Deep learning-based interior inversion method and related device

Abstract

The invention discloses a deep learning-based interior inversion method and a related device, and relates to the technical field of geophysical geological resource exploration, and the method comprises the steps: inputting actual measurement magnetic anomaly data into a trained interior prediction model, and obtaining interior fluctuation data, the interior prediction model is a deep learning network model obtained by training with interior fluctuation simulation data as feature data and magnetic anomaly simulation data as label data. Compared with a traditional Parker-Oldenburg interface inversion method, a deep learning method is adopted to invert the interior, calculation is convenient and fast, high calculation precision can be obtained, and the interior with the high precision value can be rapidly obtained by conducting interior inversion through the method.

Description

Technical Field

[0001] The present application relates to the technical field of geophysical geological resource exploration, and in particular to a Curie point inversion method and related devices based on deep learning. Background Art

[0002] With the development of regional geophysical work, geologists and geophysicists have paid more and more attention to studying regional and deep geological structures and finding and developing deep mineral resources. The application field of magnetic exploration is also expanding. Determining the depth of the Curie isotherm is of great significance to studying regional and deep geological structures.

[0003] The iterative inversion method for fluctuating density interfaces in the prior art has the defect of low accuracy of interface inversion results. Summary of the invention

[0004] The purpose of this application is to provide a Curie point inversion method and related devices based on deep learning, which can improve the inversion accuracy of Curie point.

[0005] To achieve the above objectives, this application provides the following solutions:

[0006] In a first aspect, the present application provides a Curie point inversion method based on deep learning, comprising:

[0007] Obtain measured magnetic anomaly data within the study area;

[0008] The measured magnetic anomaly data is used as input, and the Curie point fluctuation data is output using a trained Curie point prediction model, wherein the Curie point prediction model is a deep learning network model trained with magnetic anomaly simulation data as feature data and magnetic anomaly simulation data as label data.

[0009] Optionally, the Curie prediction model is a U-Net network model.

[0010] Optionally, the training process of the Curie prediction model specifically includes:

[0011] The prism model is used to simulate the fluctuation of the Curie surface in the study area and obtain the Curie surface fluctuation simulation data;

[0012] Calculating magnetic anomaly simulation data according to the Curie point fluctuation simulation data and magnetic susceptibility, wherein the magnetic susceptibility is a random value;

[0013] Normalizing the Curie point fluctuation simulation data and the magnetic anomaly simulation data to obtain normalized Curie point fluctuation simulation data and normalized magnetic anomaly simulation data;

[0014] The normalized Curie point fluctuation simulation data is used as feature data, and the normalized magnetic anomaly simulation data is used as label data to train a Curie point prediction model.

[0015] Optionally, the number of input channels in the Curie prediction model is 1, and the number of output channels is 1.

[0016] Optionally, the activation function in the Curie prediction model uses a ReLU function in the middle layer and a Tanh function in the last layer.

[0017] Optionally, the upsampling process in the Curie prediction model is batch normalized.

[0018] Optionally, the optimizer in the Curie prediction model uses the Adam algorithm.

[0019] In a second aspect, the present application provides a computer device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement any of the Curie point inversion methods based on deep learning described above.

[0020] In a third aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the above-mentioned Curie point inversion methods based on deep learning.

[0021] In a fourth aspect, the present application provides a computer program product, including a computer program, which, when executed by a processor, implements any of the above-mentioned deep learning-based Curie point inversion methods.

[0022] According to the specific embodiments provided in this application, this application discloses the following technical effects:

[0023] The present application provides a Curie point inversion method and related device based on deep learning, which obtains Curie point fluctuation data by inputting measured magnetic anomaly data into a trained Curie point prediction model, wherein the Curie point prediction model is a deep learning network model trained with Curie point fluctuation simulation data as feature data and magnetic anomaly simulation data as label data. The use of deep learning method to invert the Curie point avoids the technical defects existing in the traditional Parker-Oldenburg interface inversion method, that is, "the high-frequency amplification effect of the exponential factor during inversion causes the calculation results to diverge violently, even when the iteration remainder is very small, making the stable convergence of the inversion problem particularly difficult, and the low noise resistance of this interface inversion method will affect the final inversion result". Compared with the traditional Parker-Oldenburg interface inversion method, the calculation is convenient and higher calculation accuracy can be obtained. The inversion of the Curie point using this method can quickly obtain the Curie point with higher accuracy value. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0025] Figure 1 This is an application environment diagram of a Curie point inversion method based on deep learning in Example 1 of the present application;

[0026] Figure 2 A schematic flow chart of a Curie point inversion method based on deep learning provided in Example 1 of the present application;

[0027] Figure 3 (a) is a schematic diagram of Curie surface fluctuations simulated by the prism model in Example 1 of the present application, Figure 3 (b) is a schematic diagram of magnetic anomaly simulated by the prism model in Example 1 of the present application;

[0028] Figure 4 (a) is a schematic diagram of Curie surface fluctuations simulated by the original prism in Example 1 of the present application, Figure 4 (b) is a schematic diagram of Curie surface fluctuations predicted by the U-net network in Example 1 of the present application, Figure 4 (c) is a schematic diagram showing the difference between the Curie surface fluctuation value simulated by the original prism in Example 1 of the present application and the Curie surface fluctuation value predicted by the U-net network. Figure 4 (d) is a schematic diagram of the percentage error between the Curie point fluctuation value simulated by the original prism in Example 1 of the present application and the Curie point fluctuation value predicted by the U-net network;

[0029] Figure 5 (a) is a schematic diagram of Curie surface fluctuations of the prism simulation in Example 1 of the present application, Figure 5 (b) is a schematic diagram of Curie surface fluctuations calculated using the Parker-Oldenburg method in Example 1 of the present application; Figure 5 (c) is a schematic diagram showing the difference between the Curie surface fluctuation calculated by the Parker-Oldenburg method in Example 1 of the present application and the Curie surface fluctuation simulated by the prism; Figure 5 (d) is a schematic diagram of the percentage error of Curie point fluctuation calculated using the Parker-Oldenburg method in Example 1 of the present application;

[0030] Figure 6 A schematic diagram of the structure of a computer device provided in Example 2 of the present application. DETAILED DESCRIPTION

[0031] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0032] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0033] Example 1

[0034] The Curie point inversion method based on deep learning provided in the embodiment of the present application can be applied to Figure 1In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be set up separately, integrated on the server 104, or placed on the cloud or other servers. The terminal 102 can send the measured magnetic anomaly data to be processed to the server 104. After the server 104 receives the measured magnetic anomaly data to be processed, it uses the trained Curie point prediction model to process the measured magnetic anomaly data to obtain the Curie point fluctuation data. The server 104 can feed back the obtained Curie point fluctuation data to the terminal 102. In addition, in some embodiments, the Curie point inversion method based on deep learning can also be implemented separately by the server 104 or the terminal 102, such as the terminal 102 can directly use the Curie point inversion method based on deep learning to process the measured magnetic anomaly data to be processed, or the server 104 can obtain the measured magnetic anomaly data from the data storage system and process it using the Curie point inversion method based on deep learning.

[0035] The terminal 102 may be, but is not limited to, various desktop computers, laptop computers, smart phones, tablet computers, IoT devices, and portable wearable devices. The IoT devices may be smart speakers, smart TVs, smart air conditioners, smart vehicle-mounted devices, etc. The portable wearable devices may be smart watches, smart bracelets, head-mounted devices, etc. The server 104 may be implemented as an independent server or a server cluster consisting of multiple servers, or may be a cloud server.

[0036] Relevant scholars have proposed various methods for calculating interface fluctuations based on known potential field anomalies, including inversion using equivalent prisms to fit the interface. This method first calculates the anomaly generated by each prism, and then sums it up to obtain the anomaly generated by the corresponding interface, but this method has not been widely used due to its slow calculation speed. In 1973, Parker introduced fast Fourier transform in the forward calculation of the potential field interface, making this method fast and powerful, and becoming a tool for quantitative interpretation of gravity and magnetic anomalies, suitable for comprehensive interpretation and joint inversion of geophysical data in oil and gas structural areas. On this basis, Oldenburg proposed an iterative inversion method for determining the fluctuating density interface. However, this method converges very slowly during the calculation process and takes a lot of time. Secondly, the high-frequency amplification effect of the exponential factor during inversion will cause the calculation results to diverge violently, even if the iterative residual is very small. This makes the stable convergence of the inversion problem particularly difficult. In addition, this interface inversion method has low noise resistance, which will affect the final inversion results.

[0037] With the continuous advancement of deep learning technology, some new network structures and training methods have also been proposed to improve the performance and efficiency of the model. For example, the U-Net network is a deep learning model that is mainly used for image segmentation tasks, especially in the field of medical image processing. The core feature of U-Net is its encoder-decoder structure, which contains dense skip connections. These connections can directly connect the feature maps of the upper layer with the feature maps of the lower layer, thereby retaining more contextual information and details.

[0038] This embodiment aims to solve the problems existing in the prior art and combines deep learning technology to provide a Curie point inversion method, which has practical significance for improving the accuracy of Curie point inversion. Figure 2 As shown, a Curie point inversion method based on deep learning is provided. The method is executed by a computer device, and can be executed by a computer device such as a terminal or a server alone, or by a terminal and a server together. In the embodiment of the present application, the method is applied to Figure 1 The server 104 in the example is used as an example to illustrate the method, which includes the following steps 201 to 208. Among them:

[0039] Step 201, obtaining measured magnetic anomaly data in the study area.

[0040] Step 202: Taking the measured magnetic anomaly data as input, the trained Curie point prediction model is used to output Curie point fluctuation data.

[0041] The Curie prediction model of this embodiment is a neural network model, preferably a U-net network model, which is a deep learning model mainly used for image segmentation tasks, especially in the field of medical image processing. The core feature of U-Net is its encoder-decoder structure, which contains dense skip connections that can directly connect the feature maps of the upper layer to the feature maps of the lower layer, thereby retaining more context information and details.

[0042] In order to make those skilled in the art more aware of the specific implementation process of the Curie point inversion method based on deep learning provided by this embodiment, a specific explanation is given below.

[0043] S1: Use a prism (i.e., prism model) to simulate the Curie surface fluctuation and calculate the magnetic anomaly of the prism.

[0044] The Curie depth varies in different regions, so the Curie depth fluctuation is set to a random value based on the global Curie depth.

[0045] The magnetic anomaly data caused by Curie point fluctuations are calculated. During the calculation process, the uncertainty of the magnetic susceptibility of underground rocks (the ability of rocks to be magnetized) is taken into account, and the magnetic susceptibility value will be included in the calculation model through random selection.

[0046] like Figure 3 (a) shows the Curie surface fluctuation diagram of the prism simulation. The top of the prism is set to a fixed -3 km, the bottom interface fluctuation is -5 km to -35 km, and the area of ​​the prism base is 400 km. 2 , a total of 1024 prisms are generated within the range of 640x640, the magnetic susceptibility is set to 0.4, the magnetic inclination is set to 90°, and the magnetic anomaly generated is as follows Figure 3 As shown in (b) in .

[0047] By simulating and analyzing the magnetic characteristics of the underground structure in a specific area, the underground geological conditions can be inferred. This method is very important in geophysical exploration and helps to find mineral resources or understand geological structures.

[0048] S2: Grid and normalize the Curie surface fluctuation data and magnetic anomaly simulated in S1 (i.e., Curie surface fluctuation simulation data and magnetic anomaly simulation data) so that the data values ​​are mapped between [-1, 1] to speed up the subsequent training process.

[0049] The data normalization calculation formula is as follows:

[0050]

[0051] Among them, d n is the normalized data, d is the original data of the forward modeling, and d max is the maximum value in the original data, d min is the minimum value in the original data.

[0052] S3: The U-net network structure and parameters are modified, that is, the number of input channels is changed to 1, and the number of output channels is changed to 1; batch normalization is performed in each upsampling process; the Leaky ReLU function is used in the middle layer of the activation function, the Tanh function is used in the last layer, Dropout regularization is added, and the Adam algorithm is used as the optimizer.

[0053] The Curie point fluctuation value simulated by the prism is used as the label data, and the corresponding magnetic anomaly calculated by the Curie point is used as the feature data, which is input into the U-net network model for training.

[0054] According to the error size of each training result, relevant parameters such as learning rate, batch size, weight decay, regularization weight (dropoutprob), etc. are adjusted so that the train loss and test loss curves show a downward trend and eventually stabilize.

[0055] S4: Input the measured magnetic anomaly data of the specific study area into the U-net network model trained in S3 to calculate the Curie point fluctuation value of the study area;

[0056] S5: Plot the results obtained in S4 to intuitively display the Curie point data obtained in the study area.

[0057] The U-net network model of this embodiment has been trained and only needs to be run in the MATLAB platform. The MATLAB used is matlab2024a.

[0058] In this embodiment, the magnetic anomaly generated by the prism forward modeling is first normalized to a value between [-1, 1], and then input into the trained network to obtain the undulation of the lower interface of the prism predicted by the U-net, that is, the undulation of the Curie surface. Figure 4 As shown, Figure 4 (a) in the figure simulates the Curie surface fluctuation. Figure 4 (b) in the figure is the Curie surface fluctuation predicted by the U-net network. By comparison Figure 4 (a) and 4(b) show that the Curie surface predicted by the U-net network has a high degree of similarity with the Curie surface simulated by the original prism, and the details of most areas are well presented. Figure 4 (c) is the difference between the Curie surface area simulated by the original prism and the Curie surface area predicted by the U-net network. Figure 4 In (c), most of the area differences are less than 5 km, which shows that the U-net network prediction effect is good. Figure 4 (d) is a percentage error diagram of the Curie point simulated by the prism and the Curie point predicted by the U-net network. Most of the errors are controlled below 20%, which is sufficient to prove the high-precision advantage of the Curie point inversion method based on deep learning provided by this embodiment.

[0059] In order to clarify the advantages of the method of this embodiment, the traditional Parker-Oldenburg method is used for comparison, and the same set of energy prisms is used to simulate the Curie point and the corresponding magnetic anomaly, and the Curie point is calculated to obtain Figure 5 The comparison results are shown.

[0060] Figure 5 (a) is the Curie surface fluctuation diagram simulated by the prism. Its simulation method and specific parameters are the same as Figure 3 is exactly the same as (a) in . Figure 5 (b) is the Curie surface plot calculated by the Parker-Oldenburg method. Comparing it with (a) in 5, we can see that only some areas of the Curie surface plot calculated by the Parker-Oldenburg method correspond to the Curie surface plot simulated by the prism. Some trends can correspond, but the resolution is low and cannot show more details. Figure 5 (b) and Figure 4 From (b) in the figure, it can be clearly seen that this embodiment has a higher accuracy in calculating the Curie point. Figure 5 (c) in the figure shows the difference between the Curie point calculated by the Parker method and the simulated Curie point. The analysis results show that the error of the Parker method is mostly within 8 kilometers, but some errors exceed 8 kilometers, and the overall accuracy is not high. Figure 5 (d) shows the percentage of error calculated by the Parker method. The data shows that most errors are controlled below 40%, but some errors exceed 60%. Figure 5 (c) and Figure 4 In (c), it can be seen that the error in calculating the Curie point in this embodiment is smaller. Figure 5 (d) and Figure 4 From (d) in the figure, it can be seen that the error ratio of the Curie point calculation in this embodiment is controlled below 20%, which means that the Curie point calculation of the present invention is more stable and reliable, and its calculation method is more convenient and less time-consuming than that of Parker-Oldenburg, and has the advantages of high efficiency and high precision.

[0061] In this embodiment, a Curie point inversion method based on a U-net network is obtained by simulating the Curie point with a prism and calculating magnetic anomalies as training data to train the U-net network model. Compared with the traditional Parker-Oldenburg interface inversion method, the calculation is convenient and higher calculation accuracy can be obtained. The inversion of the Curie point using this method can quickly obtain the Curie point with higher accuracy.

[0062] The present application also provides an application scenario, which applies the above-mentioned Curie point inversion method based on deep learning. Specifically: The Curie point inversion method based on deep learning provided in this embodiment can be applied in geophysical geological resource exploration scenarios. The geophysical geological resource exploration scenario includes a data acquisition link, a data processing link, and a resource prediction link. The Curie point inversion method based on deep learning provided in this embodiment belongs to the key technology in the data processing link.

[0063] Example 2

[0064] This embodiment provides a computer device, which may be a server or a terminal. Its internal structure diagram may be as follows: Figure 6 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. Among them, the processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store all data in the Curie point inversion method based on deep learning. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a Curie point inversion method based on deep learning provided in Example 1 is implemented.

[0065] Those skilled in the art will understand that Figure 6 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0066] Example 3

[0067] This embodiment provides a computer device, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, a Curie point inversion method based on deep learning provided in Embodiment 1 is implemented.

[0068] Example 4

[0069] This embodiment provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, the Curie point inversion method based on deep learning provided in Embodiment 1 is implemented.

[0070] Example 5

[0071] This embodiment provides a computer program product, including a computer program, which, when executed by a processor, implements a Curie point inversion method based on deep learning provided in Embodiment 1.

[0072] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0073] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0074] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.

[0075] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0076] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A Curie point inversion method based on deep learning, characterized in that: The Curie point inversion method based on deep learning includes: Obtain measured magnetic anomaly data within the study area; The measured magnetic anomaly data is used as input, and the Curie point fluctuation data is output using a trained Curie point prediction model, wherein the Curie point prediction model is a deep learning network model trained with Curie point fluctuation simulation data as feature data and magnetic anomaly simulation data as label data.

2. The Curie point inversion method based on deep learning according to claim 1, characterized in that: The Curie prediction model is a U-Net network model.

3. The Curie point inversion method based on deep learning according to claim 1, characterized in that: The training process of the Curie prediction model includes: The prism model is used to simulate the fluctuation of the Curie surface in the study area and obtain the Curie surface fluctuation simulation data; Calculating magnetic anomaly simulation data according to the Curie point fluctuation simulation data and magnetic susceptibility, wherein the magnetic susceptibility is a random value; Normalizing the Curie point fluctuation simulation data and the magnetic anomaly simulation data to obtain normalized Curie point fluctuation simulation data and normalized magnetic anomaly simulation data; The normalized Curie point fluctuation simulation data is used as feature data, and the normalized magnetic anomaly simulation data is used as label data to train a Curie point prediction model.

4. The Curie point inversion method based on deep learning according to claim 1, characterized in that: The number of input channels in the Curie prediction model is 1, and the number of output channels is 1.

5. The Curie point inversion method based on deep learning according to claim 1, characterized in that: The activation function in the Curie prediction model uses the ReLU function in the middle layer and the Tanh function in the last layer.

6. The Curie point inversion method based on deep learning according to claim 1, characterized in that: The upsampling process in the Curie prediction model is batch normalized.

7. The Curie point inversion method based on deep learning according to claim 1, characterized in that: The optimizer in the Curie prediction model uses the Adam algorithm.

8. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement a Curie point inversion method based on deep learning as described in any one of claims 1 to 7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, a Curie point inversion method based on deep learning as described in any one of claims 1 to 7 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, a Curie point inversion method based on deep learning as described in any one of claims 1 to 7 is implemented.

Images