Time domain aviation electromagnetic inversion method based on deep learning

Abstract

The invention belongs to the technical field of crossing of geophysical exploration and artificial intelligence, and discloses a time domain aviation electromagnetic inversion method based on deep learning, which comprises the following steps of: performing forward modeling on a recursive algorithm, a received stratum thickness, a random flight height and an acquired actually measured stratum resistivity to obtain magnetic field response of each frequency point; performing transformation processing on the magnetic field response through Hankel transformation kernel function integration to obtain a frequency domain magnetic induction intensity component; generating a time domain step response for the frequency domain magnetic induction intensity component by using a Hankel transformation coefficient of a preset order, and performing convolution processing on the time domain step response through a Gaussian integral node to obtain electromagnetic response data in a vertical direction; using the electromagnetic response data, the stratum thickness and the random flight height as a training set to train a long short-term memory network to be trained to obtain the long short-term memory network; and inputting actually measured electromagnetic response data into the long short-term memory network for inversion to obtain a predicted target flight height and target stratum resistivity.

Description

Technical Field

[0001] The present invention relates to the cross - technical field of geophysical exploration and artificial intelligence, and relates to, but is not limited to, the time - domain airborne electromagnetic inversion method based on deep learning. Background Technique

[0002] The airborne electromagnetic method is a geophysical exploration method with mobile platforms such as airplanes and airships as system carriers. Due to its advantages of high data acquisition efficiency, high detection resolution, and the need for no ground staff to approach the target area, the airborne electromagnetic method has been widely used in geological surveys, mineral resource exploration, underground water resource assessment, and environmental monitoring in areas with complex topographic and geological conditions. When the airborne electromagnetic method is collecting data, it has been in a bumpy state, resulting in errors in the recorded flight altitude of the data acquisition system; in addition, some survey areas of the airborne electromagnetic are covered by vegetation, which may cause the loss of flight altitude data for some survey lines. Considering that the flight altitude is a first - level influencing parameter of airborne electromagnetic data and has a very strong impact on the data, it is necessary to restore the system flight altitude during the inversion and interpretation of airborne electromagnetic data.

[0003] In related technologies, the meter - level flight altitude recording error is restored based on the obtained measured flight altitude, and a high - precision fine - modeling result of electromagnetic data is obtained. However, when the flight altitude data is lost or the recording error is large, these methods often cannot obtain accurate inversion results of airborne electromagnetic data.

[0004] Therefore, how to quickly and accurately determine the flight altitude and formation resistivity has become an urgent problem to be solved. Summary of the Invention

[0005] In view of this, the time - domain airborne electromagnetic inversion method based on deep learning provided by the embodiments of the present invention at least solves the problem that related technologies cannot effectively and accurately invert the flight altitude and formation resistivity.

[0006] According to the first aspect of the embodiments of the present invention, a time - domain airborne electromagnetic inversion method based on deep learning is provided, including: Calculating the magnetic field response at each frequency point through a frequency - domain forward - modeling engine based on a recurrence algorithm, the received formation thickness, the random flight altitude, and the measured formation resistivity data obtained; Performing a transformation process on the magnetic field response through the Hankel transform kernel function integral to obtain the frequency - domain magnetic induction intensity component; Generating a time - domain step response for the frequency - domain magnetic induction intensity component by using a Hankel transform coefficient of a preset order, and performing a convolution process on the time - domain step response based on Gaussian integration nodes to obtain the electromagnetic response data in the vertical direction; Using the electromagnetic response data, the formation thickness, and the random flight altitude as a training set, and training a long short-term memory network to be trained using the training set to obtain a long short-term memory network; Inputting the measured electromagnetic response data into the long short-term memory network for inversion to obtain the predicted target flight altitude and target formation resistivity.

[0007] According to the second aspect of the embodiments of the present invention, there is provided an electronic device, including: a processor, a memory, a communication interface, and a communication bus, where the processor, the memory, and the communication interface complete mutual communication through the communication bus; the memory is used to store at least one executable instruction, and the executable instruction causes the processor to perform the operations corresponding to the method described in the first aspect.

[0008] According to the third aspect of the embodiments of the present invention, there is provided a computer storage medium, on which a computer program is stored, and when the program is executed by a processor, it implements the method described in the first aspect.

[0009] According to the solution provided by the embodiments of the present invention, the magnetic field response at each frequency point is calculated by a frequency-domain forward modeling engine based on a recursive algorithm, the received formation thickness, the random flight altitude, and the data of the measured formation resistivity; the magnetic field response is transformed by integrating the Hankel transform kernel function to obtain the frequency-domain magnetic induction intensity component; the frequency-domain magnetic induction intensity component is used to generate a time-domain step response using a preset-order Hankel transform coefficient, and the time-domain step response is convolved based on Gaussian integration nodes to obtain the electromagnetic response data in the vertical direction; using the electromagnetic response data, the formation thickness, and the random flight altitude as a training set, and training a long short-term memory network to be trained using the training set to obtain a long short-term memory network; inputting the measured electromagnetic response data into the long short-term memory network for inversion to obtain the predicted target flight altitude and target formation resistivity. In this process, a training set composed of the random flight altitude, the measured formation resistivity, and the calculated airborne electromagnetic response is established. Finally, based on the multi-task learning model of the long short-term memory network, through a network architecture that separates shared feature extraction from task branches, the altitude and resistivity are inversely calculated synchronously, improving the prediction accuracy and efficiency and reducing the risk of overfitting. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings, where: Figure 1Schematic flow chart of a time-domain airborne electromagnetic inversion method based on deep learning provided by an embodiment of the present invention; Figure 2 Schematic diagram showing the effects of training loss and validation loss during the training process of the model provided by an embodiment of the present invention; Figure 3 Schematic diagram showing the effect of the loss corresponding to the predicted flight altitude during the training process of the model provided by an embodiment of the present invention; Figure 4 Schematic diagram showing the effect of the loss corresponding to the predicted formation resistivity during the training process of the model provided by an embodiment of the present invention; Figure 5 Schematic diagram showing the effect of the inversion result of the target formation resistivity predicted at the first measuring point of the airborne electromagnetic provided by an embodiment of the present invention; Figure 6 Schematic diagram showing the effect of the inversion result of the target formation resistivity predicted at the second measuring point of the airborne electromagnetic provided by an embodiment of the present invention; Figure 7 Schematic diagram showing the effect of the inversion result of the target formation resistivity predicted at the third measuring point of the airborne electromagnetic provided by an embodiment of the present invention; Figure 8 Schematic diagram showing the effect of the inversion result of the target flight altitude predicted at the first measuring point of the airborne electromagnetic provided by an embodiment of the present invention; Figure 9 Schematic diagram showing the effect of the inversion result of the target flight altitude predicted at the second measuring point of the airborne electromagnetic provided by an embodiment of the present invention; Figure 10 Schematic diagram showing the effect of the inversion result of the target flight altitude predicted at the third measuring point of the airborne electromagnetic provided by an embodiment of the present invention; Figure 11 Schematic diagram showing the effect of the inversion result of the target flight altitude predicted on a measuring line of the airborne electromagnetic provided by an embodiment of the present invention; Figure 12 Thermal sense diagram of the inversion result of the target formation resistivity predicted on a measuring line of the airborne electromagnetic provided by an embodiment of the present invention; Figure 13 Thermal sense diagram of the inversion result of the formation resistivity at the known measuring line with the flight altitude provided by an embodiment of the present invention; Figure 14 Schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. Detailed implementation manners

[0011] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Apparently, the described embodiments are some, but not all, of the embodiments of the present invention. The following embodiments are used to illustrate the present invention, but are not used to limit the scope of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0012] In the following description, reference is made to "some embodiments", which describe a subset of all possible embodiments. However, it can be understood that "some embodiments" can be the same subset or different subsets of all possible embodiments, and can be combined with each other without conflict.

[0013] It should be noted that the terms "first / second / third" involved in the embodiments of the present invention are only used to distinguish similar objects and do not represent a specific order for the objects. It can be understood that "first / second / third" can be interchanged with a specific order or sequence when permitted, so that the embodiments of the present invention described herein can be implemented in an order other than that illustrated or described herein.

[0014] Those skilled in the art of the present technology can understand that, unless otherwise defined, all terms (including technical terms and scientific terms) used herein have the same meaning as the general understanding of those of ordinary skill in the art in the field to which the embodiments of the present invention belong. It should also be understood that terms such as those defined in a general dictionary should be understood to have a meaning consistent with the meaning in the context of the prior art, and will not be interpreted in an idealized or overly formal sense unless specifically defined as herein.

[0015] Figure 1 As a schematic flowchart of a time-domain airborne electromagnetic inversion method based on deep learning, the time-domain airborne electromagnetic inversion method based on deep learning provided by the embodiments of the present invention can be executed by an electronic device, and the electronic device can be, for example, a computer, a server, etc.

[0016] As Figure 1 shown, the time-domain airborne electromagnetic inversion method based on deep learning includes: S101. Calculate the magnetic field responses at each frequency point based on a recursive algorithm, the received formation thickness, the random flight altitude, and the data of the obtained formation resistivity.

[0017] In an embodiment of the present invention, the formation thickness is the thickness of each geological layer, the random flight height refers to the height of the measuring instrument relative to the ground, and the formation resistivity data refers to the resistivity value of each layer of formation. The magnetic field response at each frequency point can be calculated in the frequency-domain forward modeling engine based on the recursive algorithm, formation thickness, random flight height, and the obtained formation resistivity data.

[0018] In an embodiment of the present invention, an efficient one-dimensional time-domain airborne electromagnetic forward model can be proposed. The model includes an input module, which can receive the input formation parameters (number of layers, resistivity, layer thickness, Cole-Cole parameters), device parameters (transmitter-receiver distance, height, waveform type), and the setting of time. The frequency-domain forward modeling engine of the model calculates the magnetic field response at each frequency point based on the recursive algorithm and the received parameters.

[0019] S102. Perform a transformation process on the magnetic field response through the Hankel transform kernel function integration to obtain the frequency-domain magnetic induction intensity component.

[0020] In an embodiment of the present invention, the Hankel transform is a method to achieve this conversion. It completes this process by using the Bessel function as the kernel function, converting the data describing the magnetic field response from the spatial domain to the frequency domain using the Hankel transform kernel function integration to obtain the frequency-domain magnetic induction intensity component.

[0021] S103. Generate a time-domain step response from the frequency-domain magnetic induction intensity component using the Hankel transform coefficients of a preset order, and perform convolution processing on the time-domain step response based on the Gaussian integration nodes to obtain the electromagnetic response data in the vertical direction.

[0022] In an embodiment of the present invention, the Gaussian integration nodes refer to the sampling points and weights used in numerical integration, which are selected to accurately approximate the integration result. The frequency-domain magnetic induction intensity component is converted into a time-domain step response using the 0.5-order Hankel transform coefficients, and then precise convolution processing is performed on this time-domain step response based on the Gaussian integration nodes to simulate the propagation process of electromagnetic waves in the medium, and finally detailed electromagnetic response data in the vertical direction (i.e., the detection depth direction) is obtained, providing a basis for analyzing the underground geological structure. This process combines the conversion from the frequency domain to the time domain, the application of numerical integration methods, and signal processing techniques, aiming to accurately reflect the electromagnetic characteristic changes at different depths. The convolution operation allows simulating the propagation effect of the signal through the system, thus better understanding the variation of the electromagnetic response over time.

[0023] S104. Use the electromagnetic response data, formation thickness, and random flight height as the training set data, and use the training set data to train the long short-term memory network to be trained to obtain the long short-term memory network.

[0024] S105. Input the measured electromagnetic response data into a long short-term memory network for inversion to obtain the predicted target flight altitude and target formation resistivity.

[0025] In an embodiment of the present invention, training set data is composed of electromagnetic response data, formation thickness, and random flight altitude (10–120 m). Then, the long short-term memory network to be trained is trained with the training set data to obtain the long short-term memory network. Finally, the measured electromagnetic response data is acquired and input into the long short-term memory network for inversion to obtain the predicted target flight altitude and target formation resistivity (subsurface medium electrical structure).

[0026] Specifically, the measured electromagnetic response data (100×1) can be input into the long short-term memory network for forward propagation to obtain the predicted target flight altitude and target formation resistivity at 29 layers.

[0027] Among them, in airborne electromagnetic data inversion, the electromagnetic response data is essentially a time series signal with the following characteristics: 1. Temporal dependence: The electromagnetic field changes at different time points are closely related to the physical evolution process of formation parameters. 2. Long-distance correlation: Shallow formation parameters may affect the early electromagnetic response, while deep parameters affect the late response, and it is necessary to capture the long-term dependencies across time steps. 3. Noise sensitivity: Measured data often contains random noise, and the model needs to have a robust feature extraction ability.

[0028] It can be understood that in an embodiment of the present invention, forward modeling is performed based on a recursive algorithm, the received formation thickness, random flight altitude, and the measured formation resistivity obtained to obtain the magnetic field response at each frequency point; the magnetic field response is transformed through the Hankel transform kernel function integral to obtain the frequency-domain magnetic induction intensity component; the frequency-domain magnetic induction intensity component is used to generate a time-domain step response with a preset order of Hankel transform coefficients, and the time-domain step response is convolved based on Gaussian integration nodes to obtain the electromagnetic response data in the vertical direction; the electromagnetic response data, formation thickness, and random flight altitude are used as the training set, and the long short-term memory network to be trained is trained with the training set to obtain the long short-term memory network; the measured electromagnetic response data is input into the long short-term memory network for inversion to obtain the predicted target flight altitude and target formation resistivity. In this process, a training set composed of random flight altitude, measured formation resistivity, and the airborne electromagnetic response calculated therefrom is established. Finally, based on the multi-task learning model of the long short-term memory network, through a network architecture that separates shared feature extraction from task branches, the altitude and resistivity are inversely modeled synchronously to improve the prediction accuracy and efficiency and reduce the risk of overfitting.

[0029] In some embodiments of the present invention, S104 can be implemented through S1041 to S1042, and the following steps are used for illustration.

[0030] S1041. Preprocess the training data set to obtain a preprocessed data set, and based on the preprocessed data set, the first long short-term memory layer, the second long short-term memory layer, the third long short-term memory layer, the fourth long short-term memory layer, the fifth long short-term memory layer, the dropout layer, and the fully connected layer, obtain the predicted flight altitude and the predicted formation resistivity.

[0031] In some embodiments of the present invention, the long short-term memory network is composed of five long short-term memory layers, a dropout layer, and a fully connected layer. The number of hidden units in the first long short-term memory layer is 32, the number of hidden units in the second long short-term memory layer is 64, the number of hidden units in the third long short-term memory layer is 128, the number of hidden units in the fourth long short-term memory layer is 256, and the number of hidden units in the fifth long short-term memory layer is 512. First, preprocess the training data set, and input the preprocessed training data set into the five-layer LSTM, dropout layer, and fully connected layer for processing. Finally, the fully connected layer outputs the predicted flight altitude and the predicted formation resistivity. Among them, the dropout layer is the random inactivation layer.

[0032] S1042. Train the long short-term memory network to be trained based on the predicted flight altitude and the predicted formation resistivity to obtain the long short-term memory network.

[0033] In some embodiments of the present invention, calculate the loss value based on the predicted flight altitude and the predicted formation resistivity in combination with the loss function, and use the loss value to adjust the parameters of the long short-term memory network to be trained to achieve the training purpose until the training conditions are met, and finally obtain the long short-term memory network.

[0034] In some embodiments of the present invention, the preprocessing of the training set in S1041 to obtain the preprocessed training set can be implemented through S201 to S202, and the following steps are used for illustration.

[0035] S201. Normalize the electromagnetic response data, formation thickness, and random flight altitude in the training set to obtain the first electromagnetic response data, the first formation thickness, and the first random flight altitude.

[0036] In some embodiments of the present invention, perform standardization processing (mean-variance normalization) on the electromagnetic response data, formation thickness, and random flight altitude in the training set to obtain the processed first electromagnetic response data, the first formation thickness, and the first random flight altitude.

[0037] S202. Normalize the tag height using a standard scaler to obtain a first height, and normalize the tag resistivity using a min-max scaler to obtain a first formation resistivity.

[0038] In some embodiments of the present invention, the training set further includes tag data, which includes 29 layers of tag heights and 29 layers of tag resistivities. Use a standard scaler, that is, use StandardScaler to standardize the 29 layers of tag heights to obtain a first height, and use a min-max scaler, that is, use MinMaxScaler to perform normalization on the 29 layers of tag resistivities to obtain a first formation resistivity.

[0039] In some embodiments of the present invention, obtaining the predicted flight height and the predicted formation resistivity based on the preprocessed training set, the first long short-term memory layer, the second long short-term memory layer, the third long short-term memory layer, the fourth long short-term memory layer, the fifth long short-term memory layer, the dropout layer, and the fully connected layer in S1041 can be implemented through S301 to S303, and will be described through the following steps.

[0040] S301. Input the first electromagnetic response data, the first formation thickness, and the first random flight height into the first long short-term memory layer to obtain a first temporal feature, and input the first temporal feature into a batch normalization layer to obtain a second temporal feature.

[0041] In some embodiments of the present invention, input the first electromagnetic response data (shape=(100,1)), the first formation thickness, and the first random flight height into the first long short-term memory layer, and output a temporal feature of 100×32, retaining all time steps. Subsequently, normalize the feature dimension (32 dimensions) through a BatchNorm layer (batch normalization layer) to accelerate convergence and reduce the risk of gradient disappearance, and obtain a second temporal feature.

[0042] S302. Input the second temporal feature into the second long short-term memory layer to obtain a third temporal feature, and input the third temporal feature into the batch normalization layer to obtain a fourth temporal feature.

[0043] In some embodiments of the present invention, input the second temporal feature into the second long short-term memory layer, output a feature of 100×64, and normalize the feature dimension (64 dimensions) through the BatchNorm layer again to obtain a fourth temporal feature.

[0044] S303. Perform task-independent branch prediction based on the fourth temporal feature, the third long short-term memory layer, the fourth long short-term memory layer, and the fifth long short-term memory layer to obtain the predicted flight height and the predicted formation resistivity.

[0045] In some embodiments of the present invention, the fourth temporal feature is input into the third long short-term memory layer to predict the height, and the fourth temporal feature is input into the fourth long short-term memory layer and the fifth long short-term memory layer to predict the resistivity. By allocating independent branches for different tasks, both the shared information between tasks is retained and the interference between tasks is reduced, thereby improving the overall performance of the model.

[0046] In some embodiments of the present invention, S303 can be implemented through S3031 to S3032, and the following steps are used for illustration.

[0047] S3031. Obtain the predicted flight height based on the fourth temporal feature, the third long short-term memory layer, the batch normalization layer, and the dropout layer.

[0048] In some embodiments of the present invention, the fourth temporal feature is input into the third long short-term memory layer, and 100×128 features are output. After the BatchNorm layer normalizes the feature dimension (128 dimensions), the fifth temporal feature is obtained. The fifth temporal feature passes through the dropout layer to randomly mask 20% of the neurons to prevent overfitting, and the predicted flight height is obtained.

[0049] S3032. Obtain the predicted formation resistivity based on the fourth temporal feature, the fourth long short-term memory layer, the fifth long short-term memory layer, the batch normalization layer, and the dropout layer.

[0050] In some embodiments of the present invention, the fourth temporal feature is input into the fourth long short-term memory layer, and 100×256 features are output. After the BatchNorm layer normalizes the feature dimension (256 dimensions), the sixth temporal feature is obtained. The sixth temporal feature is input into the fifth long short-term memory layer for further processing, and 100×512 features are obtained. After the BatchNorm layer normalizes the feature dimension (512 dimensions), the seventh temporal feature is obtained. The seventh temporal feature passes through the dropout layer to randomly mask 20% of the neurons to prevent overfitting, and the predicted formation resistivity is obtained.

[0051] In some embodiments of the present invention, S1042 can be implemented through S401 to S402, and the following steps are used for illustration.

[0052] S401. Obtain the first loss by using the predicted flight height and the first height, and obtain the second loss by using the predicted formation resistivity and the first formation resistivity.

[0053] S402. Obtain the total loss from the first loss and the second loss, and train the long short-term memory network to be trained with the total loss to obtain the long short-term memory network.

[0054] In some embodiments of the present invention, the mean square error between the predicted flight altitude and the first altitude is calculated as the first loss, and the mean absolute error between the predicted formation resistivity and the first formation resistivity is calculated as the second loss. The first loss and the second loss are summed to obtain the total loss, and the long short-term memory network to be trained is trained through the total loss to obtain the long short-term memory network.

[0055] Among them, the calculation formula of the total loss loss is as follows: ; In the above formula, h_pred is the predicted flight altitude, h_true is the first altitude, r_pred is the predicted formation resistivity, and r_true is the first formation resistivity.

[0056] Among them, α = 1 and β = 0.5 (initial value). L2 regularization (λ = 0.0001) is added to constrain the weights of the fully connected layer. The Adam optimizer is used with an initial learning rate of 0.001, combined with ReduceLROnPlateau for dynamic decay. An early stopping mechanism is set, and the training is terminated if the validation set loss does not decrease for 30 epochs.

[0057] Among them, , ; In the above formula, MSE represents the mean square error, MAE represents the mean absolute error, m and n are both the number of samples, and y i is the true value of the th sample, is the predicted value of the th sample, and m and n are the same.

[0058] As Figure 2 shown, Figure 2 is a schematic diagram of the training loss and validation loss effects of the model provided by the embodiments of the present invention during the training process. Figure 2 The abscissa in

[0059] is the number of training epochs, and the ordinate is the loss (mean square error). This model and the following models are long short-term memory networks. The training loss refers to the performance of the model on the training set, that is, the degree of fitting of the model to the training data. It is measured by calculating the difference between the predicted value of the model and the actual label. The validation loss refers to the performance of the model on the validation set. The validation set is a part of the data divided from the training set and does not participate in the process of updating the model parameters. The validation loss is used to evaluate the prediction ability of the model for unseen data, that is, the generalization ability of the model. Figure 3 shown, Figure 3This is a schematic diagram showing the effect of the loss corresponding to the predicted flight altitude during the training process of the model provided by the embodiments of the present invention. After each training iteration, the model makes a prediction on the samples in the training set and calculates the loss between these predicted values and the actual flight altitude. This loss value reflects the performance of the model in its current state.

[0060] As Figure 4 shown, Figure 4 This is a schematic diagram showing the effect of the loss corresponding to the predicted formation resistivity during the training process of the model provided by the embodiments of the present invention. After each training iteration, the model makes a prediction on the samples in the training set and calculates the loss between these predicted values and the actual formation resistivity. This loss value reflects the performance of the model in its current state.

[0061] As Figure 5 , Figure 6 and Figure 7 shown, Figure 5 This is a schematic diagram showing the inversion result of the target formation resistivity predicted at the first measurement point of the airborne electromagnetic provided by the embodiments of the present invention, Figure 6 This is a schematic diagram showing the inversion result of the target formation resistivity predicted at the second measurement point of the airborne electromagnetic provided by the embodiments of the present invention, Figure 7 This is a schematic diagram showing the inversion result of the target formation resistivity predicted at the third measurement point of the airborne electromagnetic provided by the embodiments of the present invention. Figure 5 , Figure 6 and Figure 7 The difference between Figure 5 is that they are different at the three measurement points. After the model is trained, the model starts to perform inversion. The input is the electromagnetic response (dBz / dt) at 100 time points. The inversion effect of the target formation resistivity predicted at a single measurement point is as Figure 8 , Figure 9 and Figure 10 shown, Figure 8 This is a schematic diagram showing the inversion result of the target flight altitude predicted at the first measurement point of the airborne electromagnetic provided by the embodiments of the present invention, Figure 9 This is a schematic diagram showing the inversion result of the target flight altitude predicted at the second measurement point of the airborne electromagnetic provided by the embodiments of the present invention, Figure 10 This is a schematic diagram showing the inversion result of the target flight altitude predicted at the third measurement point of the airborne electromagnetic provided by the embodiments of the present invention. Figure 8 , Figure 9 and Figure 10 The difference between Figure 8 , Figure 9 and Figure 10, where the actual height can be set randomly, such as from 10m to 120m.

[0062] Such as Figure 11 shown Figure 11 is a schematic diagram of the effect of the inversion result of the predicted target flight height on an airborne electromagnetic survey line provided by an embodiment of the present invention. Among them, a survey line can be a series of measurements along a specific path, where the actual height can be set randomly, such as from 10m to 120m. By Figure 11 the actual height and the predicted target flight height in can calculate the height difference. Figure 12 is a thermal sense schematic diagram of the inversion result of the predicted target formation resistivity on an airborne electromagnetic survey line provided by an embodiment of the present invention. Figure 12 is Figure 11 a thermal sense schematic diagram of the inversion result of the predicted target formation resistivity on the section where the survey line is located in.

[0063] Such as Figure 13 shown Figure 13 is a thermal sense schematic diagram of the inversion result of the formation resistivity at the flight height on a known survey line provided by an embodiment of the present invention. The flight height can be set randomly, such as from 10m to 120m.

[0064] In summary, the deep learning method based on long short-term memory network of the present invention aims to solve the problem that in the process of inverting airborne electromagnetic data by traditional methods, due to terrain undulation or ground obstacles, the flight height record is incorrect or missing. Traditional methods usually rely on accurate records of flight height during the inversion process, but due to the irregularity of the flight path or environmental factors, the flight height often cannot be accurately recorded or not recorded at all, thus affecting the accuracy of data inversion and subsequent 3D modeling.

[0065] To solve this problem, the present invention adopts a long short-term memory network, which can dynamically restore the missing flight height information (and effectively integrate it into the inversion process. By introducing deep learning technology, the long short-term memory network can automatically estimate and restore the flight height according to the long-term dependence relationship of time series data during the inversion process. This ability is particularly important for areas with complex terrain or many obstacles, which can ensure that even when the flight height data is incomplete or inaccurate, accurate inversion can still be completed.

[0066] In addition, through the gating mechanism and memory units of the long short-term memory network, the model can effectively handle the temporal dependencies in the time series, reduce the impact caused by highly incomplete records, and thus improve the accuracy of the inversion results. This not only improves the recovery effect of the flight altitude but also enhances the stability of the overall inversion. Through the long short-term memory network learning method of the present invention, not only the flight altitude information is effectively restored, but also the inversion accuracy, stability, and processing efficiency are improved. Especially under complex geological and environmental conditions, it has significant advantages. Refer to Figure 14 , a schematic structural diagram of an electronic device according to an embodiment of the present invention is shown. The specific implementation of the electronic device in the specific embodiments of the present invention is not limited.

[0067] As Figure 14 shown, the electronic device may include: a processor 502, a communication interface 504, a memory 506, and a communication bus 508.

[0068] Among them: The processor 502, the communication interface 504, and the memory 506 communicate with each other through the communication bus 508.

[0069] The communication interface 504 is used to communicate with other electronic devices or servers.

[0070] The processor 502 is used to execute the program 510, and specifically can execute the relevant steps in the above method embodiments.

[0071] Specifically, the program 510 may include program code, and the program code includes computer operation instructions.

[0072] The processor 502 may be a central processing unit CPU, or a specific integrated circuit ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the embodiments of the present invention. One or more processors included in the intelligent device may be of the same type of processor, such as one or more CPUs; or may be of different types of processors, such as one or more CPUs and one or more ASICs.

[0073] The memory 506 is used to store the program 510. The memory 506 may include a high-speed RAM memory, and may also include non-volatile memory, such as at least one disk memory.

[0074] The program 510 is specifically used to cause the processor 502 to execute the operations corresponding to the methods described in the above method embodiments.

[0075] For the specific implementation of each step in Program 510, reference may be made to the corresponding steps and the corresponding descriptions in the units in the foregoing method embodiments, which will not be elaborated herein. Those skilled in the art can clearly understand that for the convenience and conciseness of description, the specific working processes of the devices and modules described above may refer to the corresponding process descriptions in the foregoing method embodiments, which will not be elaborated herein again.

[0076] It should be noted that according to the needs of implementation, each component / step described in the embodiments of the present invention may be split into more components / steps, or two or more components / steps or partial operations of components / steps may be combined into new components / steps to achieve the objectives of the embodiments of the present invention.

[0077] The method according to the embodiments of the present invention described above may be implemented in hardware, firmware, or may be implemented as software or computer code stored in a recording medium (such as a CD ROM, RAM, floppy disk, hard disk, or magneto-optical disk), or may be implemented as computer code originally stored in a remote recording medium or a non-transitory machine-readable medium and downloaded through a network and to be stored in a local recording medium, so that the method described herein may be stored in such software processes on a recording medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware (such as an ASIC or FPGA). It can be understood that a computer, a processor, a microprocessor controller, or programmable hardware includes a storage component (such as a RAM, a ROM, a flash memory, etc.) capable of storing or receiving software or computer code, and when the software or computer code is accessed and executed by the computer, the processor, or the hardware, the method described herein is implemented. In addition, when a general-purpose computer accesses the code for implementing the method shown herein, the execution of the code converts the general-purpose computer into a dedicated computer for implementing the method shown herein.

[0078] Those of ordinary skill in the art can realize that the units and method steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or by a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. A professional technician may use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the embodiments of the present invention.

[0079] The above embodiments are only used to illustrate the embodiments of the present invention, rather than to limit the embodiments of the present invention. Those of ordinary skill in the relevant technical field can make various changes and modifications without departing from the spirit and scope of the embodiments of the present invention. Therefore, all equivalent technical solutions also belong to the scope of the embodiments of the present invention. The patent protection scope of the embodiments of the present invention shall be defined by the claims.

Claims

1. A time-domain airborne electromagnetic inversion method based on deep learning, characterized in that, Including: Forward modeling is performed based on a recursive algorithm, the received formation thickness, the random flight height, and the measured formation resistivity obtained, to obtain the magnetic field response at each frequency point; The magnetic field response is transformed by integrating with the Hankel transform kernel function to obtain the magnetic induction intensity component in the frequency domain; The magnetic induction intensity component in the frequency domain is used to generate a step response in the time domain using the Hankel transform coefficient of a preset order, and the step response in the time domain is convolved based on Gaussian integration nodes to obtain electromagnetic response data in the vertical direction; The electromagnetic response data, the formation thickness, and the random flight height are used as a training set, and the training set is used to train a long short-term memory network to be trained, to obtain a long short-term memory network; The measured electromagnetic response data is input into the long short-term memory network for inversion, to obtain the predicted target flight height and target formation resistivity.

2. The method according to claim 1, wherein The long short-term memory network is composed of five long short-term memory layers, a dropout layer, and a fully connected layer, where the number of hidden units in the first long short-term memory layer is 32, the number of hidden units in the second long short-term memory layer is 64, the number of hidden units in the third long short-term memory layer is 128, the number of hidden units in the fourth long short-term memory layer is 256, and the number of hidden units in the fifth long short-term memory layer is 512; The training of the long short-term memory network to be trained using the training set to obtain a long short-term memory network includes: Preprocessing the training set to obtain a preprocessed training set, and obtaining a predicted flight height and a predicted formation resistivity based on the preprocessed training set, the first long short-term memory layer, the second long short-term memory layer, the third long short-term memory layer, the fourth long short-term memory layer, the fifth long short-term memory layer, the dropout layer, and the fully connected layer; Training the long short-term memory network to be trained based on the predicted flight height and the predicted formation resistivity to obtain the long short-term memory network.

3. The method according to claim 2, wherein The training set further includes label data, and the label data includes a label flight height and a label formation resistivity; The preprocessing of the training set to obtain a preprocessed training set includes: Normalizing the electromagnetic response data, the formation thickness, and the random flight height in the training set to obtain first electromagnetic response data, a first formation thickness, and a first random flight height; Normalizing the label flight height using a standard scaler to obtain a first height, and normalizing the label formation resistivity using a min-max scaler to obtain a first formation resistivity.

4. The method according to claim 3, characterized in that, The obtaining of the predicted flight height and the predicted formation resistivity based on the preprocessed training set, the first long short-term memory layer, the second long short-term memory layer, the third long short-term memory layer, the fourth long short-term memory layer, the fifth long short-term memory layer, the dropout layer, and the fully connected layer includes: Inputting the first electromagnetic response data, the first formation thickness, and the first random flight height into the first long short-term memory layer to obtain a first time series feature, and inputting the first time series feature into a batch normalization layer to obtain a second time series feature; Input the second timing feature into the second long short-term memory layer to obtain a third timing feature, and input the third timing feature into the batch normalization layer to obtain a fourth timing feature; Based on the fourth timing feature, the third long short-term memory layer, the fourth long short-term memory layer, and the fifth long short-term memory layer, perform task-independent branch prediction to obtain the predicted flight altitude and the predicted formation resistivity.

5. The method according to claim 4, wherein The performing task-independent branch prediction based on the fourth timing feature, the third long short-term memory layer, the fourth long short-term memory layer, and the fifth long short-term memory layer to obtain the predicted flight altitude and the predicted formation resistivity includes: Obtain the predicted flight altitude based on the fourth timing feature, the third long short-term memory layer, the batch normalization layer, and the dropout layer; Obtain the predicted formation resistivity based on the fourth timing feature, the fourth long short-term memory layer, the fifth long short-term memory layer, the batch normalization layer, and the dropout layer.

6. The method according to claim 3, wherein The training the long short-term memory network to be trained based on the predicted flight altitude and the predicted formation resistivity to obtain the long short-term memory network includes: Obtain a first loss through the predicted flight altitude and the first altitude, and obtain a second loss through the predicted formation resistivity and the first formation resistivity; Obtain a total loss through the first loss and the second loss, and train the long short-term memory network to be trained through the total loss to obtain the long short-term memory network.

Images