively using the low magnetic moment excitation data and the high magnetic moment excitation data to train a deep learning network model to obtain a forward simulation model further includes:

[0018] The high magnetic moment excitation data is used to train the deep learning network model, and the mean square error is used as the loss function of the deep learning network model training for iterative training until the iteration stop condition is met to obtain a forward simulation model.

[0019] Optionally, the normalization formula of the resistivity model is:

[0020]

[0021] Among them, ρ nis the normalized resistivity used to construct the composite network model; ρ is the resistivity used to construct the composite network model; ρ max and ρ min are the maximum and minimum values ​​of resistivity used to construct the composite network model; a is -1; b is 1;

[0022] The normalization formula for the transmitting coil height is:

[0023]

[0024] Among them, r n is the normalized flight altitude of the airborne transient electromagnetic system; r is the flight altitude of the airborne transient electromagnetic system used to construct the composite network model; r max and r mi n is the maximum or minimum value of the flight altitude of the aviation transient electromagnetic system.

[0025] Optionally, before taking the resistivity model and the transmitting coil height as input data and the electromagnetic response data as output data to train the deep learning network model to obtain the forward simulation model, the method further includes:

[0026] Inputting the resistivity model into two convolutional pooling layers respectively, extracting characteristic maps from the observed airborne transient electromagnetic data used to construct the forward simulation model;

[0027] The characteristic map and the transmitting coil height are combined and input into a long short-term memory neural network to generate electromagnetic response data.

[0028] Optionally, the parallel Bayesian inversion framework is based on the Matlab open source code SIPPI.

[0029] The second aspect of the embodiment of the present application provides an airborne transient electromagnetic fast Bayesian inversion device, including: an acquisition module, a training module, a generation module and an analysis module, wherein:

[0030] The acquisition module is configured to acquire a resistivity model and a transmitting coil height; the resistivity model represents the resistivity distribution of the underground medium, and the transmitting coil height represents the flight altitude of the airborne transient electromagnetic system;

[0031] The training module is configured to use the resistivity model and the transmitting coil height as input data and the electromagnetic response data as output data to train the deep learning network model to obtain a forward simulation model;

[0032] The generation module is configured to couple the forward simulation model to a parallel Bayesian inversion framework to generate a target network model;

[0033] The analysis module is configured to perform resistivity inversion and quantitative uncertainty analysis on the airborne transient electromagnetic data to be analyzed based on the target network model to obtain analysis results.

[0034] A third aspect of an embodiment of the present application provides an electronic device, comprising a processor and a memory; the memory stores a computer program, wherein the computer program, when executed by the processor, implements the airborne transient electromagnetic fast Bayesian inversion method described in the first aspect.

[0035] A fourth aspect of an embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described in the first aspect.

[0036] Compared with the prior art, the technical solution provided by this application has the following beneficial effects:

[0037] The present invention provides a method, device and equipment for rapid Bayesian inversion of airborne transient electromagnetic. By acquiring a resistivity model and a transmitting coil height, and using the resistivity model and the transmitting coil height as input data, and using electromagnetic response data as output data, a deep learning network model is trained to obtain a forward simulation model; the forward simulation model is coupled to a parallel Bayesian inversion framework to generate a target network model; resistivity inversion and quantitative uncertainty analysis are performed on the airborne transient electromagnetic data to be analyzed based on the target network model to obtain analysis results. By integrating the deep learning forward simulator into the Bayesian inversion framework, rapid Bayesian inversion and quantitative uncertainty estimation of airborne transient electromagnetic data are achieved, greatly improving the inversion efficiency, so that it has broad application prospects in various geophysical exploration tasks, significantly improving the exploration efficiency and the reliability of the results, and providing strong technical support for geological research and resource development. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 A schematic diagram of a flow chart of a method for rapid Bayesian inversion of airborne transient electromagnetics provided in an embodiment of the present application;

[0039] Figure 2 A schematic diagram of the structure of the deep learning network model provided in the embodiment of the present application;

[0040] Figure 3 A schematic diagram of the structure of an airborne transient electromagnetic fast Bayesian inversion device provided in an embodiment of the present application;

[0041] Figure 4 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0042] Below, embodiments of the present application will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the present application. In the following detailed description, for ease of explanation, many specific details are set forth to provide a comprehensive understanding of the embodiments of the present application. However, it is apparent that one or more embodiments may also be implemented without these specific details. In addition, in the following description, descriptions of known structures and technologies are omitted to avoid unnecessary confusion of the concepts of the present application.

[0043] The terms used herein are only for describing specific embodiments and are not intended to limit the present application. The terms "include", "comprising", etc. used herein indicate the existence of the features, steps, operations and / or components, but do not exclude the existence or addition of one or more other features, steps, operations or components.

[0044] All terms (including technical and scientific terms) used herein have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification, and should not be interpreted in an idealized or overly rigid manner.

[0045] Some block diagrams and / or flow charts are shown in the accompanying drawings. It should be understood that some blocks or combinations thereof in the block diagrams and / or flow charts may be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that these instructions, when executed by the processor, may create a device for implementing the functions / operations described in these block diagrams and / or flow charts.

[0046] In some embodiments, see Figure 1 , Figure 1 A schematic diagram of a flow chart of a method for rapid Bayesian inversion of airborne transient electromagnetic provided in an embodiment of the present application; the method for rapid Bayesian inversion of airborne transient electromagnetic provided in an embodiment of the present application comprises:

[0047] S110, obtaining a resistivity model and a transmitting coil height; the resistivity model represents the resistivity distribution of the underground medium, and the transmitting coil height represents the flight altitude of the airborne transient electromagnetic system.

[0048] The resistivity model can be a two-dimensional or three-dimensional grid that represents the resistivity distribution of the underground medium, which can be converted into vector or matrix form. The transmitting coil height is a scalar value that represents the height of the transmitting coil from the ground. Here, data can be collected in the survey area by dividing the survey line and the detailed survey line. The survey line layout is relatively sparse, aiming to fully cover the survey area, while the detailed survey line layout is more dense, ensuring detailed coverage of specific areas.

[0049] S120, using the resistivity model and the transmitting coil height as input data and the electromagnetic response data as output data, training the deep learning network model to obtain a forward simulation model.

[0050] The electromagnetic response data can be time series or frequency domain data, such as voltage, current or other electromagnetic field parameters measured at the receiving coil. The deep learning network model includes an input layer, a hidden layer and an output layer. The input layer receives the encoding of the resistivity model and the height of the transmitting coil. According to the complexity of the problem, multiple hidden layers can be designed, each containing a certain number of neurons. The output layer is used to output the predicted value of the electromagnetic response.

[0051] In some embodiments, S120, using the resistivity model and the transmitting coil height as input data and the electromagnetic response data as output data, training the deep learning network model to obtain a forward simulation model includes:

[0052] Normalizing the resistivity model, the transmitting coil height and the electromagnetic response data to generate normalized data;

[0053] The normalized data is divided into low magnetic moment excitation data and high magnetic moment excitation data based on a preset magnetic field excitation intensity threshold;

[0054] The deep learning network model is trained using low magnetic moment excitation data and high magnetic moment excitation data respectively to obtain a forward simulation model.

[0055] In this embodiment, normalizing the input data and output data to the same scale can be more helpful for network training. If the resistivity model is two-dimensional or three-dimensional, it is flattened into a one-dimensional array, or a convolutional neural network (CNN) is used to directly process the two-dimensional / three-dimensional data. High magnetic moment excitation can produce a stronger electromagnetic field, thereby having a greater exploration depth. The electromagnetic field generated by low magnetic moment excitation is relatively weak, and the exploration depth is small, which is more suitable for fine detection of shallow geological structures or abnormal bodies.

[0056] In some embodiments, the deep learning network model is trained using low magnetic moment excitation data and high magnetic moment excitation data to obtain a forward simulation model, further comprising:

[0057] The deep learning network model is trained using low magnetic moment excitation data, and the mean absolute error is used as the loss function for deep learning network model training for iterative training until the iteration stopping condition is met to obtain the forward simulation model.

[0058] In this embodiment, the iterative training process can be to calculate the predicted value through forward propagation, update the weight through back propagation, and repeat this process until the stop condition is met (such as reaching the maximum number of iterations, the loss value no longer decreases significantly, etc.). Here, the mean absolute error is used as the loss function for deep learning network model training to ensure the effect of using low magnetic moment excitation data to train the deep learning network model.

[0059] In some embodiments, the deep learning network model is trained using low magnetic moment excitation data and high magnetic moment excitation data to obtain a forward simulation model, further comprising:

[0060] The deep learning network model is trained using high magnetic moment excitation data, and the mean square error is used as the loss function of the deep learning network model training for iterative training until the iteration stopping condition is met to obtain the forward simulation model.

[0061] In this embodiment, the iterative training process can be the same as the above embodiment. Here, the mean square error is used as the loss function of the deep learning network model training to ensure the effect of using high magnetic moment excitation data to train the deep learning network model.

[0062] In some embodiments, the normalized formula of the resistivity model is:

[0063]

[0064] Among them, ρ n is the normalized resistivity used to construct the composite network model; ρ is the resistivity used to construct the composite network model; ρ max and ρ min are the maximum and minimum values ​​of resistivity used to construct the composite network model; a is -1; b is 1;

[0065] The normalized formula for the transmitting coil height is:

[0066]

[0067] Among them, r n is the normalized flight altitude of the airborne transient electromagnetic system; r is the flight altitude of the airborne transient electromagnetic system used to construct the composite network model; r max and r mi n is the maximum or minimum value of the flight altitude of the aviation transient electromagnetic system.

[0068] In some embodiments, at S120, before training the deep learning network model using the resistivity model and the transmitting coil height as input data and the electromagnetic response data as output data to obtain the forward simulation model, the method further includes:

[0069] The resistivity model is fed into two convolutional pooling layers to extract feature maps from the observed airborne transient electromagnetic data used to construct the forward simulation model.

[0070] The feature map and the transmitting coil height are combined and input into the long short-term memory neural network to generate electromagnetic response data.

[0071] In this embodiment, a sliding window operation can be performed on the resistivity model through multiple convolution kernels (filters) to extract local features. Each convolution kernel generates a feature map, which contains different feature information of the model. On this basis, the pooling layer is used to reduce the dimension (spatial size) of the feature map while retaining important features, reducing the amount of calculation and preventing overfitting. After extracting the feature map of the resistivity model, the height of the transmitting coil is added to the feature map as an additional feature dimension, or it is implemented as part of the input in the subsequent neural network layer. The merged feature map and the height of the transmitting coil are used as the input of the LSTM. The LSTM network generates the corresponding electromagnetic response data by learning the mapping relationship between the input data (i.e., the feature map of the resistivity model and the height of the transmitting coil) and the electromagnetic response data. These electromagnetic response data may include information such as the electromagnetic field strength and phase at different time points.

[0072] In an example, see Figure 2 , Figure 2 Structural diagram of the deep learning network model provided in the embodiment of the present application; The resistivity model is processed by two convolutional layers to obtain a feature map focused on the resistivity model. Then, the dimension of the flight height data is adjusted to match the characteristics of the resistivity model, and the two are spliced ​​into new features to ensure that both the resistivity and the flight height are effectively incorporated into the network information. Finally, the synthesized features are input into the bidirectional long short-term memory layer and output through the fully connected layer. The main function of the convolutional neural network (CNN) is to extract spatial features from the resistivity model. CNN can effectively identify and extract local patterns and features in resistivity data through its convolutional layer, which is crucial for capturing the details of underground structures. The main function of the bidirectional long short-term memory network (Bi-LSTM) is to process and memorize long-term dependencies in time series data. In ATEM data analysis, LSTM helps to understand and predict changes in electromagnetic response over time, especially the continuity and changes when processing HM data and LM data.

[0073] S130, coupling the forward simulation model to a parallel Bayesian inversion framework to generate a target network model.

[0074] In some embodiments, the parallel Bayesian inversion framework is based on the Matlab open source code SIPPI.

[0075] The parallel Bayesian inversion framework can process multiple sets of data simultaneously. SIPPI is a powerful Matlab open source toolbox that can be used to solve inverse problems with complex prior information. It provides a variety of methods and tools to sample posterior probability density functions and supports the solution of linear and nonlinear inverse problems.

[0076] In one example, multiple different prior information are defined, each of which corresponds to a set of data, and the number can range from 1 to 2000, where 2000 is the limit for achieving maximum efficiency. The resistivity models obtained by sampling are combined into a matrix and input into a forward modeling simulator based on a neural network to calculate the corresponding electromagnetic response. Subsequently, the likelihood function value and the corresponding acceptance rate are calculated for each set of data. Finally, by comparing with the randomly generated values, it is determined whether the current model is accepted and added to the respective Markov chain. This parallel processing method can greatly improve the efficiency of Bayesian inversion.

[0077] S140, based on the target network model, resistivity inversion and quantitative uncertainty analysis are performed on the airborne transient electromagnetic data to be analyzed to obtain analysis results.

[0078] Inputting the airborne transient electromagnetic data to be analyzed into the above-mentioned trained target network model can obtain resistivity distribution maps, uncertainty assessment results and geological interpretations. By analyzing the resistivity distribution map obtained by inversion, it is possible to identify abnormal resistivity areas of underground media, such as low-resistance bodies (such as aquifers, minerals, etc.) and high-resistance bodies (such as rocks, dry formations, etc.). According to the results of quantitative uncertainty analysis, the reliability and accuracy of the inversion results can be determined. For areas with high uncertainty, careful interpretation and further verification are required. Combined with geological background knowledge, other geophysical data and exploration experience, the inversion results can be geologically interpreted and reasonable geological models and exploration suggestions can be proposed.

[0079] The embodiment of the present application integrates the deep learning forward modeling simulator into the Bayesian inversion framework to achieve fast Bayesian inversion and quantitative uncertainty estimation of airborne transient electromagnetic data, greatly improving the inversion efficiency. This enables it to have broad application prospects in various geophysical exploration tasks, significantly improves the exploration efficiency and the reliability of the results, and provides strong technical support for geological research and resource development.

[0080] In some embodiments, see Figure 3 , Figure 3 The schematic diagram of the structure of an airborne transient electromagnetic fast Bayesian inversion device provided in the embodiment of the present application. The airborne transient electromagnetic fast Bayesian inversion device 300 provided in the embodiment of the present application includes: an acquisition module 310, a training module 320, a generation module 330 and an analysis module 340, wherein:

[0081] The acquisition module 310 is configured to acquire a resistivity model and a transmitting coil height; the resistivity model represents the resistivity distribution of the underground medium, and the transmitting coil height represents the flight altitude of the airborne transient electromagnetic system;

[0082] The training module 320 is configured to use the resistivity model and the transmitting coil height as input data and the electromagnetic response data as output data to train the deep learning network model to obtain a forward simulation model;

[0083] A generation module 330 is configured to couple the forward simulation model to a parallel Bayesian inversion framework to generate a target network model;

[0084] The analysis module 340 is configured to perform resistivity inversion and quantitative uncertainty analysis on the airborne transient electromagnetic data to be analyzed based on the target network model to obtain analysis results.

[0085] In some embodiments, the training module 320 is specifically configured as follows:

[0086] Normalizing the resistivity model, the transmitting coil height and the electromagnetic response data to generate normalized data;

[0087] The normalized data is divided into low magnetic moment excitation data and high magnetic moment excitation data based on a preset magnetic field excitation intensity threshold;

[0088] The deep learning network model is trained using low magnetic moment excitation data and high magnetic moment excitation data respectively to obtain a forward simulation model.

[0089] In some embodiments, the training module 320 is further configured to:

[0090] The deep learning network model is trained using low magnetic moment excitation data, and the mean absolute error is used as the loss function of the deep learning network model training for iterative training until the iteration stopping condition is met to obtain the forward simulation model.

[0091] In some embodiments, the training module 320 is further configured to:

[0092] The deep learning network model is trained using high magnetic moment excitation data, and the mean square error is used as the loss function of the deep learning network model training for iterative training until the iteration stopping condition is met to obtain the forward simulation model.

[0093] In some embodiments, the normalized formula of the resistivity model is:

[0094]

[0095] Among them, ρ nis the normalized resistivity used to construct the composite network model; ρ is the resistivity used to construct the composite network model; ρ max and ρ min are the maximum and minimum values ​​of resistivity used to construct the composite network model; a is -1; b is 1;

[0096] The normalized formula for the transmitting coil height is:

[0097]

[0098] Among them, r n is the normalized flight altitude of the airborne transient electromagnetic system; r is the flight altitude of the airborne transient electromagnetic system used to construct the composite network model; r max and r min They are respectively the maximum or minimum flight altitude of the aviation transient electromagnetic system.

[0099] In some embodiments, the airborne transient electromagnetic fast Bayesian inversion device further includes an extraction module; the extraction module is specifically used for:

[0100] The resistivity model is fed into two convolutional pooling layers to extract feature maps from the observed airborne transient electromagnetic data used to construct the forward simulation model.

[0101] The feature map and the transmitting coil height are combined and input into the long short-term memory neural network to generate electromagnetic response data.

[0102] In some embodiments, the parallel Bayesian inversion framework is based on the Matlab open source code SIPPI.

[0103] The airborne transient electromagnetic fast Bayesian inversion device provided in the embodiment of the present application can implement each process in the embodiment corresponding to the above-mentioned airborne transient electromagnetic fast Bayesian inversion method, and will not be described again here to avoid repetition.

[0104] It should be noted that the airborne transient electromagnetic fast Bayesian inversion device provided in the embodiment of the present application and the airborne transient electromagnetic fast Bayesian inversion method provided in the embodiment of the present application are based on the same application concept. Therefore, the specific implementation of this embodiment can refer to the implementation of the aforementioned airborne transient electromagnetic fast Bayesian inversion method, and the repeated parts will not be repeated.

[0105] In some embodiments, see Figure 4 , Figure 4 The electronic device 400 provided in the embodiment of the present application includes a processor 410 and a memory 420; the memory 420 stores a computer program, wherein the computer program implements the above-mentioned airborne transient electromagnetic fast Bayesian inversion method when executed by the processor.

[0106] Specifically, the processor 410 may include, for example, a general-purpose microprocessor, an instruction set processor and / or a related chipset and / or a special-purpose microprocessor (e.g., an application-specific integrated circuit (ASIC)), etc. The processor 410 may also include an onboard memory for cache purposes. The processor 410 may be a single processing unit or multiple processing units for executing different actions of the method flow according to the embodiment of the present application.

[0107] The memory 420 may be any medium capable of containing, storing, conveying, propagating or transmitting instructions. For example, the memory 420 may include, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared or semiconductor system, device, component or propagation medium. Specific examples of the memory 420 include: a magnetic storage device, such as a magnetic tape or a hard disk (HDD); an optical storage device, such as a compact disk (CD-ROM); a random access memory (RAM) or flash memory; and / or a wired / wireless communication link.

[0108] The present application also provides a computer-readable medium on which a computer program is stored, and when the program is executed by a processor, the above-mentioned method for fast Bayesian inversion of airborne transient electromagnetic is implemented. The computer-readable medium may be included in the device / apparatus / system described in the above-mentioned embodiment; or it may exist independently without being assembled into the device / apparatus / system. The above-mentioned computer-readable medium carries one or more programs, and when the above-mentioned one or more programs are executed, the method according to the embodiment of the present application is implemented.

[0109] According to an embodiment of the present application, a computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium or any combination of the two. A computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present application, a computer-readable storage medium may be any tangible medium containing or storing a program that may be used by or in combination with an instruction execution system, device or device. In the present application, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, wherein a computer-readable program code is carried. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer-readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wired, optical cable, radio frequency signal, etc., or any suitable combination of the above.

[0110] Those skilled in the art will appreciate that the features described in the various embodiments and / or claims of the present application may be combined and / or combined in a variety of ways, even if such combinations or combinations are not explicitly described in the present application. In particular, without departing from the spirit and teachings of the present application, the features described in the various embodiments and / or claims of the present application may be combined and / or combined in a variety of ways. All of these combinations and / or combinations fall within the scope of the present application. Therefore, the scope of the present application should not be limited to the above-described embodiments, but should be determined not only by the appended claims, but also by the equivalents of the appended claims.

Claims

1. A fast Bayesian inversion method for airborne transient electromagnetic, characterized in that: include: Obtain resistivity model and transmitting coil height; The resistivity model represents the resistivity distribution of the underground medium, and the transmitting coil height represents the flight altitude of the airborne transient electromagnetic system; The resistivity model and the transmitting coil height are used as input data, and the electromagnetic response data are used as output data to train the deep learning network model to obtain a forward simulation model; Coupling the forward simulation model to a parallel Bayesian inversion framework to generate a target network model; Based on the target network model, resistivity inversion and quantitative uncertainty analysis are performed on the airborne transient electromagnetic data to be analyzed to obtain analysis results; The method uses the resistivity model and the transmitting coil height as input data and the electromagnetic response data as output data to train the deep learning network model to obtain a forward simulation model, including: Normalizing the resistivity model, the transmitting coil height and the electromagnetic response data to generate normalized data; Dividing the normalized data into low magnetic moment excitation data and high magnetic moment excitation data based on a preset magnetic field excitation intensity threshold; The deep learning network model is trained using the low magnetic moment excitation data and the high magnetic moment excitation data respectively to obtain a forward simulation model; the deep learning network model is trained using the low magnetic moment excitation data and the high magnetic moment excitation data respectively to obtain a forward simulation model, further comprising: The deep learning network model is trained using the low magnetic moment excitation data, and the mean absolute error is used as the loss function of the deep learning network model training for iterative training until an iteration stop condition is met to obtain a forward simulation model; The high magnetic moment excitation data is used to train the deep learning network model, and the mean square error is used as the loss function of the deep learning network model training for iterative training until the iteration stop condition is met to obtain a forward simulation model.

2. The method for rapid Bayesian inversion of airborne transient electromagnetic according to claim 1, characterized in that: The normalized processing formula of the resistivity model is: Among them, ρ n is the normalized resistivity used to construct the composite network model; ρ is the resistivity used to construct the composite network model; ρ max and ρ min are the maximum and minimum values ​​of resistivity used to construct the composite network model; a is -1; b is 1; The normalization formula for the transmitting coil height is: Among them, r n is the normalized flight altitude of the airborne transient electromagnetic system; r is the flight altitude of the airborne transient electromagnetic system used to construct the composite network model; r max and r min They are the maximum and minimum values ​​of the flight altitude of the aviation transient electromagnetic system respectively.

3. The method for rapid Bayesian inversion of airborne transient electromagnetic according to claim 1, characterized in that: Before the resistivity model and the transmitting coil height are used as input data and the electromagnetic response data are used as output data to train the deep learning network model to obtain the forward simulation model, the method further includes: Inputting the resistivity model into two convolutional pooling layers respectively, extracting characteristic maps from the observed airborne transient electromagnetic data used to construct the forward simulation model; The characteristic map and the transmitting coil height are combined and input into a long short-term memory neural network to generate electromagnetic response data.

4. The method for rapid Bayesian inversion of airborne transient electromagnetic according to claim 1, characterized in that: The parallel Bayesian inversion framework is based on the Matlab open source code SIPPI.

5. An airborne transient electromagnetic fast Bayesian inversion device, characterized in that: include: Acquisition module, training module, generation module and analysis module, among which, The acquisition module is configured to acquire a resistivity model and a transmitting coil height; the resistivity model represents the resistivity distribution of the underground medium, and the transmitting coil height represents the flight altitude of the airborne transient electromagnetic system; The training module is configured to use the resistivity model and the transmitting coil height as input data and the electromagnetic response data as output data to train the deep learning network model to obtain a forward simulation model; The generation module is configured to couple the forward simulation model to a parallel Bayesian inversion framework to generate a target network model; The analysis module is configured to perform resistivity inversion and quantitative uncertainty analysis on the airborne transient electromagnetic data to be analyzed based on the target network model to obtain analysis results; The method uses the resistivity model and the transmitting coil height as input data and the electromagnetic response data as output data to train the deep learning network model to obtain a forward simulation model, including: Normalizing the resistivity model, the transmitting coil height and the electromagnetic response data to generate normalized data; Dividing the normalized data into low magnetic moment excitation data and high magnetic moment excitation data based on a preset magnetic field excitation intensity threshold; The deep learning network model is trained using the low magnetic moment excitation data and the high magnetic moment excitation data respectively to obtain a forward simulation model; The deep learning network model is trained by respectively using the low magnetic moment excitation data and the high magnetic moment excitation data to obtain a forward simulation model, and further includes: The deep learning network model is trained using the low magnetic moment excitation data, and the mean absolute error is used as the loss function of the deep learning network model training for iterative training until an iteration stop condition is met to obtain a forward simulation model; The high magnetic moment excitation data is used to train the deep learning network model, and the mean square error is used as the loss function of the deep learning network model training for iterative training until the iteration stop condition is met to obtain a forward simulation model.

6. An electronic device comprising a processor and a memory; the memory stores a computer program, wherein: When the computer program is executed by the processor, the computer program implements the airborne transient electromagnetic fast Bayesian inversion method according to any one of claims 1 to 4.

7. A computer-readable storage medium having a computer program stored thereon, wherein when the program is executed by a processor, the steps of the method according to any one of claims 1 to 4 are implemented.

Images