Noise suppression method and system used for transient electromagnetic data

Abstract

To solve the problem in the prior art in which regarding a technical field of noise suppression, in particular, regarding a noise suppression method and system used for transient electromagnetic data, some of them have lack adaptability, some of them have poor noise elimination effect, and some of them have a complicated operation process.SOLUTION: The method comprises the steps of: superposing forward modeling data and noise data based on different weights; taking the superposed data as a data set after superposing; and training a pre-constructed noise suppression network based on the data set to obtain a trained noise suppression network, where the data set comprises a training set and a test set; and inputting TEM data to be subjected to noise suppression into the trained noise suppression network to obtain the TEM data after subjected to noise suppression, where the noise suppression network is constructed based on a recurrent neural network and an LSTM network.SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] The present invention relates to the technical field of noise suppression, and in particular to noise suppression methods and systems used on transient electromagnetic data. [Background technology]

[0002] Transient electromagnetic (TEM) is a time-domain electromagnetic exploration method based on the principle of electromagnetic induction and widely used in mineral and hydrological resource exploration. TEM uses a step wave or other pulsed current field source to emit an initial field underground. When the initial field is interrupted, the time-varying decay characteristics of the secondary induced electromagnetic field generated by the underground geological body are measured. The electrical conductivity and burial depth of the target geological body are determined based on the signal received by the analytical instrument, thereby achieving the objective of underground geological exploration. Transient electromagnetic signals are secondary field signals that decay over time. When operating in a highly noisy environment, such as a mining or industrial area, the late signal received by the instrument is significantly affected by various natural and human interference noises, resulting in poor resolution of the deep geological body. Therefore, when using TEM, the presence of noise has a significant adverse effect on late data processing and geological analysis. These interference signals may mask the necessary useful signals. Without effective noise suppression and extraction of useful signal information, the transient electromagnetic data cannot be fully utilized for subsequent data processing and inverse interpretation.

[0003] Patent application publication number CN117056777A proposes an improved variational mode decomposition (VMD) method for transient electromagnetic signal denoising based on the Sparrow search algorithm. This method uses the Sparrow search algorithm to globally optimise the penalty coefficient α and the number of modes K in VMD to obtain the optimal parameters [K,α]. The optimal parameter combination obtained after global optimisation is then used to perform VMD on the collected transient electromagnetic signals, achieving the denoising process for the original signals. This method avoids the mode mixing problem caused by artificial parameter selection and improves the signal denoising effect.

[0004] Patent application publication number CN111650655A discloses a method for supervised noise suppression of transient electromagnetic signals using non-negative matrix factorization, in which, in a training stage, a pure signal is subjected to short-time Fourier transform and non-negative matrix factorization to obtain an atomic dictionary of each feature of the characteristic signal; then, in a noise suppression stage, the noisy signal is processed using the atomic dictionary and a noise suppression model to obtain a preliminary estimated transient electromagnetic signal; finally, the above steps are repeated multiple times to accumulate the late data of the preliminary estimated transient electromagnetic signal and the original early and mid-term data of the noisy signal, respectively, to obtain the arithmetic mean values ​​of each, and then combined to obtain a final complete estimated transient electromagnetic signal. This invention can effectively remove noise from actual transient electromagnetic signals and improve the accuracy of inversion of transient electromagnetic signals, but However, the above-mentioned conventional techniques are sensitive to parameter selection and have relatively high computational complexity, especially for long-term sequences or high-dimensional data. Furthermore, the variational mode decomposition method is sensitive to noise in the input signal, which may be decomposed into modes. This may require additional post-processing steps to deal with the noise.

[0005] The initial time-domain electromagnetic noise removal method used wavelet transform, Kalman filter, singular value decomposition, etc. Each of these methods has its own drawbacks, some of which lack adaptability, some of which have poor noise removal effect, and some of which have complicated operation processes.

[0006] Accordingly, the present invention proposes a method and system for noise suppression for use with transient electromagnetic data. Summary of the Invention [Means for solving the problem]

[0007] Prior art techniques are sensitive to parameter selection and have relatively high computational complexity, especially for long-term sequences or high-dimensional data. Furthermore, the variational mode decomposition method is sensitive to noise in the input signal, which can be decomposed into various modes. Initial time-domain electromagnetic noise reduction techniques have employed methods such as wavelet transform, Kalman filter, and singular value decomposition. Each of these methods has its own drawbacks, including a lack of adaptability, poor noise reduction, and complex operational processes. To address these challenges, the present invention proposes a noise suppression method and system for transient electromagnetic data.

[0008] In one aspect of the present invention, a noise suppression method for use with transient electromagnetic data is proposed, The method comprises: Stacking the forward modeling data and the noise data according to different weights, forming a dataset after stacking, and training a pre-constructed noise suppression network based on the dataset to obtain a trained noise suppression network, where the dataset includes a training set and a test set, and the different weights are obtained according to a preset target signal-to-noise ratio; The TEM data to be noise-suppressed is input to the trained noise suppression network, and the noise-suppressed TEM data is obtained. Here, the noise suppression network was constructed based on a recurrent neural network and an LSTM network.

[0009] In some preferred embodiments, the method for constructing the dataset is as follows: constructing a plurality of different one-dimensional resistivity models and performing forward modeling; and obtaining forward modeling data corresponding to the plurality of one-dimensional resistivity models; Artificially simulating noise data of different types and different intensities, the noise data including atmospheric noise, power frequency noise and Gaussian noise; A plurality of the forward modeling data and a plurality of the noise data are stacked with different weights, and the stacked plurality of data are used as a data set; The data set is divided into a training set and a test set at a set ratio.

[0010] In some preferred embodiments, the method for obtaining the data set is represented by the following formula:

[0011]

number

[0012] where D final is the stacked dataset, D i forward is the i-th forward modeling data, D i noise is the i-th noise data, ω forward and ω noise are weight values ​​for the forward modeling data and noise data, which are used to adjust the ratio of the forward modeling data to the noise data when stacking, and N is the amount of forward modeling data.

[0013] In some preferred embodiments, ω forward and ω noise is obtained by the following method:

[0014]

number

[0015]

number

[0016] where SNR target is the preset target signal-to-noise ratio.

[0017] In some preferred embodiments, the pre-constructed noise suppression network includes an input layer, an encoder, a decoder, and an output layer connected in series; the input layer is used to receive input data including the dataset or the TEM data to be noise suppressed; The encoder and decoder are both constructed by two sequentially connected LSTM layers, and the encoder is used to learn and convert time-series feature data of the input data into abstract data; the decoder is used to learn the abstract data and restore the time-series feature data; The output layer includes a sequentially connected fully connected layer and a recurrent output layer, and is used to output the restored time-series feature data as denoised TEM data.

[0018] In some preferred embodiments, the method for training the post-training noise suppression network is as follows: pre-processing the training set and the test set, the pre-processing method including normalization and folding reconstruction; Input the pre-processed training set into a pre-constructed noise suppression network to obtain initial denoising data, adjust initial parameters of the noise suppression network according to the initial denoising data and the actual label, obtain optimal parameters, and replace the initial parameters of the noise suppression network with the optimal parameters; The preprocessed training set is input again in batches to the noise suppression network with optimal parameters to obtain training denoised data, and multiple iterations are performed. In each iteration, the mean squared error between the training denoised data and the actual labels is calculated, and the iterations are stopped until the mean squared error converges; The noise suppression network after the iterations have stopped is designated as the trained noise suppression network.

[0019] In some preferred embodiments, over the course of multiple iterations, the preprocessed test set is simultaneously input to the noise suppression network with optimal parameters to obtain test data, and the test data is compared with the corresponding actual labels to obtain a validation error, and the iterations are stopped when the validation error reaches convergence and then starts to grow again.

[0020] In some preferred embodiments, at each iteration, an Adam optimizer is used to adjust the parameters of the noise suppression network to minimize the mean squared error between the training data and the actual labels until the mean squared error converges.

[0021] In some preferred embodiments, the method for refolding comprises: The training set and the test set of one-dimensional data are collapsed and reconstructed into two-dimensional data, and the two-dimensional data is used as the input of a pre-constructed noise suppression network.

[0022] Another aspect of the present invention provides a noise suppression system for transient electromagnetic data based on a noise suppression method for transient electromagnetic data, The system is The method includes a network construction and training module, stacking the forward modeling data and the noise data based on different weights, forming a dataset after stacking, training a pre-constructed noise suppression network based on the dataset, and obtaining a trained noise suppression network, wherein the dataset includes a training set and a test set, and the different weights are obtained based on a preset target signal-to-noise ratio; A noise reduction module is included, and the TEM data to be noise-suppressed is input into a trained noise suppression network to obtain noise-suppressed TEM data; Here, the noise suppression network was constructed based on a recurrent neural network and an LSTM network. [Effects of the Invention]

[0023] The beneficial effects of the present invention are as follows: (1) It has excellent denoising effect on contaminated late-stage data. The noise suppression network based on the LSTM-Autoncoder neural network we constructed has excellent denoising effect on contaminated late-stage TEM data, which can greatly enhance the utilization efficiency of late-stage TEM data and increase the exploration depth. (2) High versatility and ability to combat various types of noise. Our method excels in processing power frequency noise that is difficult to eliminate using traditional methods. We have achieved accurate removal of common power frequency noises such as square waves, charge / discharge triangular waves, power frequency interference, step waves, and spike pulses. This innovation demonstrates stronger adaptability and practicality in complex real-world environments. (3) Improved efficiency through intelligent denoising. By avoiding the need for artificial denoising parameter selection and other complex artificial operations such as wavelet denoising, variational mode decomposition, and singular value decomposition, its concise and efficient denoising efficiency far exceeds that of other methods. [Brief explanation of the drawings]

[0024] Other features, objects and advantages of the present application will become more apparent upon reading the following detailed description of non-limiting embodiments, made with reference to the drawings, in which: [Figure 1] 1 is a flowchart of a noise suppression method for use on transient electromagnetic data of the present invention. [Figure 2] 1 is a schematic diagram of the training process of the noise suppression network of the noise suppression method used for transient electromagnetic data of the present invention; FIG. [Figure 3] FIG. 2 is a block diagram of a noise suppression network for the noise suppression method used on transient electromagnetic data of the present invention. [Figure 4] 1 is a block diagram of a computer system for implementing a server according to an embodiment of the method, system, and apparatus of the present application. DETAILED DESCRIPTION OF THE INVENTION

[0025] The present application will be described in more detail below with reference to the accompanying drawings and examples. It is understood that the specific embodiments described herein are used only to explain the related invention, rather than to limit the present invention. For the sake of convenience, only parts related to the invention are shown in the drawings.

[0026] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in the present application may be combined with each other. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, the present application will be described in detail with reference to the accompanying drawings in conjunction with the embodiments.

[0027] As a first embodiment of the present invention, referring to Figs. 1 to 3, a noise suppression method for transient electromagnetic data is proposed, The method comprises: Stacking the forward modeling data and the noise data according to different weights, forming a dataset after stacking, and training a pre-constructed noise suppression network based on the dataset to obtain a trained noise suppression network, where the dataset includes a training set and a test set, and the different weights are obtained according to a preset target signal-to-noise ratio; The TEM data to be noise-suppressed is input to the trained noise suppression network, and the noise-suppressed TEM data is obtained. Here, the noise suppression network was constructed based on a recurrent neural network and an LSTM network.

[0028] Here, the network used in this embodiment is preferably a recurrent neural network (RNN), a neural network structure primarily used for time series processing. Based on the traditional backpropagation neural network, a time backpropagation algorithm is introduced to realize inter-time point connections in the hidden layer. The output of hidden layer neurons depends not only on the input at the current time but also on the output of hidden layer neurons at the previous time. This design allows the RNN to fully utilize the historical information in the time series, while maintaining a certain memory capacity and being able to process time-related data.

[0029] When the information sequence is long, RNN is prone to long-term dependency problems, making it difficult to efficiently space distant sequence information using time. To overcome the problems of gradient vanishing and gradient explosion, the present invention uses LSTM to optimize the network configuration. LSTM stabilizes gradient changes by adding cell states and information gating. The cell states are long-span information transmission channels that run directly through the entire information transmission chain and involve only a small amount of linear operations, while gating selectively passes information. The combined effect of these two alleviates the long-term dependency problems of RNN.

[0030] LSTM first uses a sigmoid function σ and a weight function W f By calculation, the forget gate f t to control the information that needs to be forgotten in the cell state,

[0031]

number

[0032] where h t-1 ,xt is the network input, and b f is the first offset, calculated f t is a forget gate, which takes a value between 0 and 1 and reflects the cell state C t-1 Determine what information needs to be forgotten,

[0033]

number

[0034] Furthermore, the forget gate f t and the cell state C at the previous time t-1 Multiplying by C t-1 Determine the information that needs to be forgotten.

[0035] Similarly, i t and the following equation (Equation 6) together determine the information that needs to be held in the total input of the storage gate.

[0036]

number

[0037]

number

[0038] The above h t-1 ,x t is the network input, and b i is the second offset, the sigmoid function σ and the first weight function W i Long-term memory by t Calculate.

[0039]

number

[0040] The above h t-1 ,x t is the network input, and b cis the third offset, tanh function and the second weighting function W c Calculate short-term memory (Equation 9) by:

[0041]

number

[0042]

number

[0043] long term memory i t and short-term memory (Equation 11) to obtain the memory gate, which determines the part of the cell's state that needs to be retained.

[0044]

number

[0045] Furthermore, the forget gate and the memory gate are added to obtain the updated value C t and

[0046]

number

[0047] Ultimately, O t is used as the gate, and the neuron output value h t Determine

[0048]

number

[0049] The above h t-1 ,x t is the network input, and b o The fourth offset, sigmoid function and the third weight function W o By output gate O t is calculated and the part that needs to be output to the cell state is determined.

[0050]

number

[0051] And the updated cell state C t is processed by the tanh function and the output gate O t is multiplied to obtain the output value of the hidden layer at that time.

[0052] In order to more clearly explain the noise suppression method used for transient electromagnetic data of the present invention, the embodiment of the present invention will be described in detail below in conjunction with FIG. 1, the details of which are as follows: Stacking the forward modeling data and the noise data according to different weights, forming a dataset after stacking, and training a pre-constructed noise suppression network based on the dataset to obtain a trained noise suppression network, where the dataset includes a training set and a test set, and the different weights are obtained according to a preset target signal-to-noise ratio; In the present invention, the method for constructing the dataset is as follows: constructing a plurality of different one-dimensional resistivity models and performing forward modeling; and obtaining forward modeling data corresponding to the plurality of one-dimensional resistivity models; Artificially simulating noise data of different types and intensities, the noise data including atmospheric noise, power frequency noise and Gaussian noise; A plurality of the forward modeling data and a plurality of the noise data are stacked with different weights, and the stacked plurality of data are used as a data set; The data set is divided into a training set and a test set at a set ratio.

[0053] Here, the method for obtaining the data set is represented by the following formula:

[0054]

number

[0055] where D final is the stacked dataset, D i forward is the i-th forward modeling data, D i noise is the i-th noise data, ω forward and ω noise are the weighting values ​​of the forward modeling data and noise data, which are used to adjust the ratio of the forward modeling data and noise data when stacking, respectively, and N is the amount of forward modeling data. By adjusting the weighting values, it is possible to obtain synthesized data with different signal-to-noise ratios. The signal-to-noise ratio (SNR) represents the relative strength of the signal and noise, and is usually expressed in decibels (dB). The weighting values ​​can be adjusted according to the required signal-to-noise ratio. Specifically, assuming a preset target signal-to-noise ratio (unit: decibel), the corresponding weighting value can be calculated using the following formula: ω forward and ω noise is obtained by the following method:

[0056]

number

[0057]

number

[0058] Here, different SNR target By setting , we can obtain the corresponding weight value and generate synthetic data with different signal-to-noise ratios. For example, if we want to obtain data with a higher signal-to-noise ratio, we can set a larger SNR target Conversely, if you want to acquire data with a lower signal-to-noise ratio, you can select a smaller SNR targetIn this way, the signal-to-noise ratio of the synthesized data can be flexibly controlled.

[0059] Specifically, to train a neural network, we need to construct a dataset with a sufficiently large number of samples. However, because ideal natural TEM data are unavailable, we use TEM data obtained from 1D forward modeling combined with an artificially simulated noise synthetic dataset to train an LSTM-Autoncoder neural network. We constructed 1,000 different 1D resistivity models and forward modeled 1,000 forward modeling data. These forward modeling data were temporal data obtained at a 100 MHz sampling rate with 1,000 sampling points each, ranging from 0 to 0.001 s. We also artificially simulated different types and intensities of noise, including atmospheric noise, power frequency noise, and Gaussian noise. We stacked the mixed noise with different interference intensities and the pure forward modeling data to obtain 20,000 noise-containing data samples, which were then divided into training and test sets at a 4:1 ratio: 16,000 samples for the training set and 4,000 samples for the test set.

[0060] Referring to Figure 3, the pre-constructed noise suppression network of the present invention includes an input layer, an encoder, a decoder, and an output layer connected in sequence, the input layer is used to receive input data including the dataset or the TEM data to be noise suppressed; The encoder and decoder are both constructed by two sequentially connected LSTM layers, and the encoder is used to learn and convert time-series feature data of the input data into abstract data; the decoder is used to learn the abstract data and restore the time-series feature data; The output layer includes a sequentially connected fully connected layer and a recurrent output layer, and is used to output the restored time-series feature data as denoised TEM data.

[0061] Here, in this embodiment, the two LSTM layers of the encoder have 128 and 64 neurons, respectively, and the two LSTM layers of the decoder have 64 and 128 neurons, respectively.

[0062] Specifically, see Figure 3, the two LSTM layers in the encoder are LSTM 1 and LSTM 2, respectively, and the two LSTM layers in the decoder are LSTM 3 and LSTM 4, respectively.

[0063] Referring to FIG. 2, the training method of the post-training noise suppression network of the present invention is as follows: pre-processing the training set and the test set, the pre-processing method including normalization and folding reconstruction; Input the pre-processed training set into a pre-constructed noise suppression network to obtain initial denoising data, adjust initial parameters of the noise suppression network according to the initial denoising data and the actual label, obtain optimal parameters, and replace the initial parameters of the noise suppression network with the optimal parameters; The preprocessed training set is input again in batches to the noise suppression network with optimal parameters to obtain training denoised data, and multiple iterations are performed. In each iteration, the mean squared error between the training denoised data and the actual labels is calculated, and the iterations are stopped until the mean squared error converges; The noise suppression network after the iterations have stopped is designated as the trained noise suppression network.

[0064] Here, the folding and reconstruction method is as follows: The training set and the test set of one-dimensional data are collapsed and reconstructed into two-dimensional data, and the two-dimensional data is used as the input of a pre-constructed noise suppression network.

[0065] Here, for the input data set, normalization is first performed due to the peculiarities of the data exponential decay. The data of the data set is taken logarithmically and then max-min normalized. In the following formula, x is the data set data (raw data), and logx mean is the logarithmic mean value of the data set, ,logx max is the maximum value in the data set, and logx min is the minimum value in the dataset, x' is the normalized data,

[0066]

number

[0067] After normalizing the data, considering the relatively poor training effect of 1D data input into the neural network, each data is folded and reconstructed into 2D data before input into the network, i.e., the 1000x1 1D data is converted into 40x25x1 2D data.

[0068] In this invention, after data preprocessing is completed, the model is pre-trained and the model parameters are adjusted. In addition to the fixed input and output dimensions, other parameters are usually adjusted based on the training and validation effects to find the optimal values. After parameter adjustment is completed, the model is trained. During training, data is input in batches, and the MSE between the training denoised data and the actual labels is calculated at each iteration, and the Adam optimizer is used to optimize the model parameters until the MSE converges. Finally, the model is saved after training.

[0069] In a further interpretation of the present invention, during the model training process, the training effect must be verified to avoid overfitting. Specifically, during multiple iterations, the preprocessed test set is simultaneously input into the noise suppression network with optimal parameters to obtain test data, and the test data is compared with the corresponding actual labels to obtain a validation error. When the validation error reaches convergence and then starts to grow again, it indicates that the network has reached an overfitting state, and the iteration is stopped.

[0070] As a further illustration of the present invention, when the mean squared error between the training denoised data and the actual labels is calculated for each iteration, an Adam optimizer is used for each iteration to adjust the parameters of the noise suppression network to minimize the mean squared error between the training data and the actual labels until the mean squared error converges.

[0071] In the above embodiment, the steps are described in the above order, but those skilled in the art will understand that in order to achieve the effects of this embodiment, different steps do not need to be performed in this order, and may be performed simultaneously (in parallel) or in the reverse order. All of these simple modifications are within the scope of protection of the present invention.

[0072] A second embodiment of the present invention is a noise suppression system for use with transient electromagnetic data based on a noise suppression method for use with transient electromagnetic data, comprising: The system is The method includes a network construction and training module, stacking the forward modeling data and the noise data based on different weights, forming a dataset after stacking, training a pre-constructed noise suppression network based on the dataset, and obtaining a trained noise suppression network, wherein the dataset includes a training set and a test set, and the different weights are obtained based on a preset target signal-to-noise ratio; A noise reduction module is included, and the TEM data to be noise-suppressed is input into a trained noise suppression network to obtain noise-suppressed TEM data; Here, the noise suppression network was constructed based on a recurrent neural network and an LSTM network.

[0073] Those skilled in the art can clearly understand that the specific operating procedures and explanations of the above-mentioned system can refer to the corresponding procedures in the above-mentioned method embodiments for convenience and conciseness of explanation, and will not be further described here.

[0074] It should be noted that the above embodiments provide a noise suppression system for transient electromagnetic data, which is described by taking only the division of each functional module as an example. In actual applications, the above functions can be distributed by different functional modules as needed. That is, the modules or steps in the embodiments of the present invention can be re-decomposed or combined, for example, the modules in the above embodiments can be integrated into one module, or further divided into multiple sub-modules to complete all or part of the above functions. The names of the modules and steps in the embodiments of the present invention are only for the purpose of distinguishing each module or step, and are not to be considered as an undue limitation of the present invention.

[0075] An electronic device according to a third embodiment of the present invention includes: At least one processor, and a memory communicatively coupled to the at least one processor; where: The memory stores instructions that are executed by the processor to implement the noise suppression method for use with transient electromagnetic data described above.

[0076] A computer-readable storage medium according to a fourth embodiment of the present invention, having computer instructions stored thereon, the computer instructions being executed by the computer to implement the noise suppression method for use with transient electromagnetic data as described above.

[0077] For convenience and simplicity of explanation, the specific operation procedures and explanations of the above-mentioned storage device and processing device may refer to the corresponding procedures in the above-mentioned method embodiments, and the explanations thereof will be omitted here.

[0078] Those skilled in the art will appreciate that the modules and method steps of each example described in connection with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. The programs corresponding to the software modules and method steps may be stored in random access memory (RAM), memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, a hard disk, removable media, a CD-ROM, or any other form of storage medium known in the art. To clearly illustrate the compatibility of electronic hardware and software, the above description generally describes the components and steps of each example functionally. These functions may be implemented in electronic hardware or software depending on the specific application and design constraints of the technology. Those skilled in the art will appreciate that different methods may be used to implement the described functions for each specific application, and this implementation is not considered to be beyond the scope of the present invention.

[0079] 4 shows a configuration diagram of a computer system for realizing a server of the method, system, and device embodiments of the present application. The server shown in FIG. 4 is merely an example and does not limit the functionality and scope of use of the embodiments of the present application.

[0080] As shown in Figure 4, the computer system includes a central processing unit (CPU) 401 that can perform various appropriate operations and processes in accordance with programs stored in a read-only memory (ROM) 402 or programs loaded from a storage unit 408 into a random access memory (RAM) 403. The RAM 403 also stores various programs and data necessary for the operation of the system. The CPU 401, ROM 402, and RAM 403 are interconnected via a bus 404. An input / output (I / O) interface 405 is also connected to the bus 404.

[0081] The I / O interface 405 includes an input unit 406 including a keyboard, a mouse, etc., an output unit 407 including a cathode ray tube (CRT), a liquid crystal display (LCD), a speaker, etc., a storage unit 408 including a hard disk, etc., and a communication unit 409 including a network interface card such as a LAN (Local Area Network) card and a modem. The communication unit 409 performs communication processing via a network such as the Internet. A driver 410 is also connected to the I / O interface 405 as needed. Removable media 411 such as a magnetic disk, optical disk, magneto-optical disk, or semiconductor memory is installed in the drive 410 as needed, and computer programs read from the storage unit 408 are installed as needed.

[0082] In particular, according to embodiments of the present disclosure, the processes described above with reference to the flowcharts can be implemented as a computer software program. For example, embodiments of the present disclosure include a computer program product, which includes a computer program carried on a computer-readable medium, the computer program including program code for executing the methods illustrated in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via the communication unit 409 and / or from removable media 411. When the computer program is executed by the central processing unit (CPU) 401, the functions defined in the methods of the present disclosure are performed. Note that the computer-readable medium described herein may be a computer-readable signal medium, a computer-readable storage medium, or any combination of the above. The computer-readable storage medium may be, for example, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. Note that the computer-readable medium described herein may be a computer-readable signal medium, a computer-readable storage medium, or any combination of the above. The computer-readable storage medium can be, for example, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of the computer-readable storage medium include, but are not limited to, an electrical connection having 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 compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.In this application, a computer-readable storage medium may be any tangible medium that contains or stores a program that can be instructed to execute or used in conjunction with a system, apparatus, or device. In this application, a computer-readable storage medium may be any tangible medium that contains or stores a program that can be instructed to execute or used in conjunction with a system, apparatus, or device. In contrast, in this application, a computer-readable signal medium may include a propagating data signal in baseband or as part of a carrier wave that carries computer-readable program code. Such propagating data signals may take various forms, including, but not limited to, electromagnetic signals, optical signals, or any suitable combination of the foregoing. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium that can transmit, propagate, or transmit a program used by or used in conjunction with an instruction execution system, apparatus, or device. The program code contained in the computer-readable medium may be transmitted over any suitable medium, including, but not limited to, wireless, electrical wire, optical cable, RF, etc., or any suitable combination of the foregoing.

[0083] Computer program code for carrying out the operations of the present application can be written in one or more programming languages ​​or combinations thereof. Such programming languages ​​include object-oriented programming languages ​​such as Java, Smalltalk, C++, "C," and general procedural programming languages. The program code can run entirely on the user computer, partially on the user computer, as a separate package, partially on the user computer and partially on a remote computer, or entirely on a remote computer or server. When referring to a remote computer, the remote computer can be connected to the user computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computer (e.g., via the Internet using an Internet Service Provider).

[0084] Flowcharts and block diagrams in the figures illustrate possible architectures, functions, and operations of systems, methods, and computer program products according to various embodiments of the present application. In this regard, each block in the flowcharts or block diagrams may represent a module, program segment, or portion of code, including one or more executable instructions for implementing a given logical function. It should also be noted that in alternative implementations, the functions displayed in the blocks may occur in a different order than the order displayed in the figures. For example, two blocks shown in succession may actually be executed essentially in parallel or may be executed in the reverse order depending on the functionality involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented in a dedicated hardware-based system that performs the given functions or operations, or may be implemented in a combination of dedicated hardware and computer instructions.

[0085] The terms "first," "second," etc. do not describe or represent a particular order or priority, but rather are intended to distinguish between similar objects.

[0086] The term "comprises" or any other similar term is intended to encompass a non-exclusive inclusion, such that a process, method, article or apparatus / device comprising a set of elements includes not only those elements but also other elements not expressly listed, or elements inherent in those processes, methods, articles or devices.

[0087] Although the technical aspects of the present invention have been described above in connection with the preferred embodiments shown in the accompanying drawings, it will be readily understood by those skilled in the art that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art may make equivalent modifications or substitutions to the relevant technical features, and all technical proposals after these modifications or substitutions fall within the protection scope of the present invention.

Claims

1. 1. A noise suppression method for use on transient electromagnetic data, comprising: The method comprises: Stacking the forward modeling data and the noise data according to different weights, forming a dataset after stacking, and training a pre-constructed noise suppression network based on the dataset to obtain a trained noise suppression network, where the dataset includes a training set and a test set, and the different weights are obtained according to a preset target signal-to-noise ratio; The TEM data to be noise-suppressed is input to a trained noise suppression network to obtain noise-suppressed TEM data; Wherein the noise suppression network is constructed based on a recurrent neural network and an LSTM network; The method for obtaining the data set is shown by the following formula: [Equation 19] Here, D ｆｉｎａｌ is the stacked data set, D ｉ ｆｏｒｗａｒｄ is the i-th forward modeling data, D ｉ ｎｏｉｓｅ is the i-th noise data, ω ｆｏｒｗａｒｄ and ω ｎｏｉｓｅ are weight values ​​of the forward modeling data and noise data, which are used to adjust the ratio of the forward modeling data to the noise data when stacking, and N is the amount of forward modeling data. ω ｆｏｒｗａｒｄ and ω ｎｏｉｓｅ is obtained by the following method: [Equation 20] [Equation 21] Here, SNR ｔａｒｇｅｔ is a preset target signal-to-noise ratio.

2. The method for constructing the dataset is as follows: constructing a plurality of different one-dimensional resistivity models and performing forward modeling; and obtaining forward modeling data corresponding to the plurality of one-dimensional resistivity models; Artificially simulating noise data of different types and different intensities, the noise data including atmospheric noise, power frequency noise and Gaussian noise; A plurality of the forward modeling data and a plurality of the noise data are stacked with different weights, and the stacked plurality of data are used as a data set; 2. The noise suppression method for transient electromagnetic data according to claim 1, wherein the data set is divided into a training set and a test set at a set ratio.

3. The pre-constructed noise suppression network includes an input layer, an encoder, a decoder, and an output layer connected in series; the input layer is used to receive input data including the dataset or the TEM data to be noise suppressed; The encoder and decoder are both constructed by two sequentially connected LSTM layers, and the encoder is used to learn and convert time-series feature data of the input data into abstract data; the decoder is used to learn the abstract data and restore the time-series feature data; 2. The noise suppression method for transient electromagnetic data according to claim 1, wherein the output layer includes a fully connected layer and a recurrent output layer connected in series, and the output layer is used to output the restored time-series feature data as noise-removed TEM data.

4. The training method of the noise suppression network after the training is as follows: pre-processing the training set and the test set, the pre-processing method including normalization and folding reconstruction; Input the pre-processed training set into a pre-constructed noise suppression network to obtain initial denoising data, adjust initial parameters of the noise suppression network according to the initial denoising data and the actual label, obtain optimal parameters, and replace the initial parameters of the noise suppression network with the optimal parameters; The preprocessed training set is input again in batches to the noise suppression network with optimal parameters to obtain training denoised data, and multiple iterations are performed. In each iteration, the mean squared error between the training denoised data and the actual labels is calculated, and the iterations are stopped until the mean squared error converges; The noise suppression method for transient electromagnetic data according to claim 3, wherein the noise suppression network after stopping the iteration is the trained noise suppression network.

5. The noise suppression method for transient electromagnetic data described in claim 4, characterized in that, during the process of multiple iterations, the preprocessed test set is simultaneously input into the noise suppression network with optimal parameters to obtain test data, and the test data is compared with the corresponding actual labels to obtain a verification error, and the iteration is stopped when the verification error reaches convergence and then begins to grow again.

6. 6. The noise suppression method for transient electromagnetic data of claim 5, wherein at each iteration, an Adam optimizer is used to adjust the parameters of the noise suppression network to minimize the mean squared error between the training data and the actual labels until the mean squared error converges.

7. The folding and reconstruction method is as follows: The noise suppression method for transient electromagnetic data described in claim 6, characterized in that the training set and the test set of one-dimensional data are collapsed and reconstructed into two-dimensional data, and the two-dimensional data is used as input to a pre-constructed noise suppression network.

8. A noise suppression system for use with transient electromagnetic data based on the noise suppression method for use with transient electromagnetic data according to any one of claims 1 to 7, comprising: The system is The method includes a network construction and training module, stacking the forward modeling data and the noise data based on different weights, forming a dataset after stacking, training a pre-constructed noise suppression network based on the dataset, and obtaining a trained noise suppression network, wherein the dataset includes a training set and a test set, and the different weights are obtained based on a preset target signal-to-noise ratio; A noise reduction module is included, and the TEM data to be noise-suppressed is input to a trained noise suppression network to obtain noise-suppressed TEM data; Wherein the noise suppression network is constructed based on a recurrent neural network and an LSTM network; The method for obtaining the data set is shown by the following formula: [Equation 22] Here, D ｆｉｎａｌ is the stacked data set, D ｉ ｆｏｒｗａｒｄ is the i-th forward modeling data, D ｉ ｎｏｉｓｅ is the i-th noise data, ω ｆｏｒｗａｒｄ and ω ｎｏｉｓｅ are weight values ​​of the forward modeling data and noise data, which are used to adjust the ratio of the forward modeling data to the noise data when stacking, and N is the amount of forward modeling data. ω ｆｏｒｗａｒｄ and ω ｎｏｉｓｅ is obtained by the following method: [Equation 23] [0000] Here, SNR ｔａｒｇｅｔ is a preset target signal-to-noise ratio.

Images